Citation
Risk Factors for Opioid Use Patterns Assessed through a Community Engagement Program in North Central Florida

Material Information

Title:
Risk Factors for Opioid Use Patterns Assessed through a Community Engagement Program in North Central Florida
Creator:
Serdarevic, Mirsada
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (180 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
Cottler,Linda B
Committee Co-Chair:
Gurka,Kelly
Committee Members:
Striley,Catherine L
Leeman,Robert Francis
Graduation Date:
5/3/2019

Subjects

Subjects / Keywords:
community -- florida -- opioid
Epidemiology -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Epidemiology thesis, Ph.D.

Notes

Abstract:
Prescription opioid use has reached epidemic levels in the United States with 130 individuals dying daily due to opioid overdose. Research investigating risk factors for prescription opioid use among community members is important to identify areas where prevention strategies can be implemented, helping to reduce the significant burden of overdose in the US. While there have been some studies that focus on risk factors for prescription opioid use, studies examining sex differences in prescription opioid use in the community with large sample sizes have been limited. Specifically, there has been a lack of attention on women in addressing prescription opioid use. Data from HealthStreet, a community outreach program that is based out of the University of Florida Clinical and Translational Science Institute, was used to assess risk factors for varying patterns of prescription opioid use (N=9,785). Through HealthStreet, Community Health Workers (CHWs) directly engage with community members in locations such as parks, libraries, health fairs, and grocery stores and assess their demographics, social determinants of health, medical conditions, and drug use through a health intake form (the Health Needs Assessment). Using a cross-sectional study design, the specific objectives of this dissertation research are to: 1) characterize prescription opioid use patterns and examine sex specific risk factors, among women, 2) examine risk factors for prescription opioid use patterns by age, and 3) identify geospatial clusters for prescription opioid use patterns. A socio-ecological model framework was used to examine individual, relationship, and community level factors which may provide additional risk of prescription opioid use. The outcomes of this dissertation will enhance our understanding of prescription opioid use patterns and risk factors for prescription opioid use among an out of treatment sample in the community. In addition, risk factors for prescription opioid use among women specifically were examined, which have been largely understudied. With this knowledge, we are better equipped for designing and targeting prevention strategies in reducing the harmful consequences of prescription opioid use in similar communities. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2019.
Local:
Adviser: Cottler,Linda B.
Local:
Co-adviser: Gurka,Kelly.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2020-05-31
Statement of Responsibility:
by Mirsada Serdarevic.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
5/31/2020
Classification:
LD1780 2019 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

RISK FACTORS FOR OPIOID USE PATTE RNS ASSESSED THROUGH A COMMUNITY ENGAGEMENT PROGRAM I N NORTH CENTRAL FLOR IDA By MIRSADA SERDAREVIC A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2019

PAGE 2

© 2019 Mirsada Serdarevic

PAGE 3

To Mama and Babo

PAGE 4

4 ACKNOWLEDGMENTS First, I would like to thank Dr. Linda B. Cottler for her mentorship throughout the PhD program and for pushing me to challenge myself. I have learned a great deal during my time in the Department of Epidemiology and the expertise I have gained from Dr.Cot tler will stay with me throughout my career. I also want to thank the other members of my dissertation committee. This dissertation would not have been possible without guidance from Dr. Catherine W. Striley, Dr. Kelly K. Gurka, and Dr. Robert F. Leeman. I especially want to thank my parents for their unconditional support and for leaving everything they had behind in search of better opportunities for my sisters and me. Without their support, my academic success would not have been possible. In addition, I thank my sisters, Suada and Medina, for always being kind and encouraging. I have made a number of friends and collaborated on numerous projects with my colleagues during my time in the Epidemiology Department, but through it all has been my childhood friend Selma Basic. I also want to thank Denisse Ballinas for her unwavering support throughout my studies and for taking time to visit me. To all the friends I have made during my time in the program Sadaf Milani, Vicki Osborne, Sonam Lasopa, Jae Min, Shivani Khan, Tessa Frohe, and everyone else from the Department of Epidemiology, thank you. My experience would not have been the same without you all. I would like to thank Dr. Tara L. Fazzino for her unconditional mentorship and Dr. Gail L. Rose who was the first person to give me a job in substance use research. Without them I would not have followed the career pat h that I am on. Finally, I would like to acknowledge Tupac Amaru Shakur and Afeni Shakur who sparked my initial interest in substance use research and health disparities.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURES ................................ ................................ ................................ ....................... 10 LIST OF ABBREVIATIONS ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................................ ... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 14 Prescription Opioid Use within the Community ................................ ................................ .... 16 Sex Differences in Prescription Opioid Use ................................ ................................ ........... 18 Frequent Emergency Department Utilization and Prescription Opioid Use .......................... 19 Theoretical Framework ................................ ................................ ................................ ........... 20 Aims and Hypotheses ................................ ................................ ................................ ............. 24 Aim 1 ................................ ................................ ................................ ............................... 24 Aim 2 ................................ ................................ ................................ ............................... 26 Aim 3 ................................ ................................ ................................ ............................... 27 Potential Implications ................................ ................................ ................................ ............. 28 2 DATA SOURCE ................................ ................................ ................................ .................... 33 HealthStreet ................................ ................................ ................................ ............................ 33 Recruitment ................................ ................................ ................................ ..................... 33 Health Needs Assessment ................................ ................................ ................................ 34 Variables ................................ ................................ ................................ ................................ . 35 Outcome: Prescription Opioid Use Patterns ................................ ................................ .... 36 Individual level Variables ................................ ................................ ............................... 36 Relationship level Variables ................................ ................................ ........................... 37 Community level Variables ................................ ................................ ............................. 37 Preliminary Studies ................................ ................................ ................................ ......... 38 3 SEX DIFFERENCES IN PRESCRIPTION OPIOID USE PATTERNS ASSESSED THROUGH A COMMUNITY ENGAGEMENT PROGRAM IN NORTH CENTRAL FLORIDA ................................ ................................ ................................ ............................... 46 Background ................................ ................................ ................................ ............................. 46 Aims and Hypotheses ................................ ................................ ................................ ............. 50 Methods ................................ ................................ ................................ ................................ .. 51 Study Population ................................ ................................ ................................ ............. 51

PAGE 6

6 Measurements ................................ ................................ ................................ .................. 52 Analysis ................................ ................................ ................................ ........................... 53 Results ................................ ................................ ................................ ................................ ..... 55 Characteristics among all community members, stratified by prescription opioid use pattern ................................ ................................ ................................ .. 56 Characteristics and prescription opioid use among HealthStreet members, stratified by sex ................................ ................................ ................................ ..... 58 Risk factors and self reported prescription opioid use pattern among HealthStreet members ................................ ................................ ........................... 59 Risk factors and self reported prescription opioid use pattern among HealthStreet members, stratified by sex ................................ ............................... 60 Discussion ................................ ................................ ................................ ............................... 63 Strengths and Limitations ................................ ................................ ................................ 68 Conclusions ................................ ................................ ................................ ..................... 70 4 PATTERNS OF PRESCRIPTION OPIOID USE AND RISK FACTORS FOR USE AMONG OLDER AND YOUNGER WOMEN IN THE COMMUNITY ............................ 83 Background ................................ ................................ ................................ ............................. 83 Aims and Hypotheses ................................ ................................ ................................ ............. 88 Methods ................................ ................................ ................................ ................................ .. 89 Study Population ................................ ................................ ................................ ............. 89 Me asurements ................................ ................................ ................................ .................. 90 Analysis ................................ ................................ ................................ ........................... 91 Results ................................ ................................ ................................ ................................ ..... 93 Characteristics of women, stra tified by prescription opioid use pattern .................. 94 Characteristics and prescription opioid use among women, stratified by age ......... 95 Risk factors and self reported prescription opioid use pattern among women ........ 96 Risk factors and self reported prescription opioid use pattern among women, stratified by age ................................ ................................ ................................ ..... 98 Discussion ................................ ................................ ................................ ............................. 100 Strengths and Limitations ................................ ................................ .............................. 104 Conclusions ................................ ................................ ................................ ................... 106 5 GEOSPATIAL CLUSTERS OF PRESCRIPTION OPIOID USE AMONG WOMEN IN A COMMUNITY SAMPLE ................................ ................................ ........................... 119 Background ................................ ................................ ................................ ........................... 119 Aims and Hypotheses ................................ ................................ ................................ ........... 124 Methods ................................ ................................ ................................ ................................ 125 Study Population ................................ ................................ ................................ ........... 125 Measurements ................................ ................................ ................................ ................ 126 Analysis ................................ ................................ ................................ ......................... 127 Results ................................ ................................ ................................ ................................ ... 129 Factors by ED utilization among lifetime not past 30 opioid users ....................... 129 Geospatial cluster analysis among lifetime opioid users ................................ ....... 130 Geospatial cluster analysis among past 30 day opioid users ................................ . 131

PAGE 7

7 Discussion ................................ ................................ ................................ ............................. 131 Strengths and Limitations ................................ ................................ .............................. 137 Conclusions ................................ ................................ ................................ ................... 138 6 FINAL CONCLUSIONS ................................ ................................ ................................ ...... 147 Main Findings ................................ ................................ ................................ ....................... 148 Future Work ................................ ................................ ................................ .......................... 152 APPENDIX: INTERACTION TERMS ................................ ................................ ...................... 155 LIST OF REFERENCES ................................ ................................ ................................ ............. 159 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 179

PAGE 8

8 LIST OF TABLES Table page 2 1 Individual level variables and the corresponding HealthStreet health assessment questions from which they were derived ................................ ................................ ........... 43 2 2 Relationship level variables and the corresponding HealthStreet health assessment questions from which they were derived ................................ ................................ ........... 45 2 3 The community level variable and the corresponding HealthStreet health assessment question from which it was derived ................................ ................................ ................... 45 3 1 Association between individual , relationship , and community level characteristics of prescription opioid use among HealthStreet members, 2011 2018, N=9,221. ........... 74 3 2 Association between individual , relationship , and community level characteristics and prescription opioid use among HealthStreet members , stratified by sex, 2011 2018, N=9,221. ................................ ................................ ................................ .................. 76 3 3 Association between select risk factors and self reported prescription opioid use patt ern among HealthStreet members, 2011 2018 (N = 9,221) ................................ ........ 79 3 4 Association between select risk factors and self reported prescription opioid use pattern among HealthStreet members, by sex, 2011 2018 (N = 9,221) ........................... 81 4 1 Association between individual , relationship , an d community level characteristics of prescription opioid use among women enrolled in HealthStreet, 2011 2018, N=5,549. ................................ ................................ ................................ .......................... 110 4 2 Associ ation between individual , relationship , and community level characteristics and prescription opioid use among women enrolled in HealthStreet, stratified by age, 2011 2018, N=5,549 ................................ ................................ ................................ ..... 112 4 3 Association between various risk factors and self reported prescription opioid use among women enrolled in HealthStreet, 2011 2018 N=5,549 ................................ ........ 115 4 4 Association between select risk factors and self reported prescription opioid use pattern among women, by sex, 2011 2018 ................................ ................................ ..... 117 5 1 HealthStreet characteristics by opioid use pattern and self report frequent ED utilization, (n=3,098) ................................ ................................ ................................ ....... 141 5 2 HealthStreet characteristics by opioid use pattern among frequent ED utilizers, (n=618) ................................ ................................ ................................ ............................. 143 A 1 Table 3 3 with interaction terms fitted into the model ................................ .................... 155

PAGE 9

9 A 2 Table 4 3 with interaction terms fitted into the model ................................ .................... 157

PAGE 10

10 LIST OF FIGURES Figure page 1 1 The Socio ecological Model ................................ ................................ .............................. 31 1 2 HealthStreet variables to be examined within the context of the socio ecological model ................................ ................................ ................................ ................................ .. 32 2 1 An infographic depicting the HealthStreet mission to community members who enroll in HealthStreet ................................ ................................ ................................ ......... 40 2 2 Enrollment in HealthStreet and eligibility for current study, November 2011 June 2018. ................................ ................................ ................................ ................................ ... 41 2 3 Portion of the HealthStreet Health Needs Assessment instrument on which interview ................................ ................................ ................................ ............................ 42 3 1 The frequency of the distribution of age among HealthStreet members ........................... 72 3 2 The frequency of the distrib ution of the number of children among HealthStreet members ................................ ................................ ................................ ............................. 73 3 3 Enrollment in HealthStreet and eligibility for current study, as of Ju ne 2018 .................. 73 4 1 The frequency of the distribution of age among HealthStreet female members ............. 108 4 2 The frequency of the distribution of the number of children among HealthStreet female members ................................ ................................ ................................ ............... 108 4 3 Enrollment in HealthStreet and eligibility for women in current study, as of June 2018. ................................ ................................ ................................ ................................ . 109 5 1 Women enrolled in HealthStreet and ED utilization by self reported prescription opioid use ................................ ................................ ................................ ......................... 140 5 2 Map representing cluster s of frequent ED utilization among lifetime but not past 30 day prescription opioid users. ................................ ................................ .......................... 145 5 3 Map representing cluster s of frequent ED utilization among past 30 day users. ............ 146

PAGE 11

11 LIST OF ABBREVIATIONS AIC Akaike information criterion aOR Adjusted odds ratio CE Community engagement CHWs Community Health Workers CI Confidence interval ED Emergency department LRT Likelihood ratio test N Sample size NAS Neonatal abstinence syndrome NMU Non medical use PDMP Prescription drug monitoring program Ref Reference group Rx opioids Prescription opioids SD Standard deviation US United States VIF Variance inflation factor

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RISK FACTORS FOR OPIOID USE PATTE RNS ASSESSED THROUGH A COMMUNITY ENGAGEMENT PROGRAM I N NORTH CENTRAL FLOR IDA By Mirsada Serdarevic May 2019 Chair: Linda B. Cottler Major: Epidemiology Prescription opioid use has reached epidemic leve ls in the United States with 130 individuals dying daily due to opioid overdose. Research investigating risk factors for prescription opioid use among community members is important to identify areas where prevention strategies can be implemented, helping to reduce the significant burden of overdose in the US. While there have been some studies that focus on risk factors for prescription opioid use, s tudies examining sex differences in p rescription opioid use in the community with large sample sizes have been limited. Specifically, there has been a lack of attention on women in addressing prescription opioid use . Data from HealthStreet, a community outreach program that is based out of th e University of Florida Clinical and Translational Science Institute , was used to assess risk factors for varying patterns of prescription opioid use (N=9,785 ). Through HealthStreet, Community Health Workers (CHWs) directly engage with community members in locations such as park s , libraries, health fairs, and grocery store s and assess their demographics, social determinants of health, medical conditions, and drug use through a health intake form (the Health Needs Assessment).

PAGE 13

13 Using a cross sectional study design, the specific objectives of this dissertation research are to : 1) characterize prescription opioid use patterns and examine sex specific risk factors , among women, 2) examine risk factors for prescrip tion opioid use patterns by age, and 3) identify geospatial clusters for prescription opioid use patterns. A socio ecological model framework was used to examine individual, relationship, and community level factors which may provide additional risk of prescription opioid use. The outcomes of this disse rtation will enhance our understanding of prescription opioid use patterns and risk factors for prescription opioid use among a n out of treatment sample in the community. In addition, risk factors for prescription opioid use among women specifically were e xamined, which have been largely understudied . With this knowledge, we are better equipped for designing and targeting prevention strategies in reduci ng the harmful consequences o f prescription opioid use in similar communities .

PAGE 14

14 CHAPTER 1 INTRODUCTION Prescription opioids ( e . g . OxyContin ® , Vicodin ® Darvocet ® , Lortab ® , Percocet ® , oxycodone, and hydrocodone) are analgesics typically prescribed to t reat both chronic and acute pain (Beaudoi n, Lin, Guan, & Merchant, 2014; Caudill Slosberg, Schwartz, & Woloshin, 2004) , though alternative euphorigenic motivations for use other than pain exist (McCabe, West, & Boyd, 2013c) . Prevention efforts for prescription opioid use have focused on prescribing and d ue to the euphoric effects produced by opioids and the possibility of psychological or physical dependence, the Drug Enforceme nt Administration (DEA) has listed most prescription opioids as controlled substances. Schedule I drugs have a high potential for abuse and are defined by the DEA as drugs with no current ly accepted medical use such as heroin , also an opioid . Similar to Sc hedule I drugs, Schedule II drugs have high abuse potential. Examples of Schedule II drugs include Vicodin ® , cocaine, and Adderall ® . While the abuse potential for Schedule III drugs is less than Schedule I and Schedule II, Schedule III drugs still have a moderate to low potential for physical and psychological dependence and include drugs such as Tylenol with codeine. Schedule IV drugs have low potential for a buse and dependence and include drugs such as Ativan, Ambien and Tramadol (Drug Enforcement Administration, 2017 ) . In spite of their abuse potential, opioids attained a peak prescription rate in 2012 (Centers for Disease Control and Prevention, 2018e) and prescriptions more than doubled in the US, from 107.3 million prescriptions written in 1992 to 246.2 million in 2015 (Pezalla, Rosen, Erensen, Haddox, & Mayne, 2017) . Higher prescribing rates subsequently led to higher rates of use. An estimated 17.4% of the U.S. population received one or more opioid prescription in 2017 (Centers for Disease Control and Prevention, 2018a) . More recently, the overall national prescribing rate has decreased to the lowest it has been in the last decade to 58.5 prescriptions

PAGE 15

15 per 100 persons from 81.3 per 100 persons , however certain regions across the US continue to have high prescribing rates (Centers for Disease Control and Prevention, 2018e) . P rescription opioid use is also a concern due to the potential for non medical use (i.e., u se of higher doses, use longer than prescribed, or use of someone else's medication (McCabe, West, & Boyd, 2013b; Osborne, Serdarevic, Crooke, Striley, & Cottler, 2017) . T he non medical use of prescription opioids has been identified as a risk factor for heroin use, which is also an important public health concern as deaths related to heroin use have increased over fivefold between 2010 and 2016 (Centers for Disease C ontrol and Prevention, 2018c; Cerdá, Santaella, Marshall, Kim, & Martins, 2015; Compton, Jones, & Baldwin, 2016a; National Institute on Drug Abuse, 2015; Palamar, Shearston, Dawson, Mateu Gelabert, & Ompad, 2016) . Overdose deaths from opioids (including prescription opioids , heroin , and fentanyl ) in the US have risen sharply; approximately 70,200 people died from a drug overdose in 2017, with 68% of all overdose deaths involving a n opioid (Centers for Disease Control and Prevention, 2018b) . Current estimates indicate that approximately 130 people in the US die every day from drug overdose involving an opioid (Centers for Disease Control and Prevention, 2018g, 2018h; Christie et al., 2017) . Specifically, in Florida, emergency department (ED) visits and fatal overdose deaths related to synthetic opioids have increased substantially (Prekupec, Mansky, & Baumann, 2017) . Data from the Healthcare Cost and Utilization Pro ject (HCUP) shows Florida had approximately 71 ED visits per 100,000 population in 2008 related to opioid use, whereas in 2016 there were 217 ED visits per 100,000 population (HCUP, 2018) . Though deaths related to fentanyl use increased in the entire nation, deaths increased 250% b etween 2010 and 2014 in Florida (Aschenbrenner, 2017) . More recently, data from the Florida Drug Related Outcomes

PAGE 16

16 Surveillance and Tracking System (FROST) reporte d that there were 1,685 deaths that in c luded a fentanyl analog in 2017 (FROST, 2018) . While fatal overdoses are obviously detrimental, there are other consequences related to prescription opioid use . They have put a large burden on health care resources used to combat this epidemic (affecting o r tending to affect a large number of individuals within a population, community, or region at the same time; Merriam Webster , 2018) . The National Institute on Drug Abuse ( NIDA) estima tes that for every death due to prescription drug overdose there are an additional 22 treatment admissions and 119 emergency room visits (National Institute on Drug Abuse, 2016) . The estimated annual healthcare cost ( excess medical costs, substance abuse treatment, prevention, and research costs ) associated with prescription opioid abuse is estimated to be $78.5 billion (National Institute on Drug Abuse, 2019) , which highlights the need to reduce both the economic and personal burden of the current opioid epidemic and generate evidence to design effective prevention strategies. To reduce this burden related to the use of opioids and to inform prevention efforts, information from communities with high rates of prescription opioid use need s to be examined. Prescription Opioid Use within the Community Though prescription opioid use and consequences related to use have in creased in the last 15 years, limited studies have examined data from a community sample and have focused instead on data at the national level. Data from national agencies such as the C enters for Disease Control and Prevention (CDC) and Substance Abuse and Mental Health Services Administration (SAMHSA) have examined prescription opioid use extensively. An estimated 6.9% of the US adult population used prescription opioids in the past 30 day s (Centers for Disease Control and Prevention, 2015) , and 38 % of the US adult population used prescription opioids in the past 12 months (Han et al., 2017a) . These national studies contribute to the field by providing an

PAGE 17

17 overview of prevalence rates; however they do not provide information on the fluctuations in prevalence among community dwelling people within specific geographic regions. The burden of the opioid crisis is not evenly distributed throughout the country; certain regions are experiencing far higher rates of overdose than the national average (Behe shti et al., 2015; Rudd et al., 2014) . Aside from the fatal consequences of prescription opioid us e, there are other consequences which may impact the community. For example, the increased burden on the healthcare system by those prescribed opioids in a specific community (e.g. utilization of emergency care services to obtain prescriptions) could lead to poor health outcomes for all community members who require emergency care at hospitals for more urgent medical conditions, regardless of whether they us e prescription opioids (Trzeciak & Rivers, 2003) . There is also the issue of safely disposing of prescription opioids that are no longer needed, that may contaminate the water suppl y if they are flushed down the toilet (Bates, Laciak, Southwick, & Bishoff, 2011; Rodayan et al., 2016) . Because there are limited ways to safely dispose of prescription opioids, unused prescriptions left in , for example, could result in high levels of prescript ion opioids being readily available or diverted to friends and relatives who were not prescribed them (Inciardi, Surratt, Kurtz, & Cicero, 2007; Ruhoy & Daughton, 2008) or they could be used to treat pain symptoms in the future . Thus, more granul ar and specific data from a community setting is necessary in orde r to better understand prescription opioid use. In addition to the focus on national level data, recent studies have focused on opioid depe ndent populations in treatment. R elatively few individuals who are in need of treatment for substance use receive tr eatment. In 2016, an estimated 48.5 million people in the US reported use of illicit drugs or misuse of prescription drugs in the past year, however in 2016 only 2.2 million

PAGE 18

18 people reported that they had received any treatment in the past year for illicit drug use or misuse of prescription drug s (Centers for Disease Control and Prevention, 2018a) . Individuals who are in the need of treatment but do not seek treatment may not be included in national studies or in studies that focus on populations in treatment. Therefore , data collected from a sample of community dwelling people can be used to represent those who are not traditionally represented in research a nd a sample not specifically selected for its access to car e . This suggests that data from multiple levels , including the community level , are needed in order examine prescription opioid use . Sex Differences in Prescription Opioid Use General risk factors for prescription opioid use regardless of sex have been examined. Known risk factors for prescription opioid use include older age (those 40 years of age and older are more likely to use prescription opioids than adults 20 39 years of age ) that may be due to chronic pain which is age relat ed , and race /ethnicity ( whites more likely to use prescription opioids than other races ; CDC, 2017) which is related to the disparities in health care and access to opioid prescriptions between races. In addition, pain (Blanco et al., 2016) and specifically cancer pain (Pinkerton & Hardy, 2017) are indicators for prescription opioid use . P rescription opioid use may occur due to influences that change behavior s which differ by sex. Thus, examining sex differences may inform us of risk factors specific to sex (Serdarevic, Striley, & Cottler, 2017) . The few s tudies that have focused on sex differences in prescription opioid use have found that women are more likely to use p rescription opio ids than men (Manubay et al., 2015; Serdarevic, Striley, et al., 2017) , partially attributed to women being more likely to report pain and subse quently being prescribed opioids. Specifically, w omen are 50% more likely to be prescribed opioids than men in the US (Simoni Wastila, 2000) and to generally use prescription

PAGE 19

19 drugs more frequently than men (Manteuffel et al., 2014; Olfson, King, & Schoenbaum, 2015) . Other studies have found men are more likely than women to overdose on drugs (Unick, Rosenblum, Mars, & Ciccarone, 2013) , but this gend er gap is closing as opioid related overdoses are increasing among women (Unick et al., 2013) . This emp hasizes the importance of examining prescription opioid use and risk factors for us e among women and men separately so we can reduce health dipartites related to opioid use. In addition, there are physiologic differences including the rate of drug absorption and metabolism between men and women that affect drug activity (Whitley & Lindsey, 2009) . The negative consequences such as risk of addiction, and vulnerability to adverse outcomes related to prescription opioid use are grea ter for women than men (Darnall, St acey, & Chou, 2012) . Regardless of these health disparities related to prescription opioid use among women, there has been a lack of studies and publications regarding women . W omen differ physiologically and socially from men regarding prescription opioid use, which may put women at greater risk for non medical use. Frequent Emergency Department Utilization and Prescription Opioid Use As described above, opioid prescribing rates have decreased nationally, though there are regions in the US including Florida that continue to have high rates. This may partially be due to high prescribing rates in EDs . Specifically, from 1999 to 2005, there was an increase in pain related complaints in US EDs (from 23% of visits to 37% of visits) which was associated wi th a 1.5 fold increase in opioid prescribing (Cantrill et al., 2012) . More recently, studies have found that despite the increase in adverse events rel ated to prescription opioid use, there is wide variation in opioid prescribing rates among emergency medicine physicians (B arnett, Olenski, & Jena, 2017; Hoppe, McStay, Sun, & Capp, 2017) . Though studies have shown that EDs are an inadequate resource for pain management (Baehren et al., 2010) b ecause chronic pain requires

PAGE 20

20 extended treatment, utilization of the ED for pain management may be more common in communities in which significant portions of the population do not have health insurance and consequently do not seek care from a primary care provider. Pain management through primary care is likely to be far more effective and may result in a lower demand for pain medications through the ED (Dowell, Haegerich, & Chou, 2016; Elder et al., 2017; Schneiderhan, Clauw, & Schwenk, 2017) . Appropriate pain manage ment is important , especially considering the high rates of opioid misuse and dependence the US is currently experiencing. S ince analyses at the national level may mask important findings at the community level, i t is important to understand the relations hip between access to care and prescription opioid use , especially within community settings. G eospatial methods can be used to identify clusters of prescription opioid use and examine the association between geographic clusters and ED utilization. For example, one study identified cancer hot spots through spatial statistics in Florida. Results from thi s study found disparities by minority status in regard geographic differences , which further helped inform community intervention efforts (Ruktanonchai, Pindolia, Striley, Odedina, & Cottler, 2014) . Another study conducted in Massachusetts that focused on prescription opioid use identified hot spots of opioid related emergency needs (i.e., n aloxone distribution) through spatial statistics (Dworkis, Weiner, Liao, Rabickow, & Goldberg, 2018) . Si milar methods can be used to examine prescription opioid use and identify cluster s of frequent ED users within a community setting. Theoretical F ramework To better understand sex differences and geographical variation in patterns of prescription opioid use, it is important to identify risk factors across multiple levels. Prescription opioid use cannot be understood by individual le vel risk factors alone. Rather, examination of risk factors f or prescription opioid use need to focus on interpersonal, organ izational,

PAGE 21

21 community, and policy level factors at the same time. By considering multiple levels, prevention efforts are more likely to be sustained over time compared to using single level interventions (Davis, Dumas, Wagner, & Merrin, 2016; Richard, Potvin, Kishchuk, Prlic, & Green, 1996; Stokols, 1996) . In order to examine prescription opioid use at multiple levels within a communit y sample, this dissertation utilize d the socio ecological model to examine risk factors for varying patterns of prescription opioids on the individual, relationship, and community level. There are many theories that attempt to explain substance use behavior ; however, this dissertation will focus on the socio ecological model , which is commonly used for health prevention. This multi level mode l can be used to understand individual, relationship , and community level factors (CDC, 2017) . Individual factors include characteristics such as knowledge, attitudes, sex , age, religious identity, race/ethnicity, sexual ori entation, and socio economic status . Relationship fact ors include social networks, social support systems , and employment (i.e., family, friends, peers, co workers, cu stoms or traditions). While community level factors include relationships among organizations, institutions, and the built environment ( i.e., parks, transportation, etc.). The s ocio ecological model has evolved over time and has been adapted based on the conceptual framework proposed by Urie Bronfenbrenner (Bronfenbrenner, 1977; McLeroy, Bibeau, Steckler, & Glanz, 1988) original ecological model explains that behavior is affected by and, in turn, affect s multiple levels of influence through environmental factors on micro , meso , exo , and macro system levels (Bronfenbrenner, 1977) . Jay early work that included both individual and family factors , was used to further a framework to account for individual, family, social, and cultural influences (Belsky, 1980) . Thus, the socio ecological model that is most

PAGE 22

22 commonly used aimed for health prevention including substance use . Multiple studies have e xa mined substance use within the context of the socio ecological model. For example, risk factors in the individual, family, peer, and community domains were examined to predict substance use patterns among youth (Connell, Gilreath, Aklin, & Brex, 2010) . One study assess ed risk and protective factors at the individual, parental, peer and neighborhoo d level for predictors of substance use among juvenile offenders using the soci o ecological f ramework (Davis et al., 2016) . Anoth er study used this theoretical model to examine risk factors for non medical opioid use behavior among youth (Osborne, Striley, Nixon, Winterstein, & Cottler, 2019) . T he rings in the mode l , depicted in Figure 1 1 , illustrate the different levels of the sociological model. In addition, work conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA), uses this model as their framework for prevention strategies for non medical use of prescription opioids (Substance Abuse and Mental Health Services Administration, 2016b) . This model is used to illustrate that people are not solely influenced by their own individual traits but also by their rel ationships with others and the community where they live (Substance Abuse and Mental Health Services Administration, 2016b) . Altho ugh all levels of the socio ecological model can be applied to th e use of prescription opioids, this dissertation will examine risk factors for prescription opioid use at the individual, relationship, and community levels only (Figure 1 2 ). The first level (individual) consists of biological and personal factors that are hypothesized to increase the likelihood of prescription opioid use. One of most important factor s to be examined at this level is sex. Other factors to be examined at this level include social determinants of health ( i.e., access to educational opportunities and health care services) , chronic

PAGE 23

23 health conditions, and other substance use. The findings of this dissertation research can inform interventions that target this level through educa tion and behaviors that prevent prescription opioid use. The second level (relationships) consists of relationships that either increase or decrease risk of the outcome . Factors to be examined at this level are marital status, having children, employment, and access to social media . Based on the findings from this dissertation, prevention strategies could focus on designing programs that build social support and integrate social media and other communication efforts regarding prescription opioid use. The t hird level in the social ecological model is community, which focuses on the settings in which individual behaviors and social relationships occur. We have interpreted this to be rural area of residence through zip code . A number of studies have found that health is a ffected and influenced by geographic location (Dummer, 2008; Dwyer Lindgren et al., 2017; Graham, 2016) . Similarly, rural area of residence has been identified as a risk factor for prescriptio n opioid use (Cicero, Surratt, Inciardi, & Munoz, 2007; Keyes, Cerdá, Brady, Havens, & Galea, 2014) , which can be further ex amined using spatial statistics. B ased on this information, prevention efforts and harm reduction can be focused on specific geographic regions within a community that m ay have higher rates of prescription opioid use compared to areas in the community with lower rates of use . Finally, the fourth level of the model is the societal level, which includes factors such as health and education policy . Policy changes that may influence opioid use include the implementation of prescription drug monitoring programs (PDMPs ) . E FORC S E ( Electronic Florida Online Reporting of Controlled Substance Evaluation Program electronic databases used to provide information to physicians and to identify individuals at

PAGE 24

24 increased risk to misuse or abuse opioid s . This informati on can be used to help target responses to epidemics (Centers for Disease Control and Prevention, 2017d) . Utilization of PDMPs has been associated with fewer opioid related overdose deaths (Patrick, Fry, Jones, & Buntin, 2016) . In Florida, u tilization of E FORCSE among emergency medicine physicians has been low (H. W. Young, Tyndall, & Cottler, 2017) . However, enhancements to PDMPs and accessibility can lead to an increase in utilization and subsequent reduction in opioid prescribing. Education policy includes recomm endations for prescribing opioids from na tional agencies such as the CD C which can assess risks and address harm from opioid use among patients (Dowell et al., 2016) . H owever, since we do not have data on the societal level , this dissertation focus es on the first three levels only individual, rel ationship, and community levels. Aims and Hypotheses The specific objective s of this dissertation were to 1) characterize opioid use patterns and examine risk factors by sex. Among women only, 2) examine risk factors for prescription opioi d use patterns for older and younger women, and 3) identify geospatial clusters for prescription opioid use . To achieve these objectives, data from a community sample in North Central Florida were analyzed. A socio ecological model was used to form the fra mework for constructs that will examine individual, relationship, and community level factors , which will provide additional information on risk of prescription opioid use. Aim 1 Characterize prescription opioid use patterns by sex [ past 30 day use, life time but not past 30 day use, and no use of prescription opioids] among adult (>18 years) community members and examine risk factors for these prescription opioid use patterns at the individual, relationship, and community level.

PAGE 25

25 This analysis quantified l ifetime and recent prescription opioid use in the community as well as sex differences in lifetime and recent opioid use. Previous research in the US has shown that women are more likely than men to be prescribed opioids in the US (Centers for Disease Control and Prevention, 2015 ; Han et al., 2017a; Simoni Wastila, 2000) and generally use prescription drugs more frequently compared to men (Manteuffel et al., 2014; Olfson et al., 2015) . W e hypothesize d that community dwelling women would be more likely than men to endorse prescription opioid use . The increased rate of opioid prescription use for women may be due to increased subjective pain ratings because pain sensitivi ty varies between men and women, with w omen reporting higher incidence of chronic conditions that cause pain (Darnall et al., 2012) which corresponds to more reports of pain, com pared to men (Barsky, Peekna, & Borus, 2001; Robinson et al., 2001) . Further, chronic pain is common among older adults which accounts for studies reporting an age related increase in the prevalence of pain . Thus, we theorize d that risk facto rs would differ by sex and age , and that i ndividual level risk factors , specif ically, self reported pain, would be associated with pre scription opioid use among both men and women, but would have a greater magnitude of effect among women compared to men . We hypothesized that : Hypothesis 1.1 . Women will be more likely to endorse past 30 day us e and lifetime, but not past 30 day use (vs. never use ), than men, after controlling for factors at the individual, relationship, and community level. Hypothesis 1.2. Lifetime prescription opioid use (both past 30 day use and lifetime, but not past 30 day use ) will have a stronger association among older individuals (50+ years) than among younger individuals (18 49 years), after adjusting for factors at the individual, relationship, and community level.

