Citation
Long-Term Survival and the Use of Health Care Services among Veterans with AIDS in the Era of Highly Active Antiretroviral Therapy

Material Information

Title:
Long-Term Survival and the Use of Health Care Services among Veterans with AIDS in the Era of Highly Active Antiretroviral Therapy
Creator:
MKANTA, WILLIAM NEHEMIAH ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
AIDS ( jstor )
Comorbidity ( jstor )
Diseases ( jstor )
Health care industry ( jstor )
Health care services ( jstor )
Highly active antiretroviral therapy ( jstor )
HIV ( jstor )
Information use ( jstor )
Predisposing factors ( jstor )
Veterans ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright William Nehemiah Mkanta. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
12/31/2006

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

LONG-TERM SURVIVAL AND THE US E OF HEALTH CARE SERVICES AMONG VETERANS WITH AIDS IN THE ERA OF HIGHLY ACTIVE ANTIRETROVIRAL THERAPY By WILLIAM NEHEMIAH MKANTA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by William Nehemiah Mkanta

PAGE 3

This dissertation is dedicated to my parent s, Nehemiah and Agnes, who did not live long enough to celebrate my achievement, but w hose love and caring made it possible.

PAGE 4

iv ACKNOWLEDGMENTS This dissertation is the p eak of a lifelong process of learning, understanding, and growing that took me across continents, and th at would have been difficult to achieve without the guidance, assistan ce, and support of several individuals and organizations. My primary school teacher, Mr. Kabengula, a nd my undergraduate professor, Dr. Mbago, believed in my language and mathematics skill s and abilities and set high standards of excellence for my performance. The standard continued with my major professor and dissertation chair, Dr. R. Paul Duncan, who also challenged me to work independently and reach goals beyond my expectations. I express sincere gratitude and appreciat ion to Dr. Duncan for his assistance, generosity, and advice throughout my doctoral work at the University of Florida. Throughout my doctoral work, he encouraged me to develop independent thinking and research skills. He continually stimulated my analytical th inking and greatly assisted me with American and scientific writing. I am also very grateful for having an exceptional doctoral committee and wish to thank Dr. Constance Uphold for her expertise in HIV/AIDS research, Dr. Christy Lemak for her expertise in disease management a nd willingness to read and make constructive suggestions on this dissertation, Dr. Neale Chumbler for his a ssistance in the concepts of access and utilization and content of the Depart ment of Veterans Affairs databases, and Dr. Nabih Asal for his expertise in epidem iology, especially in methods on infectious

PAGE 5

v disease. Each of them, in his/her own way, helped make this a valuable and enjoyable learning experience. I wish to thank the U.S. Department of Veterans Affairs, through the Public Health Strategy Health Care Group, for allowing me to use its data for this dissertation. Special recognition goes to Ms. Jude Lopez and Mr. Larry Mole of the VA Palo Alto Health Care System for their specific gui dance on the content and the use of the Immunology Case Registry database. I owe a special note of gr atitude to the Sherri Aversa Memorial Foundation for awarding me the dissertation completion gr ant for the year 2005. I thank Dr. Lee Crandall, president of the foundation, and his board members for their efforts to encourage doctoral research in behavioral and social scienc e research on health, part of the late Dr. Sherri Aver sa’s research interest. I extend many thanks to the administrativ e staff in the Department of Health Services Research, Management and Policy. In this regard, I wish to recognize the efforts of Mrs. Juanita Cooper and Mrs. Diane J ohns in making my four-year stay in the department manageable in academic and financial matters. More than a few times, as things looked too much for me or too late to accomplish, they were able to pitch in and restore everything to normal. My doctoral program was partially f unded by the link project between the University of Dar es Salaam, Tanzania, and the University of Flor ida. I thank Dr. Todd Leedy, assistant director, Center for African Studies at the University of Florida, for doing an excellent job with managing the link affa irs. I am very gratef ul for the support I received from the Department of Statistics and University of Dar es Salaam as whole

PAGE 6

vi through the Links and Projects O ffice in facilitating my studies . Special thanks go to Mrs. Mshigeni and Mrs. Kaaya, who were in char ge of the links office at different times during my studies. Finally, I would like to thank God and my family. No one deserves more thanks than my lovely wife, Susan, whose love, support, encouragement, and belief in me finally made this Ph.D. possible. In many ways she and our children, Nehemiah and Faith, made me realize my potential for strength and resili ence. I thank God for his blessings in my family and the spirit of understanding a nd supporting each other in times of need.