PAGE 26

26 Hypothesis 1. 3. Self reported p ain will be the strongest risk for opioid use a mong both men and women; pain will be more strongly associated with prescription opioid use among women compared to men. Aim 2 S tratifying by age [younger ( < 49 years) vs older ( > 50 years)] among women only , examine individual, relationship, and community l evel risk factors for prescription opioid use patterns. This analysis examine d risk factors for prescription opioid use among women only , stratified by age. Given that women are more likely to be prescribed and use prescription opioids, they subsequently carry a heavier burden of the consequences related to prescription opioid use compared to men Serdarevic, Striley, et al., 2017) . Thus, prescription opioid use among women , specifically, should be further examined . More specifically, studies have foun d that older women had higher prevalence of continued opioid use compared to younger women (Campbell et al., 2010) . In aim 2.1, we hypothesize d that older age w ould be related t o prescription opioid use, especially among women. In this aim focusing on women only , we hypothesize d that prescription opioid use (both past 30 day us e and lifetime, not past 30 day use ) would be more prevalent among older women than younger women. Chronic conditions are strong predictors of prescription opioid use , and women are more likely to be prescribed opioids (Barsky et al., 2001; Darnall et al., 2012; Robinson et al., 2001) . In ad dition, women have a higher prevalence of comorbid health conditions (Darnall et al., 2012; Malatesta, 2007) , which require different prescription drugs fo r treatment. For example, sedatives are more commonly prescribed to women than men (Kroll, Nieva, Barsky, & Linder, 2016) , and are even more commonly prescribed among older women compared to younger

PAGE 27

27 women (Nordon, Akamine, Novo, & Hübner, 2009) seeking treatment for mental h ealth conditions such as anxiety and depression. Due to chronic conditions that develop over the lifespan that require prescription drugs for treatment , we hypothesize that individual level risk factors for pre scription opioid use including painful chronic conditions , prescription sedative use, and substance use wil l differ among older and younger women. Two hypotheses are proposed for the second aim among women : Hypothesis 2. 1. Risk factors at the individual level (chronic conditions and co morbid substa nce use) compared to risk factors at the relationship level will have a stronger association for prescription opioid use (both for past 30 day use and lifetime, not past 30 day use ) among older women compared to younger women. Hypothesis 2. 2. Prescription sedative use will be the strongest risk factor for prescription opioid use among both older and younger women; the strength of association between sedative use and prescription opioid use will be stronger among older women compared to younger women. Aim 3 Identify community clusters of frequent female ED users (frequent ED user >2 vs non frequent user <2 in the past 6 months ) among women who have endorsed lifetime prescription opioid use (both past 30 day use and lifetime, not past 30 day use ) . This analysis identified clusters of past 30 day and lifetime ( not past 30 day use ) prescription opioid use by frequency of ED utilization. Using longitude and latitude point data corresponding to home address , spatial statistics were used to identify and analy ze clusters. We examined t he relationship between prescription opioid use and frequent ED utilization within a community because frequent ED use is a known risk factor for prescription opioid use (Doran, Raven, & Rosenheck, 2013; Mazer Amirshahi, Mullins, Rasooly, Anker, & Pines, 2014) . The

PAGE 28

28 increase in opioid prescribing rates in the US in the past two decades may partially be attributed to emergency medicine (Cantri ll et al., 2012) . Despite the adverse events related to prescription opioid use, prescribing rates in EDs remains high in certain regions of the country (Levy, Paulozzi, Mack, & Jones, 2015; Mazer Amirshahi et al., 2014) . Furthermore, a number of studies have shown that trea tment for pain in the ED by physicians is not adequate or appropriate (Baehren et al., 2010; Cordell et al., 2002) . However, frequent ED visits ( 4 or more visits in a one y ear period) account for 21% to 28% of all ED visits (LaCalle & Rabin, 2010) . Although appropriate pain management is important, opioid use in the U S is affecting a large number of people utilizing healthcare services . Thus, we hypothesized that ED visits would be significantly associated with prescription opioid use and women who endorsed recent opioid use and frequent ED visits would geospatially cluster within the community . I n aim 3, geospatial analyses were used to identify clusters of participants and to map prescription opioid use among women in the community b y their ED utilization. There is one hypothesis proposed for the third aim among women who have endorsed prescription opioid use i n the past 30 day s and lifetime, but not past 30 day s : Hypothesis 3. 1. W omen who are both frequent ED users and who used prescription opioids in the past 30 day s will geospatially cluster more compared to women who were frequent ED users and who used in their lifetime, not past 30 day use. Potential Implications As outlined above, this project will help increase our und erstanding of prescription opioid use and risk factors for prescription opioid s within a community dwelling sample . Specifically, this dissertation will examine lifetime and recent prescription opioid use within a community setting in North Central Florida, an area where opioid prescribing is common and is not necessarily reflective of data at the state or national level (Centers for Disease Control and

PAGE 29

29 Prevention, 2014, 2016) . These findings may re enfor ce exiting knowledge and serve as a framework for future studies interested in designing and implementing interventions at the individual, relationship, and community level, which are focused on reducing prescription opioid use in community settings . By ta rgeting high risk communities for intervention, the overall impact on the opioid crisis is likely to be greater and more efficient than attempting to design strategies for the US as a whole. At the individual level, prevention strategies that target indivi dual factors could be implemented through education al efforts aimed at prevent ing prescription opioid use . At the relationship level, prevention strategies could focus on designing programs that facilitate relationships and integrate social media and other communication efforts regarding prescription opioid use. Finally, at the community level, prevention strategies could be targeted to geographical areas in the community where prevalence of prescription opioid use is the highest (i.e., areas with high freq uency of ED utilization). Based on the identified risk factors, strategies separately tailored for women and men and for specific geographic locations may increase efficiency and reduce negative consequences associated with prescription opioid use. Us ing these strategies to address prescription opioid use in the community may impact the opioid epidemic by reducing : 1) prescription opioid use among women, who are disproportion ately being prescribed and using prescribed opioids compared to their male counter parts, 2) subsequent opioid abuse and overdose conditionally related t o prescription opioid use, and 3) psychiatric co morbidities including addiction and abuse related to prescription opioid use which can reduce the societal costs associated with prescrip tion opioid use overall. To carry out these aims and to address these gaps in the literature, we analyzed data from

PAGE 30

30 HealthStreet, an on going community engagement program that collects real time data on prescription opioid use.

PAGE 31

31 Figure 1 1. The Socio ecological Model Individual Relationship Community Societal

PAGE 32

32 Figure 1 2. HealthStreet variables to be examined within the context of the socio ecological model Prescription opioid use Relationship level risk factors: Marital status Children Employment Social media use Community level risk factors Zip code: Rural/urban Individual level risk factors: Socio demographics: Race Age Sex Education Health Insurance Medical visits: ED MD Chronic health conditions: Cancer Pain Anxiety/ depression Insomnia Substance use Alcohol Rx sedatives Marijuana Cigarettes

PAGE 33

33 CHAPTER 2 DATA SOURCE HealthStreet HealthStreet is a community engagement program at the University of Florida that addresses gaps in health care and health research in the community . First funded in 1989 with a National Institute on Drug Abuse award, t he HealthStreet model reach ed out to community members in St. Louis to link them to HIV prev ention and to drug and alcohol related research at Washington University (WU) . In 2008 as part of the Clinical and Translational Science award (CTSA) at WU, the model was expanded to focus on all health problems and was funded by the Natio nal Institutes o f Health (NIH) National Center for Research Resources (NCRR ), which later became the National Center for Advancing Translational Sciences (NCATS) . The Founding Director of HealthStreet, Dr. Linda B. Cottler, relocated to the University of Florida in 2011 and established HealthStreet as a major effort of the CTSA at the University of Florida. HealthStreet link s community members to services based on their assessed needs and concerns, provides opportunities for members to learn about health research, engages the community in bidirectional health promotional research, and works to build trust between the community and researchers . The honey comb model is depicted in Figure 2 1. The data collected from UF Healt hStreet participants are being used in this dissertation. Recruitment Through HealthStreet, Community Health Workers (CHWs) are trained and certified to directly engage community members at parks, grocery stores, churches, laundromats, health fairs, and in areas where community members recreate concerns . CHWs visit different locations in the community on a daily basis at different times of the day and meet people where they are. As people are engaged in conversation, th ey are

PAGE 34

34 requested to provide written conse nt then are interviewed with a health needs assessment . Once they have undergone the health needs assessment they become members of the HealthStreet community cohort. The data collected by the CHWs at HealthStreet i s then enter ed into REDCap (Harris et al., 2009) . A study navigator reviews this information and then refers communi ty members to social and medical services and research opportunities. To date, HealthStreet has over 10,800 members. The HealthStreet catchment includes areas covered by the Gainesville site (69% enrollment) , the Jacksonville site (24% enrollment) , and the Miami site (2% enrollment). Members come from the surrounding areas in North Central Florida (5% enrollment) with the majority identifying as African American (58.1%) and female (59.3%). The sample includes a total of 12,986 community members who had a meaningful contact ( a 3 minute conversation) with a CHW , among whom 10,293 (79.2%) completed a health needs assessment. Of those, 9,785 (95.1%) were eligible for the present study, of whom 49% reported no lifetime use of prescription opioids, over a t hird (37%) reported lifetime but not past 30 day use of prescription opioids, and the remaining 14% reported pa st 30 day opioid use (Figure 2 2 ). Health Needs Assessment After signed, CHWs interview participants using the HealthStreet Health Needs Assessm ent. The needs assessment is a face to face interview that takes approximately 30 minutes to complete (CHWs record the responses using a paper copy of the health needs assessmen t during the interview) and includes demographics, willingness to participate in health research, health conditions including mental health , medical health concerns, and history of substance use . This dissertation utilized data collected during health need s assessments conducted between November 2011 and June 2018 , that fit within the socio ecological model, a commonly used framework for health prevention, individual , relationship , and community level factors.

PAGE 35

35 Variables HealthStreet provides a large community sample with comprehensive information on health conditions, health concerns, and history of substance use including prescription opioid use. To examine prescription opioid use among this community sample we used variables at the individual, relat ionship, and community level from the health needs assessment in our analyses. The v ariables selected at each level include r elevant predictors of prescription opioid use that were based on the literature. Demographic variables at the individual level that were examined include sex , age, and race. Specifically, age was categorized into younger ( < 49 years ) and older ( > 50 years) adults because this is typically the age at which chronic disease s and pain become more prevalent (Centers for Disease Control and Prevention, 2018) . In addition to chronic diseases becoming more prevalent at this age and because we focused specifically on opioid use among women in aims 2 and 3 , t he cut off at age 50 years was used to stratify by estimated pre and postmenopausal status . Menopause typically occurs in women between 45 and 55 years of age, and thus, examining those aged 50 years and over (the midpoint of the range) (Food and Drug Administration, 2018; Gold, 2011) could provide additional information on risk, since physiological changes occur after menopause which may affect prescription opioid use. Other f actors at the individual level that are associated with increased prescription opioid use included having health insurance , going to the doctor , and visiting the emergency department (Cantrill et al., 2012; Chang, Kharrazi, Bodycombe, Weiner, & Alexander, 2018; Hahn, 2011; Steinmetz, Zheng, Okunseri, Szabo, & Okunseri, 2017) . In addition, chronic conditions including pain, cancer, insomnia, and depression ( (Blanco et al., 2016; Pinkerton & Hardy, 2017; Serdarevic, Osborne, Striley, & Cottler, 2017; Sullivan, Edlund, Zhang, Unützer, & Wells, 2006) ) as well other substance use (Gudin, Mogali, Jones, & Comer, 2013; Stein et al., 2017; Sullivan et al., 2006) have been associated with increased prescrip tion opioid use at the individual level. M arital

PAGE 36

36 status, having children, employment (i.e., access to health care and prescription drugs from colleagues , and financial resources ), and social media use have also been associated with an increase in prescript ion opioid use at the relat ionship level (Fillingim et al., 2005; Inciardi et al., 2007; Piko, 2000; Rabinovitch, Cassidy, Schmitz, Joober, & Malla, 2013; Rönkä & Katainen, 2017; Rozenbroek & Rothstein, 2011) . Finally, on the community level, i n creases in morbidity and mortality related to prescription opioid use have been found to be associated with rurality (Keye s et al., 2014) . Tables 2 1 through 2 3 below list question numbers, the question ( s ) from which the variable is derived , the range of responses allowed for each variable , and the way each variable was coded in the analysi s . Outcome: Prescription Opioid Use Patterns ever used prescription pain medications like Vicodin®, oxycodone, codeine, Demerol®, ted that they used a prescription opioid, they were subsequently asked if they used one of these prescription medications in the past 30 day s. The HealthStreet needs assessment questions regarding drug use are shown in Figure 2 3 . Based on responses to thi s question (#87), past 30 day use was defined as any reported opioid use in the 30 days preceding interview, lifetime use was defined as any reported opioid use that occurred more than 30 days prior to interview, and no use was defined as no reported prescription opioid use ever (Figure 2 2 ) . Individual level Variables The following variables from the health needs assessment were used to represent factors at the individual level of the socio ecological model: age (continuous variable ranging from 18 t o 94 years of age ; ), race (white, black/African American, other), sex (male or female), educational attainment (more than high school or high school and less), health insurance (yes or no), doctor

PAGE 37

37 visit in the last 6 months (yes or no), any emergency depa rtment (ED) visit in the last 6 months (yes or no), mental health conditions ( coded as yes or no: depression, anxiety ), chronic health conditions ( coded as yes or no: back pain, cancer , insomnia ), and substance use ( coded as yes or no: cigarette smoking , h azardous alcohol use, marijuana use, sedative use ). Individual level variables and how they were coded are outlined in Table 2 1. Relationship level Variables Several relationship level risk factors were assessed using the health needs assessment: marital status (never married; currently married; or separated, divorced, or widowed), number of children (continuous variable), employment (full time or part time), and use of social media (Twitter, Facebook, or Instagram; yes or no) . Relationship level variables are outlined in Table 2 2. Community level Variables The community level risk factor that was assessed using the health needs assessment was A variable for r urality was created using US census tract data for the zip code . The data from the census showed that for each zip code, there are populations living in urbanized areas, urban cluster s , and rural areas. Urban areas were designated based on the census designation of ur banized area greater than 50,000 people or urban cluster greater than 2 , 500 people (US Census Bureau, 2018) . Zip codes from question 16 f rom the health needs assessment were merged to designate each person as living in an urban or rural residence. The geolocation (longitude and latitude point data) of each HealthStreet Interface (API). The one community level variable is outlined in Table 2 3.

PAGE 38

38 Preliminary Studies Data from HealthStreet in this community sample have been used for a number of studies and publications on prescription opioid use. Specifically, sex differences in patterns of bing e drinking and opioid use have been observed among middle aged community members (Serdarevic & Cottler, 2017) . Males were more likely to engage in binge drinking without prescription opioid use, while females were more likely to report higher rates of prescription opioid use . A nother stud y found that women were significantly more likely to report lifetime use of prescription opioids and cancer than men ; yet , women with cancer had a significantly reduced risk of using prescription opioids in their lifetime compared to men with cancer (Serdarevic, Striley, et al., 2017) . In addition, when we examined prescription opioid and hazardous alcohol use among older women, we found approximately 33 % of women r eported prescription opioid and/ or hazardous alcohol use in the past 30 day s. We also found concurrent prescription opioid and hazardous alcohol use were significantly associated with comorbid depression and anxiety , while women who endorsed prescription opioid use only were significantly more likely to report a history of back pain and cancer in this community sample (Serdarevic, Gurka, Striley, Vaddiparti, & Cottler, 2018) . Further, chronic health conditions such as insomnia have also been associated with prescription opioid use; opioid users were significantly more likely to report insomnia than non users (Serdarevic, Osborne, et al., 2017) . Overall, the results from these studies using HealthStreet data on prescription opioid use have provided a preliminary framework for targeted recommendations for interventions in the community. As such, results from HealthStreet have had an important impact, and data from this community sample on prescription opioid use are yielding equally important results. The underlying focus of HealthStreet is to include underrepresented populations in research. This makes HealthStreet a valuable data source given that the current literature include s

PAGE 39

39 limited epidemiologic data for addressing prescription opioid use among underrepresented populations in the community . Specifically, this community cohort includes individuals who are not tied to one health system and are geographically diverse allowing us to examine a heterogeneous and community dwelling population.

PAGE 40

40 Figure 2 1. An infographic depicting the HealthStreet mission to community members who enroll in HealthStreet

PAGE 41

41 Figure 2 2 . Enrollment in HealthStreet and eligibility for current study, November 2011 June 2018. Participants < 18 years, transgendered participants, and participants with missing Rx opioid use , sex, or zip code data were excluded (n=1,072 ) Total participan ts included in current study (n=9,221) Participants from the c ommunity c ontacted by or at HealthStreet (n=12,986) Completed intake questionnaire (n= 10,293 ) No Rx opioid use ( n=4,472 , 49%) Lifetime Rx opioid use, not past 30 days (n=3,463, 37 %) Past 30 day Rx opioid use ( n=1,286 , 14%) Lifetime use (n=4,749, 51%)

PAGE 42

42 Figure 2 3. Portion of the HealthStreet Health Needs Assessment instrument on which Community Health Workers record interview

PAGE 43

43 Table 2 1. Individual level variables and the corresponding HealthStreet health assessment questions from which they were derived Variable Question No. Questions Potential Responses Dissertation Coding Age 14 Age: _ _ Open ended in years 18 49 years= 0 50 + years = 1 Race 11 Race/ethnicity: __ _ American Indian/Alaskan N ative, Asian, Black/African American, Na tive Hawaiian/Pacific Islander, White, Other White= 0 Black= 1 Other= 2 Sex 9 Gender: __ _ Male, Female, Transgender Male= 0 Female= 1 Educational attainment 43 Last grade completed: _ _ Continuous Grades 0 12= 0 Grade 13+= 1 Health insurance 53 Do you have any type of medical insurance? No, Yes No= 0 Yes= 1 Doctor visits 51 Have you seen a doctor for any reason in the last 6 months? No, Yes No= 0 Yes= 1 ED visits 53b How many times have you been to the ER in the last 6 months for your own injury, illness, or condition? Continuous Non frequent (0 1 visits)= 0 Frequent (2+ visits)=1 Depression 68e Have you ever been told you had, or have you ever had a problem with depression ? No, Yes No= 0 Yes= 1 Anxiety 68b Have you ever been told you had, or have you ever had a problem with anxiety ? No, Yes No= 0 Yes= 1 Back pain 69a Have you ever been told you had, or have you ever had a problem with back pain? No, Yes No= 0 Yes= 1 Cancer 60 Have you ever been told you had, or have you ever had a problem with cancer? No, Yes No= 0 Yes= 1

PAGE 44

44 Table 2 1 . C ontinued Variable Question No. Questions Potential Responses Dissertation Coding Insomnia 72a Have you ever been told you had, or have you ever had a problem with insomnia? No, Yes No= 0 Yes= 1 Cigarettes use 92a Have you ever smoked cigarettes? No, Yes No= 0 Yes= 1 Hazardous alcohol use 81a FOR MEN: Within the last 30 days, have you had more than 4 drinks like beer, wine, liquor in a single day? No, Yes No= 0 Yes= 1 82b FOR WOMEN: Within the last 30 days, have you had more than 3 drinks like beer, wine, liquor in a single day? No, Yes No= 0 Yes= 1 Marijuana use 84 Have you ever used marijuana? No, Yes No= 0 Yes= 1 Prescription sedative use 89 Have you ever used prescription medications for anxiety or sleep like Valium, Xanax, Ambien? No, Yes No= 0 Yes= 1

PAGE 45

45 Table 2 2. Relationship level variables and the corresponding HealthStreet health assessment questions from which they were derived Variable Question No. Questions Potential Responses Dissertation Coding Marital status 42 Marital status Never married, Married, Separated, Divorced, Widowed Never married= 0 Married= 1 Separated/Divorced/ Widowed= 2 Children 47 How many children do you have? Continuous No= 0 Yes= 1 Employment 44 Are you employed full time or part time? No, Yes No= 0 Yes= 1 Social media use 33b Do you use social media like Twitter, Facebook, Google Plus, Instagram, etc.? No, Yes No= 0 Yes= 1 Table 2 3. The c ommunity level variable and the corresponding HealthStreet health assessment questio n from which it was derived Variable Question no. Questions Potential Responses Dissertation Coding Rurality 16 Street Address Open ended response Rural= 0 Urban= 1

PAGE 46

46 CHAPTER 3 SEX DIFFERENCES IN PRESCR IPTION OPIOID USE PATTERNS AS S ESSED THROUGH A COMMUNITY ENGAGEMENT PROGRAM IN NORTH CENTRAL FLORIDA Background Prescription opioids ( e.g. OxyContin®, Vicodin® and Percocet®) are typically prescribed to treat either chronic or acute pain (Beaudoin et al., 2014; Caudill Slosberg et al., 2004) , though there is the potential for misuse and abuse which raises concerns about overprescribing (Kaye et al., 2017; Strand, Eukel, & Burck, 2018; A. G. White, Birnbaum, Schiller, Tang, & Katz, 2009) . Nevertheless , opioid prescriptions more than doubled in the United States (US), from 107.3 million prescriptions written in 1 992 to 246.2 million in 2015 (Pezalla et al., 2017) . More recently , the overall national prescribing rate has decreased to the lowest it has been in the last decade, however certain regions across the US continue to have high prescribing rates (Centers for Disease Control and Prevention, 2018e) . Prescription opioid use in the US has led to concerns about the drug being used non medically (i.e., use of hig her doses, use longer than prescribed, use for the experience or feeling caused or use of someone else's medication ) (McCabe et al., 2013b; Osborne et al., 2017) ) and the associated potential consequences of use, including overdose. Prescription opioid overdose deaths more than quadrupled between 1999 and 2015 (Compton, Boyle, & Wargo, 2015; Compton & Volkow, 2006; Han, Compton, Jones, & Cai, 2015) and in 2017 opioids were involved in approximately 47,700 overdose deaths , which is more than any ye ar on record (Centers for Disease Control and Prevention, 2017a, 2018g) . Current estimates indicate that 130 people die every day from drug overdose involving an opioid (Centers for Disease Control and Prevention, 2018g) . The over prescription of opioids has been implicated in the overall overdose deaths ; it is also believed to be a risk f actor for heroin use (Compton, Jones, & Baldwin, 2016b) . Deaths due to heroin use

PAGE 47

47 have also risen sharply in recent years (Centers for Disease Control and Prevention, 2017c; Rudd et al., 2014) . Given the extent of the overdose crisi s and the US opioid epidemic, the factors associated with prescription opioid use are not well documented. Most studies examine use with data from national surveys including the National Survey on Drug Use and Health (NSDUH) and the Centers for Disease Con trol and Prevention (CDC) . Data from NSDUH shows approximately 38 % of the US adult population used prescription opioids in the past 12 months (Han et al., 2017a) . In addition, NSDUH also reports men have a lower prevalence of past 12 month prescription opioid use than women (35.3% vs. 40.2%) (Han et al., 2017a) . The CDC reports 6.9% of US adults used prescription opioids in the past 30 day s , and similar to data from NSDUH, men were less likely to use prescription opioids in the past 30 day s compared to women (6.3% vs 7.2%) (Centers for Disease Control and Prevention, 2015) . While national studies can provide a useful high level overview of prevalence, they do not provide granular data in prevalence at the community level within specific geographic regions. Such fluctuations can provide important data to target prevention programs a nd crisis response efforts. In certain US cities and regions higher prevalence estimates and overdose rates have been reported (Beheshti et al., 2015; Rudd et al., 2014) . How usage varies within cities and counties is more difficult to understand. Studies that examine specific settings (e.g. schools) within a sma ller geographic region or that only examine specific populations within such regions (e.g. young adults) are rare (Boyd, Esteban McCabe, & Teter, 2006; Boyd, McCabe, Cranford, & Young, 2007; Edlund et al., 2015; McCabe, West, & Boyd, 2013a) . Most studies utilize opioid use data from treatment centers or EDs. They represent only those who were able to afford treatment or seek care. These studies often suffer from differential mortality.

PAGE 48

48 T here is also limited epidemiologic data available to address sex differences for prescription opioid use in a community setting (Serdarevic, Striley, et al., 2017) , though sex differences at the national level have been examined previously (Back et al., 2011; Han et al., 2017a) . Although opioid prescribing has decreased recently due to local and statewide legislation, data examining patt erns of prescription opioid use f rom a community where overdoses are high could inform efforts to reduce prescription opioid use and related morbidity and mortality (Centers for Disease Control and Prevention, 2016; Flo rida drug Related Outcomes Surveillance and Tracking System (FROST), 2018) . Among the many theoretical models which explain prescription opioid use behavior is the socio ecological model that includes four levels (individual, relationship, community, and societ y ) which may influence behavior (McLeroy et al., 1988) . This model postulates that people are not only influenced by their individual traits, but are also influenced by relationships with others, the commu nity in which they reside, and the societies in which those communities are supported (Substance Abuse and Mental Health Services Administration, 2016a) . Given that the focus of this research is on the community setting, the socio ecological model concepts align with this. The individual level includes biological and personal factors that influence behavior while the relationship level consists of close relationship s with others who influence behavior. The community level refers to where people live and recreate. All of these levels may influence prescription opioid use and by considering multiple levels, prevention efforts are more likely to be sustained over time compared to using single level interventions (Richard et al., 1996; Stokols, 1996) . Sex is one factor within the individual level of the socio ecological model tha t needs to be examined further. P rescription opioid use may occur due to influences that change behavior

PAGE 49

49 which differ by sex, even though sex alone may not be a risk factor for prescription opioid use. Starting at the individual level, research has reported sex differences related to substance use. For example, w omen experience higher pain sensitivity than men (Darnall et al., 2012) , which subsequently may be one reason for higher rates of opioid prescribing among women. In thi s analysis, we will characterize prescription opioid use by sex and examine risk factors at the individual, relationship, and community level. Based on the theoretical fra mework and previous research on prescription opioid use, our hypothesis is that women will be significantly more likely to endorse prescription opioid use than males. This could be due to higher prevalence of chronic conditions among women (Darnall et al., 2012) which cause pain , and might su bsequently require opioids for treatment. An additional hypothesis is that prescription opioid use will increase wi th age. C hronic pain is more common among older adults ; thus, we expect an age related increase in prescription opioid use . These hypotheses can be tested within a socio ecological framework . To determine that any identified effect is not a result of confounding we must control for factors at the individual, relationship, and community level. R elevant predictors of prescription opioid use from the literature were used to assess risk (and protective) factors at the individual, relationship, and community level. Factors on the i ndividual level, included having health insurance (Chang et al., 2018) , visits to the doctor (Hahn, 2011) and emergency depa rtment (ED; Cantrill et al., 2012 ) , which all have been associated with increased prescription opioid use. In addition, chronic conditions including pain, cancer, insomnia, and depression (Blanco et al., 2016; Pinkerton & Hardy, 2017; Serdarevic, Osborne, et al., 2017; Sullivan et al., 2006) as well other substance use (Gudin et al., 2013; Stein et al., 2017; Sullivan et al., 2006) have been associated with prescription opioid use on the individual level. On the relationship level, factors including marital st atus, children, employment

PAGE 50

50 (i.e., access to health care and getting prescription drugs from colleagues), and social media use (Fillingim et al., 2005; Inciardi et al., 2007; Piko, 2000; Rabinovitch et al., 2013; Rönkä & Katainen, 2017; Rozenbroek & Rothstein, 201 1) . Finally, on the community level rurality has been previously associated with prescription drug use (Keyes et al., 2 014) . Increases in morbidity and mortality related to prescription opioid use has been found to be concentrated in states with large rural populations (Keyes et al., 2014) . Studies on s ex differe nces in prescription opioid use have been conducted previously, though they have often been limited to national level prevalence estimates. However, these few studies have foun d w omen are prescribed opioids more frequently than men in the US (Simoni Wastila, 2000) and generally use prescription drugs more frequently than men (Manteuffel et al., 2014; Olfson et al., 2015) . Women are approximately 50% more likely than men to be prescribed and to use prescription opioids (Simoni Wastila, 2000) . Though men are more likely to overdose from prescriptions compared to women (Unick et al., 2013) , recent literature shows that this gender gap is closing (Unick et al., 2013) and we are seeing more overdoses among women than ever before . In addition, there are physiological differences between men and women that affect drug activity, which can alter the effectivene ss of opioids (Whitley & Lindsey, 2009) . There is a major gap in the literature relating to patterns of prescription opioid use and risk factors for use by sex within a community, rather than within the US as a whole. As part of a community engagement program (HealthStreet) within Florida, we cha racterize prescription opioid use by sex and examine risk factors for such use. Aims and Hypotheses The aim of this analysis was to examine patterns of prescription opioid use among men and women separately within a community sample. Three patterns of use were examined in detail: past 30 day use, lifetime but not past 30 day use, and no use ever of presc ription opioids .

PAGE 51

51 The influence of individual , relationship , and community level risk factors for each pattern of prescription opioid use was examined by sex. Three hypotheses were proposed among 9, 221 community members (1, 286 with past 30 day use, 3, 463 with lifetime but not past 30 day use, and 4, 472 with no use of prescription opioids): Hypothesis 1. 1. Women will be more likely to endorse past 30 day us e and lifetime, but not past 30 day use (vs. never use ) , than men, after controlling for factors at the individual, relationship, and community level. Hypothesis 1.2. Lifetime prescription opioid use (both past 30 day use and lifetime, but not past 30 day use ) will have a stronger association among older individuals (50+ years) than among younger individuals (18 49 years), after adjusting for factors at the individual, relationship, and community level. Hypothesis 1. 3. Self reported p ain will be the strongest risk for opioid use among both men and women; pain will be mor e strongly associated with prescription opioid use among women compared to men. Methods To test these hypotheses , data f rom assessment tool ( the health needs assessment ) were used to construct variables in the analyses and are outlined below . HealthStreet provides a rich data source with information on health conditions, health concerns, and history of substance use including the use of prescription opioids. Study Population Community Health Workers (CHWs) from HealthStreet, a community engag ement program at the University of Florida, conditions and concerns, then link ed community members to health research opportunities and medical and

PAGE 52

52 social services. CHWs are trained and certified to directly engage commu nity members in the community at parks, grocery stores, churches, laundromats, health fairs, and in areas where community members recreate to meet people where they are . T his me thod of recruitment provides CHWs with the opportunity to engage with underrepr esent ed populations within their own communities. Using this CHW model , over 10,293 community members Gainesvil le, Jacksonville, and Miami) since it began at the Univers ity of Florida. A fter CHWs introduce themselves and explain the HealthStreet model, written consent is obtained and a health needs assessment is completed. The health needs assessment is a 30 minute , face to face interview, conducted by the CHWs to assess each new HealthStreet . Responses to the health needs assessment are recorded on paper copies during outreach at va rious locations (listed above) in the community. Once the participant has undergone the health needs assessment they become a member of the HealthStreet community cohort. HealthStreet members who self identified as female or male , aged 18 years and older, and who enrolled between November 2011 and June 2018 were included in this analysis . This study was approved by the University of Florida Institutional Review Board. Measurements To focus ever used prescription pain medications like Vicodin®, oxycodone, codeine, Demerol®, morphine, prescription opioid, they were asked subsequently if they used one of these pr escription medications in the past 30 day s. The three patterns for prescription opioid use were (1) past 30 day use (reported use of prescription opioids in the past 30 day s ), (2) lifetime but not past 30 day

PAGE 53

53 use (reported using prescription opioids, but n ot in the past 30 day s ) , and (3) no use (reported no use ever). Risk f actors at the individual level included age, race (white, black/African American, other), sex, educational attainment (more than high school or high school and less), health insurance (yes or no), doctor visit in the last 6 months (yes or no), frequent ED visits in the last 6 months (>2 visits; yes or no), a diagnosis of depression or anxiety from a health professional ( yes or no). C hr onic health conditions included ba ck pain ( yes or no) , cancer ( yes or no) , insomnia ( yes or no), and substance use [ cigarette smoking (yes or no), hazardous alcohol use (yes or no), marijuana use (yes or no), and prescription sedati ve use ( yes or no) ] . Data regarding several relationship level risk factors were collected during the health needs assessment: marital status (never married; currently married; or separated, divorced, or widowed), number of children, employment (full time/ part time), and use of social media (Twitter, Facebook, o r Instagram; yes or no). The community level risk factor was assessed using urban/rural) from the health needs assessment . Specifically, zip codes for each participant were merged with US census tract data to d esignate each person as living in an urban or rural residence. Analysis Participants less than 18 years of age (n=1,042), persons for whom prescription opioid use was missing due to it not being included in the health needs assessment yet (n=24) or who did not identify their biological sex (n=6) were excluded. Descriptive statistics were calculated to summarize patterns of p rescription opioid use (past 30 da y use, lifetime but not past 30 day use, and no reported lifetime use). The association betwe en i ndividual , relationship , and community level variables and the outcome , prescription opioid use , w as examined using chi square tests for categorical variables (significance was determined based on Bonferroni corrected

PAGE 54

54 p value of 0.002 after dividing alpha 0.05 by 21 , the number of variables). Age and number of children were categorized into groups due to their non normality and non unimodal distribution (Figure 3 1 and 3 2) . Age was categorized into two groups (18 49 years and 50+ years) and the numbe r of children was categorized as yes ( > 1) or no (0). To test the moderating effect of age on prescription opioid use (categorized as no use, past 30 day and lifetime) , the first model was built with age and sex, but without any interaction terms. The seco nd model included age, sex, and the interaction term between age and sex. W hen interaction terms were added to the model, there was no appreciable difference in the odds ratios compared to the model without the interaction terms (these results are displaye d in the Appendix, Table A 1). For ease of interpretation, the final model without the interaction terms was used. In addition, the moderating effect of pain between sex and prescription opioid use was also assessed and the final model without interaction terms was used . Importantly, it is known that tests to assess interaction have low power and so even though no evidence of interaction was found, we cannot confirm it is not present (Cronbach & Snow, 1977; Greenland, 1983; Marshall, 2007) . Consequently, all variables were fitted individually in the final model with no interaction term. Subsequently, analyses were conducted for men and women separately to examine sex specific effects driven by our hypotheses, based on knowledge from the liter ature and expected sex differences arising from the socio ecological model. Multinomial logistic regression was used to calculate adjusted odds ratios (ORs) and 95% confidence intervals to examine risk factors for prescription opioid use (with no use of p rescription opioids as the referen t group). For all community members, the first set of models examined sex as a risk factor for lifetime no past 30 day use and past 30 day prescription opioid use. The first model was built using factors only at the indivi dual level. Variables at the

PAGE 55

55 individual level that did not reach significance were not included in the second model , which added relationship level factors. Significance was determin ed by examining the corresponding confidence intervals (one for lifetime a nd one for past 30 day opioid use) for each variable. The final model excluded non significant variables from the individual, relationship, and community levels. Moderation of the effect of sex by pain, and age by pain, was assessed and no evidenc e of mode ration was found (p>0. 0 5 ). As such, all variables were fitted individually in the final model with no interaction term. Similar methods were applied after stratifying by sex to build two specific models for men and women. Model fit was assessed using the l ikelihood ratio test comparing the null model to each of our models. All models were statistically significant and fit the data better than the null model. The variance inflation factor (VIF) was assessed for severity of multicollinearity. A ll VIF values w ere under 10 ; thus , no variables were removed . All statistical analyses were c onduct ed using SAS® 9.4 (SAS Institute Inc., 9.4, Car y, NC: SAS Institute Inc., 2011 ) . For a complete case analysis, people with missing variables were excluded which provided a total sample size of 9,221 participants. The calculation of power to detect a difference as little as 1% in opioid use between males and females was conducted. Given the sample was comprised of 5, 549 women and 3, 672 men, statistical power exceeded 96% at a 0.05 significance level. Thus, t he sample size was sufficient to detect an ex pected odds ratio as small as 1.12 (or 0.89) . Results A total of 12,986 community members had a meaningful contact ( considered to be at least a 3 minute conversation) with a CHW, and 10,293 (79.2%) completed a health needs assessment (Figure 3 3 ). Of th ose, 9, 221 ( 90 %) were eligible for the present study, of whom 60% were female, 5 8 % were black, and 44% were 50 years or older. About half of these (n= 4,472 ) reported no lifetime us e of prescription opioids, another 37% (n=3, 463 ) reported

PAGE 56

56 lifetime but not past 30 day use of prescription opioids, and the remaining 14% (n=1, 286 ) reported past 30 day opioid use. Among 3, 672 men, 2, 021 (55%) reported no use of prescription opioids, 1, 96 1 (33%) reported lifetime but not past 30 day use, and 4 55 (12%) reported pa st 30 day use (Figure 3 3 ). Among 5, 549 women, 2, 451 (44%) reported no use, 2 , 267 (41%) reported lifetime but not past 30 day use , and 8 31 (15%) reported past 30 day prescription opioid use. All factors examined on the individual level were significantly a ssociated with prescription opioid use with the exception of past 30 day hazardous alcohol use (T able 3 1 ). These risk factors included older age, female sex , white race, higher education, health insurance, doctor visits, ED visits, having a history of dep ression , anxiety , back pain, cancer, and insomnia, and reporting cigarettes, marijuana , and sedative use. Marital status, having children, being employed, and social media use were significantly associated with opioid use at the relationship level. Ruralit y, the community level risk factor, was also found to be significantly associated . Characteristics among all community members, strati fied by prescription opioid use pattern On the individual level, those who were older (50+ years) reported higher rates of both lifetime but not past day use and past 30 day opioid use compared to those who were younger (18 49 years; <.0001). Women compared to men reported higher rates of past 30 day use and use in their lifetime but not past 30 day s (<.0001). T hose who used in the past 30 day s were less educated compared to those who endorsed lifetime but not past 30 day use . Linear associations between having health insurance, visiting the doctor in the past 6 months, frequent ED use in the past 6 months, and r ecency of prescription opioid use were observed. For example, p revale nce of health insurance was highest among those who used prescription opioids in the past 30 day s (71.1%), followed by those who used in their lifetime, but not past 30 day s (6 5.1 %), and lowest among those who reported no use (5 9.3 %). A similar linear association for doctor visits ( past 30 -

PAGE 57

57 day use= 89. 6 %, lifetime but not past 30 day use = 74. 5 %, no use= 58.7%) and frequent ED use ( past 30 day use=31. 6 %, lifetime but not past 30 day use= 1 5.0%, no use= 10. 0 %) was observed (Table 3 1) . Those who reported a history of depression, anxiety, back pain, cancer, and insomnia had higher rates of past 30 day and lifetime but not past 30 day prescription opioid use (p<0.0001). A linear association be tween cigarette, sedative, and prescription opioid use was observed. Cigarette and sedative use was most prevalent among those who used prescription opioids in the past 30 day s (cigarette= 59.9 %, sedatives=44. 6 %), followed by lifetime, but not past 30 day use (cigarette=56. 6 %, sedatives= 34. 7 %), and least prevalent among those who did not use prescription opioids (cigarette=4 5.7 %, sedatives=10. 5 %). Marijuana use significantly differed by prescription opioid use group. Those who endorsed any use, reported a higher rate of marijuana us e compared to those who did not. Regarding relationship level factors (Table 3 1), those who were previously married (separated, divorced, or widowed) reported a higher rate of use (p < 0.0001) with recency . Those with children and those who used social media reported higher rates of prescription opioids in the past 30 day s (children: N=79 .4 %, social media: N= 36 .9 %) and in their lifetime, but not past 30 day s (children: N=73 .0 %, social media: N=4 7.7 %) compared to those who repo rted no use (children=64 .1 %, social media=3 6.2 %; all p < 0.0001). Those who endorsed past 30 prescription opioid use reported higher rates of unemployment compared to those who used in lifetime but not past 30 days and non users . For the community level fa ctor rural area of residence, living in a rural area was approximately twice the rate among those with past 30 day prescription use (4. 7 %) and lifetime not pas t 30 day use (4. 4 %) compared to non prescription opioid users (2. 6 %; p < 0.0001).

PAGE 58

58 Characteristic s and prescription opioid use among HealthSt reet members, stratified by sex Women, as outlined in Table 3 2, had a higher proportion of past 30 day (1 4.9 %) and lifetime (40. 9 %) prescription opioid use than men (pas t 30 day= 12.1% ; lifetime=32. 6 %). On the individual level, men who used prescription opioids ( past 30 day s and in their lifetime) were older (p<0.0001). Older age among men and women was associated with prescription opioid use. For men, having health insurance, visiting the doctor or frequent ED visits were more prevalent among those who endorsed past 30 day use (61. 5 , 86. 1 and 30. 6 %, respectively) and lifetime not past 30 day use (54. 4 , 6 9.4 and 14. 1 %, respectively) compared to those who did not use (50. 4 %, 5 1.1 and 8. 8 %, respectively; all p<0.0 001). The same was observed among women, with higher prevalence of health insurance, doctor visits, and ED visits among those who endorsed past 30 day and lifetime no past 30 day use compared to no use . Both men and women who endorse d prescription opioid u se reported depression, anxiety, back pain, cancer, and insomnia at higher rates (all p<0.0001). Similarly, for both men and women, higher prevalence of cigarette, marijuana, and sedative use was observed among those who used in the past 30 days and in the ir lifetime; however, after adjusting for multiple comparisons , hazardous alcohol use was not associated with prescription opioid use for men (p=0.002 5 ) or women (p=0. 3789 ). A mong men, those who reported any prescription opioid use, regardless of recency , reported higher rates of marriage or separation, divorce , or being widowed. Men who reported no use of prescription opioid use reported higher rates of be ing unmarried and reported lower rates of having children (p<0.0001). For men and women , unemploymen t was more prevalent among those who reported past 30 day prescrip tion opioid use compared to those who did not endorse prescription opioid use . Past 30 day prescription opioid use and lifetime not past 30 day use was more prevalent among men and women who reported having children . Social media use significantly differed among prescription opioid use group for both men and women (p<0.0001).

PAGE 59

59 Examination of the community level factor of rural area of residence remained significant when stratified by sex (Tabl e 3 2) . The prevalence of living in a rural area was higher among men who reported past 30 day (4.0%) and lifetime but not past 30 day use (2. 9 %) compared to those who reported no use (1.6%; p=0.00 33 ). Similar patterns were found among women ( 5.0 , 5. 2 and 3. 3 %, respectively; p=0.00 52 ). Risk factors and self reported prescription opioid use pattern among HealthStreet members Table 3 3 provides the final model, model 3, which includes all significant risk factors on the individual, relationship, and community level ( with depression, cigarette use, and rurality eliminated from this model). After adjustments for covariates, women were significantly more likely than men to report past 30 day and lifetime not past 30 day prescription opioid use . Additional risk factors for lifetime but not past 30 day prescription opioid use included past 6 month doctor visits ( adjusted Odds Ratio (aOR)= 1.45 ; 95% Confidence Interval (CI) : 1.30, 1.62 ) , frequent ED visits (aOR= 1.24; 95% CI, 1.06 1.44) , history of cancer (aOR= 1.5 7; 95% CI, 1.29 1. 90 ) and insomnia (aOR= 1.26; 95% CI, 1.11 1. 42 ) . At the relationship level, risk factors for lifetime but not past 30 day prescription opioid use included having children (aOR= 1.26; 95% CI, 1.12 1. 42 ) and social media use (aOR= 1.32; 95% CI, 1.19 1. 46 ). T hose who endorsed sedative use (vs. those who did not use prescription sedatives) had a 2.6 7 times greater likelihood of using prescription opioids in their lifetime but not past 30 day s compared to non prescription opioid users (CI : 2.34 , 3.0 4 ). The second strongest association for lifetime but not past 30 day prescription opioid use was lifetime marijuana use ( aOR = 1.76; 95% CI: 1.59, 1.95). Following marijuana use, history of back pain was the third strongest risk factor for lifetime no t past 30 day use of prescription opioids (aOR= 1.6 2 ; 95% CI, 1.4 7 1. 79 ). Protective factors for lifetime not past 30 day use of prescription opioids at the individual level include d being black

PAGE 60

60 (aOR = 0.64 ; 95% CI: 0.58, 0.72 ) and having higher educational attainment (aOR= 0.72; 95% CI: 0.65, 0.80) , and at the relationship level included never being married (aOR = 0.72 ; 95% CI: 0.63 , 0.83 ) . Similar results were observed for past 30 day use of prescription opioids , which included: sex, frequent ED visits, his tory of cancer and insomnia, and lifetime marijuana use at the individual level, and having children and unemployment at the relationship level . Past 30 day hazardous alcohol use, a factor at the individual level, was risk a factor for past 30 day prescrip tion opioid use (aOR= 1.48; 95% CI: 1.25, 1.74), but this relationship was not observed for lifetime no past 30 day prescription opioid use. Prescription sedative use was also the str ongest risk factor for past 30 day use (aOR= 3.9 6 ; 95% CI, 3.3 5 4. 68 ). Ma rijuana use was the second largest risk factor for lifetime not past 30 day use of prescription opioids, but for past 30 day prescription opioid use, past 6 month doctor visits the second strongest association (aOR= 3.32; 95% CI: 2.71, 4.07) . Back pain remained the third largest risk factor, with those who had a history of back pain (vs. no back pain) having a 2.8 2 times greater likelihood of past 30 day prescription opioid use (95% CI, 2.4 4 3.2 6 ) compared to non prescription opioid users . Among those wh o reported past 30 day prescription opioid use, h aving health insurance (aOR= 0.72; 95% CI: 0.62, 0.84) and being another race were protective factors at the individual level (aOR= 0.69; 95% CI: 0.51, 0.94). At the relationship level, never being married ( aOR = 0.78 ; 95% CI: 0.63, 0.95 ) was a protective factor for past 30 use day of prescription opioids. Risk factors and self reported prescription opioid use pattern among HealthStreet members, stratified by sex After controlling for significant covariates, s ex specific estimates were observed among risk factors for past 30 day and lifetime no t past 30 day use of prescription opioids with never use of prescription opioids as the reference group (Table 3 4 ). History of depression and cigarette use were not foun d to be significant risk factors at the individual level for either men or

PAGE 61

61 women and thus were eliminated. Rural area of residence was also found not to be a significant risk factor for either men or women after controlling for other covariates and was eliminated . Among women, t he strongest risk factor for past 30 day and lifetime use of prescription opioids was presc ription sedative use, which was also found to be the strongest association among men. Women who used sedatives in their lifetime (vs no us e of prescription sedatives) had a 4.2 7 times greater likelihood of using prescriptions opioids in the past 30 day s (95% CI, 3. 48 5.2 4 ) compared to women with no use of prescription opioids. Women who used sedatives (vs women who did not use sedatives) had a 2. 83 times greater likelihood of using opioid in their lifetime no t past 30 day s use compared to women who did not use opioids (aOR= 2. 83 ; 95% CI, 2. 41 3. 33 ). The second strongest risk factor for past 30 day prescription opioid use among women was past 6 month doctor visits. Women who visited the doctor (vs no doctor visi ts ) had a 3.1 4 times greater likelihood of using prescription opioids in the past 30 d ay s use compared to women who did not use prescription opioids (aOR= 3.1 4 ; 95% CI: 2.39, 4.14). Older women were significantly more likely than younger women to report past 30 day prescription opioid use. Additional risk factors for past 30 day prescription opioid use at the individual level included being black (aOR= 1.24; 95% CI, 1.01 1. 5 2) , freque nt ED visits (aOR= 2.73; 95% CI, 2.19 3.39 ) , history of back pai n (aOR= 2.93; 95% CI, 2.44 3.52 ) and cancer (aOR= 2.16; 95% CI, 1.63 2.88 ) , and lifetime marijuana use (aOR= 1.43; 95% CI, 1.19 1. 71 ) . At the relationship level, risk factors for past 30 day p rescription opioid use included having children (aOR= 1.67; 95% CI, 1.31 2.14 ) and being employed (aOR= 1.48; 95% CI, 1.20 1. 81 ) . The second strongest association for lifetime but no t past 30 day prescription opioid use for women was lifetime use of marijuana (aOR= 1.86; 95% CI, 1.63 2.11). In addition, risk factors for lifetime but not past 30 day use at the individual level included visits to the doctor (aOR= 1.32; 95% CI, 1.14 1. 53 ) ,

PAGE 62

62 frequent ED visits (aOR= 1.27; 95% CI, 1.05 1. 53 ) , and a history of back pain (aOR= 1.83; 95% CI, 1.61 2.08 ) and cancer (aOR= 1.81; 95% CI, 1.44 2.30 ) . Having children (aOR= 1.46; 95% CI, 1. 2 5 1. 7 2) and social media use (aOR= 1.41; 95% CI, 1.24 1. 60 ) were risk factor at the relationship level for lifetime but not past 30 day use . For past 30 day use of prescription opioids, not having health insurance (vs having insurance) decreased the odds of past 30 day prescription opioid use by 29% compared to no use (aOR= 0.71; 95% CI, 0.58 0.87). Black race was a protective factor at the individual level and never being married, having children, and unemployment were protective factors at the relationship level for lifetime but not past 30 day use of prescription opioids. Among men, health insur ance, history of depression, cancer, and cigarette use were non significant risk factors at the individual level and were eliminated. Age was also found not to be a significant risk factor among men after controlling for other covariates. At the relationship level, having children was not a significant risk factor for prescription opioid use among men and was eliminated. The final model (Table 3 3) revealed that prescription sedative use in lifetime was the strongest risk factor for both p ast 30 and lifetime use of prescription opioids. Men who had ever used prescription sedatives in their lifetime (vs no use of prescription sedatives) were over 4 times as likely to report opioids in the past 30 day s as those who did not use prescription opioids (aOR= 4.02; 95% CI, 3.06 5.29 ). The same effect was observed for lifetime not past 30 day use of prescription opioids (aOR= 2.71; 95% CI, 2.18 3.36). The second strongest risk factor for past 30 day prescription opioid use was past 6 month doctor v isits, with those who went to a doctor visit (vs. no doctor visit) having a 3.74 times greater likelihood of past 30 day use of prescription opioids compared to those who did not use prescription opioids (aOR= 3.74; 95% CI, 2.76 5.05 ). Other risk factors f or past 30 day prescription opioid use at the individual level

PAGE 63

63 among men included frequent ED visits (aOR= 2.35; 95% CI, 1.77 3.11 ), history or back pain (aOR= 2.81; 95% CI, 2.22 3.56 ), and past 30 day hazardous alcohol use (aOR= 1.66; 95% CI, 1.3 1 2.11 ). At the relationship level, social media use (aOR= 1.50; 95% CI, 1.14 1. 97 ) was the only risk factor for past 30 day prescription opioid use among men. In contrast to past 30 day use , the second strongest risk factor of lifetime no t past 30 day prescription opioid use was history of marijuana use (aOR= 1.79; 95% CI, 1.52 2.12). Additionally, visiting the doctor (aOR= 1. 68; 95% CI, 1.42 2.00 ), history of back pain (aOR= 1. 3 9 ; 95% CI, 1.18 1.63 ) and insomnia (aOR= 1.40; 95% CI, 1.16 1.70 ) were risk factors at the individual level for lifetime no t past 30 day use of prescription opioids . Protective factors for both past 30 day and lifetime use of prescription opioid use were observed . These includ e being black or other race, p reviously never being married , being separated, divorced or windowed , and less education. Men who completed high school or less (vs . more than high school) had a 30% decrease in odds of lifetime prescription opioid use compared to no use of prescription opioid s in lifetime but no t past 30 da y use (aOR= 0.70; 95% CI, 0.59 0.83). Discussion Prescription opioids and consequences related to use such as non medical use and overdose continue to be a significant public health problem in the US . Though health problems attributed to prescription opio id use continue, recent studies have focused little on sex diff erences and prescription opioid use . The current study characterize d opioid use by se x and examined risk factors for prescription opioid use at the individual, relationship, and community level . After adjustment for other covariates, we observed sex specific risk factors at the individual level as well as relationship level. On the community level factor we did not find differences between men and women.

PAGE 64

64 We hypothesized that women compared to m en would be more likely to endorse bo th lifetime use but not past 30 day use a nd past 30 day use of prescription opioids compared to never use after controlling for factors within the context of the socio ecological model. Examination of these risk factors regarding prescription opioid use revealed differences among women and men. Overall, approximately 37% of the entire sample reported lifetime but not past 30 day use and about 14% reported past 30 day use of prescription opioids. After stratifying by sex we found a higher proportion of women who endorsed past 30 day (1 4.9 %) and lifetime (40. 9 %) prescription opioid use compared to men (12.1%, 32. 6 %, respectively). The prevalence rate for past 30 day prescription opioid use was higher than the national rate. The CDC reports 6.9 % of the US adult population used prescription opioids in the past 30 day s (Centers for Disease Control and Prevention, 2015) , half of our rate. The CDC has also found women to be more likely to use prescription opioids in the past 30 day s compared to men (7.2% vs 6.3%; Centers for Disease Control and Prevention, 2015) , and we similarly found hig her rates of past 30 day opioid use among women compared to men (1 4.9% vs 12.1%). Further examination , which allowed us to control for factors at different levels of the socio ecological model , revealed women , compared to men , were more likely to endorse any use of prescription opioids , which confirmed our first hypothesis. This may be partly due to the variation in pain sensitivity between men and women . Women have a higher prevalence of chronic conditions that cause pain and they report pain more frequently than men (Bartley & Fillingim, 2013; Darnall et al., 2012; Houghton et al., 2016) that subsequently may lead to higher opioid prescribing which may also put women at great er risk for adverse events related to prescription opioid use.

PAGE 65

65 Additionally we found that risk factors and protective factors for prescription opioid use varied by sex at the individual level. We found cancer wa s a risk factor for any prescription opioid use. This relationship may have been observed because chronic conditions such as cancer are associated with pain and may require opioids to treat pain (Boland & Pockley, 2018; Bruera & Kim, 2003) . However, when we stratified by sex, we found history of cancer was a significant risk factor for women but not men. Interestingly, males have a higher lifetime probability of developing cancer and have a higher cancer mortality rate compared to women (Dorak & Karpuzoglu, 2012) . Due to the association between cancer pain and prescription opioid use, we would expect to see an association not only among women but also men who have also reported a history of cancer in our sample . Further research is needed to investigate this relationship , specifically the potential dev elopment of opioid use disorders among cancer patients . In concordance with the literature, we found insomnia was more preval ent among women compared to men (Suh, Cho, & Zhang, 2018) . Though insomnia has previously been associated with prescription opioid use (Serdarevic, Osborne, et al., 2017) wh en we stratified by sex we found insomnia to be a risk factor for lifetime use of prescription opioids among men only. At the relatio nship level, having a child was a risk factor for prescription opioid use (both lifetime and past 30 day use) among women but not among men. This may be due to biological differences women who giv e birth may suffer from pain during or after labor which increases their exposure to opioid use in order to alleviate pain (Shah, Hayes, & Martin, 2017) . A nother risk regarding opioid use among women is the development of neonatal abstinence syndrome (NAS ) among their infants (Jansson, Velez, & Harrow, 2009; Krans & Patrick, 2016) which we did not elicit . However, p revalence of NAS has increased dramatically; specifically in Florida NAS increased from 1.6 to 25.2 per 1000 live births between 2000 and 2010 (Wang et al., 2017) .

PAGE 66

66 These findings demonstrate that differences in prescription opioid use are pres ent between sexes and may be important consider in the context of prevention strategies . We also hypothesized lifeti me prescription opioid use would have a greater association among older individuals than among younger individuals, after adjusting for fac tors at the individual , relationship , and community level. This effect was only observed among women for past 30 day prescription opioid use, which we did not hypothesize , and may be due to the way age was categorized in our analyses . Some studies have s hown chronic conditions to be strong predictors for prescription opioid use and found women are more likely to be prescribed opioids for chronic conditions (Barsky et al., 2001; Darnall et al., 2012; Robinson et al., 2001) . We may have observed this relationship because of the link between chronic conditions that develop through the lifespan and because women are more likely to seek health care (Pinkhasov et al., 2010) in which they subsequently may be prescribed opioids more fre quently than men to treat pain associated with chronic conditions. We also explored the relationship between pain and sex separately and found there was no moderating effect of pain between sex on prescription opioid use. Further examination of prescriptio n opioid use among women only is needed to get a better u nderstanding of this relationship. Finally, we hypothesized that s elf reported pain would be the strongest risk factor for opioid us e among both men and women and that pain would be more strongly associated with prescription opioid use among women compared to men. Our findings partially confirmed this hypothesis; women who self reported pain had higher odds of endorsing prescription opioid use in their lifetime (OR=1.8 3 vs 1.39) a nd in the past 30 day s (OR=2.9 3 vs 2.8 1 ) compared to men . Interestingly, w e found that the strongest risk factor for both past 30 day and lifetime prescription opioid use was lifetime sedative use. This effect was observed overall and this

PAGE 67

67 remained the str ongest risk factor even after we stratified by sex to examine prescription opioid use in two separate models. Men who had ever used prescription sedatives in their lifetime (vs. those who never used prescription sedatives) had over 4 times the odds of usin g prescription opioids in the past 30 day s and almost 3 times the odds of using prescription opioids in their lifetime prior to the past 30 day s compared to men who did not use prescription opioids . Women who used sedatives in their lifetime (vs. those who never used prescription sedatives) were more than 4 times as likely to use prescriptions opioids in the past 30 day s as women who did not endorse prescription opioid use and almost 3 times as likely to use opioids in their lifetime not including the past 30 days . This finding is of importance as the combined use of prescription drugs such as opioids and sedatives ca n be lethal due to the effects these drugs produce when they are ingested together. Specifically, the pharmacodynamics of opioids and sedatives change when combined and work synergistically to reduce respiratory function (i.e., slowed or shallow breathing) which heightens the risk of overdose (Dowell et al., 2016; Sun et al ., 2017) . Both opioids and sedatives are central nervous system ( CNS ) depressants and when used simultaneously they are processed using two different pharmacologic mechanisms resulting in significant respiratory depression (Jann, Kennedy, & Lopez, 2014) . In addition, the drug classes of both opioids and sedatives are associated with risks of withdrawal and then dependence subsequently . Unfortunately, the combination of pr escribing opioids and sedatives is common and subsequently overdose death due to concurrent use of these two pr escription drugs has increased in the United States (Hwang et al., 2016; Paulozzi, Weisler, & Patkar, 2011; Sun et al., 2017) . One retrospective study found approximately 30% of fatal opioid ove rdoses involved sedatives between 2001 and 2013 among 315,428 patients enrolled in private health insurance (Sun et al., 2017) . Though the combined use

PAGE 68

68 of opioids and sedatives is one of the motivations for using Prescription Drug Monitoring P rograms (PDMPs) (Centers for Disease Control and Prevention, 2017d; Winstanley et al., 2018) , t he negative and potentially fatal consequence ou tlined above due to simultaneous use of these two drugs signifies the need to prevent such use in the community. Strengths and Limitations There are a few limitations to t hese analyses . The cross sectional data limits the ability to establish temporality between the risk factors examined and prescription opioid use. For this reason, we cannot establish causality in relation to certain risk factors because we do not know if these precede d the outcome. We found that the relationship between sex and opioid use does not depend on age or pain based on our interaction analyses. However, tests for interaction are under power ed and so even though no evidence of interaction was found, we cannot confirm it is not present. Participants were only asked about any prescription opioid use during two time frames ( past 30 day s and prior to the past 30 day s ); information regarding dose, f requency , duration of opioid use , sour ce of the opioids, and whether there was misuse was not ascertained . S ince non medical use of prescription opioids was not assessed, the potential for participant responses to be influenced by perceived desirability of their response by the interviewer to bias the findings was reduced. In addition , self report drug use could be over or under reported, which may result in misclassification and causes a bias of the odds ratio towards or away from the null. P revious studi es have found self report data regarding chronic conditions (i.e., diabetes and hypertension), is often consistent with medical records (Johansson, Hellénius, Elofsson, & Krakau, 1999; Okura, Urban, Mahoney, Jacobsen, & Rodeheffer, 2004) , and thus, the p otential for misclassification is unlikely in our community sample. Poor recall and intentional misreporting are still possible though. Additional self report data regarding other covariates such as pain may also be under or over reported. This may be due to how the questions were asked

PAGE 69

69 result, though care was taken in the design of the questionnaire to aid understanding and CHWs were able to clarify the meaning of a question if required. Residual confounding by factors not measured in this study remains a possibility, though we have attempted to include a sufficient number of factors in each model to account for confounding where possible. In addition, effect estimates may have been affected by the inclusion of select variables in the model; however, based on the number of variables included in the model and the total sample size for each strata, it is unlikely that any imprecision in effect estimates and 95% c onfidence intervals would change the conclusions drawn. The sample size was sufficiently large and number of variables included were sufficiently parsimonious to provide robust estimates. CHWs try to ensure all community members they approach participate and visit locations within the community at different times and days during the week. The sampling technique may still be susceptible to selection bias though if those who choose not to participate are systematically different t han those who choose to part icipate. This is unlikely to be the case with respect to the outcome of prescription opioid use, given that members are not made aware of the questions on the health intake before deciding to join HealthStreet. Overall, the relationships observed between t he risk factors we examined and prescription opioid use may be generalizable to similar communities. It should be noted that the data collected regarding prescription opioid use covers a broad time period (November 2011 June 2018) and may not reflect curre nt prescribing practices that have changed in Florida . Finally, we acknowledge that true sex differences may not be present where the confidence intervals for estimates of effect overlap for men and women . However, when we assessed sex and prescription opioid use as an interaction

PAGE 70

70 term we found no appreciable differences in odds ratios between the model that included the interaction terms and the model without the interaction terms. These sex specific effect estimates provide a useful indication of which factors at individual, relationship and community level are important among men and women separately. There are many strengths of this analysis , including use of a community based sample from North Central Florida which is typically underrepresented in re search . The use of community based samples facilitates bidirectional research which can strengthen ties within the community to further reduce health disparities. This study provides data from a large s ample that addressed risks for recent and past use . In pa rticular, the larg e sample size likely provided sufficien t power to detect true effe cts. This community based sample also provides variables at multiple levels of the socio ecological model and was conducted in a diverse population among individuals who are not traditionally represented in research . This is more likely to produce results which are generalizable to similar communities in the US, potentially providing effective strategies for reducing prescription opioid use in these communities. The data collected includes a wide variety of medical and drug use information for each participant in addition to social and behavioral information, allowing adjustment for many different confounding factors which are often not available in other data sources . Fin ally, the non anonymous design of the HealthStreet model allows for the potential to follow up with community members in the future for further studies regarding prescription opioid use and possible intervention strategies. Conclusions We found higher rate s of prescription opioid use in this community sample compared to national rates suggesting that examination of prescription opioid use within a community sample is crucial . Our data suggest women are more likely to use prescription opioids compared to men and this may be due to prescribing rates or higher prevalence of chronic conditions that require

PAGE 71

71 opioid use to treat pain among wome n which supports previous findings . Further, we found older age was a risk factor for prescription opioid use among women b ut the same was not observed among men which warrants further examination among women specifically . We were able to provide information on risk factors specific to sex regarding prescription opioid use , and found prescription sedative use was the strongest risk factor regardless of sex. This is especially important as the concurrent use of sedatives with opioids increases the risk of overdose. Prescription opioid use prevention should be considered on t he individual and relationship level within communities and targeted efforts should be considered for men and women separately. Risk factors identified from national studies may not be applicable to all communities, findings from this study provide informa tion form a unique sample which should be considered in prevention strategies.

PAGE 72

72 Figure 3 1. The frequency of the distribution of age among HealthStreet members

PAGE 73

73 Figure 3 2. The f requency of the distribution of the number of children among HealthStreet members Figure 3 3. Enrollment in HealthStreet and eligibility for current study, as of June 2018 Participants < 18 years, and participants with missing Rx opioid use, sex and gender , or zip code data excluded ( N=1,072 ) Total participan ts included in current study N=9,221 Participants from the community c ontacted by or at HealthStreet N =12, 986 Completed intake questionnaire N = 10,293 Men N=3,672 (40 % ) Women N=5,549 (60 %) Past 30 day (n=831 , 15%) Life time but no past 30 day use (n=2,267 , 41%) N one (n=2,451 , 44%) L ifetime but not past 30 day use (n=1,196, 33%) Past 30 day (n=455 , 12%) N one (n=2,021 , 55%)

PAGE 74

74 Table 3 1. Association between individual , relationship , and community level characteristics of prescription opioid use among HealthStreet members, 2011 2018, N=9,221. Characteristic Overall N=9,221 N (%) No Rx opioid use N=4,472 N (%) Lifetime, no t past 30 day Rx opioid use N=3,463 N (%) Past 30 day Rx opioid use N=1,286 N (%) P value INDIVIDUAL MODEL FACTORS Age 18 49 years 50+ years 5143 (55.8) 4078 (44.2) 2781 (62.2) 1691 (37.8) 1788 (51.6) 1675 (48.4) 574 (44.6) 712 (55.4) <.0001 Sex Male Female 3672 (39.8) 5549 (60.2) 2021 (45.2) 2451 (54.8) 1196 (34.5) 2267 (65.5) 455 (35.4) 831 (64.6) <.0001 Race Black Other White 5370 (58.2) 641 (7.0) 3210 (34.8) 2981 (66.7) 328 (7.3) 1163 (26.1) 1646 (47.5) 243 (7.0) 1574 (45.5) 743 (57.8) 70 (5.4) 473 (36.8) <.0001 Education HS or less More than HS 5246 (56.9) 3975 (43.1) 2786 (62.3) 1686 (37.7) 1690 (48.8) 1773 (51.2) 770 (59.9) 516 (40.1) <.0001 Health insurance No Yes 3401 (36.9) 5820 (63.1) 1822 (40.7) 2650 (59.3) 1208 (34.9) 2255 (65.1) 371 (28.9) 915 (71.1) <.0001 Doctor visits (past 6 months) No Yes 2961 (31.0) 6360 (69.0) 1845 (41.3) 2627 (58.7) 882 (25.5) 2581 (74.5) 134 (10.4) 1152 (89.6) <.0001 ED visits (past 6 months) 0 1 >2 7849 (85.1) 1372 (14.9) 4025 (90.0) 447 (10.0) 2945 (85.0) 518 (15.0) 879 (68.4) 407 (31.6) <.0001 Depression No Yes 6499 (70.5) 2722 (29.5) 3559 (79.6) 913 (20.4) 2228 (64.3) 1235 (35.7) 712 (55.4) 574 (44.6) <.0001 Anxiety No Yes 6881 (74.6) 2340 (25.4) 3772 (84.4) 700 (15.6) 2331 (67.3) 1132 (32.7) 778 (60.5) 508 (39.5) <.0001 Back pain No Yes 5096 (55.3) 4125 (44.7) 2992 (66.9) 1480 (33.1) 1694 (48.9) 1769 (51.1) 410 (31.9) 876 (68.1) <.0001 Cancer No Yes 8416 (91.3) 805 (8.7) 4266 (95.4) 206 (4.6) 3052 (88.1) 411 (11.9) 1098 (85.4) 188 (14.6) <.0001 Insomnia No Yes 6823 (74.0) 2398 (26.0) 3691 (82.5) 781 (17.5) 2338 (67.5) 1125 (32.5) 794 (61.7) 492 (38.3) <.0001 Cigarette use (lifetime use) No Yes 4449 (48.3) 4772 (51.7) 2427 (54.3) 2045 (45.7) 1506 (43.5) 1957 (56.5) 516 (40.1) 770 (59.9) <.0001 Hazardous alcohol use ( past 30 day s) No Yes 7035 (76.3) 2186 (23.7) 3439 (76.9) 1033 (23.1) 2646 (76.4) 817 (23.6) 950 (73.9) 336 (26.1) 0.0779

PAGE 75

75 Table 3 1. Continued. Characteristic Overall N=9,221 N (%) No Rx opioid use N=4,472 N (%) Lifetime, no t past 30 day Rx opioid use N=3,463 N (%) Past 30 day Rx opioid use N=1,286 N (%) P value Marijuana use (lifetime) No Yes 4565 (49.5) 4656 (50.5) 2544 (56.9) 1928 (43.1) 1425 (41.2) 2038 (58.8) 596 (46.4) 690 (53.6) <.0001 Rx sedative use (lifetime) No Yes 6976 (75.7) 2245 (24.3) 4003 (89.5) 469 (10.5) 2260 (65.3) 1203 (34.7) 713 (55.4) 573 (44.6) <.0001 RELATIONSHIP MODEL FACTORS Marital status Never married Married Separated, divorced, or . widowed 4216 (45.7) 2013 (21.8) 2992 (32.5) 2442 (54.6) 863 (19.3) 1167 (26.1) 1312 (37.9) 860 (24.8) 1291 (37.3) 462 (35.9) 290 (22.6) 534 (41.5) <.0001 Children No Yes 2805 (30.4) 6416 (69.6) 1606 (35.9) 2866 (64.1) 934 (27.0) 2529 (73.0) 265 (20.6) 1021 (79.4) <.0001 Employment No Yes 5905 (64.0) 3316 (36.0) 2732 (61.1) 1740 (38.9) 2167 (62.6) 1296 (37.4) 1006 (78.2) 280 (21.8) <.0001 Social media use No Yes 5474 (59.4) 3747 (40.6) 2852 (63.8) 1620 (36.2) 1811 (52.3) 1652 (47.7) 811 (63.1) 475 (36.9) <.0001 COMMUNITY MODEL FACTORS Rurality No Yes 8894 (96.4) 327 (3.6) 4357 (97.4) 115 (2.6) 3311 (95.6) 152 (4.4) 1226 (95.3) 60 (4.7) <.0001

PAGE 76

76 Table 3 2 . Association between individual , relationship , and community level characteristics and prescription opioid use among HealthStreet members, stratified by sex, 2011 2018, N = 9,221 . Characteristic Males (n=3,672) Females (n=5,549) No Rx opioid use N=2,021 N (%) Lifetime, no t past 30 day Rx opioid use N=1,196 N (%) Past 30 day Rx opioid use N=455 N (%) P value No Rx opioid use N=2,451 N (%) Lifetime, no t past 30 day Rx opioid use N=2,267 N (%) Past 30 day Rx opioid use N=831 N (%) P value INDIVIDUAL MODEL FACTORS Age 18 49 years 50+ years 1221 (60.4) 800 (39.6) 589 (49.3) 607 (50.7) 197 (43.3) 258 (56.7) <.0001 1560 (63.7) 891 (36.3) 1199 (52.9) 1068 (47.1) 377 (45.4) 454 (54.6) <.0001 Race Black Other White 1394 (69.0) 155 (7.7) 472 (23.4) 549 (45.9) 80 (6.7) 567 (47.4) 260 (57.1) 28 (6.2) 167 (36.7) <.0001 1587 (64.8) 173 (7.1) 691 (28.2) 1097 (48.4) 163 (7.2) 1007 (44.4) 483 (58.1) 42 (5.1) 306 (36.8) <.0001 Education HS or less More than HS 1340 (66.3) 681 (33.7) 636 (53.2) 560 (46.8) 296 (65.1) 159 (34.9) <.0001 1446 (59.0) 1005 (41.0) 1054 (46.5) 1213 (53.5) 474 (57.0) 357 (44.0) <.0001 Health insurance No Yes 1003 (49.6) 1018 (50.4) 545 (45.6) 651 (54.4) 175 (38.5) 280 (61.5) <.0001 819 (33.4) 16632 (66.6) 663 (29.3) 1604 (70.7) 196 (23.6) 635 (76.4) <.0001 Doctor visits (past 6 months) No Yes 989 (48.9) 1032 (51.1) 366 (30.6) 830 (69.4) 63 (13.9) 392 (86.1) <.0001 856 (34.9) 1595 (65.1) 516 (22.8) 1751 (77.2) 71 (8.5) 760 (91.5) <.0001 ED visits (past 6 months) 0 1 >2 1843 (91.2) 178 (8.8) 1028 (85.9) 168 (14.1) 316 (69.4) 139 (30.6) <.0001 2182 (89.0) 269 (11.0) 1917 (84.6) 350 (15.4) 563 (67.8) 268 (32.2) <.0001 Depression No Yes 1666 (82.4) 355 (17.6) 815 (68.1) 381 (31.9) 277 (60.9) 178 (39.1) <.0001 1893 (77.2) 558 (22.8) 1413 (62.3) 854 (37.7) 435 (52.4) 396 (47.6) <.0001 Anxiety No Yes 1763 (87.2) 258 (12.8) 855 (71.5) 341 (28.5) 318 (69.9) 137 (30.1) <.0001 2009 (82.0) 442 (18.0) 1476 (65.1) 791 (34.9) 460 (55.3) 371 (44.7) <.0001 Back pain No Yes 1365 (67.5) 656 (32.5) 629 (52.6) 567 (47.4) 149 (32.8) 306 (67.2) <.0001 1627 (66.4) 824 (33.6) 1065 (47.0) 1202 (53.0) 261 (31.4) 570 (68.6) <.0001

PAGE 77

77 Table 3 2. Continued Characteristic Males (n=3,672) Females (n=5,549) No Rx opioid use N=2,021 N (%) Lifetime, no t past 30 day Rx opioid use N=1,196 N (%) Past 30 day Rx opioid use N=455 N (%) P value No Rx opioid use N=2,451 N (%) Lifetime, no t past 30 day Rx opioid use N=2,267 N (%) Past 30 day Rx opioid use N=831 N (%) P value Cancer No Yes 1946 (96.3) 75 (3.7) 1093 (91.4) 103 (8.6) 408 (89.7) 47 (10.3) <.0001 2320 (94.7) 131 (5.3) 1959 (86.4) 308 (13.6) 690 (83.0) 141 (17.0) <.0001 Insomnia No Yes 1693 (83.8) 328 (16.2) 831 (69.5) 365 (30.5) 308 (67.7) 147 (32.3) <.0001 1998 (81.5) 453 (18.5) 1507 (66.5) 760 (33.5) 486 (58.5) 345 (41.5) <.0001 Cigarette use (lifetime use) No Yes 899 (44.5) 1122 (55.5) 390 (32.6) 806 (67.4) 140 (30.8) 315 (69.2) <.0001 1528 (62.3) 923 (37.7) 1116 (49.2) 1151 (50.8) 376 (44.2) 455 (54.8) <.0001 Alcohol use (past 30 days) No Yes 1441 (71.3) 580 (28.7) 829 (69.3) 367 (30.7) 287 (63.1) 168 (36.9) 0.0025 1998 (81.5) 453 (18.5) 1817 (80.2) 450 (19.8) 663 (79.8) 168 (20.2) 0.3789 Marijuana use (lifetime) No Yes 930 (46.0) 1091 (54.0) 354 (29.6) 842 (70.4) 158 (34.7) 297 (65.3) <.0001 1614 (65.9) 837 (34.1) 1071 (47.2) 1196 (52.8) 438 (52.7) 398 (47.3) <.0001 Rx sedative use (lifetime) No Yes 1841 (91.1) 180 (8.9) 826 (69.1) 370 (30.9) 276 (60.7) 179 (39.3) <.0001 2162 (88.2) 289 (11.8) 1434 (63.3) 833 (36.7) 437 (52.6) 394 (47.4) <.0001 RELATIONSHIP MODEL FACTORS Marital status Never married Married Separated, divorced, or widowed 1187 (58.7) 360 (17.8) 474 (23.5) 509 (42.6) 268 (22.4) 419 (35.0) 178 (39.1) 109 (24.0) 179 (36.9) <.0001 1255 (51.2) 503 (20.5) 693 (28.3) 803 (35.4) 592 (26.1) 872 (38.5) 284 (34.2) 181 (21.8) 366 (44.0) <.0001 Children No Yes 899 (44.5) 1122 (55.5) 461 (38.6) 735 (61.4) 146 (32.1) 309 (67.9) <.0001 707 (28.9) 1744 (71.1) 473 (20.9) 1794 (79.1) 119 (14.3) 712 (85.7) <.0001 Employment No Yes 1230 (60.9) 791 (39.1) 766 (64.1) 430 (35.9) 356 (78.2) 99 (21.8) <.0001 1502 (61.3) 949 (38.7) 1401 (61.8) 866 (38.2) 650 (78.2) 181 (21.8) <.0001

PAGE 78

78 Table 3 2. Continued Characteristic Males (n=3,672) Females (n=5,549) No Rx opioid use N=2,021 N (%) Lifetime, no t past 30 day Rx opioid use N=1,196 N (%) Past 30 day Rx opioid use N=455 N (%) P value No Rx opioid use N=2,451 N (%) Lifetime, no t past 30 day Rx opioid use N=2,267 N (%) Past 30 day Rx opioid use N=831 N (%) P value Social media use No Yes 1394 (69.0) 627 (31.0) 719 (60.1) 477 (39.9) 323 (71.0) 132 (29.0) <.0001 1458 (59.5) 993 (40.5) 1092 (48.2) 1175 (51.8) 488 (58.7) 343 (41.3) <.0001 COMMUNITY MODEL FACTORS Rurality No Yes 1988 (98.4) 33 (1.6) 1161 (97.1) 35 (2.9) 437 (96.0) 18 (4.0) 0.0033 2369 (96.7) 82 (3.3) 2150 (94.8) 117 (5.2) 789 (95.0) 42 (5.0) .0052

PAGE 79

79 Table 3 3. Association between select risk factors and self reported prescription opioid use pattern among HealthStree t members, 2011 2018 (N = 9,221 ) Characteristic Lifetime, no t past 30 day Rx opioid use (vs. no use) N=3, 463 P ast 30 day Rx opioid use (vs. no use) N=1,2 86 aOR 95% CI aOR 95% CI INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years Ref 1.09 0.98 1.23 Ref 1.17 0.99 1.37 Sex Male Female Ref 1.36 1.22 1.51 Ref 1.18 1.01 1.37 Race White Black Other Ref 0.64 0.69 0.58 0.72 0.56 0.84 Ref 1.01 0.69 0.86 1.19 0.51 0.94 Education More than HS HS or less Ref 0.72 0.65 0.80 Ref 0.94 0.81 1.09 Health insurance Yes No Ref 0.95 0.86 1.06 Ref 0.72 0.62 0.84 Doctor visits (past 6 months) No Yes Ref 1.45 1.30 1.62 Ref 3.32 2.71 4.07 ED visits (past 6 months) 0 1 >2 Ref 1.24 1.06 1.44 Ref 2.57 2.16 3.05 Depression No Yes eliminated eliminated Back pain No Yes Ref 1.62 1.47 1.79 Ref 2.82 2.44 3.26 Cancer No Yes Ref 1.57 1.29 1.90 Ref 1.87 1.48 2.36 Insomnia No Yes Ref 1.26 1.11 1.42 Ref 1.27 1.08 1.49 Cigarette use (lifetime) No Yes eliminated eliminated Hazardous alcohol use ( past 30 day s) No Yes Ref 1.07 0.95 1.21 Ref 1.48 1.25 1.74 Marijuana use (lifetime) No Yes Ref 1.76 1.59 1.95 Ref 1.28 1.10 1.48

PAGE 80

80 Table 3 3. Continued Characteristic Lifetime, no t past 30 day Rx opioid use (vs. no use) N=3,463 Past 30 day Rx opioid use (vs. no use) N=1,286 aOR 95% CI aOR 95% CI Rx sedative use (lifetime) No Yes Ref 2.67 2.34 3.04 Ref 3.96 3.35 4.68 RELATIONSHIP LEVEL FACTORS Marital status Married Never married Separated, divorced, or widowed Ref 0.72 0.98 0.63 0.83 0.85 1.12 Ref 0.78 0.96 0.63 0.95 0.79 1.16 Children No Yes Ref 1.26 1.12 1.42 Ref 1.37 1.14 1.63 Employment Yes No Ref 0.90 0.81 1.00 Ref 1.47 1.24 1.73 Social media use No Yes Ref 1.32 1.19 1.46 Ref 0.92 0.79 1.07 COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated a OR= adjusted Odds Ratio; CI= Confidence Interval ; ref= reference group; N= sample size AIC= Akaike information criterion; intercept only 18325.975; intercept with covariates 15859.524 2 (df=38)= 2542.4512; p<.0001

PAGE 81

81 Table 3 4. Association between select risk factor s and self reported prescription opioid use pattern among HealthStreet members, by sex, 2011 2018 (N = 9,221 ) Characteristic Women (N=5,549 ) Men (N =3,672 ) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) P ast 30 day Rx opioid use (vs. no use) aOR (95% CI) Lifetime, no t past 30 day Rx opioid U se (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years Ref 1.11 (0.96 1.28) Ref 1.31 (1.07 1.61) Ref 1.07 (0.90 1.28) Ref 1.08 (0.84 1.39 ) Race White Black Other Ref 0.80 (0.69 0.92) 0.89 (0.68 1.15 ) Ref 1.24 (1.01 1.52 ) 0.82 (0.55 1.23 ) Ref 0.47 (0.39 0.55) 0.48 (0.35 0.67 ) Ref 0.79 (0.61 1.01 ) 0.61 (0.38 0.99) Education More than HS HS or less Ref 0.69 (0.61 0.79) Ref 0.87 (0.72 1.05 ) Ref 0.70 (0.59 0.83) Ref 0.99 (0.78 1.26 ) Health insurance Yes No Ref 0.92 (0.80 1.05) Ref 0.71 (0.58 0.87) e liminated eliminated Doctor visits (past 6 months) No Yes Ref 1.32 (1.14 1.53) Ref 3.14 (2.39 4.14) Ref 1.68 (1.42 2.00 ) Ref 3.74 (2.76 5.05 ) ED visits (past 6 months) 0 1 >2 Ref 1.27 (1.05 1.53 ) Ref 2.73 (2.19 3.39 ) Ref 1.20 (0.93 1.53 ) Ref 2.35 (1.77 3.11) Depression No Yes eliminated eliminated eliminated eliminated Back pain No Yes Ref 1.83 (1.61 2.08 ) Ref 2.93 (2.44 3.52 ) Ref 1.39 (1.18 1.63 ) Ref 2.81 (2 .22 3.56 ) Cancer No Yes Ref 1.81 (1.44 2.30 ) Ref 2.16 (1.63 2.88 ) eliminated eliminated Insomnia No Yes eliminated eliminated Ref 1.40 (1.16 1.70 ) Ref 1.14 (0.88 1.49 ) Cigarette use (lifetime) No Yes eliminated eliminated eliminated eliminated Hazardous alcohol use ( past 30 day s) No Yes eliminated eliminated Ref 1.13 (0.95 1.35 ) Ref 1.66 (1.31 2.11 ) Marijuana use (lifetime) No Yes Ref 1.86 (1.63 2.11) Ref 1.43 (1.19 1.71 ) Ref 1.79 (1.52 2.12 ) Ref 1.23 (0.97 1.56)

PAGE 82

82 Table 3 4 Continued Characteristic Women (N=5,549) Men (N=3,672) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) Rx sedative use (lifetime) No Yes Ref 2.83 (2.41 3.33) Ref 4.27 (3.48 5.24) Ref 2.71 (2.18 3.36) Ref 4.02 (3.06 5.29) RELATIONSHIP LEVEL FACTORS Marital status Married Never married Separated, divorced, . or Widowed Ref 0.74 (0.62 0.88) 1.00 (0.84 1.18) Ref 0.92 (0.71 1.19) 1.08 (0.85 1.36) Ref 0.63 (0.51 0.78) 0.94 (0.75 1.18) Ref 0.52 (0.38 0.70) 0.72 (0.53 0.98) Children No Yes Ref 1.46 (1.25 1.72) Ref 1.67 (1.31 2.14) eliminated eliminated Employment Yes No Ref 0.87 (0.76 0.99) Ref 1.48 (1.20 1.81) Ref 0.95 (0.80 1.13) Ref 1.50 (1.14 1.97) Social media use No Yes Ref 1.41 (1.24 1.60) Ref 1.02 (0.85 1.22) eliminated eliminated COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated eliminated eliminated aOR= adjusted Odds Ratio; CI= Confidence Interval; ref= reference group; N= sample size Male model AIC= Akaike information criterion; intercept only 7001.148; intercept with covariates 6084.750 LRT= Likelihood ratio 2 (df= 28)= 972.3983; p<.0001 Female model AIC= Akaike information criterion; intercept only 11223.899; intercept with covariates 9787.928 2 (df= 32)= 1499.9705; p<.0001

PAGE 83

83 CHAPTER 4 PATTERNS OF PRESCRIPTION OPIOID USE AND RISK FACTORS FOR USE AMONG OLDER AND YOUNGER WOMEN IN THE COMMUNITY Background The prevalence of past 12 month opioid use among adults in the United States (US) is 38%, representing approximately 92 million users (Han et al., 2017b) . Though prescription opioids are commonly prescribed to treat pain (Beaudoin et al., 2014; Caudill Slosberg et al., 2004) , use has become a public health concern in the US due to an increase in related morbidity and mortality (Compton et al., 2016a) . In the last two decades, there was an increase in prescribing from 107.3 million prescriptions written in 1992 to 246.2 million in 2015 (Pezalla et al., 2017) . This increase had led to a number of negative consequences related to prescription opioid use including non medical use (i.e., use of higher doses, use longer than prescribed, use for the experience or feeling caused , use of someone else's medication (McCabe et al., 2013b) ) and overdose. Though more recently the overall national prescribing rate has decreased to the lowest it has been in the last decade , certain regions across the US continue to have to have high prescribing rates (Centers for Disease Control and Prevention, 2018e) . An estimated 2 million people have been diagnosed with an opioid use disorder (Han et al., 2017b) and approximately 130 people d ie every day from drug overdoses involving an opioid in the US (Centers for Disease Control and Prevention, 2018g; National Institute on Drug Abuse, 2018b) . Heroin, prescription opioids, and synthetic opioids including fentanyl are responsible for the increase in overall overdose deaths in the last decade (Compton, Jones, & Baldwin, 2016b; Centers for Disease Control and Prevention, 2017; Rudd et al., 2014) . Though prescription opioid use is affecting the health of many people living in the US , women bear a h eavier burden of the consequences related to prescription opioid use compared to their male counterparts. The Centers for Disease Control and Prevention reports a higher national

PAGE 84

84 prevalence rate for past 30 day prescription opioid use for women co mpared to men (7.2% vs 6.3%; Centers for Disease Control and Prevention, 2015) , and the National Survey on Drug Use and Health (NSDUH) reports higher p ast 12 month use for women compared to men (35.3% vs. 40.2%; Han et al ., 2017b) . Although men are more likely to overdose from drugs than women (Calcaterra, Glanz, & Binswanger, 2013) , the gender gap is closing and opioid overdoses have greatly increased among women. Fatal opioid overdo ses increased 400% among women compared to 265% for men from 1999 to 2010 (Centers for Disease Control and Prevention, 2018d) . O ur previous work conducted among community members in Florida has also shown women have a higher lifetime prevalence rate of prescription opioid use compared to men (54.9 vs. 42.2; Serdarevic, Striley, & Cottler, 2017) . This may partly be due to higher prescribing rat es (Simoni Wastila, 2000) . The a dverse consequences amo ng women may be partially explained by the higher doses being prescribed to women (Campbell et al., 2010; Williams, Sampson, Kalilani, Wurzelmann, & Janning, 2008) . Further, s ex specific health risks associated with prescription opioid use among women include amenorrhea and infertility (Daniell, 2008; Fillingim et al., 2005) . However, the role of opioid use in fe male fertility has not been well established (Alom, Wymer, & Trost, 2018) and warrants further examination . This may suggest that the negative consequences related to prescripti on opioid use are not just life threatening but can also have life changing impacts for women Abdolahifard, & Jahromi, 2013) . Additional consequences of prescription opioid use among women include a greater risk of addiction (Darnall et al., 2012) and a risk that their infants may develop N eonatal Abstinence Syndrome (NAS) (Patrick et al., 2015) . Though these health disparities are present, there h as been little attention in the literature regarding prescription

PAGE 85

85 opioid use among women and this further highlights the need to examine pres cription opioid use among women specifically . Age has also been associated with prescription opioid use with o lder adults represent ing the largest consumers of prescription medications (Han et al., 2017b; Olfson et al., 2015; Schepis & McCabe, 2016) . C hronic pain , another risk factor for prescription opioid use, is common among older adults with some epidemiologic studies reporting an age related increase in the prevalence of pain with age (Herr, 2002; Parsells Kelly et al., 2008) . Specifically, o lder women have a higher prevalence of opioid use , and adults who are 65 and older ( 19% of men and 23% of women ) take at least five prescription drugs a week (Campbell et al., 2010) . In general, opioid use is understudied in older adults and information on such use is limited (US Health and Human Services, 2017) . Prescription opioid u se among younger women is also associated with adverse consequ ences. S pecifically , consequences such as NAS and infertility affect younger women which can have life changing outcomes. A previous study conducted in Florida found that prevalence of NAS increased dramatically between 2000 and 2010 , from 1.6 to 25.2 per 1000 live births (Wang et al., 2017) . In addition, it is known that the most serious adverse consequences related to opioid use , such as overdose, are prevalent among young adults (Martins & Ghandour , 2017; White, Hingson, Pan, & Y i, 2011) . In 2016, the highest rate of fatal overdose occurred among 25 34 year olds (Centers for Disease Control and Prevention, 2017b) . Young adults also have the highest prevalence of non medical use of prescription opioids in the US (National Institute on Drug Abuse, 2018d; A. Young, McCabe, Cranford, Ross Durow, & Boyd, 2012) . In addition to prescription opioid use, women ar e more likely than men to be prescribed benzodiazepine s or other sedatives (Hearon et al., 2011) which may also be partially due to

PAGE 86

86 higher prevalence of mood disorders reported by women (Bodnar & Wisner, 2005; Kessler, 2006) . More troubling is the concurrent use of opioid s and sedative s , which is also more prevalent among women (Hwang et al., 2016; Saunders et al., 2012) . Approximately 30% of fatal opioid overdoses involved a benzodiazepine , which may suggest some of the increase in opioid related deaths might be caused by increas es in concurrent benzodiazepine and opioid use (Saunders et al., 2012; Sun et al., 2017) . This ma y suggest that there is a n eed to address opioid use among women specifi cally , and that prescribers should be more vigilant of the concurrent use of benzodiazepines and opioids among women . One approach to examine prescription opioid use behavior among women is within the context of the socio ecological model. This model post ulates that factors on the individual, relationship, community, and societal levels interact to influence behavior (McLeroy et al., 1988) . Though behaviors on the individual level are influenced through personal tra its, behaviors are also influenced by relationships with others, and by communities in which people live (Substance Abuse and Mental Health Services Administration, 2016a) . Examples of factors on the individual level include biolo gical and personal factors , while relationship level factors consist of close relationships with others that influence behavior. Finally, factors on the community level refer to where people live and recreate. Factors at all levels influence behavior such as prescription opioid use, and to sustain prevention efforts over time multiple levels are considered within the context of the model instead of focusing on a single level intervention (Richard et al., 1996; Stokols, 1996) . D ifference s in prescription opioid use among women may arise because behavior is influenced by factors that differ by age which can be explained w ithin the contex t of the socio ecological model . At the individual level, for example, the prevalence of chronic health

PAGE 87

87 conditions that cause pain differ s by age, with chronic conditions that cause pain being more prevalent among older adults (Reid, Eccleston, & Pillemer, 2015) which may subsequently lead to higher rates of prescription opioid use in older adults. Other factors on the individual level include health insurance (Chang et al., 2018) , and medical care visits (i.e., doctor, emergency depa rtment (ED ) ) which have been associated with an increase in prescription opioid use. In addition, other chronic conditions including a history of insomnia and depression ( (Blanco et al., 2016; Pinkerton & Hardy, 201 7; Serdarevic, Osborne, et al., 2017; Sullivan et al., 2006) ) as well as substance use (Gudin et al., 2013; Stein et al., 2017; Sullivan et al., 2006) have been associated with increased prescription opioid use on the individual l evel. Marital status (being single) , having children, being employed , and social media use all have been associated with increased prescription opioid use (Fillingim et al., 2005; Inciardi et al., 2007; Piko, 2000; Rabinovitch et al., 2013; Rönkä & Katainen, 2017 ; Rozenbroek & Rothstein, 2011) at the relationship level. On the community level, rurality has been previously associated with increased prescription drug use (Keyes et al., 2014) . To further examine prescription opioid use among women, we examined individual, relationship and community level risk factors for prescription opioid use among older and younger women separa tely. Based on previous research on prescription opioid use among women and the theoretical framework, we hypothesized that risk factors at the individual level (chronic conditions and co morbid substance use) compared to risk factors at the relationship l evel would be more strongly associated with prescription opioid use (both for past 30 day use and lifetime, not past 30 day use ) among older women compared to younger women. We also hypothesize d that prescription sedative use would be the strongest risk factor for opioid use and that the magnitude of effect would be stronger among older women compared to younger

PAGE 88

88 women. This is based our work in Chapter 3 of this dissertation and may be due to the higher prevalence of comorbid health conditions that occur among older women compared to younger women such as depression, anxiety, and pain (Darnall et al., 2012; Koons, Rayl Greenberg, Cannon, & Beauchamp, 2018; Malatesta, 2007) which subsequently may require sedatives and opioids for treatment. To determine if any effects are not due to confounding we control led for f actors at the individual, relationship, and community level and test ed these hypotheses within a theoretical framework. Though some previous studies have found women are more likely than men to endorse prescription opioid use and that use increases with ag e for women, opioid use among women remains understudied in the literature . In particular, many studies have not focused on data at the community level, which may provide more granular information on the risks associated with prescription opioid use among women . Thus, these current analyses will examine factors for varying patterns of prescription opioid use on individual, relationship and the community level among a community sample of women in North Central Florida. Aims and Hypotheses The aim of this analysis was to examine risk factors among older and younger community dwelling women in North Central Florida . Specifically, this analysis stratified by age [younger (< 50 years ) vs older ( 50 + years )] among women only, to examine individual , relationship , and community level risk factors for prescription opioid use patterns ( past 30 day use, lifetime but not past 30 day use, and no use). The cut off for age (50 years) was used to categorize women into pre and postmenopausal. Menopause typically occurs in women between 45 and 55 years of age, thus, examining those aged 50 years and over (the midpoint of the range) is appropriate (Food and Drug Administration, 2018; Gold, 2011) . Two hypotheses were proposed among 5 ,549 women (15% (n=8 31 ) with past 30 day use, 41% (n= 2, 267 ) with lifetime but not past 30 day use, and 44% (n= 2, 451 ) with no use of prescription opioids).

PAGE 89

89 Hypothesis 1.1. Risk factors at the individual level (chronic conditions and co morbid substance use) compared to risk factor s at the relationship level would have a stronger association for prescription opioid use (both for past 30 day use and lifet ime u se, but not past 30 day s ) among older women compared to younger women. Hypothesis 1.2. Prescription sedative use would be the strongest risk factor for prescription opioid use among both older and younger women; the association between sedative use and pre scription opioid use will be stronger among older women compared to younger women. Methods To focus on prescription opioid use among women, a large community sample with comprehensive data on demographic information, health conditions, and history of drug use was used . V ariables collected from a community engagement program, HealthStreet, are included in our analyses and are outline below . Study Population HealthStreet was established at the University of Florida as a major effort of the Clinical Translati onal Science Institute (CTSI). HealthStreet was first developed in St. Louis at Washington University to link participants to drug and alcohol related research and then was relocated to the University of Florida in 2011, which has since expanded to focus on all health problems in the community. HealthStreet utilizes a Community Health Worker model. Community Health Workers ( CHWs) are trained and certified to directly engage community members at parks, grocery stores, churches, laundromats, and health fairs health needs and concerns. Since November 2011, using this CHW model (Cottler et al., 2011) , over 10,293 community members in the HealthStreet catchment area (Gainesville, Jacksonville and Miami) have been dire ctly engaged.

PAGE 90

90 CHWs approach people in various locations in the community on a daily basis at different times and introduce themselves and the HealthStreet model. Those who agree to become members of HealthStreet provide written consent and are then assesse d using a health needs assessment. The health needs assessment is conducted by the CHWs in a 30 minute face to face interview and is recorded on a paper copy in the field during outreach. During the interview, CHWs collect information regarding health cond itions, health concerns, and history of substance from each new participant. Once they have undergone the health needs assessment they become members of the HealthStreet community cohort . HealthStreet members who self identified as female, aged 18 years an d older, and enrolled between November 2011 and June 2018 were eligible for this analysis. This study was approved by the University of Florida Institutional Review Board. Measurements Responses from the interview are recorded on a paper copy of the health needs assessment. The health needs assessment includes a section that focuses on questions regarding drug use including prescription drug use. The outcome variable, prescription opioid use, was ications like Vicodin®, oxycodone, codeine, Demerol®, morphine, Percocet®, omen who indicated that they used a pre scription pain medication were subsequently asked if they used one of those prescription medications in the past 30 da y s. The three patterns for prescription opioid use were (1) past 30 day use (those who reported using prescription opioids in the past 30 day s), (2) lifetime but not past 30 day use (women who reported using prescription opioids, but not in the past 30 day s), and (3) no use (women who reported no use in their lifetime). Risk factors at the indi vidual level elicited from the Health Needs A ssessment include d age (continuous variable ranging from 18 to 94 years of age), race (white, black/African

PAGE 91

91 American, ot her), level of educational attainment (more than high school or high school and less), health insurance (yes or no), doctor visit in the last 6 months (yes or no), fre quent emergency department visits in the last 6 months (>2 v isits; yes or no), depression (yes or no), anxiety (yes or no), back pain (yes or no), cancer (yes or no), insomnia ( yes or no), and substance use including lifetime cigarette smoking (yes or no), past 30 day hazardous al cohol use (yes or no), lifetime marijuana use (yes or no), and lifetime prescription sedative use (yes or no). Several relationship level risk factors were elicited through the health needs assessment: marital status (never married; currently married; or separated, divorced, or widowed), number of children (continuous variable), employment (full time/ part time ; yes or no ), and use of any social media (Twitter, Facebook, or Instagram; yes or no) . The community level risk factor was assessed using the he alth needs assessment . Analysis Participants under 18 years of age, and men, and women for whom prescription opioid use or zip code data were missing were excluded from these analyses (n=4, 744 ). Descriptive statistics were calculated to summarize lifetime prescription opioid use (past 30 day use, lifetime but not past 30 day use, and no use). The association between individual , relationship , and community level variables and opioid use were a nalyzed using chi square tests for categorical variables (significance was determined based on Bonferroni corrected p value of 0.002 after dividing alph a 0.05 by 20 , the number of variables) . Due to their non normality and non unimodal distribution, age an d number of children were transformed into categorical groups (Figure 4 1 and 4 2) . Age was categorized into two groups (18 49 years and 50+ years) and the number of children were categorized as yes (>1) or no ne (0).

PAGE 92

92 To test t he moderating effect of pain on the relationship between age (c ategorized as 18 49 years and 50+ years ) and prescription opioid use the first model was built with age and pain, but without any interaction terms. The second model included pain, age, and the interaction term between pa in and age. When interaction terms were added to the model, there was no appreciable difference in the odds ratios compared to the model without the interaction terms (these results are displayed in the Appendix, Table A 2). For ease of interpretation, the final model without the interaction terms was used. In addition, the moderating effect of prescription sedative use (categorized as ever or never) on the relationship between age and prescription opioid use was also assessed using this approach and the fi nal model without the interaction term was used . Importantly, tests that assess interaction have low power and so even though no evidence of interaction was found, we cannot confirm it is not present (Cronbach & Snow, 1977; Greenland, 1983; Marshall, 2007) . In the end , all variab les were fitted individually in the final model with no interaction term. Analyses were conducted for older and younger women separately to examine age specific effects driven by our hypotheses, based on knowledge from the literature and expected age relat ed differences arising from the socio ecological model. Multinomial logistic regression was used to estimate adjusted odds ratios (ORs) and 95% confidence intervals (with no use of prescription opioids as the referent group). Among all women, the first se t of models examined age as a risk factor for lifetime and past 30 day prescription opioid use. The first model was built using factors only at the individual level. Variables at the individual level that were not significant were eliminated from the secon d model, which added relationship level factors. Significance was determined by examining each The final model with community level factors excluded non significant variables from both the individual and relationship levels.

PAGE 93

93 Moderation of the effect of age by pain, and age by sedative use, was assessed and no evidence for moderation was observed ( p> 0.05 ). The same methods were then applied after stratifying by age to build two specific models for younger women and older women. Model fit was assessed using the likelihood ratio test comparing the null model to each of our models. All model s were statis tically significant and fit the data better than the null model. The variance inflation factor (VIF) was assessed for severity of multicollinearity; all VIF values were under 10 thus no variables were removed. All statistical analyses were conducted using SAS® 9.4 (SAS Institute Inc., 9.4, Cary, NC: SAS Institute Inc., 2011) . The calcu lation of power was based on effect sizes that might explain as little as 1% of the difference in opioid use between younger and o lder women. Given the sample had 3,136 younger women and 2, 413 older women, we had statistical power greater than 98% at a 0.0 5 significance level. Therefore, the sample size had sufficient power to detect an expected odds ratio minimum of 1.17. Results As of June, 2018, o f the 10,293 members of HealthStreet, 5, 549 women qualified for inclusion in the current study (Figure 4 1), 57. 1 % of whom were black , 4 6.4 % of whom c ompleted m ore than a high school degree, 3 6.0 % of whom were employed, and 2 3.0 % of whom wer e currently married (Table 4 1). Among all women, 2, 451 (44 .1 %) reported no us e of prescription opioids, 2, 267 (4 0.9 %) reported lifetime but not past 30 day use of pre scription opioids, and 8 31 (15 .0 %) reported past 30 day opioid use. Among 3,327 younger women, 1, 560 (50 .0 %) reported no us e of prescription opioids, 1, 199 (38 .0 %) reported lifetime use but not in the past 30 days, and 377 (12 .0 %) reported past 30 day use. Among 2, 413 older women, 891 (37 .0 %) reported no use, 1, 068 (44 .0 %) reported lifetime u se but not in the past 30 days, and 4 54 (19 .0 %) reported past 30 day prescription opioid use (Table 4 2) .

PAGE 94

94 All examined risk factors were significantly associated with prescription opioid use outcome before stratification by age with the exception of hazard ous alcohol use . Race, education, health insurance, past 6 month doctor and ED visits, history of anxiety, depression, back pain, cancer, insomnia, and lifetime cigarette, marijuana, and sedative use were all significant individual level risk factors. On t he relationship level, marriage, having children , being employed and using social media were significant risk factors . Rural area of residence, the community level risk factor, was also found to be a significant risk factor for prescription opioid use (Tab le 4 1). Characteristics of women, stratified by prescription opioid use pattern As outlined in Table 4 1, age was significantly associated with prescription opioid use at the individual level . Specifically, both lifetime and past 30 day opioid use rs were older than non opioid users ( p < 0 .0001). A larger proportion of women who endorsed pas t 30 day and lifetime prescription opioid use were white compared to non users . Women who endorsed prescription opioids in the past 30 day s reported higher rates o f insurance , past 6 month doctor visits, and frequent ED visits compared to those who used in their lifetime or those who did not use prescription opioids (<.0001). A linear association between d epression, anxiety, back pain, cancer, insomnia , and recency of prescription opioid use was observed . P revalence of depression was the least prevalent among those who reported no use (22. 8 %), followed by those who used in their lifetime, but not past 30 days (37. 7 %), and most prevalent among those who used in the pa st 30 days (47. 6 %). Similar linear associations were observed for anxiety (past 30 day use= 44. 6 %, lifetime no past 30 day use = 34. 9 %, no use= 18.0 %) , back pain (past 30 day use= 68. 6 %, lifetime no past 30 day use = 53.0 %, no use= 33.6 %) , cancer (past 30 day use= 17.0 %, lifetime no past 30 day use= 13. 6 %, no use= 5.3 %) , and insomnia (past 30 day use = 4 1.5 %, lifetime no past 30 day us e = 33. 5 %, no use= 18. 5 %) . Cigarette, marijuana, and sedative use significantly

PAGE 95

95 differed by prescription opioid use, with thos e who endorsed past 30 d ay and lifetime prescription opioid use reporting higher rates of them compared to those who reported no use of prescription opioids (p< 0 .0001). In terms of relationship level factors, being separated, divorced, or widowed was more prevalent among those who used prescription opioids regardless of recency compare d to women who reported no use ( p< 0 .0001) . Having children was more prevalent among those who used prescription opioids in the past 30 day s (children=85. 7 %) and used in their lifetime (children=7 9.1 % ) compare d to women who reported no use (children=71. 2 %). Employment was less prevalent among those who reported past 30 day use compared to those who endorsed lifetime but not past 30 day use and no use ( p< 0 .0001) . Social media use was more prevalent among those who reported lifetime but not past 30 day use compared to those who reported past 30 day or no use. At the community level , after correcting for multiple comparisons, living in a rural area was not associated with a pres cription opioid use. Characteristics and prescription opioid use among women, stratified by age When women were stratified by age, a higher proportion of older women were found to endorse past 30 day use (1 9.0 %) and lifetime use of prescription opioids ( 44. 0 %) compared to younger women (12. 0 %, 38. 0 %, respectively) (Table 4 2) . For older and younger women , a higher p r o po rtion of past 30 day and lifetime but not past 30 day user s were white compared to non users . Both older and younger women who endorsed lifetime but not past 30 day prescription opioid use were more educated than women who endorsed past 30 day and no use . Among older women, as recency increased so did prevalence in which past 30 day users were more likely to report having health insurance, visiting the doctor in the past 6 mon ths, and frequent ED visits . The same pattern was observed for younger women. Linear associations between having depression, anxiety, back pain, cancer, insomnia, cigarette, and sedative use , and

PAGE 96

96 recency of prescriptio n opioid use were observed among both older and younger women. Among both younger and older women , a higher proportion of those who reported past 30 day and lifetime not past 30 day use endorsed marijuana compared to those who did not use prescription opio ids. In regards to relationship level factors, younger women who reported no use were more likely to have never been married compared to women who endorsed lifetime not past 30 day use and past 30 day use . O lder women who reported past 30 day use had highe r rates of being separated, divorced, or widowed compared to women who reported no use . Having children was more prevalent among younger women who used prescription opioids in the past 30 day s and lifetime not past 30 day use compared to those who reported no use (< 0 .0001). This relationship was not observed among older women (p=0.0858). Regardless of age, there was reduced employment among women who endorsed past 30 day use compared to lifetime not past 30 day use and no use. Both older and younger women w ho endorsed lifetime not past 30 day use compared to no use had higher rates of social media use. Upon examining the community level fa ctor of rural area of residence, we found rurality was non sig nificant when stratified by a ge regardless of opioid use pattern. Risk factors and self reported prescription opioid use pattern among women The final model , outlined in Table 4 3 included all significant risk factors on the individual, relationship, and community level ( with depress ion, insomnia, cigarette use , past 30 day hazardous alcohol use , and rurality eliminated from this model). The strongest association for lifetime no t past 30 day use of prescription opioids was prescription sedative use in lifetime , with those endorsing pr es cription sedative use (vs no prescription sedative use) having a 2.83 times greater likelihood of lifetime use of prescription opioids (95% Confidence Interval (CI): 2.41 3.33 ) than non prescription opioid users . The second largest effect observed was li fetime

PAGE 97

97 marijuana use ( adjusted Odds Ratio (aOR)= 1.86; 95% CI, 1.63 2.11 ). Following marijuana use , history of back pain was the third strongest association for lifetime no t past 30 day use of prescription opioids ( aOR= 1.83 ; 95% CI, 1 .62 2.08 ). After adju stment for covariates, additional risk factors at the individual level for lifetime but not past 30 day prescription opioid use included visiting t he doctor (aOR = 1.32; 95% CI, 1.14 1.53), frequent ED visits (aOR= 1.27; 95% CI, 1.05 1.53), and a history of cancer (aOR= 1.82; 95% CI, 1.44 2.30 ). At the relationship level , factors included having children (aOR= 1.46; 95% CI, 1.25 1.72) and social media use (aOR= 1.41; 95% CI, 1.24 1.60 ). At the individual level, black race and less educational attainment were protective factors for lifetime not past 30 day prescription opioid use. Never being married and employment were protective factors at the relationship level. For past 30 day use, o lder age was a risk factor. Older women were s ignificantly more likely to report past 30 day prescription opioid u se compared to younger women . P rescription sedative use was also the strongest risk factor for past 30 day prescription opioid use ( aOR= 4.27 ; 95% CI, 3.48 5.24 ). The second strongest association for past 30 day prescription opioid use was past 6 month doctor visits (aOR= 3.14; 95% CI, 2.39 4.14) while history of b ack pain was the third strongest association , with those who had a history of back pain (vs. no back pain) having a 2.93 times greater likelihood of past 30 day presc ription opioid use (95% CI, 2.44 3.52 ) compared to those who did not use prescription opioids . Additional risk factors for past 30 day prescription opioid use at the indivi dual level include d black race ( aOR= 1.24 ; 95% CI, 1.01 1.52), frequent ED visits ( aOR= 2.73 ; 95% CI, 2.19 3.39 ), history of cancer ( aOR= 2.16 ; 95% CI, 1.63 2.88 ), and lifetime marijuana use ( aOR= 1.43 ; 95% CI , 1.19 1.71 ). At the relationship level, having children ( aOR= 1.67 ; 95% CI, 1.31 2.14 ) and being employed ( aOR= 1.48 ; 95% CI, 1.20 1.81 ) were risk factors for past 30 day use. Not h aving health insurance was a protective

PAGE 98

98 factor for past 30 day prescription opioid use. Women who did not have health ins urance (vs health insurance) had a 29% decrease in odds of past 30 day prescription opioid use . Risk factors and self reported prescription opioid use pattern among women, stratified by age A fter controlling for other significant covariates, differences in risk factors among older and younger women for prescription opioid use at the individual and relationship level were observed , but not at the community level (Table 4 4). At the individual level , history of depression, insomnia, and cigarette use were non significant risk factors and thus were eliminated for both older and younger women. At the relationship level, marital status was not significant regardless of age, thus , it was eliminated. At the individual level, hazardous alcoh ol use , and at the relationship level , having children were non significant risk factors for older women and were eliminated from the final model . Rural area of residence was not found to be a significant risk factor for either older or younger women after controlling for ot her covariates. Among younger women (Table 4 4), lifetime prescript ion sedative use was the strongest risk factor for lifetime not past 30 day use of prescription opioids. Younger women who endorsed any lifetime prescription sedative use (vs. no prescrip tion sedative use) had a 2.96 times greater likelihood of prescription opioid use in their lifetime but not past 30 days compared to their counterparts who endorsed no prescription opioid use (aOR= 2.96; 95% CI, 2.37 3.70 ). T he second strongest risk factor for lifetime but no t past 30 day prescription opioid use was having children ( aOR= 1.9 1 ; 95% CI, 1. 5 8 2.3 0 ). At the individual level, visiting the doctor (aOR= 1.27; 95% CI, 1.06 1.53 ) , frequent ED visits (aOR= 1.26; 95% CI, 1.00 1.59 ) , history of back pain (aOR= 1.80; 95% CI, 1.52 2.13 ) and cancer (aOR= 1.58; 95% CI, 1.05 2.36 ) , and marijuana use (aOR= 1.88; 95% CI, 1.58 2.23 ) were also found to be risk factors for lifetime but not past 30 day use. At the relationship level, in additio n to having children, the use of social

PAGE 99

99 media was a risk factor for lifetime but not past 30 day use (aOR= 1.31; 95% CI, 1.11 1.54 ) . Protective factors for lifetime but not past 30 day use included black race, less educational attainment, and un employment. The strongest risk factor for past 30 day prescription opioid use among younger women was sedative use ( aOR= 4.84 ; 95% CI, 3.59 6.51 ) . In contrast to lifetime use, the second strongest association for past 30 day use was back pain. Those with back pain (v s. no back pain) had a 3.01 times greater likelihood of past 30 day use of prescription opioids compared to those who did not use prescription opioids ( aOR= 3.01 ; 95% CI, 2.3 2 3.9 0 ). In addition, doctor (aOR= 2.72; 95% CI, 1.92 3.84 ) and frequent ED visits (aOR= 2.84; 95% CI, 2.14 3.78 ) , history of cancer (aOR= 1.58; 95% CI, 1.05 2.36 ) , past 30 day hazardous alcohol use (aOR= 1.57; 95% CI, 1.17 2.10 ) , and lifetime marijuana use (aOR= 1.41; 95% CI, 1.08 1.83 ) were risk factors at the individual level. Having children (aOR= 1.93; 95% CI, 1.42 2.64 ) and social media use (aOR= 1.31; 95% CI, 1.00 1.73 ) were risk factors at the relationship level for past 30 day use. Protective factors for past 30 day prescription opioid use included less education and having no health insurance at the individual level. Among older women, t he strongest association for lifetime but not past 30 day use was prescription sedative use (aOR= 2.70 ; 95% CI, 2.13 3.41) while the second largest association was a history of cance r (aOR= 1.91 ; 95% CI, 1.43 2.5 6). Risk factors that were found to be significant for lifetime ( not past 30 day use ) prescription opioid use after adjustment for other covariates included doctor visits (aOR= 1.45 ; 95% CI, 1.14 1.86 ) , history of back pain (aOR= 1.86 ; 95 % CI, 1.53 2.26 ) , and marijuana use (aOR= 1.74 ; 95% CI, 1.42 2.13 ) . At the relationship level, social media use was a risk factor for lifetime but not past 30 day use (aOR= 1.69 ; 95% CI, 1.38 2.07) . Protective factors for lifetime but not past 30 day use were black race and less educational attainment. T he strongest risk factor for past 30 day use of prescription

PAGE 100

100 opioids was past 6 month doctor visits; older women who visited the doctor in the past 6 months (vs. no doctor visits in the past 6 months) had a 4.1 5 times greater likelihood of prescriptions opioids in the past 30 day s (95% CI, 2.6 2 6. 60 ) compared to older women who did not use prescription opioids . The second strongest association for past 30 day prescription opioid use was prescription sedative us e (aOR= 3. 76 ; 95% CI, 2.8 2 5.0 1 ). Additional risk factors at the individual level included frequent ED visits (aOR= 2.72; 95% CI, 1.91 3.88 ), history of back pain (aOR= 2.91; 95% CI, 2.24 3.78 ) and cancer (aOR= 2.29; 95% CI, 1.61 3.24 ) , and lifetime mar ijuana use (aOR= 1.32; 95% CI, 1.02 1.72 ) . At the relationship level we found unemployment (aOR= 1.85; 95% CI, 1.34 2.55 ) and social media use (aOR= 1.43; 95% CI, 1.09 1.87 ) were risk factors for past 30 day use. The only protective factor for past 30 day prescription opioid use among older women was lack of health insurance (aOR= 1.85; 95% CI, 1.34 2.55 ) . Discussion The aim of this analysis was to examine risk factors on the individual, relationship, and co mmunity level for patterns of prescription opioid use (past 30 day use, lifetime but not past 30 day use, and no use) among older and younger women separately. Prior to adjustment for other covariates, most factors associated with prescription opioid use p atterns remained significant after stratifying by age. Two separate models (one for older women and one for younger women) allowed us to adjust for other covariates . After controlling for other factors, we found risk factors for prescription opioid use amo ng older and younger women differed . Among the entire sample of women, 15 .0 % reported past 30 day use, 4 0.9 % re ported lifetime but not past 30 day use and, and 44 .1 % reported no use of prescription opioids. Compared to the national rate for women, the pre valence rat e of past 30 day prescription opioid use in our community sample of women was more than double ( 15 .0 % vs 7.2%; Centers for

PAGE 101

101 Disease Control and Prevention, 2015) . Further examination allowed us to control for factors at different levels of the socio ecological model and revealed older women compared to younger women were more likely to endorse past 30 day use of prescription opioids compared to never use ( aOR= 1.3 1 ; 95% CI, 1.07 1.6 1 ). However, this association w as not observed for lifetime ( no t past 30 day use ) . This finding is in line with the existing literature ; older women have higher rates of pres cription opioid use compared to younger women (Campbell et al., 2010; US Health and Human Services, 2017) . This relationship may be partly due to the natural onset of painful chronic conditions that occur as individuals age, thus women are more likely to be exposed to and to use prescription opioids to treat pain as they age. In addition, some studies have found older women are more likely to utilize healthcare services (Ma ckenzie, Gekoski, & , which may provide access to prescription opioids . O ur current results also show an association between prescription opioid use, ED visits, and doctor visit, which may further expos e older women to medical professionals who prescribe opioids . Within the context of the socio ecological model, we hypothesized risk factors at the individual level compared to risk factors at the relationship level will have a greater association for prescription opioid use for older women compared to younger women. Examination of risk factors on the individual, relationship and community level for prescription opioid use r evealed risk factors among older and younger women differed, partially conf irming our hypothesis . Specifically, we found that past 30 day hazardous alcohol use, a risk factor on the individual level, was not significant for older women but was found to be a risk factor among younger women for past 30 day prescription opioid use. Our finding is concordant with the literature; many studies have found alcohol use to be more prevalent among younger populations (Chan,

PAGE 102

102 Neighbors, Gilso n, Larimer, & Alan Marlatt, 2007; National Institute on Alcohol Abuse and Alcoholism, 2018) . However, concurrent use of opioids and alcohol is associated with serious health consequences and has been outlined in our previous work (Serdarevic et al., 2018) . Further examination of such use among younger women is warranted. On the relationship level, having children was not associated with prescription opioid us e among older women. However, among younger women, having children was a risk factor for use. The difference we observed in risk factors at the relationship level among younger and older women may be in part due to the different stages in life these women are transitioning through . For example , r elationship level factors may be more impactful among younger women compared to older women, and have a more integral part of th e lives for young er women who are still of child bearing age and have not yet been married or are newly married. In addition , younger women that are routinely visiting their healthcare provider for prenatal care may have access to prescription drugs such as opioids. Older women may have already had children or are no longer of child bearing age thus; individual level factors (i.e., chronic conditions ) compared to relationship level factors may have a greater association with prescription opioid use among older women . These risk factors for use, which differed at the individual and relationship level among older and younger women, should be considered in the implementation and prevention of prescription opioid use among women. We also hypothesized pr escrip tion sedative use would be the strongest risk factor for prescription opioid use for both older and younger women, with the strength of association being stronger among older women compared to younger women. Though p rescription sedative use in lifetime was the strongest risk factor for both younger and older women, the magnitude of effect was greater among younger women (aOR=2.9 6 ) who used prescription opioids in their lifetime

PAGE 103

103 compared to older women (aOR=2.7 0 ). The same effect was observed for past 30 day use of prescription opioids (aOR=4.8 4 among younger women , and aOR= 3. 76 among older women), therefore only part ially confirming our hypothesis ( t hough the magnitude of effect was larger among younger women than among older women we did not compare them d irectly and cannot say they are significantly different from each other) . We did not hypothesize that this association would be greater among younger women. Nonetheless, this finding is of importance as the combined use of prescription drugs (i.e., opioids and sedatives) can be fatal. The risks associated with combined use of prescription opioids and sedatives includes reduced respiratory function which heightens the risk of overdose (Dowell et al., 2016; Jann et al., 2014; Sun et al., 2017) , increases the risk of ED visits (Day, 2013) , and is associated with a higher all cause mortality (Gaither et al., 2016) . Though the combination of these drugs is dangerous and both the Centers for Disease Control and Prevention (CDC) and the Department of Veterans Affairs/Department of Defense (VA/DoD) recommends against prescribing these drugs together (Depart ment of Veterans Affairs/Department of Defense, 2017; Dowell et al., 2016) , the co prescribing of opioids and sedatives is common in the US (National Institute on Drug Abuse, 2018a) . Several studies have found that co prescribing of sedative and opioid medications is common and that rates range between 12% to 80% (Dasgupta et al., 2016; Paulozzi, Mack, Hockenberry, & Division of Unintentional Injury Prevention, National Center for Injury Prevention and Control, CDC, 2014; Sun et al., 2017) . Specifically, the concurrent use of opioids and sedatives among women is prevalent (Saunders et al., 2012) . We found a strong association between prescription sedative and opioid use which may suggests that wome n in this community sample in Florida may be at increased risk from the concomitant use of opioids and sedatives. T he combined use of opioids and sedatives is one of the motivations for using Presc ription Drug

PAGE 104

104 Monitoring Programs (PDMP) (Centers for Disease Control and Prevention, 2 017d; Winstanley et al., 2018) . However, co prescribing is relatively common and due to the negative consequences related to the combined use of opioids and sedatives the re is a need for continued surveillance and for prevention of such use, especially in this community setting in Florida. Specifically, this further confirms that concurrent use of prescription drugs and additional substance use such as hazardous alcohol use among women in the younger age ranges needs to be further examined. Even without considering the more serious consequence of overdose, there is a risk of long term consequences such as infertility, NAS and depression, which may place increased burden on the healthcare system in such communities, in addition to the effect on quality of life for these younger women. Strengths and Limitations There are a few limitations to t hese analyses. We were not able to establish temporality between prescription opioid use and the other covariates due to the cro ss sectional nature of the data. Thus, causation cannot be established because we do not know if the risk factors preceded the outcome. In addition, prescription opioid use was self reported and we only assessed use of prescription pain medications. Inform ation regarding non medical use, frequency , and duration of opioid use was not available for further examination . However, the potential for social desirability bias was reduced since we did not assess the misuse of prescription opioids. Self reported pres cription drug use could have been over or under reported, which would result in misclassification and bias our results. However, p revious studies have found self report data regarding chronic conditions (i.e., diabetes and hypertension), are often consist ent with medical records (Johansson et al., 1999; Okura et al., 2004) , thus, the potential for misclassification may also be re duced in our sample. Self report data collected for other covariates may be under or over

PAGE 105

105 information bias cannot be ruled out. However, CHWs are able to clarify the m eaning of questions while conducting the assessment and the simplicity we used in our assessment further reduces the potential of information bias. We have included a sufficient number of factors in each model to account for confounding, but confoundin g remains a possibility. Further, the sample size and number of variables included in the regression analyses was sufficient ly large and the number of variables were parsimonious enough to provide robust estimates. The sampling method used to recruit parti cipants to become members of HealthStreet may be affected by selection bias if those who choose to become members are systematically different t han those who choose not to become members of HealthStreet. This may be unlikely because CHWs try to ensure all potential members they approach participate and visit locations within the community at different times and days during the week. The relationship we observed between the risk factors we examined and prescription opioid use may be generalizable to women in other similar communities. However, it should be noted that the data pertaining to prescription opioid use that we analyzed w ere collected over a long time period (November 2011 through June 2018), which may not reflect th e current state of prescribing practices. Finally, we acknowledge that age differences may not be present where the confidence intervals for estimates of effect overlap for younger and older women. Though we found that the relationship between age and opio id use does not depend on pain or sedative use based on our interaction analyses, tests for interaction have low power and even though no evidence of interaction was found, we cannot confirm that it is not present. These age specific effect estimates still provide a useful indication of which factors at individual, relationship and community level are important among younger and older women separately.

PAGE 106

106 There are many strengths of this analysis which includes the examination of differences in risk factors fo r prescription opioid use among women specifically, which is often understudied. These findings not only address risk factors for recent and past use of prescription opioids from a large sample of women but examined risk factors among older and younger wom en separately. This study also provides consistency, a criteria for causation (Hill, 1965) , and has provided information on risk regarding use within this s pecific community based population of women. This sample of women provided a number of variables related to prescription opioid use that we were able to examine at the individual, relationship, and community level within the context of a theoretical model. In addition, the data collected cover a wide range of information on medical and drug use for each woman from a diverse population who are not traditionally represented in research. These findings may be generalizable to similar communities of women in the US to inform strategies and prevention efforts for reducing prescription opioid use. The HealthStreet model also allows for future follow up with community members regarding prescription drug use. Th is allow s us to use the findings from this study to affect positive change on prescription opioid use among women within this community in North Central Florida. Initiatives such as distribution of Deterra® bags (drug deactivation pouches) are already being implemented within this community to combat the ongoing opioid crisis (Cottler, Otufowora, Baker, & Egan, 2018) . Conclusions Compared to national rates, we found higher rates of prescription opioid use in this community sample of women from North Centra l Florida . We also found older women were significantly more likely to report both past 30 day and lifetime use of prescription opioid s compared to younger women. This may be partially due to access. O lder women utilize health care services more frequently which exposes them to medical professionals who prescribe

PAGE 107

107 opioids. Specifically, we observed risk factors for older and younger women. Risk factors for prescription opioid use differed among older women and younger women . Regardless of age, sedative use w as the strongest risk factor for lifetime use of prescription opioids. While any visits to the doctor in the past 6 month s was the strongest association for past 30 day use among older women, sedative use remained the strong est risk factor for past 30 day use of opioids among younger women. Consequently, while prevalence of use may be greater among older women, younger women may be at risk of adverse consequences as well. When considering the prevention of non medical prescription opioids, clinicians shoul d avoid co prescribing opioids and sedatives and i ntervention efforts should not only target older and younger women separately but should consider different levels of the socio ecological model for the intervention.

PAGE 108

108 Figure 4 1. The frequency of the distribution of age among HealthStreet female member s Figure 4 2. The frequency of the distr ibution of the number of children among HealthStreet female members

PAGE 109

109 Figure 4 3. Enrollment in HealthStreet and eligibility for women in current study, as of June 2018. Tot al women included in analysis N=5,549 Participants < 18 years, men, and women with Rx opioid use or zip code data missing excluded Participants from the c ommunity c ontacted by or at HealthStreet N =12, 986 C ompleted intake questionnaire N = 10,293 Older women > 50 years N=2,413 (43 %) Y ounger women <49 years N=3,136 (57 %) Past 30 day (n=454, 19%) L ifetime but not past 30 day use ( n=1,068 , 44 %) None (n=891, 37%) L ifetime but not past 30 day use (n= 1 , 199 38 %) Past 30 day (n=377, 12%) None (n=1,560, 50%)

PAGE 110

110 Table 4 1. Association between individual , relationship , and community level characteristics of prescription opioid use among women enrolled in HealthStreet, 2011 2018, N=5,549 . Characteristic Overall N=5,549 N (%) No Rx opioid use N=2,451 N (%) Lifetime, no t past 30 day Rx opioid use N=2,267 N (%) Past 30 day Rx opioid use N=831 N (%) P value INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years 3136 (56.5 ) 2413 (43.5 ) 1560 (63.7 ) 891 (36.3 ) 1199 (52.9 ) 1068 (47.1 ) 377 (45.4 ) 454 (54.6 ) <.0001 Race Black Other White 3167 (57.1 ) 378 (6.8 ) 2004 ( 36.1 ) 1587 (64.7 ) 173 (7.1 ) 691 (28.2 ) 1097 (48.4 ) 163 (7.2 ) 1007 (44.4 ) 483 (58.1 ) 42 (5.1 ) 306 (36.8 ) <.0001 Education HS or less More than HS 2974 (53.6 ) 2575 (46.4 ) 1446 (59.0 ) 1005 (41.0 ) 1054 (46.5 ) 1213 (53.5 ) 474 (57.0 ) 357 (43.0 ) <.0001 Health insurance No Yes 1678 (30.2 ) 3871 (69.8 ) 819 (33.4 ) 1632 (66.6 ) 663 (29.2 ) 1604 (70.8 ) 196 (23.6) 635 (76.4 ) <.0001 Doctor visits (past 6 months ) No Yes 1443 (26.0 ) 4 106 (74.0 ) 856 (34.9 ) 1595 (65.1 ) 516 (22.8) 1751 (77.2) 71 (8.5) 760 (91.5) <.0001 ED visits (past 6 months) 0 1 >2 4662 (84.0) 887 (16.0) 2182 (89.0 ) 269 (11.0 ) 1917 (84.6 ) 350 (15.4 ) 563 (67.8 ) 268 (32.2 ) <.0001 Depression No Yes 3741 ( 67.4 ) 1808 (32.6 ) 1893 (77.2 ) 558 (22.8 ) 1413 (62.3 ) 854 (37.7 ) 435 (52.4 ) 396 (47.6 ) <.0001 Anxiety No Yes 3945 (71.1 ) 1604 (28.9 ) 2009 (82.0 ) 442 (18.0 ) 1476 (65.1 ) 791 (34.9 ) 460 (55.4 ) 371 (44.6 ) <.0001 Back pain No Yes 2954 (53.2 ) 2596 (46.8 ) 1627 (66.4) 824 (33.6) 1065 (47.0) 1202 (53.0) 261 (31.4 ) 570 (68.6 ) <.0001 Cancer No Yes 4969 (89.6) 580 (10.4) 2320 (94.7) 131 (5.3) 1959 (86.4) 308 (13.6) 690 (83.0) 141 (17.0) <.0001

PAGE 111

111 Table 4 1. Continued Characteristic Overall N=5,549 N (%) No Rx opioid use N=2,451 N (%) Lifetime, no t past 30 day Rx opioid use N=2,267 N (%) Past 30 day Rx opioid use N=831 N (%) P value Insomnia No Yes 3991 (71.9) 1558 (28.1) 1998 (81.5) 453 (18.5) 1507 (66.5) 760 (33.5) 486 (58.5) 345 (41.5) <.0001 Cigarette use (lifetime use) No Yes 3020 (54.4) 2529 (45.6) 1528 (62.3) 923 (37.7) 1116 (49.2) 1151 (50.8) 393 (45.2) 455 (54.8) <.0001 Hazardous alcohol use (past 30 days) No Yes 4478 (80.7) 1071 (19.3) 1998 (81.5) 453 (18.5) 1817 (80.1) 450 (19.9) 663 (79.8) 168 (20.2) 0.3789 Marijuana use (lifetime) No Yes 3123 (56.3) 2426 (43.7) 1614 (65.9) 837 (34.1) 1071 (47.2) 1196 (52.8) 438 (52.7) 393 (47.3) <.0001 Rx sedative use (lifetime) No Yes 4033 (72.7) 1516 (27.3) 2162 (88.2) 289 (11.8) 1434 (63.3) 833 (36.7) 437 (52.6) 394 (47.4) <.0001 RELATIONSHIP LEVEL FACTORS Marital status Never married Married Separated, divorced, or widowed 2342 (42.2) 1276 (23.0) 1931 (34.8) 1255 (51.2) 503 (20.5) 693 (28.3) 803 (35.4) 592 (26.1) 872 (38.5) 284 (34.2) 181 (21.8) 366 (44.0) <.0001 Children No Yes 1299 (23.4) 4250 (76.6) 707 (28.8) 1744 (71.2) 473 (20.9) 1794 (79.1) 119 (14.3) 712 (85.7) <.0001 Employment No Yes 3553 (64.0) 1996 (36.0) 1502 (61.3) 949 (38.7) 1401 (61.8) 866 (38.2) 650 (78.2) 181 (21.8) <.0001 Social media use No Yes 3038 (54.8) 2511 (45.2) 1458 (59.5) 993 (40.5) 1092 (48.2) 1175 (51.8) 488 (58.7) 343 (41.3) <.0001 COMMUNITY LEVEL FACTORS Rurality No Yes 5308 (95.7) 241 (4.3) 2369 (96.7) 82 (3.3) 2150 (94.8) 117 (5.2) 789 (94.9) 42 (5.1) 0.0052

PAGE 112

112 Table 4 2 . Association between individual , relationship , and community level characteristics and prescription opioid use among women enrolled in HealthStreet, stratified by age, 2011 2018, N=5, 549 Characteristic Younger Women (n=3,136) Older Women (n=2,413) No Rx opioid use (n=1,560, 50%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,199, 38%) n (%) Past 30 day Rx opioid use (n=377, 12%) n (%) P value No Rx opioid use (n=891, 37%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,068, 44%) n (%) Past 30 day Rx opioid use (n=454, 19%) n (%) P value INDIVIDUAL LEVEL FACTORS Race Black Other White 1047 (67.1 ) 128 (8.2 ) 385 (24.7 ) 636 (53.0 ) 107 (8.9 ) 457 (38.1 ) 232 (61.5 ) 25 (6.6 ) 120 (31.8 ) <.0001 540 (60.6 ) 45 (5.1 ) 306 (34.3 ) 462 (43.3 ) 56 (5.2 ) 5 50 (51.5 ) 251 (55.3 ) 17 (3.7 ) 186 (41.0 ) <.0001 Education HS or less More than HS 948 (60.8 ) 612 (39.2 ) 598 (49.9 ) 601 (50.1 ) 224 (59.4 ) 153 (40.6 ) <.0001 498 (55.9 ) 393 (44.1 ) 456 (42.7 ) 612 (57.3 ) 250 (55.1 ) 204 (44.9 ) <.0001 Health insurance No Yes 556 (35.6 ) 1004 (64.4 ) 410 (34.2 ) 789 (65.8 ) 106 (28.1 ) 271 (71.9 ) <.0001 263 (29.5 ) 628 (70.5 ) 253 (23.7 ) 815 (76.3 ) 90 (19.8 ) 364 (80.2) .0002 Doctor visits (past 6 months) No Yes 618 (39.6 ) 942 (60.4 ) 338 (28.2 ) 861 (71.8 ) 47 (12.5 ) 330 (87.5 ) <.0001 238 (26.7 ) 653 (73.3 ) 178 (16.7 ) 890 (83.3 ) 24 (5.3 ) 430 (94.7 ) <.0001 ED visits (past 6 months) 0 1 >2 1360 (87.2) 200 (12.8) 976 (81.4 ) 223 (18.6 ) 227 (60.2 ) 150 (39.8 ) <.0001 822 (92.3 ) 69 (7.7 ) 941 (88.1) 127 (11.9) 336 (74.0 ) 118 (26.0 ) <.0001 Depression No Yes 1194 (76.5) 366 (23.5) 755 (63.0 ) 444 (37.0 ) 197 (52.2 ) 180 (47.8 ) <.0001 699 (78.5 ) 192 (21.5 ) 658 (61.6 ) 410 (38.4 ) 238 (52.4 ) 216 (47.6 ) <.0001 Anxiety No Yes 1257 (80.6 ) 303 (19.4 ) 756 (63.1 ) 443 (36.9 ) 198 (52.5 ) 179 (47.5 ) <.0001 752 (84.4 ) 139 (1 5.6 ) 720 (67.4 ) 348 (32.6 ) 262 (57.7 ) 192 (42.3 ) <.0001 Back pain No Yes 1090 (69.9) 470 (30.1) 611 (51.0 ) 588 (49.0 ) 131 (34.8 ) 246 (65.2 ) <.0001 537 (60.3) 354 (39.7 ) 454 (42.5 ) 614 (57.5 ) 130 (28.6 ) 324 (71.4 ) <.0001

PAGE 113

113 Table 4 2. Continued Characteristic Younger Women (n=3,136) Older Women (n=2,413) No Rx opioid use (n=1,560, 50%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,199, 38%) n (%) Past 30 day Rx opioid use (n=377, 12%) n (%) P value No Rx opioid use (n=891, 37%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,068, 44%) n (%) Past 30 day Rx opioid use (n=454, 19%) n (%) P value Insomnia No Yes 1304 (83.6) 256 (16.4) 839 (70.0) 360 (30.0) 228 (60.5) 149 (39.5) <.0001 694 (77.9) 197 (22.1) 668 (62.6) 400 (37.4) 258 (56.8) 196 (43.2) <.0001 Cigarette use (lifetime use) No Yes 1011 (64.8) 549 (35.2) 593 (49.5) 606 (50.5) 173 (45.9) 204 (54.1) <.0001 517 (58.0) 374 (42.0) 523 (49.0) 545 (51.0) 203 (44.7) 251 (55.3) <.0001 Alcohol use (past 30 days) No Yes 1233 (79.0) 327 (21.0) 912 (76.1) 287 (23.9) 274 (72.7) 103 (27.3) 0.0163 765 (85.9) 126 (14.1) 905 (84.7) 163 (15.3) 389 (85.7) 65 (14.3) 0.7622 Marijuana use (lifetime) No Yes 997 (63.9) 563 (36.1) 527 (44.0) 672 (56.0) 172 (45.6) 205 (54.4) <.0001 617 (69.3) 274 (30.7) 544 (50.9) 524 (49.1) 266 (58.6) 188 (41.4) <.0001 Rx sedative use (lifetime) No Yes 1406 (90.1) 154 (9.9) 802 (67.9) 397 (33.1) 204 (54.1) 173 (45.9) <.0001 756 (84.9) 135 (15.1) 632 (59.2) 436 (40.8) 233 (51.3) 221 (48.7) <.0001 RELATIONSHIP LEVEL FACTORS Marital status Never married Married Separated, divorced, or widowed 1075 (68.9) 248 (15.9) 237 (15.2) 647 (54.0) 286 (23.8) 266 (22.2) 207 (54.9) 69 (18.3) 101 (26.8) <.0001 180 (20.2) 255 (28.6) 456 (51.2) 156 (14.6) 306 (28.7) 606 (56.7) 77 (17.0) 112 (24.7) 265 (58.4) 0.0050 Children No Yes 582 (37.3) 978 (62.7) 312 (26.0) 887 (74.1) 70 (18.6) 307 (81.4) <.0001 125 (14.0) 766 (86.0) 161 (15.1) 907 (84.9) 49 (10.8) 405 (89.2) 0.0858 Employment No Yes 884 (56.7) 676 (43.3) 647 (54.0) 552 (46.0) 263 (69.8) 114 (30.2) <.0001 618 (69.4) 273 (30.6) 754 (70.6) 314 (29.4) 387 (85.2) 67 (14.8) <.0001

PAGE 114

114 Table 4 2. Continued Characteristic Younger Women (n=3,136) Older Women (n=2,413) No Rx opioid use (n=1,560, 50%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,199, 38%) n (%) Past 30 day Rx opioid use (n=377, 12%) n (%) P value No Rx opioid use (n=891, 37%) n (%) Lifetime, no t past 30 day Rx opioid use (n=1,068, 44%) n (%) Past 30 day Rx opioid use (n=454, 19%) n (%) P value Social media use No Yes 830 (53.2) 730 (46.8) 524 (43.7) 675 (56.3) 210 (57.7) 167 (44.3) <.0001 628 (70.5) 263 (29.5) 568 (53.2) 500 (46.8) 278 (61.2) 176 (38.8) <.0001 COMMUNITY LEVEL FACTORS Rurality No Yes 1529 (98.0) 31 (2.0) 1162 (96.9) 37 (3.1) 368 (97.6) 9 (2.4) 0.1804 840 (94.3) 51 (5.7) 988 (92.5) 80 (7.5) 421 (92.7) 33 (7.3) 0.2738

PAGE 115

115 Table 4 3. Association between various risk factors and self reported prescription opioid use among women enrolled in HealthStreet, 2011 2018 N=5,549 Characteristic Lifetime, no t past 30 day Rx opioid use , N=2,267 (vs. no use) P ast 30 day Rx opioid use, N=831 (vs. no use) aOR 95% CI aOR 95% CI INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years Ref 1.11 0.96 1.28 Ref 1.31 1.07 1.61 Race White Black Other Ref 0.80 0.89 0.69 0.92 0.68 1.15 Ref 1.24 0.82 1.01 1.52 0.55 1.23 Education More than HS HS or less Ref 0.69 0.61 0.79 Ref 0.87 0.72 1.05 Health insurance Yes No Ref 0.92 0.80 1.05 Ref 0.71 0.58 0.87 Doctor visits (past 6 months) No Yes Ref 1.32 1.14 1.53 Ref 3.14 2.39 4.14 ED visits (past 6 months) 0 1 >2 Ref 1.27 1.05 1.53 Ref 2.73 2.19 3.39 Depression No Yes eliminated eliminated Back pain No Yes Ref 1.83 1.61 2.08 Ref 2.93 2.44 3.52 Cancer No Yes Ref 1.82 1.44 2.30 Ref 2.16 1.63 2.88 Insomnia No Yes eliminated eliminated Cigarette use (lifetime) No Yes eliminated eliminated Hazardous alcohol use ( past 30 day s) No Yes eliminated eliminated Marijuana use (lifetime) No Yes Ref 1.86 1.63 2.11 Ref 1.43 1.19 1.71

PAGE 116

116 Table 4 3. Continued Characteristic Lifetime, no t past 30 day Rx opioid use, N=2,267 (vs. no use) Past 30 day Rx opioid use, N=831 (vs. no use) aOR 95% CI aOR 95% CI Rx sedative use (lifetime) No Yes Ref 2.83 2.41 3.33 Ref 4.27 3.48 5.24 RELATIONSHIP LEVEL FACTORS Marital status Married Never married Separated, divorced, or widowed Ref 0.74 1.00 0.62 0.88 0.84 1.18 Ref 0.92 1.08 0.71 1.19 0.85 1.36 Children No Yes Ref 1.46 1.25 1.72 Ref 1.67 1.31 2.14 Employment Yes No Ref 0.87 0.76 0.99 Ref 1.48 1.20 1.81 Social media use No Yes Ref 1.41 1.24 1.60 Ref 1.02 0. 85 1.22 COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated aOR= adjusted Odds Ratio; CI= Confidence Interval; ref= reference group; N= sample size AIC= Akaike information criterion; intercept only 11223.899 ; intercept with covariates 9787.928 2 (df= 32 )= 1499.9705 ; p<.0001

PAGE 117

117 Table 4 4. Association between select risk factors and self reported prescription opioid use pattern among women, by sex, 2011 2018 Characteristic Younger women (n=3,136 ) Older women (n=2,413 ) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) INDIVIDUAL LEVEL FACTORS Race White Black Other Ref 0.75 (0.62 0.91 ) 0.85 (0.62 1.17 ) Ref 1.17 (0.87 1.57 ) 0.76 (0.45 1.29 ) Ref 0.80 (0.64 0.99 ) 0.95 (0 .61 1.50 ) Ref 1.31 (0.99 1.73 ) 0.89 (0.47 1.70 ) Education More than HS HS or less Ref 0.63 (0.53 0.75 ) Ref 0.75 (0.57 0.98 ) Ref 0.76 (0.62 0.93 ) Ref 1.05 (0.81 1.37 ) Health insurance Yes No Ref 1.01 (0.85 1.21) Ref 0.70 (0.53 0.92) Ref 0.82 (0.65 1.03) Ref 0.71 (0.52 0.96 ) Doctor visits (past 6 months) No Yes Ref 1.27 (1.06 1.53 ) Ref 2.72 (1.92 3.84 ) Ref 1.45 (1.14 1.86 ) Ref 4.15 (2.62 6.60 ) ED visits (past 6 months) 0 1 >2 Ref 1.26 (1.00 1.59 ) Ref 2.84 (2.14 3.78 ) Ref 1.29 (0.92 1.80 ) Ref 2.72 (1.91 3.88 ) Depression No Yes eliminated eliminated eliminated eliminated Back pain No Yes Ref 1.80 (1.52 2.13 ) Ref 3.01 (2.32 3.90 ) Ref 1.86 (1.53 2.26) Ref 2.91 (2.24 3.78 ) Cancer No Yes Ref 1.58 (1.05 2.36) Ref 1.92 (1.16 3.17) Ref 1.91 (1.43 2.56) Ref 2.29 (1.61 3.24 ) Insomnia No Yes eliminated eliminated eliminated eliminated Cigarette use (lifetime) No Yes eliminated eliminated eliminated eliminated Hazardous alcohol use ( past 30 day s) No Yes Ref 1.05 (0.86 1.28 ) Ref 1.57 (1.17 2.10 ) eliminated eliminated Marijuana use (lifetime) No Yes Ref 1.88 (1.58 2.23 ) Ref 1.41 (1.08 1.83 ) Ref 1.74 (1.42 2.13) Ref 1.32 (1.02 1.72) Rx sedative use (lifetime) No Yes Ref 2.96 (2.37 3.70 ) Ref 4.84 (3.59 6.51 ) Ref 2.70 (2.13 3.41 ) Ref 3.76 (2.82 5.01 ) RELATIONSHIP LEVEL FACTORS

PAGE 118

118 Table 4 4. Continued Characteristic Younger women (n=3,136) Older women (n=2,413) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) Lifetime, no t past 30 day Rx opioid use (vs. no use) aOR (95% CI) Past 30 day Rx opioid use (vs. no use) aOR (95% CI) Marital status Married Never married Separated, divorced, or widowed eliminated eliminated eliminated e liminated Children No Yes Ref 1.91 (1.58 2.30) Ref 1.93 (1.42 2.64 ) eliminated eliminated Employment Yes No Ref 0.80 (0.68 0.95 ) Ref 1.31 (1.00 1.73) Ref 0.98 (0.79 1.21 ) Ref 1.85 (1.34 2.55 ) Social media use No Yes Ref 1.31 (1.11 1.54 ) Ref 0.79 (0.62 1.02 ) Ref 1.69 (1.38 2.07 ) Ref 1.43 (1.09 1.87 ) COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated eliminated eliminated aOR= adjusted Odds Ratio; CI= Confidence Interval; ref= reference group; N= sample size Younger women model AIC= Akaike information criterion; intercept only 6085.477 ; intercept with covariates 5332.523 2 (df= 28)= 808.9547 ; p<.0001 Older women model AIC= Akaike information criterion; intercept only 5037.231 ; intercept with covariates 4477.075 2 (df= 24 )= 608.1566 ; p<.0001

PAGE 119

119 CHAPTER 5 GEOSPATIAL CLUSTERS OF PRESCRIPTION OPIOID USE AMONG WOMEN IN A COMMUNITY SAMPLE Background Excessive prescription opioid use continues to be an immediate public health concern, with current es timates indicating that 130 people die daily due to opioid related drug overdose in the US (Centers for Disease Control and Prevention, 2018g, 2018h) . Though pr escription opioids are commonly prescribed to treat both chronic and acute pain (Beaudoin et al., 2014; Caudill Slosberg et al., 2004) , use has become a concern in the US due to their high potential for misuse and ab use which raises concerns about overprescribing (Drug Enforcement Administration, 2017; Kaye et al., 20 17; Strand et al., 2018; A. G. White et al., 2009) . The increase in overdose leading to increased risk of adverse consequences, including overdose (National Institute on Drug Abuse, 2018c; Pezalla et al., 2017) . Specifically, the number of deaths related to opioid use has increased in Florida. In 2017, d ata from the Florida Drug Related Outcomes Surveillance and Tracking System (FROST) reporte d that there were 1,685 deaths that in cluded a fentanyl analog in 2017 (FROST, 2018) , representing a higher rate . It is we ll known that Florida has had a pr oblem with prescription opioids including unregulated pain clinics (Kennedy Hendricks et al., 2015; Surratt et al., 2014) , which has led the state to implement major policy changes resulting in a 7.3% decrease in opioid prescriptions between 2013 and 2015 (National Institute on Drug Abuse, 2018e) . Changes include decreasing number of pills prescribed and implementation of the p resc ription drug monitoring program ( PDMP ) so physicians can check if patients have received o pioids from another prescriber. In addition to prescription opioids, overdose deaths related to synthetic

PAGE 120

120 o pioids have increased substantially in Florida (Prekupec et al., 2017) . Between 2010 and 2014, deaths related to fentanyl use increased 250% in Florida (Aschenbrenner, 2017) . Use a mong women is a growing concern as well. Opioid overdoses have increased among women at a higher rate compared to men in recent years (Centers for Disea se Control and Prevention, 2018d; Unick et al., 2013) and women have been found to be twice as likely to be prescribed opioids (Hall et al., 2008; Manubay et al., 2015; Simoni Wastila, 2000) and are subsequently more likely to use prescription opioids than men (Serdarevic, Striley, et al., 2017) . Though prevalence of use and adverse consequen ces related to use may be greater among women, there has been a lack of studies and publications regarding women and prescripti on opioid use. In this paper, we focus specifically on women because this is a population that may be at greater risk for opioid use disorder and overdose . T he negative consequences and overdose rates from prescription opioid use have dramatically increased in the US partially due to high prescribing rates in emergency departments (EDs). The i with the increase in prescribing rates in EDs (Phillips, 2000) . From 1999 to 2005, one study found an increase in pain related complaints in US EDs (from 23% of visits to 37% of visits) with an associated 1.5 fold increase in opioid prescribing (Cantrill et al., 2012) . Besides the high opioid prescribing rates in the ED, frequent ED use (defined as 4 or more ED visits in the past 12 months) is another major concern because of increasing health care costs (Vinton, Capp, Rooks, Abbott, & Ginde, 2014) related to overcrowding and longer wait times, which threaten patients with other medical health problems (Trzeciak & Rivers, 2003) . Consequently, frequent use of the ED by those seeking prescription opioids for pain management may have negative impacts on those in the community who do not use prescription opioids. Frequent ED users represent a small

PAGE 121

121 proportion of people who receive treatment at the ED but account for a disproportionate amount of health care costs . Frequent ED use is also commonly associated with social disadvantages and multi morbid chronic disease (Milani, Crooke, Cottler, & Striley, 2016; Vinton et al., 2014) . Approximately one third of patients in the ED receive an opioid for pain related complaints either administered at the time of their visit or after being discharged (Mazer Amirshahi, Mullins, Rasooly, van den Anker, & Pines, 2014) . Though the most common reason for seeking care in the ED is pain (Cordell et al., 2002) and given that there has been an increase in opioid prescribing in EDs for pain related complaints , a number of studies have found that prescribing p ain medication s by physicians in the ED to manage pain is not appropriate (Baehren et al., 2010) . Utilization of the ED for pain medication may be more common amo ng those who do not have a ccess to health insurance and where visiting a primary care physician is not an option. Regular primary care physician visits and prescribing of pain medication in primary care is likely to be far more effective though and result in a lower demand for pain medications through the ED (Dowell et al., 2016; Elder et al., 2017; Schneiderhan et al., 2017) . Visiting a primary care physician for pain medication prescriptions should resul t in less need to visit the ED. Appropriate prescribing of pain medication s is important, espec ially considering the high rate of opioids being prescribed to manage pain. Within this paper, we examine ED utilization among women who use prescription opioids, to determine the relationship between the two in a com munity setting. Though the focus within this dissertation work is not to examine pain sensitivity among women, i n the context of prescription opioid use and ED utilization , pain is an important factor to consider. Women have a higher prevalence of chronic conditions that cause pain (Darnall et al., 2012) whi ch may manifest itself in the higher prevalence of prescription opioid use among

PAGE 122

122 women compared to men. In addi tion, studies have found women are more likely to utilize healthcare services for chronic and mental health conditions et al., 2002; Thompson et al., 2016) , which is specifically relevant in terms of prescription opioid use and ED utilization. Given that some communities may access the ED for pain management frequently, it is important to focus on communities where health insurance coverage and primary care visits are not accessible in order to identify those who may be at risk. Thus, it is important to further examine the relationship between access to ca re and prescription opioid use. We do this, with a focus on a community sample of women. To further examine prescription opioid use and ED utilization, geospatial analyses can be used to identify clusters of prescription opioid use patterns and to examine th e association between geographical clusters and ED utilization. The utilization of spatial statistics has the potential not only to examine the epidemiological relationship between the environment and prescription opioid use, but can inform community effor ts in opioid prevention by focusing on specific regions or geographical areas which may o ptimize efforts. For example, one study which used a community based sample previously identified cancer hot spots through spatial statistics. Results from this study found racial disparities in cancer which could inform community intervention efforts (Ruktanonchai et al., 2014) . Another study identified hot spots of opioid related emergency needs through spatial statistics provided physicians with information on opioids in their community (Dworkis, Taylor, Peak, & Bearnot, 2017) . The current analysis differs because it focuses solely on women in the community and ED utilization . More importantly, the i nformation generated from this analysis may educate health care professionals about the dangers of opioid prescribing. This current analysis uses geospatial methods wit hin the cont ext of a theoretical

PAGE 123

123 framework, the socio ecological model, to examine pre scription opioid use among women in a community setting in North Central Florida. This analysis will explore the community level variable of area of residence and its association with the individual level variables of frequent ED utilization and prescription opioid use. The socio ecological model is preferable to the health bel ief model because it can be used to examine behaviors such as prescription opioid use (Glanz, Rimer, & Viswanath, 2015; Substance Abuse and Mental Health Services Administration, 2016a) without limiting risk factors to influences at the individual level only. The socio ecological model postulates that people are influenced not only by individual level factors but also by their relationships with others and by the communities in which they reside (McLeroy et al., 1988) . Our analysis add s to the li terature by considering prescription opioid use behavior within the context of a theoretical framework, examining factors at multiple levels of the model. Relevant factors associated with prescription opioid use from the literature were identified within o ur data and aligned with constructs from the socio ecological model. For example, at the individual level , health insurance (Chang et al., 2018) , and medical care visits such as doctor visits (Hahn, 2011) , and emergency department (ED; Cantrill et al., 2012 ) were examined . Chronic co nditions including pain and cancer and history of insomnia and depression (Blanco et a l., 2016; Pinkerton & Hardy, 2017; Serdarevic, Osborne, et al., 2017; Sullivan et al., 2006) as well other substance use (Gudin et al., 2013; S tein et al., 2017; Sullivan et al., 2006) were also added . Factors at the relationship level that we examined included marital status (being single) , having children, being employed, and social media use (Fillingim et al., 2005; Inciardi et al., 2007; Piko, 2 000; Rabinovitch et al., 2013; Rönkä & Katainen, 2017; Rozenbroek & Rothstein,

PAGE 124

124 2011) . At the community l evel, rurality has also been associated with increased morbidity and mortality related to prescription opioid use (Keyes et al., 2014) and was included in the model. We have previously examined prescription opioid use patterns among women and differences in risk fact ors for use among older and younger women (Chapter 4), within the context of the socio ecological model . Results showed no difference in prescription opioid use patterns among women at the community level by solely examining urban or rural areas of residence. Given the diverse communities contained within the state of Florida and the challenges that may arise from trying to categorize areas as simply urban or rural, it is likely that another layer of granularity is needed to examine the community level further. Geospatial analysis will allow area of residence to be examined in further detail. Aims and Hypothe ses The aim of this analysis was to identify community clusters of frequent female ED users (frequent ED user > 2 vs non frequent user <2 in the past 6 months ) among those who have endorsed prescription opioid use (both in their lifetime [not including past 30 day use] and past 30 day use). This analysis employ ed spatial scan statistics for both past 30 day and lifetime ( not past 30 day use) prescription opioid use . ED utilization (frequent user vs non frequent user) by residence (home addresses) through the longitude and latitude point data will be examined. The following hypothesis was proposed among 3, 098 community dwelling women who use d opioids (8 31 with past 30 day use and 2 267 with lifetime but not past 30 day use of prescription opioids) : Hypothesis 1. Among women who endorsed prescription opioid use , those who are frequent ED users and who used prescription opioids in the past 30 day s will have a higher number of cluster s than women who were frequent ED users and who used in their lifetime ( not past 30 day use ).

PAGE 125

125 Methods To test our hypothesis, a large community sample with extensive data regarding history of drug use, health concerns and conditions was used to examine prescription opioid use and frequent ED utilization. The variables that were includ ed in our analyses were collected through a community engagement program and are outlined below . Study Population HealthStreet is a community engagement program at the University of Florida (UF) which was established at UF in 2011 as a major effort of the Clinical and Translational Science Institute (CTSI) to include community based research. HealthStreet was first developed at Washi ngton University in St. Louis by the founding director Dr. Linda B. Cottler in order to link participants to drug and alcohol related research. HealthStreet has expanded since it was first developed to focus on all health problems and has engaged with ove r 10,293 community members using a community health worker model. Community Health Workers (CHWs) are trained and certified to directly engage community members in the community at parks, grocery stores, churches, laundromats, health fairs, and meet people where they are . CHWs are able to assess the health of members in the community as well as their health needs and concerns. Using this model (Cottler et al., 2011) , community members in the HealthStreet catchment area, including Gainesville, Jacksonville and Miami have been engaged in community health research since November 2011 and f rom the population to be examined in this analysis. After CHWs explain the HealthStreet model and participants provide written consent, CHWs interview participants using a Health Needs A ssessment. The interview takes approximately 30 minutes to complete and is conducted face to face. The responses are recorded using a paper copy of the needs assessment in the field during outreach. Through the needs

PAGE 126

126 substance use are assessed . Once members are asse ssed they become members of the HealthStreet community cohort and are linked to health research opportunities and medical and social services. HealthStreet members who self identified as female, were 18 years and older, self reported any prescription opioi d use in their lifetime , and enrolled between November 2011 and June 2018 were included in this analysis. This study was approved by the University of Florida Institutional Review Board. Measurements Questions regarding ED use and drug use were assessed as part of the health needs assessment. The outcome variable, frequent ED visits, ( > 2 visits last 6 months), was assessed by ER in the last 6 months for your own injury, he ED once or less were coded as a non frequent ED user and those who visited the ED two or more times in the past 6 months were coded as a frequent ED user (based on the literature , frequent ED use is typically defined as 4 or more ED visits in the past y ear (Hunt, Weber, Showstack, Colby, & Callaham, 2006; Kushel, Perry, Bangsberg, Clark, & Moss, 2002; LaCalle & Rabin, 2010; Vinton et al., 2014) thus, we de fined frequent ED use as > 2 visits because our data only measures past 6 month use of the ED ). Prescription opioid use status was assessed by asking medications like Vicodin®, oxycodone, codeine, Demerol®, morphine, Percocet®, Darvon®, subsequently asked if they used one o f these prescription medications in the past 30 day s. Those who reported use in the past 30 day s were coded as a past 30 day prescription opioid user , those who did not use in the past 30 day s but did indicate use in the past were coded as a for li fetime, but not past 30 day prescription opioid user.

PAGE 127

127 Additional risk factors at the individual level were elicited from the Health Needs A ssessment including: age, race (white, black/African American, other), level of educational attainment (more than hi gh school or high school and less), health insurance (yes or no), doctor visit in the last 6 months (yes or no), depression (yes or no), anxiety (yes or no), back pain (yes or no), cancer (yes or no), i nsomnia ( yes or no), and substance use such as cigaret te smoking (yes or no), hazardous alc ohol use (yes or no), marijuana use (yes or no), and prescription sedative use ( yes or no). Relationshi p level risk factors were also elicited through the health needs assessment: marital status (never married; current ly married; or separated, divorced, or widowed), number of children (continuous variable), employment (fu ll time/ part time ; yes or no ), and any use of social media (Twitter, Facebook, or Instagram; yes or no) . The community level risk factor was assessed using rurality from the Health Needs A ssessment. The variable for r urality was created using US census tract data for the zip code . Suburban ar eas were not examined because they are within the urban cluster designation. Geospatial analyses were further used to examine prescription opioid use at iden tify clusters of frequent ED users compared to non users among those who used prescription opioids. Analysis The final sample size for this ana lysis consisted of 3,098 women after excluding those with missing prescription opioid use data and those who r eported no use of prescription opioids, missing sex or transgender status, community members who self identified as male , and those who were under 18 years old (n=7,195) . Descriptive statistics and chi square tests were calculated to summarize and tabulate frequencies for individual, relationship, and community

PAGE 128

128 level variables by ED status among prescription opioid users ( significance was determined based on Bonferroni corrected p value of 0.002 after dividing alpha 0.05 by 19 , the number of variables). Ana lyses were conducted for past 30 day and lifetime prescription opioid users separately. The geolocation the Google MAPS Applicat ion Programming Interface (API). The longitude and latitude coordinates rsatscan control files. Cases represented frequent ED users and controls represented non frequent ED users (coded in binary format; cases=1, controls=0). Separa te datasets for cases, controls, and longitude and latitude information were created. After importing these three datasets into SaTScan, spatial scan statistics were conducted to create and evaluate the statistical significance of opioid use clusters. The Bernoulli method with default options within SaTScan was selected to allow for spatial analysis to scan for both high and low rates of clusters. Two separate maps for lifetime, but not past 30 day use, and past 30 day prescription opioid use were generated by SaTScan , showing clusters of frequent ED users (vs. non frequent users) . A p value below the significant threshold of 0.05 indicated that the number of case s within that geographic area was more or less than the number of cases expected. This analysis approach allow ed us to determine if recent ( past 30 day ) opioid users who were frequent utilize rs of the ED clustered at a greater rate than expected , suggesting possible problem areas which may require further investigation. For lifetime users of prescrip tion opioids, we would not expect to see the same clustering. Descriptive statistical analyses were calculated using SAS® 9.4 (SAS Institute Inc., 9.4, Cary, NC: SAS Institute Inc., 2011) while s patial analyses that included the longitude and

PAGE 129

129 latitude point data were conducted in RStudio (RStudio 1.1.456, 2016) and SaTScan TM (Kulldorff, 1997) . Results The sample of women prior to stratification by opioid use and ED visits (n=3, 098 ) was 51. 0 % black, half (50. 7 %) completed a h igh school degree or more, 49 .1 % were older (50+ ye ars), more than a third (33. 8 %) were employed part or full time, and approximately a quarter (24.9%) were mar ried (data not shown) . Among all women who had used prescription opioids, 2, 267 (73 % ) reported life time but not past 30 day use of prescription opioids and 8 31 ( 27 % ) reported past 30 day opioid use (Figure 5 1). In total, f requent ED use (defined as >2 vis its in the last 6 months) was reported by 6 18 of the women (20 %) and non frequent use was reported by 2, 480 (80%) women. Factors by ED utilization among lifetime not past 30 opioid users Table 5 1 presents individual, relationship, and community level factors by ED utilization and prescription opioid use status. When women were stratified by opioid use status, a higher proportion of past 30 day opioid users reported visit ing the ED frequently (32 % ) compared to lifetime, not past 30 day opioid users (15 %; Table 5 1) . Among those with lifetime no past 30 day use of prescription opioids, fre quent ED utilizers were younger , less educated and reported higher rates of doctor visit in the past 6 months, history of depression, anxiety, back pain, and insomnia (p<.0001). Th ey also reported higher rates of substance use including cigarette and prescription sedative use in their lifetime compared to non frequent ED users . At the relationship level , frequent E D users compared to non frequent ED users reported lower rates of employment but did not significantly differ from non frequent ED users in rurality .

PAGE 130

130 Similar to lifetime users, past 30 day prescription opioid users who were frequent ED users compared to non frequent ED utilizers were younger and less likely to be insured. In addition, at the individual level, they reported higher rates of depression and anxiety. Past 30 day opioid users who were frequent ED users were also more likely to repor t lifetime use of prescription sedatives compared to non frequent ED users (p=.0011). At the relationship and community level , past 30 day opioid users with frequent ED use did not significantly differ from non frequent ED users. Table 5 2 presents individ ual, relationship, and community level factors among opioid users by opioid use pattern among the 618 frequent ED utilizers only . After correcting for multiple comparisons , factors at the individual, relationship, and community level did not significantly different between lifetime no past 30 day users and past 30 day users who reported frequent ED utilization except for social media use (p=.001 1 ). Geospatial cluster analysis among lifetime opioid users To further examine prescription opioid use at the com munity level, t wo separate maps for prescription opioid use by frequent ED utilization were generated using the SaTScan software. Figure 5 2 represents clusters of frequent ED utilization among women who reported li fetime but mot past 30 day prescription o pioid use only using a map which cove rs a part of North Central Florida ( our catchment area ). This map includes a total of 357 cases (frequent ED users) among 2,267 HealthStreet female members who reported lifetime but no t past 30 day prescription opioid use. There were a total of seven clusters detected. Red circles indicate geospatial areas with more frequent ED users than would be expected if the risk of ED use was evenly distributed around North Central Florida (although there were three areas, they were clustered so that they were undetected in the map which showed two red circles) . Blue circles indicate geospatial areas with less frequent ED utilization than expected (although there were 4 blue areas, only 2 were

PAGE 131

131 shown on the map) . However, there was only one significant red cluster detected for frequent ED utilization among lifetime but not past 30 prescription opioid users (p<.05). This cluster of women who used opioids in their lifetime and were frequent ED users was centered in the Jacksonville, Florida area (depicted in the map in the top right corner with a black star and red circle cluster in Figure 5 2 ; p=0.017 ) . This cluster had the highest log likelihood ratio (the certainty that th ese cases constitute a cluster) of all sev en clusters observed, where there were 29 women who frequented the ED in the past 6 months out of 77 HealthStreet female members who resided in that area. Geospatial cluster analysis among past 30 day opioid users The second map (Figure 5 3) that was generated identified six geospatial clusters of those who frequently visited the ED in the past 6 months , among women who reported past 30 day prescription opioid use . This map also depicts a portion of North Central Fl o rida and includes a total of 268 cases among 831 HealthStreet female members in the community who reported past 30 day prescription opioid use. Spatial scan statistics revealed one significant cluster. This cluster of women who reported past 30 day prescr iption opioid use and less frequent ED use was also centered in the Jacksonville, Florida area (depicted on the map in the top right corner with a black star a cluster of blue in Figure 5 3; p=0.013 ). Blue circles indicate geospatial areas with less freque nt ED utilization than expected if the risk of ED use was evenly distributed around North Central Florida . Discussion In the US, prescription opioid use continues to be a significant public health problem associated with a number of poor health outcomes. Studies have found sex difference s to be an important factor in the opioid epidemic and that prescription opioid use a ffects women and men differently; women carry a higher burden of the consequences related to prescription opioid use

PAGE 132

132 y, et al., 2017) . Though prescription opioids are commonly used to treat pain (Beaudoin et al., 2014; Bonnie, Kesselheim, & Clark, 2017) and those who seek care in the ED most commonly go for pain, prescribing pain medication by emergency medicine physicians in the ED for long term management is not adequate or appropriate because chronic pain requires extended treatment (Cordell et al., 2002) . Specifically, frequent visits to the ED have been associated with higher rates of prescription opioid use (Doran et al., 2013; Neven et al., 2016) , though it is known that pain management is more effective in primary care settings (Dowell et al., 2016; Elder et al., 2017; Schneiderhan et al., 2017) . To examine the relationsh ip between frequent ED utilization and prescription opioid use , we identified a clus ter of frequent ED users (vs non frequent users) among HealthStreet women in the community who have used prescription opioids in their lifetime but not past 30 day s within the context of the socio ecological model. The area of residence of these women was the community level factor examined within the model. We found that many factors at the individual and relationship level were significantly associated with frequent ED use for women who endorsed p rescription opioid use. Overall, approximately 20% of these women visited the ED frequently in the past 6 months. A higher p rop ortion of women who were past 30 day prescription opioid users reported frequent ED visits compar ed to women who used prescription opioids in their lifetime not past 30 day s (32% vs 15 %). This may be due to recent (in the past 30 day s) prescription opioid use being more indicative of painful conditions which required immediate medical treatment compar ed to those women who used opioids at another point in their lifetime. Frequent ED users who endorsed lifetime and past 30 day prescription opioid use appeared to be younger comp ared to non frequent ED users. This may be because older women are more likely to be enrolled in a health

PAGE 133

133 insurance plan and to seek preventive care compared to younger individuals (Bonnie et al., 2015) , which subsequently reduces the chances of requiring emergency treatment. Ser ious health conditions, including painful medical conditions, can be avoided or better managed with primary prevention (Medicine, Yong, Saunders, & Olsen, 2010; Merzel . I n our sample we found non frequent ED users were more likely to have health insurance compared to frequent ED users . This was observed for both lifetime and past 30 day prescription opioid users. This finding is in concordance with the literature and our previous theory regarding reasons f or ED utilization . P revious studies have found that a lack of health insurance is associated with frequent ED visits (H ernandez Boussard, Burns, Wang, Baker, & Goldstein, 2014; Weber, Showstack, Hunt, Colby, & Callaham, 2005) . Those with health insurance are able to afford to utilize primary care services when experiencing low levels of pain which may reduce the need fo r emergency medicine to treat pain that has become intolerable. We also found doctor visit s to be associated with frequent ED use. Frequent ED users were more likely to report a doctor visit in the past 6 months compared to non frequent ED users ; this was true for lifetime not past 30 day prescription opioid use (p=<.0001). The literature regarding the relationship between ED use and primary care is inconclusive. While some studies have found that having a primary care physician can reduce or prevent ED use (Gill, Mainous, & other studies have found that a majority of those who visit the ED have another sourc e of care other Bouley, 1996). It is possible that women in this community who frequently utilize the ED also frequently utilize their primary care team for pain man agement due to inadequate management

PAGE 134

134 obtain further pain medication from more than one provider. If this is the case, then it is suggestive of either inadequate pa in management or potential prescription opioid misuse. This remains a possibility due to the addictive potential of prescription opioids. Overall, this association needs to be further examined. In addition, anxiety and depression have been commonly found to be co morbid with medical illness and those with depression and anxiety have poorer health outcomes (DiMatteo, Lepper, & Croghan, 2000; Scott et al., 2007) . Our stud y similarly found that frequent ED use, which may be related to painful medical conditions, was associated with depression and anxiety among women who reported past 30 day and lifetime not past 30 day prescriptio n opioid use . Finally, prescription sedative use was associated with frequent ED use. Frequent ED users were more likely to endorse prescription sedative use compared to non frequent ED users. This was true for both lifetime not past 30 day users and past 30 day users. Other studies have observed a similar relationship and report that approximately 20% to 40% of ED visits are related to substance us e among frequent ED users (Billings & Raven, 2013; Byrne et al., 2003; Urbanoski, Cheng, Rehm, & Ku rdyak, 2018) . Further, the community level of rural area of residence was neither associated with frequent ED use for lifetime prescription opioid use nor past 30 day use. Because the community level was categorized into only two levels, urban and rural, the influence of the community level on prescription opioid use may have been masked , thus , we further used spatial methods to examine this relationship. Though no relationship between rurality and frequent ED use was observed at the community level through univariate analyses, geospatial methods showed significant associations . Through these spatial analyses we identified clusters of frequent ED use among

PAGE 135

135 women who used prescription opioids in their lifetime and past 30 day s s eparately . We found 7 clusters of frequent ED users among women who used prescription opioids in their lifetime and 6 cluster s among past 30 day users, though only two of these were significant . We hypothesize d that women who are both frequent ED users and used prescription opioids in the past 30 day s will cluster more geospatially compared to women who were frequent ED users and used in their lifetime ( not past 30 day use ). However, this hypothesis was not supported as there was only 1 significant cluster among each opioid use group , when comparing frequent ED use to non frequent use . The clusters were centralized in the Jacksonville, Florida area. For lifetime (no past 30 day use) the cluster that was identified showed there was more frequent ED use than e xpected in that geographic area. However, among women who reported past 30 day use, the cluster that was identified showed less frequent ED visits than expected in that geographical area. The significant clusters we observed in Jacksonville may be due to urbanicity. Though we found no significant relationships between rural area of residence, previous studies have found that urban areas are associated with frequent ED utilization ( Fan, Shah, Veazie, & Friedman, 2011; Lasser, Kronman, Cabral, & Samet, 2012; Neuman et al., 2014) . A further explanation is that urbanicity alone does not influence prescription opioid use at the community level, but there are other community level fact ors which have an influence . For example, the broken windows were used as proxy for a community sense of pride and togetherness. This theory suggests that the appearance of the physical environment drives indi vidual behavior (Wilson & Ke lling, 1982) . The physical environment may not only signal disorder but it also signals to others that certain behaviors are tolerated (Cohen et a l., 2000; Sampson & Raudenbush, 2004) . This relationship needs to be further examined in community setting s using

PAGE 136

136 other community level factors besides area of residence . Anothe r explanation among past 30 day users who had fewer frequent ED visits could be because they did not visit the ED as they were already decreasing prescription opioid use or were prescribed from another source such as their primary care physician . could Reg ardless, such findings may be used not only for future outreach efforts by CHWs regarding prescription opioid use prevention in the Jacksonville (Duval County) area, but may have larger policy implications on opioid prescribing guidelines. More recently, o pioid prescribing in emergency medicine has changed . Many states have implemented emergency department specific guidelines for opioid prescribing (Broida, Gronowski, Kalnow, Little, & Lloyd, 2017) . Such guidelines are critical , especially in states such as Florida where the re are a high number of prescriptio ns written and where the number of prescri ptions written di ffers by county . For example, data from the PDMP in Florida show s that in 2017 Duval County, which includes the Jackso nville area, had a rate of 992 opioid prescriptions per 1,000 population whereas Alachua County, which includes the Gainesville area, had a lower rate of 797 opioid prescriptions per 1,000 population (Florida drug Related Outcomes Surveillance a nd Tracking System (FROST), 2018 ) . The higher prescribing ra tes in Duval County may further explain the clusters we observed in the Jacksonville area. Florida has since seen a change in prescribing practices and now places a limit on prescription opioids for acute pain and providers are required to check the drug m onitoring program before writing new prescriptions (Florida Board of Medicine, 2018) . Though the changes in prescribing practices such as these may not influence prevalence of chronic pain, they may provide different results in the f uture in regards to formations of opioid clusters which should be further monitored. Overall, it may be that we did not identify the anticipated clustering through geospatial analyses because the main source of

PAGE 137

137 prescription opioids in Florida may not be from the ED . With that in mind, future studies utilizing geospatial analyses should consider other sources of prescription opioids, such as primary care providers or dentists t o identify important clusters. Strengths and Limitations There are a few limit ations to t hese analyses . First, the data collected relating to prescription opioid use and fr equent ED use are self reported. Self reported data regarding drug use and health care utilization can be over or under reported , which can result in misclassifi cation . H owever, participants in this study were only asked about any prescription opioid use during two time frames [past 30 day s (current) and lifetime] . I nformation regarding dose, frequency, duration of use , and whether prescription opioid use was appr opriate , was not collected so the potential for misclassification was limited. However, this assessment was not designed specially to examine drug use. S ince information regarding misuse of prescription opioids was not collected, the potential for social d esirability bias was also reduced. It is also important to note that self report data regarding other covariates such as pain, which is the most common reason for seeking care in the ED , may be under or over reported. This may be due to the artifact of th e Health Needs A ssessment and the interpretation of the question, thus information bias cannot be ruled . However, the questionnaire has been carefully designed and CHWs were trained to clarify the meani ng of questions. In addition to the potential error introduced through self report, our findings are based on cross sectional data and reflect an association between frequent ED use and prescription opioid use and not causality. We al so did not elicit information regarding medic al causes related to ED visits and we cannot discern whether prescription opioid use occurred at the time of the ED visit. The relationship we observed between ED utilization and prescription opioid use may be generalizable to other similar communiti es. T he data pertaining to prescription opioid use that we

PAGE 138

138 analyzed was collected over a long time period (November 2011 through June 2018), which may not reflect the current state of prescribing practices in the ED. Though CHWs visit locations within the community multiply times a week at different times and try to engage all the community members they approach, the sampling technique may still be susceptible to selection bias. However, the strengths of this study are that this is among the first to use g eospatial analyses to assess frequent ED use among a community sample of women who have used prescription opioids. Data were available on exact coordinates of member addresses, which is uncommon in most studies. This large sample of women was racially dive rse ; the comprehensive data provided variables that we could examine at the individual, relationship, and community level within the context of a theoretical model. The data that was collected covers a wide range of data on drug use and medical information from women who may not traditionally be represented in research. The unique HealthStreet model provides for the potential for future follow up with community members which may be of specific relevance for the clusters we observed in Jacksonville, Florida. As such, the findings from this study could be translated into efforts which will directly benefit the community who participated in this research. Finally, this study examined prescription opioid use among women specifically, an area in the literature th at has had little attention. Conclusions These exploratory analyses provide important information regarding prescription opioid use and frequent ED use in the community. We foun d a high rate of women (32 %) who reported recent prescription opioid use and frequent ED visits in this community sample , which supports previous findings. Geospatial analyses also identified a cluster in the North Central region of Florida that was significantly associated with frequent ED use among opioid users, compared to

PAGE 139

139 non f requent ED users. These results can be used to provide information to community engaged programs that aim to reduce local clustering of disease and conditions that may be associated with the use of prescription opioids.

PAGE 140

140 Figure 5 1. Women enrolled in HealthStreet and ED utilization by self reported prescription opioid use Total wom en included in analysis N=3,098 No Rx opioid use, missing Rx opioid use data, ED status, age <18 , and men : excluded Participants from the Community Contacted by or at HealthStreet N=12, 986 Co mpleted intake questionnaire N= 10,293 Past 30 day Rx opioid use N=831 (27 %) Lifetime but not past 30 day use N=2,267 (73 %) Freq. ED N=268 (32 %) Non Freq. ED N=563 (68 %) Freq. ED N = 350 (15 %) Non Freq. ED N=1,917 (85 %)

PAGE 141

141 Table 5 1. HealthStreet characteristics by opioid use pattern and self report frequent ED utilization, (n= 3,098 ) L ifetime but not past 30 day Rx opioid use (n=2,267 , 73%) Past 30 day Rx opioid use only (n=831, 27%) Characteristics Non frequent ED Utilization (n=1,917 , 85% ) n (%) Frequent ED Utilization (n=350 , 15% ) n (%) P value Non frequent ED Utilization (n=563 , 68% ) n (%) Frequent ED Utilization (n=268 , 32% ) n (%) P value INDIVIDUAL MODEL FACTORS Age 18 49 years 50+ years 976 (50.9 ) 941 (49.1 ) 223 (63.7 ) 127 (36.3 ) <.0001 227 (40.3 ) 336 (59.7 ) 150 (56.0 ) 118 (44.0 ) <.0001 Race Black Other White 921 (48.0 ) 134 (7.0 ) 862 (45.0 ) 176 (50.3 ) 29 (8.3 ) 145 (41.4 ) .4001 329 (58.4 ) 25 (4.4 ) 209 (37.1 ) 154 (57.5 ) 17 (6.3) 97 (36.2 ) .5034 Education HS or less More than HS 857 (44.7 ) 1060 (55.2) 197 (56 .3 ) 153 (43.7 ) <.0001 313 (55.6 ) 250 (44.4 ) 161 (60.1 ) 107 (39.9 ) .2227 Health insurance No Yes 529 (27.6 ) 1388 (72.4 ) 134 (38.3 ) 216 (61.7 ) <.0001 111 (19.7 ) 452 (80.3 ) 85 (31.7 ) 183 (68.3 ) .0001 Doctor visits (past 6 months) No Yes 478 (24.9) 1439 (75.1) 38 (10.9) 312 (89.1 ) <.0001 59 (10.5 ) 504 (89.5 ) 12 (4.5 ) 256 (95.5 ) .0038 Depression No Yes 1248 (65.1 ) 669 (34.9 ) 165 (47.1 ) 185 (52.9 ) <.0001 319 (56.7 ) 244 (43.3 ) 116 (43.3 ) 152 (56.7 ) .0003 Anxiety No Yes 1280 (66.8 ) 637 (33.2 ) 196 (56.0 ) 154 ( 44.0 ) 0 .0001 337 (59.9 ) 226 (40.1 ) 123 (45.9 ) 145 (54.1 ) .0002 Back pain No Yes 936 (48.8 ) 981 (51.2 ) 129 (36.9 ) 221 (63.1 ) <.0001 382 (67.9 ) 181 (32.1 ) 188 (70.1 ) 80 (29.9 ) .5046 Cancer No Yes 259 (13.5 ) 1658 (86.5 ) 49 (14.0) 301 (86.0) .8059 97 (17.2 ) 466 (82.8 ) 44 (16.4 ) 224 (83.6 ) .7709 Insomnia No Yes 1313 (68.5 ) 604 (31.5 ) 194 (55.4 ) 156 (44.6 ) <.0001 342 (60.8) 221 (39.2) 144 (53.7 ) 124 (46.3 ) .0 551

PAGE 142

142 Table 5 1. Continued Lifetime but not past 30 day Rx opioid use (n=2,267, 73%) Past 30 day Rx opioid use only (n=831, 27%) Characteristics Non frequent ED Utilization (n=1,917, 85%) n (%) Frequent ED Utilization (n=350, 15%) n (%) P value Non frequent ED Utilization (n=563, 68%) n (%) Frequent ED Utilization (n=268, 32%) n (%) P value Cigarette use (lifetime use ) No Yes 973 (50.8 ) 944 (49.2 ) 143 (40.9 ) 207 (59.1 ) .0007 275 (48.9 ) 288 (51.1 ) 101 (37.7 ) 167 (62.3 ) .0025 Alcohol use (past 30 day s) No Yes 1534 (80.0 ) 383 (20.0 ) 283 ( 80.9 ) 67 (19.1 ) .7183 446 (79.2 ) 117 (20.8 ) 217 (81.0 ) 51 (19.0 ) .5567 Marijuana use (lifetime) No Yes 910 (47.5) 1007 (52.5) 161 (46.0 ) 189 (54.0 ) .6125 310 (55.1 ) 253 (44.9 ) 128 (47.8 ) 140 (52.2 ) .0488 Rx sedative use (lifetime) No Yes 1244 (64.9 ) 673 (35.1) 190 (54.3 ) 160 (45.7 ) .0002 318 (56.5 ) 245(43.5 ) 119 (44.4 ) 149 (55.6 ) .0011 RELATIONSHIP MODEL FACTORS Marital status Never married Married Separated, divorced, or widowed 671 (35.0 ) 520 (27.1 ) 726 (37.9 ) 132 (37.7 ) 72 (20.6 ) 146 (41.7 ) .0365 185 (32.9 ) 127 (22.5 ) 251 (44.6 ) 99 (36.9 ) 54 (20.2 ) 115 (42.9 ) .4763 Children No Yes 412 (21.5 ) 1505 (78.5 ) 61 (17.4) 289 (82.6 ) .0854 94 (16.7 ) 469 (83.3 ) 25 (9.3) 243 (90.4) . 0046 Employment No Yes 1148 (59.9 ) 769 (40.1 ) 253 (72.3 ) 97 (27.7 ) <.0001 439 (78.0 ) 124 (22.0 ) 211 (78.7 ) 57 (21.3 ) 0.8050 Social media use No Yes 936 (48.8 ) 981 (51.2 ) 156 (44.6 ) 194 (54.9) .1429 333 (59.1 ) 230 (40.9 ) 155 (57.8) 113 (42.2) 0.7196 COMMUNITY MODEL FACTORS Rurality No Yes 1821 (95.0) 96 (5.0) 329 (94.0 ) 21 (6.0 ) .4404 534 (94.9 ) 29 (5.1 ) 255 (95.1 ) 13 (4.9 ) .8535

PAGE 143

143 Table 5 2. HealthStreet characteristics by opioid use pattern among frequent ED utilizers, (n=618 ) Characteristics Lifetime but not past 30 day Rx opioid use (n=350 ) n (%) Past 30 day Rx opioid use (n=268 ) n (%) P value INDIVIDUAL MODEL FACTORS Age 18 49 years 50+ years 223 (63.7 ) 127 (36.3 ) 150 (56.0 ) 118 (44 .0 ) .0 511 Race Black Other White 176 (50.3 ) 29 (8.3 ) 145 (41.4 ) 154 (57.5 ) 17 (6.3) 97 (36.2 ) .1925 Education HS or less More than HS 197 (56.3 ) 153 (43.7 ) 161 (60.1 ) 107 (39.9 ) .3444 Health insurance No Yes 134 (38.3 ) 2 16 (61.7 ) 85 (31.7 ) 183 (68.3 ) .0906 Doctor visits (past 6 months) No Yes 38 (10.9) 312 (89.1) 12 (4.5 ) 256 (95.5 ) .0039 Depression No Yes 165 (47.1 ) 185 (52.9 ) 116 (43.3 ) 152 (56.7 ) .3397 Anxiety No Yes 196 (56.0 ) 154 (44.0 ) 123 (45.9 ) 147 (54.1 ) .0127 Back pain No Yes 129 (36.9 ) 221 (63.1 ) 80 (29.8 ) 1 88 (70.2 ) .0681 Cancer No Yes 301 (86.0 ) 49 (14 .0) 224 (83.6 ) 44 (16.4 ) .4048 Insomnia No Yes 194 (55.4 ) 156 (44.6 ) 144 (53.7 ) 124 (46.3 ) .6745 Cigarette use (lifetime use ) No Yes 143 (40.9 ) 207 (59.1 ) 101 (37.7 ) 167 (62.3 ) .4242 Alcohol use (past 30 day s ) No Yes 283 (80.9 ) 67 (19.1 ) 217 (81.0 ) 51 (19.0) .9717 Marijuana use (lifetime ) No Yes 161 (46.0 ) 189 (54.0 ) 1 28 (47.8 ) 140 (52.2 ) .6637 Rx sedative use (lifetime ) No Yes 190 (54.3 ) 160 (45.7 ) 119 (44.4 ) 149 (55.6 ) .0149 RELATIONSHIP MODEL FACTORS

PAGE 144

144 Table 5 2. Continued Characteristics Lifetime but not past 30 day Rx opioid use (n=350) n (%) Past 30 day Rx opioid use (n=268) n (%) P value Marital status Never married Married Separated, divorced, or widowed 132 (37.7 ) 72 (20.6 ) 146 (41.7 ) 99 (36.9 ) 54 (20.2 ) 115 (42.9 ) .9565 Children No Yes 61 (17.4) 289 (82.6 ) 25 (9.3 ) 243 (90.7 ) .0039 Employment No Yes 253 (72.3 ) 97 (27.7 ) 211 (78.7 ) 57 (21.3 ) .0664 Social media use No Yes 154 (44.6 ) 194 (54.4 ) 155 (57.8) 113 (42.2) .0011 COMMUNITY MODEL FACTORS Rurality No Yes 329 (94.0 ) 21 (6.0 ) 255 (95.2 ) 13 (4.8) .5346

PAGE 145

145 Figure 5 2. Map representing cluster s of frequent ED utilization among life time but not past 30 day prescription opioid users . = Statistically significant cluster

PAGE 146

146 Figure 5 3. Map representing cluster s of frequent ED utili zation among past 30 day users. = Statistically significant cluster

PAGE 147

147 CHAPTER 6 FINAL CONCLUSIONS Prescription opioid use is a current public health concern in the US. It is associated with serious health consequences including opioid use disorder and overdose. Opioid prescriptions have increased more than two fold in the United States (US ) between 1992 and 2015 (Pezalla et al., 2017) and overdose deaths from prescription opioid use more than quadrupled between 1999 and 2015 (Compton et al., 2015; Compton & Volkow, 2006; Han et al., 2015) . Currently, 1 30 people die daily from opioid overdose in the US, which represents over 68 % of overdose deaths every day (Centers for Disease Control and Prevention, 2018g, 2018h) . To better understand the current opioid epidemic, we must address how and by whom prescription opioids are being used in the community be fore they progress to non medical use or overdose. Prescription opioid use has been extensively examined on a national level. National surveys from NSDUH and CDC have found that the past 12 month prevalence of prescription opioid use among US adults was a pproximately 38% (Han et al., 2017a) with 6.9% of the US adult population reporting p ast 30 day prescription opioid use (Centers for Disease Control and Prevention, 2015) . Though national studies can provide an overview on prevalence and existing literature on risk factors for prescriptio n opioid use provide insight, there are a number of limitations. Gaps in the literature include a lack of information on prevalence that fluctuates in the community within specific geographic regions, a focus on opioid dependent populations which introduce bias and only represent those seeking treatment, and limited epidemiological data for addressing sex differences. Specifically, there is limited data with a focus on women and prescription opioid use within the community, even though women use and are pre scribed opioids more frequently than to men (Serdarevic, Striley, et al., 2017; Simoni Wastila, 2000) . Women subsequently face a greater bur den due to the consequences related to prescription

PAGE 148

148 opioid use. We had the opportunity to address these gaps by analyzing data from HealthStreet, an on going community engagement program that collects data in real time . The community where these data are c ollected represents an area where opioid prescribing is common and follow up among community members is possible. As such, research among this population could potentially be translated into targeted prevention and intervention efforts. The socio ecologica l model was used as a theoretical framework to explain behaviors that may influence prescription opioid use. To examine the different levels within the context of the socio ecological model, HealthStreet data were analyzed as they provided a rich data sour ce, which allowed us to examine risk factors at the individual, relationship, and community level. The aims within this dissertation were: 1) to characterize opioid use patterns and examine risk factors by sex at the individual, re lationship, and community level . 2) Among women only , examine risk factors for prescription opioid use patterns by age, and 3) identify geospatial clusters for prescription opioid use patterns among women by ED use frequency . Main Findings In Chapter 3 , we characterized patterns of prescription opioid use by sex. We found that overall 37% of the entire community sample reported lifetime but not past 30 day use and about 14% reported past 30 day use of prescription opioids for a total of 51% . We also observed a higher prevalence of prescription opioid use in our community sample than found nationally the rate was more than double for past 30 day use compared to the national rate (14% vs 6.9%; Centers for Disease Control and Prevention, 2015) . Examination of sex differences regarding prescription opioid use revealed significant differences after controlling for factors at different levels. Women were more likely to report any type of prescription opioid use compared to men. We found that factors at the individual and relationship level were significantly associated with lifetime and past 30 day prescription opioid use, however there was no significant association

PAGE 149

149 fou nd at the community level when examining urban or rural area of residence . Further, we controlled for factors at different levels of the model for men and women separately. These results also showed that risk factors at the ind ividual and relationship leve l varied by sex and magnitude of association. Women had a greater number of risk factors for prescription opioid use at the individual and relationship level compared to men, suggesting sex is a factor that must be considered in the context of opioid use. Of concern , we found that the strongest risk factor for both past 30 day and lifetime prescription opioid use was prescription sedative use. This remained true regardless of sex . This finding was specifically important due to the dangerous and potentially fatal consequences related to ingest ion of these two prescription drugs together (Dowell et al. , 2016; Jann et al., 2014) . The findings from C hapter 3 provided unique results which examined the risk factor s for three different patterns of prescription opioid use (no use, lifetime, but not past 30 day use, and past 30 day use) among a community sample where prescription opioid use is common . We found that prescription opioid use may be sex specific and that the influences and behaviors that drive prescription opioid use varied for men and women. W omen may be at gre ater risk of the consequences related to prescription opioid use, due to the high prevalence of use , yet there has been little attention on opioid use among women in the literature . For this reason, w e examined prescription opioid use among women only in C hapter 4 . In C hapter 4 , we examined risk factors at the individual, relationship, and community level for patte rns of prescription opioid use among older and younger women separately . After stratifying by age, we found a higher proportion of older women reported lif etime prescription opioid use and past 30 day use than younger women . C ontrol ling for factors at different levels of the socio ecological model revealed a ge was not a significant predictor for lifetime use of

PAGE 150

150 prescription opioids though . However, older wom en compared to younger women were more likely to report past 30 day use of prescription opioids compared to never use ( a OR: 1.31; 95% CI, 1.07 1.61 ) . This may be partly due to the natural onset of chronic conditions that occur as a person ages, which may r equire prescription opioids to treat pain . Further examination of risk factors at the individual, relationship and community level for prescription opioid use r evealed difference in risk factors for older and younger women. We found that for younger women, there were more risk factors at the individual and relationship level compared to older women. Regardless of age, we found that prescription sedative use was the strongest predictor for both younger and older women further confirming that such use needs t o be monitored due to the dangerous effects when these drugs are taken in combination especially among women who are prescribed sedatives and opioids at high rates. The findings from Chapter 4 also provide unique results as there is limited literature whic h reports on women and prescription opioids. Though we found differences at the individual and relationship level among older and younger women, the risk factor of urban or rural area of residence at the commun ity level was not significant. An association may not have been observed because rurality alone does not influence prescription opioid use at the community level, or because our binary variable (urban/rural) may have masked any true association based on area of residence. To further examine prescripti on opioid use at the community level we employed spatial analyses in Chapter 5 among women only. In C hapter 5 , we identified clusters of lifetime and past 30 day prescription opioid use among women by frequent ED utilization using spatial methods. First, in our univariate analyses, we found factors at the individual and relationship level were significantly associated with frequent ED use for women who endorsed p rescription opioid use. We found frequent ED users

PAGE 151

151 tend ed to be younger and were more likely to report a history of anxiety, depression, and sedative use compared to non frequent ED users. A higher p rop ortion of frequent ED utilization occurred among past 30 day users compared to women who endorsed prescription opioids in t heir lifetim e (32% vs 15 %). We also found that s ome factors that were significantly associated with frequent ED use among women who used prescription opioids in their lifetime were found not to be significant for women who used in the past 30 day s. This may suggest th at the relationship between frequent ED use and prescription opioid use is not well understood and warrants further examination. Second, through our spatial analysis we identified one significant cluster of frequent ED users among women who endorsed lifeti me not past 30 day prescription opioid use. We also found one significant cluster of less frequent ED utilization among women who endorsed past 30 day prescription opioid use. Both of these clusters were located in the Jacksonville, Florida area. It is pos sible that t he significant cluster we observed in Jacksonville may be partially due to the urbanicity. Urbanicity alone may not influence prescription opioid use at the community level though, and it is possible there are other community level factors whic h have an influence exist . Alternatively, if we had collected data on past 12 month prescription opioid use it may have provided us with more insight regarding opioid use and clusters of frequent ED use , rather than past 30 day use . However, th is data was not collected and may be a limitation of this study. Identification of these clusters is of specific relevance as membership in HealthStreet allows for follow up and participation in future research. These findings can be used for intervention efforts focu sing on prescription opioids and ED utilization specifically in the Jacksonville, Florida area. Specific interventions could include community outreach in which CHWs deliver educational training regarding opioid use and access from the ED as well as educat ion on the safe disposal of prescription opioids .

PAGE 152

152 Overall, the results of this dissertation show that while research among women who use prescription opioids is limited, such studies are vital in order to appropriately address the current opioid crisis . These findings from a community based study suggest that research among communities where opioid prescribing is common may provide better insight into risk factors than national studies alone. In addition, these findings could be used to implement interv ention efforts within the same community, which may provide a more targeted approach to addressing this important public health concern. Future Work Future studies are needed to address the questions that have arisen from this research. We found higher ra tes of prescription opioid use in this community sample compared to national rates suggesting that data at the national level may not account for fluctuations in certain geographical regions in the US. Specifically , though it may not be representative , a community sample such as HealthStreet , which purposefully over samples underrepresented minorities in health research, can provide more insight in regards to prescription opioid use. In addition, risk factors identified from national studies may not be app licable to all communities, which prevention strategies should account for. Further examination of prescription opioid use within community samples is crucial, specifically targeting highly affected communities such as communities in Florida, which are exp eriencing a higher rate of morbidity and mortality related to opioid use compared to the national rate (National Institute on Drug Abuse, 2018e) . I mplementing evidence based interventions across healthcare and community based settings are a necessity in prevent ing opioid m isuse and to treat opioid use disorder . For example, educating and providing basic training to health profess ionals who provide care to people with pain, and prescribers, and pharmacists to council patients who may be at risk for opioid misuse. In addition, educating the general public about the risks of prescription opioids through education

PAGE 153

153 programs at town hall meetings may increase awareness among individuals who are at risk for opioid use disorder or overdose. Further, providing access to medic ation assisted treatment (MAT) in hospitals, substance use treatment programs, and the criminal justice system while pro moting acceptance of MAT efficacy may decrease use of opioids that can be acquired in the illegal market such as heroin. Finally, allowi ng the prescribing of naloxone to laypersons and third parties can be extremely effective in reversing the effects of opioids and opioid overdose. Implementing changes such as these may reduce prescription opioid use, overdose rates related to prescription opioid use and subsequently decrease the incidence of OUD within the community which can be sustained over time. We found that women are more likely to use prescription opioids compared to men and were able to provide information on sex specific risk fact ors for prescription opioid use . We also found that older women were significantly more likely to report use compared to younger women. This may suggest that factors at the individual level such as chronic pain conditions are more likely to be associated w ith use as individuals age. Future studies need to examine this further and collect information specifically on chronic conditions that may require prescription opioid use among older women. When considering p rescription opioid use, future studies should e xamine older and younger women separately . For example, educational programs through HealthStreet can target older women specifically, while a peer mentoring system to educate and reduce unnecessary use of prescription opioids can be used for younger women (since relationship factors are important for this age group). This has also been helpful in previous work focusing on heavy alcohol use which has used a peer delivered prevention intervention method among women (Roberts et al., 2014) .

PAGE 154

154 Prescription sedative use was the strongest risk factor regardless of sex or age. This is especially important to examine further as the simultaneous use of opioids and sedatives can be fatal. Due to the limitations of time , our study could only assess the concurrent use of prescription opioids and sedatives, detail on frequency of use and dosage of these two prescriptions drugs needs to be collected to asses for simultaneous use. However, we have examined substance use, mental health , and comorbidities more extensively among this community sample previously (Dodani et al., 2016; Serdarevic et al., 2018; Serdarevic, Osborne, et al., 2017; S erdarevic, Striley, et al., 2017; Webb, Striley, & Cottler, 2015) . In addition to the concurrent use of prescription opioids and sedatives, we have found access to care including doctor and ED visits to be predictive for prescription opioid use. Healthc are providers need to be more vigilant and need to more closely monitor opioid use among their patients to further reduce the potential for non medical use . Finally, it may be important to examine a wider variety of factors at the community level, beyond a rea of residence. Societal level factors may also be of importance. Few studies have examined this previously, but this may provide further insight into the factors which influence prescription opioid use within the context of the socio ecological model. In conclusion, while there is still further work required regarding prescription opioid use among women in this community, our results provide valuable insight and have helped us to identify those who may be at risk from the adverse consequences of the cur rent opioid epidemic.

PAGE 155

155 APPENDIX INTERACTION TERMS Table A 1. Table 3 3 with interaction terms fitted into the model Characteristic Lifetime, no t past 30 day Rx opioid use N=3,463 Past 30 day Rx opioid use N=1,286 aOR 95% CI aOR 95% CI INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years Ref 1.11 0.92 1.34 Ref 0.95 0.71 1.28 Sex Male Female Ref 1.26 1.08 1.48 Ref 1.00 0.76 1.30 Race White Black Other Ref 0.64 0.68 0.58 0.72 0.56 0.83 Ref 1.02 0.68 0.87 1.19 0.50 0.93 Education More than HS HS or less Ref 0.72 0.65 0.80 Ref 0.93 0.80 1.08 Health insurance Yes No Ref 0.95 0.86 1.06 Ref 0.71 0.61 0.84 Doctor visits (past 6 months) No Yes Ref 1.45 1.30 1.62 Ref 3.36 2.74 4.12 ED visits (past 6 months) 0 1 >2 Ref 1.24 1.06 1.44 Ref 2.57 2.16 3.05 Depression No Yes eliminated eliminated Back pain No Yes Ref 1.41 1.17 1.70 Ref 2.78 2.11 3.67 Cancer No Yes Ref 1.57 1.30 1.90 Ref 1.89 1.50 2.40 Insomnia No Yes Ref 1.24 1.10 1.40 Ref 1.26 1.07 1.47 Cigarette use (lifetime) No Yes eliminated eliminated Hazardous alcohol use (past 30 day s) No Yes Ref 1.08 0.96 1.21 Ref 1.48 1.26 1.75

PAGE 156

156 Table A 1. Continued Characteristic Lifetime, no t past 30 day Rx opioid use N=3,463 Past 30 day Rx opioid use N=1,286 aOR 95% CI aOR 95% CI Marijuana use (lifetime) No Yes Ref 1.76 1.59 1.95 Ref 1.27 1.10 1.47 Rx sedative use (lifetime) No Yes Ref 2.65 2.33 3.02 Ref 3.96 3.35 4.68 INTERACTIONS Sex (male)*Pain (No) Sex (male)*Pain (Yes) Sex (female)*Pain (No) Sex (female)*Pain (Yes) Sex (male)*Age (18 49 years) Sex (male)*Age (50+ years) Sex (female)* Age(18 49years) Sex (female)*Age (50+ years) Ref 1.41 1.26 1.78 Ref 1.11 1.26 1.40 1.17 1.69 1.08 1.48 1.35 2.35 0.92 1.34 1.08 1.48 1.05 1.86 Ref 2.78 1.00 2.78 Ref 0.95 1.00 0.95 2.11 3.67 0.76 1.30 1.77 4.37 0.71 1.28 0.76 1.30 0.60 1.50 RELATIONSHIP LEVEL FACTORS Marital status Married Never married Separated, divorced, or widowed Ref 0.72 0.99 0.63 0.83 0.86 1.13 Ref 0.78 0.96 0.64 0.95 0.80 1.16 Children No Yes Ref 1.26 1.12 1.42 Ref 1.39 1.16 1.66 Employment Yes No Ref 0.90 0.81 1.00 Ref 1.48 1.26 1.74 Social media use No Yes Ref 1.33 1.20 1.47 Ref 0.92 0.80 1.07 COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated aOR= adjusted Odds Ratio; CI= Confidence Interval; ref= reference group; N= sample size

PAGE 157

157 Table A 2. Table 4 3 with interaction terms fitted into the model Characteristic Lifetime, no past 30 day Rx opioid use, N=2,267 Past 30 day Rx opioid use, N=831 aOR 95% CI aOR 95% CI INDIVIDUAL LEVEL FACTORS Age 18 49 years 50+ years Ref 1.08 0.89 1.32 Ref 1.42 1.03 1.95 Race White Black Other Ref 0.80 0.86 0.69 0.92 0.67 1.11 Ref 1.24 0.78 1.01 1.51 0.52 1.17 Education More than HS HS or less Ref 0.69 0.61 0.79 Ref 0.86 0.72 1.04 Health insurance Yes No Ref 0.91 0.79 1.05 Ref 0.71 0.58 0.87 Doctor visits (past 6 months) No Yes Ref 1.32 1.14 1.53 Ref 3.13 2.38 4.12 ED visits (past 6 months) 0 1 >2 Ref 1.26 1.04 1.52 Ref 2.70 2.17 3.36 Depression No Yes eliminated eliminated Back pain No Yes Ref 1.80 1.52 2.13 Ref 2.93 2.26 3.78 Cancer No Yes Ref 1.81 1.43 2.29 Ref 2.19 1.65 2.90 Insomnia No Yes eliminated eliminated Cigarette use (lifetime) No Yes eliminated eliminated Hazardous alcohol use (past 30 day s) No Yes eliminated eliminated Marijuana use (lifetime) No Yes Ref 1.86 1.63 2.11 Ref 1.41 1.18 1.69 Rx sedative use (lifetime) No Yes Ref 2.83 2.27 3.51 Ref 4.78 3.59 6.38 INTERACTIONS

PAGE 158

158 Table A 2. Continued Characteristic Lifetime, no past 30 day Rx opioid use, N=2,267 Past 30 day Rx opioid use, N=831 aOR 95% CI aOR 95% CI Age (18 49 years)*Pain (No) Age (18 49 years)*Pain (Yes) Age (50+ years)*Pain (No) Age (50+ years)*Pain (Yes) Age (18 49 years)*Sedative use (No) Age (50+ years)*Sedative use (No) Age (18 49 years)*Sedative use (Yes) Age (50+ years)*Sedative use (Yes) Ref 1.80 1.08 1.95 Ref 1.08 2.83 3.06 1.53 2.13 0.89 1.32 1.45 2.63 0.89 1.32 2.27 3.51 2.22 4.22 Ref 2.93 1.42 4.14 Ref 1.42 4.78 6.77 2.26 3.78 1.03 1.95 2.54 6.75 1.03 1.95 3.59 6.38 4.19 10.96 RELATIONSHIP LEVEL FACTORS Marital status Married Never married Separated, divorced, or widowed Ref 0.74 1.00 0.62 0.88 0.84 1.19 Ref 0.92 1.08 0.72 1.19 0.85 1.36 Children No Yes Ref 1.47 1.25 1.73 Ref 1.70 1.33 2.17 Employment Yes No Ref 0.87 0.76 0.99 Ref 1.47 1.20 1.81 Social media use No Yes Ref 1.43 1.26 1.62 Ref 1.03 0.86 1.24 COMMUNITY LEVEL FACTORS Rurality No Yes eliminated eliminated aOR= adjusted Odds Ratio; CI= Confidence Interval; ref= reference group; N= sample size

PAGE 159

159 LIST OF REFERENCES Alom, M., Wymer, K., & Trost, L. (2018). Chapter 6 The Impact of Lifestyle Factors on Male Reproduction: Body Composition, Diet, Exercise, and Recreational Drug Use. In S. C. Sik ka & W. J. G. Hellstrom (Eds.), Reproductive and Sexual Health (pp. 85 105). https://doi.org/10.1016/B978 0 12 801299 4.00006 2 AJN The American Journal of Nursing , 117 (10), 21. https://doi.org/10.1097/01.NAJ.0000525869.21182.72 (2011). Comparative Profiles of Men and Women with Opioid Dependence: Result s from a National Multisite Effectiveness Trial. The American Journal of Drug and Alcohol Abuse , 37 (5), 313 323. https://doi.org/10.3109/00952990.2011.596982 Baehren, D. F., Marco, C. A., Droz, D. E., Sinha, S., Callan, E. M., & Akpunonu, P. (2010). A Statewide Prescription Monitoring Program Affects Emergency Department Prescribing Behaviors. Annals of Emergency Medicine , 56 (1), 19 23.e3. https://doi.org/10.10 16/j.annemergmed.2009.12.011 Barnett, M. L., Olenski, A. R., & Jena, A. B. (2017). Opioid Prescribing Patterns of Emergency Physicians and Risk of Long Term Use. The New England Journal of Medicine , 376 (7), 663 673. https://doi.org/10.1056/NEJMsa1610524 Ba rsky, A. J., Peekna, H. M., & Borus, J. F. (2001). Somatic Symptom Reporting in Women and Men. Journal of General Internal Medicine , 16 (4), 266 275. https://doi.org/10.1046/j.1525 1497.2001.016004266.x Bartley, E. J., & Fillingim, R. B. (2013). Sex differe nces in pain: a brief review of clinical and experimental findings. BJA: British Journal of Anaesthesia , 111 (1), 52 58. https://doi.org/10.1093/bja/aet127 Bates, C., Laciak, R., Southwick, A., & Bishoff, J. (2011). Overprescription of Postoperative Narcoti cs: A Look at Postoperative Pain Medication Delivery, Consumption and Disposal in Urological Practice. The Journal of Urology , 185 (2), 551 555. https://doi.org/10.1016/j.juro.2010.09.088 Beaudoin, F. L., Lin, C., Guan, W., & Merchant, R. C. (2014). Low dos e Ketamine Improves Pain Relief in Patients Receiving Intravenous Opioids for Acute Pain in the Emergency Department: Results of a Randomized, Double blind, Clinical Trial. Academic Emergency Medicine , 21 (11), 1193 1202. https://doi.org/10.1111/acem.12510 (2015). An Evaluation of Naloxone Use for Opioid Overdoses in West Virginia: A Literature Review. American Medical Journal , 6 (1), 9 13. https://doi.org/10.3844/amjsp.2 015.9.13

PAGE 160

160 Belsky, J. (1980). Child maltreatment: An ecological integration. American Psychologist , 35 (4), 320 335. https://doi.org/10.1037/0003 066X.35.4.320 Billings, J., & Raven, M. C. (2013). Dispelling an urban legend: frequent emergency department user s have substantial burden of disease. Health Affairs (Project Hope) , 32 (12), 2099 2108. https://doi.org/10.1377/hlthaff.2012.1276 Blanco, C., Wall, M. M., Okuda, M., Wang, S., Iza, M., & Olfson, M. (2016). Pain as a Predictor of Opioid Use Disorder in a Na tionally Representative Sample. American Journal of Psychiatry , 173 (12), 1189 1195. https://doi.org/10.1176/appi.ajp.2016.15091179 Bodnar, L. M., & Wisner, K. L. (2005). Nutrition and Depression: Implications for Improving Mental Health Among Childbearing Aged Women. Biological Psychiatry , 58 (9), 679 685. https://doi.org/10.1016/j.biopsych.2005.05.009 Boland, J. W., & Pockley, A. G. (2018). Influence of opioids on immune function in patients with cancer pain: from bench to bedside. British Journal of Pharma cology , 175 (14), 2726 2736. https://doi.org/10.1111/bph.13903 Bonnie, R. J., Kesselheim, A. S., & Clark, D. J. (2017). Both Urgency and Balance Needed in Addressing Opioid Epidemic: A Report From the National Academies of Sciences, Engineering, and Medicin e. JAMA , 318 (5), 423 424. https://doi.org/10.1001/jama.2017.10046 Bonnie, R. J., Stroud, C., Breiner, H., Committee on Improving the Health, S., Board on Children, Y., Medicine, I. of, & Council, N. R. (2015). The Health Care System . Retrieved from https:/ /www.ncbi.nlm.nih.gov/books/NBK284795/ Boyd, C. J., Esteban McCabe, S., & Teter, C. J. (2006). Medical and nonmedical use of prescription pain medication by youth in a Detroit area public school district. Drug and Alcohol Dependence , 81 (1), 37 45. https://doi.org/10.1016/j.drugalcdep.2005.05.017 Boyd, C. J., McCabe, S. E., Cranford, J. A., & Young, A. (2007). Prescription Drug Abuse and Diversion Among Adolescents in a Southeast Michigan School District. Archives of Pediatrics & Adolescent Medicine , 161 (3), 276 281. https://doi.org/10.1001/archpedi.161.3.276 Broida, R. I., Gronowski, T., Kalnow, A. F., Little, A. G., & Lloyd, C. M. (2017). State Emergency Department Opioid Guidelines: Current Status. Western Journal of Emergency Medicine , 18 (3), 340 344. https://doi.org/10.5811/westjem.2016.12.30854 Bronfenbrenner, U. (1977). Toward an experimental ecology of human development. American Psychologist , 32 (7), 513 531. https://doi.org/10.1037/0003 066X.32.7.513 Bruera, E., & Kim, H. N. (2003). Cancer pai n. JAMA , 290 (18), 2476 2479. https://doi.org/10.1001/jama.290.18.2476

PAGE 161

161 Byrne, M., Murphy, A. W., Plunkett, P. K., McGee, H. M., Murray, A., & Bury, G. (2003). Frequent attenders to an emergency department: a study of primary health care use, medical profile , and psychosocial characteristics. Annals of Emergency Medicine , 41 (3), 309 318. https://doi.org/10.1067/mem.2003.68 Calcaterra, S., Glanz, J., & Binswanger, I. A. (2013). National trends in pharmaceutical opioid related overdose deaths compared to other substance related overdose deaths: 1999 2009. Drug and Alcohol Dependence , 131 (3), 263 270. https://doi.org/10.1016/j.drugalcdep.2012.11.018 Korff, M. (2010). Age and Gender Trends in Long Term Opioid Analgesic Use for Noncancer Pain. American Journal of Public Health , 100 (12), 2541 2547. https://doi.org/10.2105/AJPH.2009.180646 Cantrill, S. V., Brown, M. D., Carlisle, R. J., Delaney, K. A., Hays, D. P., Nelson, L. American College of Emergency Physicians Opioid Guideline Writing Panel. (2012). Clinical policy: critical issues in the prescribing of opioids for adult patients in the emergency department. Annals of Emergency Medicine , 60 (4), 499 525. https://doi. org/10.1016/j.annemergmed.2012.06.013 Caudill Slosberg, M. A., Schwartz, L. M., & Woloshin, S. (2004). Office visits and analgesic prescriptions for musculoskeletal pain in US: 1980 vs. 2000. Pain , 109 (3), 514 519. https://doi.org/10.1016/j.pain.2004.03.00 6 Centers for Disease Control and Prevention. (2014). Decline in Drug Overdose Deaths After State Policy Changes Florida, 2010 2012. Retrieved September 12, 2018, from https://www.cdc.gov/mmwr/preview/mmwrhtml/mm6326a3.htm Centers for Disease Control and Prevention. (2015). Prescription Opioid Analgesic Use Among Adults: United States, 1999 2012. Retrieved August 20, 2018, from https://www.cdc.gov/nchs/products/databriefs/db189.htm Centers for Disease Control and Prevention. (2016). U.S. State Prescribing Rates. Retrieved September 10, 2018, from https://www.cdc.gov/drugoverdose/maps/rxstate2016.html Centers for Disease Control and Prevention. (2017a). Drug Overdose Death Data. Retrieved October 18, 2018, from https://www.cdc.gov/drugoverdose/data/statedea ths.html Centers for Disease Control and Prevention. (2017b). Drug Overdose Deaths in the United States, 1999 2016. Retrieved October 20, 2018, from https://www.cdc.gov/nchs/products/databriefs/db294.htm Centers for Disease Control and Prevention. (2017c). Opioid Data Analysis. Retrieved March 17, 2018, from https://www.cdc.gov/drugoverdose/data/analysis.html Centers for Disease Control and Prevention. (2017d). What States Need to Know about PDMPs. Retrieved October 20, 2018, from https://www.cdc.gov/drugov erdose/pdmp/states.html

PAGE 162

162 Centers for Disease Control and Prevention. (2018a). 2018 Annual Surveillance Report Of Drug Related Risks And Outcomes . Retrieved from https://www.cdc.gov/drugoverdose/pdf/pubs/2018 cdc drug surveillance report.pdf Centers for Dise ase Control and Prevention. (2018b). Drug Overdose Deaths. Retrieved January 8, 2019, from https://www.cdc.gov/drugoverdose/data/statedeaths.html Centers for Disease Control and Prevention. (2018c, August 31). Heroin Overdose Data. Retrieved October 16, 20 18, from https://www.cdc.gov/drugoverdose/data/heroin.html Centers for Disease Control and Prevention. (2018d, September 4). Prescription Painkiller Overdoses. Retrieved September 6, 2018, from Centers for Disease Control and Prevention website: https://ww w.cdc.gov/vitalsigns/prescriptionpainkilleroverdoses/index.html Centers for Disease Control and Prevention. (2018e, October 3). U.S. Opioid Prescribing Rate Maps. Retrieved October 16, 2018, from Opioid Overdose website: https://www.cdc.gov/drugoverdose/ma ps/rxrate maps.html Centers for Disease Control and Prevention. (2018f, November 9). Adults Need More Physical Activity. Retrieved January 14, 2019, from Centers for Disease Control and Prevention website: https://www.cdc.gov/physicalactivity/inactivity am ong adults 50plus/index.html Centers for Disease Control and Prevention. (2018g, December 19). Understanding the Epidemic | Drug Overdose | CDC Injury Center. Retrieved January 10, 2019, from https://www.cdc.gov/drugoverdose/epidemic/index.html Centers for Disease Control and Prevention. (2018h, December 19). Wide ranging online data for epidemiologic research (WONDER). Retrieved January 10, 2019, from http://wonder.cdc.gov Centers for Disease Control and Prevention. (n.d.). CDC Social Ecological Model CRCCP. Retrieved July 12, 2017, from https://www.cdc.gov/cancer/crccp/sem.htm Cerdá, M., Santaella, J., Marshall, B. D. L., Kim, J. H., & Martins, S. S. (2015). Nonmedical Prescription Opioid Use in Childhood and Early Adolescence Predicts Transitions to H eroin Use in Young Adulthood: A National Study. The Journal of Pediatrics , 167 (3), 605 612.e1 2. https://doi.org/10.1016/j.jpeds.2015.04.071 Chan, K. K., Neighbors, C., Gilson, M., Larimer, M. E., & Alan Marlatt, G. (2007). Epidemiological trends in drinki ng by age and gender: Providing normative feedback to adults. Addictive Behaviors , 32 (5), 967 976. https://doi.org/10.1016/j.addbeh.2006.07.003 Chang, H. Y., Kharrazi, H., Bodycombe, D., Weiner, J. P., & Alexander, G. C. (2018). Healthcare costs and utiliz ation associated with high risk prescription opioid use: a retrospective cohort study. BMC Medicine , 16 . https://doi.org/10.1186/s12916 018 1058 y

PAGE 163

163 Christie, C., Baker, C., Cooper, R., Kennedy, P. J., Madras, B., & Bondi, P. (2017, November 11). The Preside Retrieved from https://www.whitehouse.gov/sites/whitehouse.gov/files/images/Final_Report_Draft_11 1 2017.pdf Cicero, T. J., Surratt, H., Inciardi, J. A., & Munoz, A. (2007). Relationship be tween therapeutic use and abuse of opioid analgesics in rural, suburban, and urban locations in the United States. Pharmacoepidemiology and Drug Safety , 16 (8), 827 840. https://doi.org/10.1002/pds.1452 Cohen, D., Spear, S., Scribner, R., Kissinger, P., Mas American Journal of Public Health , 90 (2), 230 236. Compton, W. M., Boyle, M., & Wargo, E. (2015). Prescription opioid abuse: Problems and responses. Preventive Medicine , 80 , 5 9. htt ps://doi.org/10.1016/j.ypmed.2015.04.003 Compton, W. M., Jones, C. M., & Baldwin, G. T. (2016a). Relationship between Nonmedical Prescription Opioid Use and Heroin Use. The New England Journal of Medicine , 374 (2), 154 163. https://doi.org/10.1056/NEJMra150 8490 Compton, W. M., Jones, C. M., & Baldwin, G. T. (2016b). Relationship between Nonmedical Prescription Opioid Use and Heroin Use. The New England Journal of Medicine , 374 (2), 154 163. https://doi.org/10.1056/NEJMra1508490 Compton, W. M., & Volkow, N. D. (2006). Major increases in opioid analgesic abuse in the United States: Concerns and strategies. Drug and Alcohol Dependence , 81 (2), 103 107. https://doi.org/10.1016/j.drugalcdep.2005.05.009 Connell, C. M., Gilreath, T. D., Aklin, W. M., & Brex, R. A. (20 10). Social Ecological Influences on Patterns of Substance Use Among Non Metropolitan High School Students. American Journal of Community Psychology , 45 (1 2), 36 48. https://doi.org/10.1007/s10464 009 9289 x Cordell, W. H., Keene, K. K., Giles, B. K., Jone s, J. B., Jones, J. H., & Brizendine, E. J. (2002). The high prevalence of pain in emergency medical care. The American Journal of Emergency Medicine , 20 (3), 165 169. HealthStreet: A Community Based Approach to Include Mental Health in Public Health Research, in: Mental Health in Public Health: The next 100 Years. New York: Oxford University Press. Cottler, L., Otufowora, A., Baker, D., & Egan, K. (2018, June). Community members are Interested in Deterra® drug deactivation pouches to dispose of unused medications . Presented at the CPDD, San Diego, CA.

PAGE 164

164 Cronbach, L. J., & Snow, R. E. (1977). Aptitudes and instructional methods: A handbook for research on interactions . In Aptitudes and Instructional Methods: A Han dbook for Research on Interactions . Oxford, England: Irvington. Daniell, H. W. (2008). Opioid endocrinopathy in women consuming prescribed sustained action opioids for control of nonmalignant pain. The Journal of Pain: Official Journal of the American Pain Society , 9 (1), 28 36. https://doi.org/10.1016/j.jpain.2007.08.005 Darnall, B. D., Stacey, B. R., & Chou, R. (2012). Medical and psychological risks and consequences of long term opioid therapy in women. Pain Medicine (Malden, Mass.) , 13 (9), 1181 1211. https://doi.org/10.1111/j.1526 4637.2012.01467.x Dasgupta, N., Funk, M. J., Proescholdbell, S., Hirsch, A., Ribisl, K. M., & Marshall, S. (2016). Cohort Study of the Impact of High Dose Opioid Analgesics on Overdose Mortality. Pain Medicine (Malden, Mass.) , 17 (1), 85 98. https://doi.org/10.1111/pme.12907 Davis, J. P., Dumas, T. M., Wagner, E. F., & Merrin, G. J. (2016). Social Ecological Determinants of Substance Use Treatment Entry Among Serious Juvenile Offenders From Adolescence Through Emerging Adulthood. Journal of Substance Abuse Treatment , 71 , 8 15. https://doi.org/10.1016/j.jsat.2016.08.004 Day, C. (2013). Benzodiazepines in Combination with Opioid Pain Relievers or Alcohol: Greater Risk of More Serious ED Visit Outcomes. In The CBHSQ Report . Retrieved from http://www.ncbi.nlm.nih.gov/books/NBK38467 2/ Department of Veterans Affairs/Department of Defense. (2017). VA/DoD Clinical Practice Guideline for Opioid Therapy for Chronic Pain: Clinician Summary . 30. DiMatteo, M. R., Lepper, H. S., & Croghan, T. W. (2000). Depression Is a Risk Factor for Noncomp liance With Medical Treatment: Meta analysis of the Effects of Anxiety and Depression on Patient Adherence. Archives of Internal Medicine , 160 (14), 2101 2107. https://doi.org/10.1001/archinte.160.14.2101 Dodani, S., Ruktanonchai, C. W., Kaeley, G. S., Vadd iparti, K., Striley, C. W., & Cottler, L. B. (2016). Clinical Comorbidities among Cocaine Users Screened in the Community through HealthStreet. Health Behavior and Policy Review , 3 (1), 54 61. https://doi.org/10.14485/HBPR.3.1.6 Dorak, M. T., & Karpuzoglu, E. (2012). Gender Differences in Cancer Susceptibility: An Inadequately Addressed Issue. Frontiers in Genetics , 3 . https://doi.org/10.3389/fgene.2012.00268 Doran, K. M., Raven, M. C., & Rosenheck, R. A. (2013). What Drives Frequent Emergency Department Use in an Integrated Health System? National Data From the Veterans Health Administration. Annals of Emergency Medicine , 62 (2), 151 159. https://doi.org/10.1016/j.annemergmed.2013.02.016

PAGE 165

165 Dowell, D., Haegerich, T. M., & Chou, R. (2016). CDC Guideline for Presc ribing Opioids for Chronic Pain United States, 2016. JAMA , 315 (15), 1624 1645. https://doi.org/10.1001/jama.2016.1464 Drug Enforcement Administration. (2017). Drug Scheduling. Retrieved July 30, 2018, from https://www.dea.gov/druginfo/ds.shtml Dummer, T. J . B. (2008). Health geography: supporting public health policy and planning. , 178 (9), 1177 1180. https://doi.org/10.1503/cmaj.071783 Dworkis, D. A., Taylor, L. A., Peak, D. A., & Bearnot, B. (2017). Geospatial an alysis of emergency department visits for targeting community based responses to the opioid epidemic. PLOS ONE , 12 (3), e0175115. https://doi.org/10.1371/journal.pone.0175115 Dworkis, D. A., Weiner, S. G., Liao, V. T., Rabickow, D., & Goldberg, S. A. (2018) . Geospatial Clustering of Opioid Related Emergency Medical Services Runs for Public Deployment of Naloxone. The Western Journal of Emergency Medicine , 19 (4), 641 648. https://doi.org/10.5811/westjem.2018.4.37054 Dwyer Lindgren, L., Bertozzi Villa, A., Stu bbs, R. W., Morozoff, C., Mackenbach, J. P., Lenthe, Counties, 1980 to 2014: Temporal Trends and Key Drivers. JAMA Internal Medicine , 177 (7), 1003 1011. https://doi.org/10.1001/ jamainternmed.2017.0918 Edlund, M. J., Forman Hoffman, V. L., Winder, C. R., Heller, D. C., Kroutil, L. A., Lipari, R. N., & Colpe, L. J. (2015). Opioid abuse and depression in adolescents: Results from the National Survey on Drug Use and Health. Drug and Alcohol Dependence , 152 , 131 138. https://doi.org/10.1016/j.drugalcdep.2015.04.010 R. A. (2017). Health Care Systems Support to Enhance Patient Centered Care: Le ssons from a Primary Care Based Chronic Pain Management Initiative. The Permanente Journal , 21 . https://doi.org/10.7812/TPP/16 101 Fan, L., Shah, M. N., Veazie, P., & Friedman, B. (2011). Factors Associated with Emergency Department Use among the Rural Eld erly. of the American Rural Health Association and the National Rural Health Care Association , 27 (1), 39 49. https://doi.org/10.1111/j.1748 0361.2010.00313.x Fillingim, R. B., Ness, T. J., Glover, T. L., Campb ell, C. M., Hastie, B. A., Price, D. D., & Staud, R. (2005). Morphine responses and experimental pain: sex differences in side effects and cardiovascular responses but not analgesia. The Journal of Pain: Official Journal of the American Pain Society , 6 (2), 116 124. https://doi.org/10.1016/j.jpain.2004.11.005 Healthcare Practitioner Licensing and Regulation. Retrieved September 20, 2018, from https://flboardofmedicine. gov/latest news/opioid epidemic in florida/

PAGE 166

166 Florida drug Related Outcomes Surveillance and Tracking System (FROST). (2018). Florida drug Related Outcomes Surveillance and Tracking System (FROST). Retrieved October 11, 2018, from Florida drug Related Outcom es Surveillance and Tracking (FROST) System website: https://frost.med.ufl.edu/frost/ Food and Drug Administration. (2018). Menopause and Hormones: Common Questions [WebContent]. Retrieved October 18, 2018, from https://www.fda.gov/forconsumers/byaudience/ forwomen/ucm118624.htm A. (2016). The Effect of Substance Use Disorders on the Association Between Guideline concordant Long term Opioid Therapy and All ca use Mortality. Journal of Addiction Medicine , 10 (6), 418 428. https://doi.org/10.1097/ADM.0000000000000255 undergoing infertility treatment. Ginekologia Polska , 88 (2), 109 112. https://doi.org/10.5603/GP.a2017.0019 Gill, J. M., Mainous, A. G., & Nsereko, M. (2000). The effect of continuity of care on emergency department use. Archives of Family Medicine , 9 (4), 333 338. Glanz, K., Rimer, B. K., & Viswanath, K. (2015). Health Behavior: Theory, Research, and Practice . John Wiley & Sons. Gold, E. B. (2011). The Timing of the Age at Which Natural Menopause Occurs. Obstetrics and Gynecology Clinics of North America , 3 8 (3), 425 440. https://doi.org/10.1016/j.ogc.2011.05.002 Graham, G. N. (2016). Why Your ZIP Code Matters More Than Your Genetic Code: Promoting Healthy Outcomes from Mother to Child. Breastfeeding Medicine , 11 (8), 396 397. https://doi.org/10.1089/bfm.2016. 0113 Greenland, S. (1983). Tests for interaction in epidemiologic studies: A review and a study of power. Statistics in Medicine , 2 (2), 243 251. https://doi.org/10.1002/sim.4780020219 Gudin, J. A., Mogali, S., Jones, J. D., & Comer, S. D. (2013). Risks, Ma nagement, and Monitoring of Combination Opioid, Benzodiazepines, and/or Alcohol Use. Postgraduate Medicine , 125 (4), 115 130. https://doi.org/10.3810/pgm.2013.07.2684 Guru, V., & Dubinsky, I. (2000). The patient vs. caregiver perception of acute pain in the emergency department Journal of Emergency Medicine. The Journal of Emergency Medicine , 18 (1). Hahn, K. L. (2011). Strategies to Prevent Opioid Misuse, Abuse, and Diversion That May Also Reduce the Associated Costs. American Health & Drug Benefits , 4 (2), 107 114.

PAGE 167

167 aulozzi, L. J. (2008). Patterns of abuse among unintentional pharmaceutical overdose fatalities. JAMA , 300 (22), 2613 2620. https://doi.org/10.1001/jama.2008.802 Han, B., Compton, W. M., Blanco, C., Crane, E., Lee, J., & Jones, C. M. (2017a). Prescription O pioid Use, Misuse, and Use Disorders in U.S. Adults: 2015 National Survey on Drug Use and Health. Annals of Internal Medicine , 167 (5), 293 301. https://doi.org/10.7326/M17 0865 Han, B., Compton, W. M., Blanco, C., Crane, E., Lee, J., & Jones, C. M. (2017b) . Prescription Opioid Use, Misuse, and Use Disorders in U.S. Adults: 2015 National Survey on Drug Use and Health. Annals of Internal Medicine , 167 (5), 293. https://doi.org/10.7326/M17 0865 Han, B., Compton, W. M., Blanco, C., Crane, E., Lee, J., & Jones, C . M. (2017c). Prescription Opioid Use, Misuse, and Use Disorders in U.S. Adults: 2015 National Survey on Drug Use and Health. Annals of Internal Medicine , 167 (5), 293 301. https://doi.org/10.7326/M17 0865 Han, B., Compton, W. M., Jones, C. M., & Cai, R. (2 015). Nonmedical Prescription Opioid Use and Use Disorders Among Adults Aged 18 Through 64 Years in the United States, 2003 2013. JAMA , 314 (14), 1468 1478. https://doi.org/10.1001/jama.2015.11859 Harris, P. A., Taylor, R., Thielke, R., Payne, J., Gonzalez, N., & Conde, J. G. (2009). Research electronic data capture (REDCap) A metadata driven methodology and workflow process for providing translational research informatics support. Journal of Biomedical Informatics , 42 (2), 377 381. https://doi.org/10.1016/j. jbi.2008.08.010 HCUP. (2018). Opioid Hospital Stays/Emergency Department Visits HCUP Fast Stats. Retrieved January 17, 2019, from https://www.hcup us.ahrq.gov/faststats/OpioidUseServlet?radio 3=on&location1=FL&characteristic1=01&setting1=ED&location2=US& characteristic2= 01&setting2=ED&expansionInfoState=hide&dataTablesState=show&definitionsState=sh ow&exportState=hide Hearon, B. A., Calkins, A. W., Halperin, D. M., McHugh, R. K., Murray, H. W., & Otto, M. W. (2011). Anxiety sensitivity and illicit sedative use among opiate dependent women and men. The American Journal of Drug and Alcohol Abuse , 37 (1), 43 47. https://doi.org/10.3109/00952990.2010.535581 Hernandez Boussard, T., Burns, C. S., Wang, N. E., Baker, L. C., & Goldstein, B. A. (2014). The Affordable Care Act Reduces Emergency Department Use By Young Adults: Evidence From Three States. Health Affairs (Project Hope) , 33 (9), 1648 1654. https://doi.org/10.1377/hlthaff.2014.0103 Herr, K. (2002). Chronic pain in the older patient: management strategies. 2. Journal of Gerontological Nursing , 28 (2), 28 34; quiz 54 55.

PAGE 168

168 Hill, A. B. (1965). The environment and disease: association or causation? Journal of the Royal Society of Medicine , 108 (1), 32 37. https://doi.org/10.1177/0141076814562718 Hoppe, J. A., McStay, C., Sun, B. C., & Capp, R. (2017). Emergency Department Attending Physician Variation in Opioid Prescribing in Low Acuity Back Pain. Western Journal of Emergency Medicine , 18 (6), 1135 1142. https://doi.org/10.5811/westjem.2017.7.33306 Houghton, L. A., Heit kemper, M., Crowell, M. D., Emmanuel, A., Halpert, A., McRoberts, J. A., Gastroenterology , 150 (6), 1332 1343.e4. https://doi.org/10.1053/j.gastro.2016.02.017 Hunt, K. A., Weber, E. J., Sh owstack, J. A., Colby, D. C., & Callaham, M. L. (2006). Characteristics of frequent users of emergency departments. Annals of Emergency Medicine , 48 (1), 1 8. https://doi.org/10.1016/j.annemergmed.2005.12.030 Hwang, C. S., Kang, E. M., Kornegay, C. J., Staf fa, J. A., Jones, C. M., & McAninch, J. K. (2016). Trends in the Concomitant Prescribing of Opioids and Benzodiazepines, American Journal of Preventive Medicine , 51 (2), 151 160. https://doi.org/10.1016/j.amepre.2016.02.014 Inciardi, J. A., Surra tt, H. L., Kurtz, S. P., & Cicero, T. J. (2007). Mechanisms of Prescription Drug Diversion Among Drug Involved Club and Street Based Populations. Pain Medicine , 8 (2), 171 183. https://doi.org/10.1111/j.1526 4637.2006.00255.x Jann, M., Kennedy, W. K., & Lo pez, G. (2014). Benzodiazepines: a major component in unintentional prescription drug overdoses with opioid analgesics. Journal of Pharmacy Practice , 27 (1), 5 16. https://doi.org/10.1177/0897190013515001 Jansson, L. M., Velez, M., & Harrow, C. (2009). The Opioid Exposed Newborn: Assessment and Pharmacologic Management. Journal of Opioid Management , 5 (1), 47 55. Johansson, J., Hellénius, M. L., Elofsson, S., & Krakau, I. (1999). Self report as a selection instrument in screening for cardiovascular disease ri sk. American Journal of Preventive Medicine , 16 (4), 322 324. L. (2017). Prescription Opioid Abuse in Chronic Pain: An Updated Review of Opioid Abuse Predictors and Strategies to Curb Opioid Abuse: Part 1. Pain Physician , 20 (2S), S93 S109. Kennedy Hendricks, A., Richey, M., McGinty, E. E., Stuart, E. A., Barry, C. L., & Webster, D. American Journal of Public Health , 106 (2), 291 297. https://doi.org/10.2105/AJPH.2015.302953 Kessler, R. C. (2006, January). The Epidemiology of Depression among Women. https://doi.org/10.1017/CBO9780511841262.004

PAGE 169

169 Keyes, K. M., Cerdá, M., Brady, J. E., Havens, J. R ., & Galea, S. (2014). Understanding the Rural Urban Differences in Nonmedical Prescription Opioid Use and Abuse in the United States. American Journal of Public Health , 104 (2), e52 e59. https://doi.org/10.2105/AJPH.2013.301709 Koons, A. L., Rayl Greenberg , M., Cannon, R. D., & Beauchamp, G. A. (2018). Women and the Experience of Pain and Opioid Use Disorder: A Literature based Commentary. Clinical Therapeutics , 40 (2), 190 196. https://doi.org/10.1016/j.clinthera.2017.12.016 Krans, E. E., & Patrick, S. W. (2016). Opioid Use Disorder in Pregnancy: Health Policy and Practice in the Midst of an Epidemic. Obstetrics and Gynecology , 128 (1), 4 10. https://doi.org/10.1097/AOG.0000000000001446 Kroll, D. S., Nieva, H. R., Barsky, A. J. , & Linder, J. A. (2016). Benzodiazepines are Prescribed More Frequently to Patients Already at Risk for Benzodiazepine Related Adverse Events in Primary Care. Journal of General Internal Medicine , 31 (9), 1027 1034. https://doi.org/10.1007/s11606 016 3740 0 Kulldorff, M. (1997). Bernoulli, Discrete Poisson and Continuous Poisson Models: A spatial scan statistic. Communications in Statistics: Theory and Methods, 26:1481 1496, 1997. Kushel, M. B., Perry, S., Bangsberg, D., Clark, R., & Moss, A. R. (2002). Eme rgency department use among the homeless and marginally housed: results from a community based study. American Journal of Public Health , 92 (5), 778 784. LaCalle, E., & Rabin, E. (2010). Frequent users of emergency departments: the myths, the data, and the policy implications. Annals of Emergency Medicine , 56 (1), 42 48. https://doi.org/10.1016/j.annemergmed.2010.01.032 Lasser, K. E., Kronman, A. C., Cabral, H., & Samet, J. H. (2012). Emergency department use by primary care patients at a safety net hospital. Archives of Internal Medicine , 172 (3), 278 280. https://doi.org/10.1001/archinternmed.2011.709 Levy, B., Paulozzi, L., Mack, K. A., & Jones, C. M. (2015 ). Trends in Opioid Analgesic Prescribing Rates by Specialty, U.S., 2007 2012. American Journal of Preventive Medicine , 49 (3), 409 413. https://doi.org/10.1016/j.amepre.2015.02.020 Mackenzie, C. S., Gekoski, W. L., & Knox, V. J. (2006). Age, gender, and th e underutilization of mental health services: the influence of help seeking attitudes. Aging & Mental Health , 10 (6), 574 582. https://doi.org/10.1080/13607860600641200 ssues. Journal of Women & Aging , 19 (1 2), 1 12. https://doi.org/10.1300/J074v19n03_01 Manteuffel, M., Williams, S., Chen, W., Verbrugge, R. R., Pittman, D. G., & Steinkellner, A. (2014). Influence of patient sex and gender on medication use, adherence, and prescribing alignment with guidelines. , 23 (2), 112 119. https://doi.org/10.1089/jwh.2012.3972

PAGE 170

170 Manubay, J., Davidson, J., Vosburg, S., Jones, J., Comer, S., & Sullivan, M. (2015). Sex differences among opioid abusing chroni c pain patients in a clinical trial. Journal of Addiction Medicine , 9 (1), 46 52. https://doi.org/10.1097/ADM.0000000000000086 Marriam Webster. (2018). Definition of EPIDEMIC. Retrieved October 18, 2018, from https://www.merriam webster.com/dictionary/epide mic Marshall, S. W. (2007). Power for tests of interaction: effect of raising the Type I error rate. Epidemiologic Perspectives & Innovations , 4 , 4. https://doi.org/10.1186/1742 5573 4 4 Martins, S. S., & Ghandour, L. A. (2017). Nonmedical use of prescript ion drugs in adolescents and young adults: not just a Western phenomenon. World Psychiatry , 16 (1), 102 104. https://doi.org/10.1002/wps.20350 Rising Opioid Prescrib ing in Adult U.S. Emergency Department Visits: 2001 2010. Academic Emergency Medicine , 21 (3), 236 243. https://doi.org/10.1111/acem.12328 Mazer Amirshahi, M., Mullins, P. M., Rasooly, I., van den Anker, J., & Pines, J. M. (2014). Rising opioid prescribing in adult U.S. emergency department visits: 2001 2010. Academic Emergency Medicine: Official Journal of the Society for Academic Emergency Medicine , 21 (3), 236 243. https://doi.org/10.1111/acem.12328 McCabe, S. E., West, B. T., & Boyd, C. J. (2013a). Leftov er Prescription Opioids and Nonmedical Use Among High School Seniors: A Multi Cohort National Study. Journal of Adolescent Health , 52 (4), 480 485. https://doi.org/10.1016/j.jadohealth.2012.08.007 McCabe, S. E., West, B. T., & Boyd, C. J. (2013b). Medical u se, medical misuse, and nonmedical use of prescription opioids: results from a longitudinal study. Pain , 154 (5), 708 713. https://doi.org/10.1016/j.pain.2013.01.011 McCabe, S. E., West, B. T., & Boyd, C. J. (2013c). Motives for Medical Misuse of Prescripti on Opioids among Adolescents. Society , 14 (10), 1208 1216. https://doi.org/10.1016/j.jpain.2013.05.004 McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An Ecological Perspective on Hea lth Promotion Programs. Health Education Quarterly , 15 (4), 351 377. https://doi.org/10.1177/109019818801500401 Medicine, I. of M. (US) R. on E. B., Yong, P. L., Saunders, R. S., & Olsen, L. (2010). Missed Prevention Opportunities . Retrieved from https://ww w.ncbi.nlm.nih.gov/books/NBK53914/ Based Health Promotion: Promise, Performance, and Potential. American Journal of Public Health , 93 (4), 557 574.

PAGE 171

171 Milani, S. A., Crooke, H., Cottler, L. B., & Str iley, C. W. (2016). Sex differences in frequent ED use among those with multimorbid chronic diseases. The American Journal of Emergency Medicine , 34 (11), 2127 2131. https://doi.org/10.1016/j.ajem.2016.07.059 Mosalanejad, L., Abdolahifard, K., & Jahromi, M. G. (2013). Therapeutic vaccines: hope therapy and its effects on psychiatric symptoms among infertile women. Global Journal of Health Science , 6 (1), 192 200. https://doi.org/10.5539/gjhs.v6n1p192 National Institute on Alcohol Abuse and Alcoholism. (2018). Alcohol Facts and Statistics. Retrieved September 9, 2018, from https://www.niaaa.nih.gov/alcohol health/overview alcohol consumption/alcohol facts and statistics National Institute on Drug Abuse. (2015). Prescription Opioids and Heroin. Retrieved April 2 , 2017, from https://www.drugabuse.gov/publications/research reports/relationship between prescription drug abuse heroin use/introduction National Institute on Drug Abuse. (2016, February 8). Abuse of Prescription (Rx) Drugs Affects Young Adults Most. Retr ieved July 30, 2018, from https://www.drugabuse.gov/related topics/trends statistics/infographics/abuse prescription rx drugs affects young adults most National Institute on Drug Abuse. (2018a). Benzodiazepines and Opioids. Retrieved September 9, 2018, fro m https://www.drugabuse.gov/drugs abuse/opioids/benzodiazepines opioids National Institute on Drug Abuse. (2018b). Opioid Overdose Crisis. Retrieved September 23, 2018, from https://www.drugabuse.gov/drugs abuse/opioids/opioid overdose crisis National Inst itute on Drug Abuse. (2018c). Prescription Opioids and Heroin. Retrieved September 12, 2018, from https://www.drugabuse.gov/publications/research reports/relationship between prescription drug abuse heroin use/introduction National Institute on Drug Abuse. (2018d). What is the scope of prescription drug misuse? Retrieved September 12, 2018, from https://www.drugabuse.gov/publications/research reports/misuse prescription drugs/what scope prescription drug misuse National Institute on Drug Abuse. (2018e, Febr uary 28). Florida Opioid Summary. Retrieved October 8, 2018, from https://www.drugabuse.gov/drugs abuse/opioids/opioid summaries by state/florida opioid summary National Institute on Drug Abuse. (2019, January 22). Opioid Overdose Crisis. Retrieved January 29, 2019, from https://www.drugabuse.gov/drugs abuse/opioids/opioid overdose crisis Berry, J. G. (2014). Characteristics of Recurrent Utilization in Pediatric Emergency Departments. Pediatrics , 134 (4), e1025 e1031. https://doi.org/10.1542/peds.2014 1 362

PAGE 172

172 (2016). A Randomized Controlled Trial of a Citywide Emergency Department Care Coordination Program to Reduce Prescription Opioid Related Emergency Department Visit s. The Journal of Emergency Medicine , 51 (5), 498 507. https://doi.org/10.1016/j.jemermed.2016.06.057 Nordon, D. G., Akamine, K., Novo, N. F., & Hübner, C. von K. (2009). Characteristics of the use of benzodiazepines by women seeking treatment in primary ca re. Revista de Psiquiatria Do Rio Grande Do Sul , 31 (3), 152 158. https://doi.org/10.1590/S0101 81082009000300004 Okura, Y., Urban, L. H., Mahoney, D. W., Jacobsen, S. J., & Rodeheffer, R. J. (2004). Agreement between self report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failur e. Journal of Clinical Epidemiology , 57 (10), 1096 1103. https://doi.org/10.1016/j.jclinepi.2004.04.005 Olfson, M., King, M., & Schoenbaum, M. (2015). Benzodiazepine use in the United States. JAMA Psychiatry , 72 (2), 136 142. https://doi.org/10.1001/jamapsyc hiatry.2014.1763 income women to cancer screening recommendations. Journal of General Internal Medicine , 17 (2), 144 154. https://doi.org/10.1046/j.1525 1497.2002.10431.x Osborne, V ., Serdarevic, M., Crooke, H., Striley, C., & Cottler, L. B. (2017). Non medical opioid use in youth: Gender differences in risk factors and prevalence. Addictive Behaviors , 72 , 114 119. https://doi.org/10.1016/j.addbeh.2017.03.024 Osborne, V., Striley, C. W., Nixon, S. J., Winterstein, A. G., & Cottler, L. B. (2019). Sex differences in patterns of prescription opioid non medical use among 10 the US. Addictive Behaviors , 89 , 163 171. https://doi.org/10.1016/j.addbeh.2018.10.009 Palamar, J. J ., Shearston, J. A., Dawson, E. W., Mateu Gelabert, P., & Ompad, D. C. (2016). Nonmedical opioid use and heroin use in a nationally representative sample of us high school seniors. Drug and Alcohol Dependence , 158 , 132 138. https://doi.org/10.1016/j.drugal cdep.2015.11.005 Parsells Kelly, J., Cook, S. F., Kaufman, D. W., Anderson, T., Rosenberg, L., & Mitchell, A. A. (2008). Prevalence and characteristics of opioid use in the US adult population. Pain , 138 (3), 507 513. https://doi.org/10.1016/j.pain.2008.01. 027 Cooper, W. O. (2015). Prescription Opioid Epidemic and Infant Outcomes. Pediatrics , 135 (5), 842 850. https://doi.org/10.1542/peds.2014 3299 Patrick, S. W., Fry , C. E., Jones, T. F., & Buntin, M. B. (2016). Implementation Of Prescription Drug Monitoring Programs Associated With Reductions In Opioid Related Death Rates. Health Affairs , 35 (7), 1324 1332. https://doi.org/10.1377/hlthaff.2015.1496

PAGE 173

173 Paulozzi, L. J., Ma ck, K. A., Hockenberry, J. M., & Division of Unintentional Injury Prevention, National Center for Injury Prevention and Control, CDC. (2014). Vital signs: variation among States in prescribing of opioid pain relievers and benzodiazepines United States, 2 012. MMWR. Morbidity and Mortality Weekly Report , 63 (26), 563 568. Paulozzi, L. J., Weisler, R. H., & Patkar, A. A. (2011). A national epidemic of unintentional prescription opioid overdose deaths: how physicians can help control it. The Journal of Clinica l Psychiatry , 72 (5), 589 592. https://doi.org/10.4088/JCP.10com06560 emergency department visits: the effect of having a regular doctor. Medical Care , 36 (8), 12 49 1255. Pezalla, E. J., Rosen, D., Erensen, J. G., Haddox, J. D., & Mayne, T. J. (2017). Secular trends in opioid prescribing in the USA. Journal of Pain Research , 10 , 383 387. https://doi.org/10.2147/JPR.S129553 Phillips, D. M. (2000). JCAHO Pain Managem ent Standards Are Unveiled. JAMA , 284 (4), 428 429. https://doi.org/10.1001/jama.284.4.423b Piko, B. (2000). Perceived Social Support from Parents and Peers: Which Is the Stronger Predictor of Adolescent Substance Use? Substance Use & Misuse , 35 (4), 617 630 . https://doi.org/10.3109/10826080009147475 Pinkerton, R., & Hardy, J. R. (2017). Opioid addiction and misuse in adult and adolescent patients with cancer. Internal Medicine Journal , 47 (6), 632 636. https://doi.org/10.1111/imj.13449 Pinkhasov, R. M., Wong, J., Kashanian, J., Lee, M., Samadi, D. B., Pinkhasov, M. M., & Shabsigh, R. (2010). Are men shortchanged on health? Perspective on health care utilization and health risk behavior in men and women in the United States. International Journal of Clinical Pr actice , 64 (4), 475 487. https://doi.org/10.1111/j.1742 1241.2009.02290.x Prekupec, M. P., Mansky, P. A., & Baumann, M. H. (2017). Misuse of Novel Synthetic Opioids: A Deadly New Trend. Journal of Addiction Medicine , 11 (4), 256. https://doi.org/10.1097/ADM. 0000000000000324 Prescribing Data | Drug Overdose | CDC Injury Center. (2017, August 31). Retrieved March 24, 2018, from https://www.cdc.gov/drugoverdose/data/prescribing.html Rabinovitch, M., Cassidy, C., Schmitz, N., Joober, R., & Malla, A. (2013). The i nfluence of perceived social support on medication adherence in first episode psychosis. Canadian Journal of Psychiatry. Revue Canadienne De Psychiatrie , 58 (1), 59 65. https://doi.org/10.1177/070674371305800111 Reid, M. C., Eccleston, C., & Pillemer, K. (2 015). Management of chronic pain in older adults. The BMJ , 350 . https://doi.org/10.1136/bmj.h532

PAGE 174

174 Richard, L., Potvin, L., Kishchuk, N., Prlic, H., & Green, L. W. (1996). Assessment of the Integration of the Ecological Approach in Health Promotion Programs. American Journal of Health Promotion , 10 (4), 318 328. https://doi.org/10.4278/0890 1171 10.4.318 Roberts, D. O., Nickel, K. B., Mattfeldt Beman, M., Fearn, N., Jones, L., & Cottler, L. B. (2014). Drinking Severity and Its Association With Nutrition Risk A mong African American Women. Journal of Ethnicity in Substance Abuse , 13 (2), 93 103. https://doi.org/10.1080/15332640.2013.847395 Robinson, M. E., Riley, J. L., Myers, C. D., Papas, R. K., Wise, E. A., Waxenberg, L. B., & Fillingim, R. B. (2001). Gender ro le expectations of pain: relationship to sex differences in pain. The Journal of Pain: Official Journal of the American Pain Society , 2 (5), 251 257. https://doi.org/10.1054/jpai.2001.24551 Rodayan, A., Afana, S., Segura, P. A., Sultana, T., Metcalfe, C. D. , & Yargeau, V. (2016). Linking drugs of abuse in wastewater to contamination of surface and drinking water. Environmental Toxicology and Chemistry , 35 (4), 843 849. https://doi.org/10.1002/etc.3085 Rönkä, S., & Katainen, A. (2017). Non medical use of prescription drugs among illicit drug users: A case study on an online drug forum. International Journal of Drug Policy , 39 , 62 68. https://doi.org/10.1016/j.drugpo.2016.08.013 Rosenblatt, R. A., Wright, G. E., Baldwin, L. M., Chan, L., Clitherow, P., Chen, F. M., & Hart, L. G. (2000). The effect of the doctor patient relationship on emergency department use among the elderly. American Journal of Public Health , 90 (1), 97 102. Roz enbroek, K., & Rothstein, W. (2011). Medical and Nonmedical Users of Prescription Drugs Among College Students. Journal of American College Health , 59 (5), 358 363. https://doi.org/10.1080/07448481.2010.512044 RStudio 1.1.456. (2016). R: A language and envi ronment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R project.org/. RStudio . A. M. (2014). Increases in Heroin Overdose Deaths 28 States, 2010 to 2012 . 63 (39), 32. Ruhoy, I. S., & Daughton, C. G. (2008). Beyond the medicine cabinet: An analysis of where and why medications accumulate. Environment International , 34 (8), 1157 1169. https://doi.org/10.1016 /j.envint.2008.05.002 Ruktanonchai, C. W., Pindolia, D. K., Striley, C. W., Odedina, F. T., & Cottler, L. B. (2014). Utilizing spatial statistics to identify cancer hot spots: a surveillance strategy to inform community engaged outreach efforts. Internatio nal Journal of Health Geographics , 13 (1), 39. https://doi.org/10.1186/1476 072X 13 39

PAGE 175

175 Sampson, R. J., & Raudenbush, S. W. (2004). Seeing Disorder: Neighborhood Stigma and the Social Psychology Quarterly , 67 (4), 319 342. https://doi.org/10.1177/019027250406700401 SAS Institute Inc., 9.4, Cary, NC: SAS Institute Inc., 2011 . (n.d.). Saunders, K., Von Korff, M., Campbell, C. I., Banta Green, C. J., Sullivan, M. D., Merrill, J. O., & Weisner, C. (2012). Concurrent Use of Alcohol and Sedatives among Persons Prescribed Chronic Opioid Therapy: Prevalence and Risk Factors. The Journal of Pain , 13 (3), 266 275. https://doi.org/10.1016/j.jpain.2011.11.004 Schepis, T. S., & McCabe, S. E. (2016). Trends in Older Adult Nonmedical Pr escription Drug Use Prevalence: Results from the 2002 2003 and 2012 2013 National Survey on Drug Use and Health. Addictive Behaviors , 60 , 219 222. https://doi.org/10.1016/j.addbeh.2016.04.020 Schneiderhan, J., Clauw, D., & Schwenk, T. L. (2017). Primary Ca re of Patients With Chronic Pain. JAMA , 317 (23), 2367 2368. https://doi.org/10.1001/jama.2017.5787 M. (2007). Depression anxiety relationships with chronic phys ical conditions: Results from the World Mental Health surveys. Journal of Affective Disorders , 103 (1), 113 120. https://doi.org/10.1016/j.jad.2007.01.015 Serdarevic, M., & Cottler, L. (2017). Gender differences in patterns of prescription opioid use and bi nge drinking among middle aged Floridians. Drug & Alcohol Dependence , 171 , e187. https://doi.org/10.1016/j.drugalcdep.2016.08.512 Serdarevic, M., Gurka, K. K., Striley, C. W., Vaddiparti, K., & Cottler, L. B. (2018). Prevalence of Concurrent Prescription O pioid and Hazardous Alcohol Use Among Older Women: Results from a Cross Sectional Study of Community Members. Journal of Community Health . https://doi.org/10.1007/s10900 018 0569 y Serdarevic, M., Osborne, V., Striley, C. W., & Cottler, L. B. (2017). The a ssociation between insomnia and prescription opioid use: results from a community sample in Northeast Florida. Sleep Health , 3 (5), 368 372. https://doi.org/10.1016/j.sleh.2017.07.007 Serdarevic, M., Striley, C. W., & Cottler, L. B. (2017). Sex differences in prescription opioid use. Current Opinion in Psychiatry , 30 (4), 238 246. https://doi.org/10.1097/YCO.0000000000000337 Shah, A., Hayes, C. J., & Martin, B. C. (2017). Factors Influencing Long Term Opioid Use Among Opioid Naive Patients: An Examination of Initial Prescription Characteristics and Pain Etiologies. The Journal of Pain , 18 (11), 1374 1383. https://doi.org/10.1016/j.jpain.2017.06.010

PAGE 176

176 Simoni Wastila, L. (2000). The use of abusable prescription drugs: the role of gender. Journal of Gender Based Medicine , 9 (3), 289 297. https://doi.org/10.1089/152460900318470 R. L. (2017). Opioid Analgesic and Benzodiazepine Prescribing Among Medicaid Enrol lees with Opioid Use Disorders: The Influence of Provider Communities. Journal of Addictive Diseases , 36 (1), 14 22. https://doi.org/10.1080/10550887.2016.1211784 Steinmetz, C. N., Zheng, C., Okunseri, E., Szabo, A., & Okunseri, C. (2017). Opioid Analgesic Prescribing Practices of Dental Professionals in the United States. JDR Clinical and Translational Research , 2 (3), 241 248. https://doi.org/10.1177/2380084417693826 Stokols, D. (1996). Translating Social Ecological Theory into Guidelines for Community Heal th Promotion. American Journal of Health Promotion , 10 (4), 282 298. https://doi.org/10.4278/0890 1171 10.4.282 Strand, M. A., Eukel, H., & Burck, S. (2018). Moving opioid misuse prevention upstream: A pilot study of community pharmacists screening for opio id misuse risk. Research in Social and Administrative Pharmacy . https://doi.org/10.1016/j.sapharm.2018.07.011 Substance Abuse and Mental Health Services Administration. (2016a). Overview of Factors and Strategies . 21. Substance Abuse and Mental Health Serv ices Administration. (2016b). Preventing Prescription Drug Misuse: Overview of Factors and Strategies . Retrieved from https://www.samhsa.gov/capt/sites/default/files/resources/preventing prescription drug misuse overview.pdf Suh, S., Cho, N., & Zhang, J. ( 2018). Sex Differences in Insomnia: from Epidemiology and Etiology to Intervention. Current Psychiatry Reports , 20 (9), 69. https://doi.org/10.1007/s11920 018 0940 9 Sullivan, M. D., Edlund, M. J., Zhang, L., Unützer, J., & Wells, K. B. (2006). Association Between Mental Health Disorders, Problem Drug Use, and Regular Prescription Opioid Use. Archives of Internal Medicine , 166 (19), 2087 2093. https://doi.org/10.1001/archinte.166.19.2087 Sun, E. C., Dixit, A., Humphreys, K., Darnall, B. D., Baker, L. C., & Mackey, S. (2017). Association between concurrent use of prescription opioids and benzodiazepines and overdose: retrospective analysis. BMJ , 356 , j760. https://doi.org/10.1136/bmj.j760 S (2014). Reductions in prescription opioid diversion following recent legislative interventions in Florida. Pharmacoepidemiology and Drug Safety , 23 (3), 314 320. h ttps://doi.org/10.1002/pds.3553

PAGE 177

177 Thompson, A. E., Anisimowicz, Y., Miedema, B., Hogg, W., Wodchis, W. P., & Aubrey Bassler, K. (2016). The influence of gender and other patient characteristics on health care seeking behaviour: a QUALICOPC study. BMC Family Practice , 17 . https://doi.org/10.1186/s12875 016 0440 0 Trends in Opioid Prescribing by Race/Ethnicity for Patients Seeking Care in US Emergency Departments | Clinical Pharmacy and Pharmacology | JAMA | JAMA Network. (n.d.). Retrieved March 17, 2018, from https://jamanetwork.com/journals/jama/fullarticle/1149438 Trzeciak, S., & Rivers, E. P. (2003). Emergency department overcrowding in the United States: an emerging threat to patient safety and public health. Emergency Medicine Journal: EMJ , 20 (5), 402 405. Unick, G. J., Rosenblum, D., Mars, S., & Ciccarone, D. (2013). Intertwined Epidemics: National Demographic Trends in Hospitalizations for Heroin and Opioid Related Overdoses, 1993 2009. PLOS ONE , 8 (2), e54496. https://doi.org/10.1371/journal.pone.0054496 Urbanoski, K., Cheng, J., Rehm, J., & Kurdyak, P. (2018). Frequent use of emergency departments for mental and substance use disorders. Emerg Med J , 35 (4), 220 225. https://doi.org/10.1136/emermed 2015 205554 US Census Bureau. (2018). Urban and Rural. Ret rieved October 17, 2018, from https://www.census.gov/geo/reference/urban rural.html US Health and Human Services. (2017). Final Report: Opioid Use, Misuse, and Overdose in Women . 87. Vinton, D. T., Capp, R., Rooks, S. P., Abbott, J. T., & Ginde, A. A. (201 4). Frequent users of US emergency departments: characteristics and opportunities for intervention. Emergency Medicine Journal: EMJ , 31 (7), 526 532. https://doi.org/10.1136/emermed 2013 202407 Wang, X., Zhu, Y., Dave, C. V., Alrwisan, A. A., Voils, S. A., & Winterstein, A. G. (2017). Trends of Neonatal Abstinence Syndrome Epidemic and Maternal Risk Factors in Florida. Pharmacotherapy , 37 (7), 806 813. https://doi.org/10.1002/phar.1947 Webb, F. J., Striley, C. W., & Cottler, L. B. (2015). Marijuana Use and It s Association with Participation, Navigation, and Enrollment in Health Research among African Americans. Journal of Ethnicity in Substance Abuse , 14 (4), 325 339. https://doi.org/10.1080/15332640.2014.986355 Weber, E. J., Showstack, J. A., Hunt, K. A., Colb y, D. C., & Callaham, M. L. (2005). Does lack of a usual source of care or health insurance increase the likelihood of an emergency department visit? Results of a national population based study. Annals of Emergency Medicine , 45 (1), 4 12. https://doi.org/1 0.1016/j.annemergmed.2004.06.023

PAGE 178

178 White, A. G., Birnbaum, H. G., Schiller, M., Tang, J., & Katz, N. P. (2009). Analytic models to identify patients at risk for prescription opioid abuse. The American Journal of Managed Care , 15 (12), 897 906. White, A. M., H ingson, R. W., Pan, I., & yi, H. (2011). Hospitalizations for Alcohol and Drug Overdoses in Young Adults Ages 18 24 in the United States, 1999 2008: Results From the Nationwide Inpatient Sample. Journal of Studies on Alcohol and Drugs , 72 (5), 774 786. Whit ley, H. P., & Lindsey, W. (2009). Sex Based Differences in Drug Activity. American Family Physician , 80 (11), 1254 1258. Williams, R. E., Sampson, T. J., Kalilani, L., Wurzelmann, J. I., & Janning, S. W. (2008). Epidemiology of opioid pharmacy claims in the United States. Journal of Opioid Management , 4 (3), 145 152. Wilson, J., & Kelling, G. (1982). Broken windows. Atlantic Monthly , 249 (3), 29 38. (2018). Mandatory review of a prescription drug monitoring program and impact on opioid and benzodiazepine dispensing. Drug and Alcohol Dependence , 188 , 169 174. https://doi.org/10.1016/j.drugalcdep.2018.03.036 Young, A., McCabe, S. E., Cranford, J. A., Ross Durow, P., & Bo yd, C. J. (2012). Nonmedical Use of Prescription Opioids among Adolescents: Subtypes Based on Motivation for Use. Journal of Addictive Diseases , 31 (4), 332 341. https://doi.org/10.1080/10550887.2012.735564 Young, H. W., Tyndall, J. A., & Cottler, L. B. (20 17). The current utilization and perceptions of prescription drug monitoring programs among emergency medicine providers in Florida. International Journal of Emergency Medicine , 10 . https://doi.org/10.1186/s12245 017 0140 0

PAGE 179

179 BIOGRAPHICAL SKETCH Mirsada Serdarevic received her Bachelor of Arts in psychology from the University of Vermont in 2013. Upon graduating she joined the Health Behavior Research Center in Burlington, Vermont as a Researc h Assistant and later progressed to Senior Research Assistant. In 2015 Mirsada began her doctoral studies at the Department of Epidemiology, College of Public Health and Health Professions and College of Medicine at the University of Florida (UF) as a NIDA T32 pre doctoral fell ow. Under the mentorship and chair of her committee, Dr. Linda B. Cottler, Mirsa da received her Ph.D. in epidemiology from UF in the spring of 2019 . , mental health, and health disparities. Her dissertati on specifically focused on prescription opioid use and risk factors for us e in the community. During her doctoral training at UF , Mirsada worked on various research projects and served as a teaching assistant from 2016 2018 in the department of epidemiolog y . She traveled to Haiti to c ollect data as part of the Haiti Health Study in 2015. In 2016 she joined the Adolescent Brain and Cognitive Development (ABCD) study as a Research Assistant at UF. This landmark study examines the health of youth over a ten ye ar period across different sites in the nation. She later also joined the Medication Use, Safety and Evidence (MUSE) study in 2016 as a Research Assistant to recruit and interview prescription opioid users in the study. Mirsada has been a member of the International Society for Pharmacoepidemiology (UF student chapter) since 2016 and the Social Media Chair of the College of Public Health and Health Professions Doctoral Student Council at UF since 2017. To date s he has published 9 papers in peer reviewed journals and 2 opinion editorial s in the Gainesville Sun , and has presented 12 posters and 2 oral presentations at national and international conferences. In addition, Mirsada has received numerous awards and travel grants including the National Center for Responsible Gaming (NCRG) Scholarship (2016), Center for Addiction Research &

PAGE 180

180 Education (CARE) Travel Award (2016 2017), the Outstanding Achievement and Excellence in Ph armacoepidemiology Award (2017), the University of Florida Graduate Student Travel Gra nt (2016 2018), and the holarship (2016 2018). After graduating from the University of Florida, Mirsada will begin her position at the Center for Health Outcomes research at JPS Health Network in Fort Worth, Texas.