PAGE 7

vii TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...............................................................................................................x LIST OF FIGURES...........................................................................................................xi ABSTRACT......................................................................................................................x ii CHAPTER 1 INTRODUCTION AND THE PROBLEM..................................................................1 1.1 Long-term Survival with AIDS.........................................................................3 1.2 Significance of the Study...................................................................................9 1.2.1 Changing Demographics of the Patients..................................................9 1.2.1.1 Age and aging................................................................................10 1.2.1.2 Race...............................................................................................11 1.2.2 Comorbidity Profile...............................................................................12 1.2.3 Disparities in Service Utilization...........................................................16 1.3 Chapter Summary............................................................................................18 2 THE IMMUNOLOGY CASE REGISTRY................................................................19 2.1 Establishment and the Purpose of the ICR......................................................19 2.2 Data Input and Extraction from the ICR..........................................................20 2.3 Sample Data and Patient Factors from the ICR for the Fiscal Year 2003.......21 2.4 Strengths and Weaknesses of the ICR.............................................................23 3 LITERATURE REVIEW...........................................................................................25 3.1 Theoretical Model............................................................................................25 3.1.1 Andersen Behavioral Model of Health Service Utilization...................25 3.2 Review of HIV/AIDS-Related Literature........................................................31 3.3 Use of HAART................................................................................................32 3.3.1 Predisposing Variables a nd the Use of HAART....................................33 3.3.2 Enabling Variables and the Use of HAART..........................................35 3.3.3 Conclusion.............................................................................................35 3.4 Effects of HAART on Se rvice Utilization.......................................................36

PAGE 8

viii 3.4.1 Effects of HAART on Service Util ization within the VA System........36 3.4.2 Effects of HAART on Service U tilization in Other Healthcare Settings.............................................................................................................37 3.4.3 The Role of Predisposing Variables in Service Utilization...................39 3.4.3.1 Age and service utilization...............................................................39 3.4.3.2 HIV risk factor and service uti lization..........................................41 3.4.3.3 Gender and service utilization.......................................................43 3.4.3.4 Race/ethnicity and service utilization...........................................44 3.4.3.5 Race/ethnicity and servic e use in the VA population....................44 3.4.3.6 Race/ethnicity and service us e in other population studies...........45 3.4.3.7 Summary on the predisposing variables........................................46 3.4.4 The Role of Enabling Variab les on Service Utilization.........................47 3.4.4.1 Insurance coverage, re gion of residence and service util ization...47 3.4.4.2 Summary on the enabling variables..............................................49 3.4.5 The Role of Need-related Vari ables on Service Utilization..................49 3.4.5.1 AIDS-defining illnesses and other comorbidities.........................49 3.4.5.2 The role of non-AIDS-defining comorbidity................................51 3.4.5.3 Influence of immune status and viral status on service utilization..................................................................................................53 3.4.5.4 Summary on need variables..........................................................55 4 RESEARCH OBJECTIVES AND METHODS.........................................................56 4.1 Introduction......................................................................................................56 4.2 Objectives........................................................................................................57 4.3 Research Questions and Hypotheses...............................................................57 4.3.1 Research Question I and Hypotheses.....................................................58 4.3.1.1 Research question I.......................................................................58 4.3.1.2 Hypotheses....................................................................................58 4.3.2 Research Question II and Hypotheses...................................................58 4.3.2.1 Research question II......................................................................58 4.3.2.2 Hypotheses....................................................................................58 4.4 Data and Methods............................................................................................59 4.4.1 Source of Data and Availability.............................................................60 4.4.2 Study Design..........................................................................................60 4.4.3 Subjects..................................................................................................61 4.5 Variables in the Study......................................................................................61 4.5.1 Dependent Variables..............................................................................61 4.5.2 Independent Variables...........................................................................62 4.6 Methods of Analysis........................................................................................63 4.6.1 Stage One: Methods for the Descript ion of the Characteristics of the Sample of Veterans with AIDS who participated in the Study.......................64 4.6.2 Stage Two: Methods for Examining the Correlates of Long-Term Survival After AIDS Diagnosis.......................................................................64 4.6.3 Stage Three: Methods for Examining the Impact of Long-Term Survival After AIDS Diagnosis on Th e Use of Healthcare Services Among Veterans Using HAART..................................................................................65

PAGE 9

ix 4.6.4 Use of the Negative Binomial M odel Based on the Nature of the Dependent Variables........................................................................................66 4.6.5 Use of the Negative Binomial Mode l to Address the Purpose of the Study ................................................................................................................69 4.6.6 Interpretation of the Regression Coefficients........................................69 4.6.7 Assessment of the Regression Models...................................................70 4.6.8 Data Processing Prior to the Fi tting of the Regression Models.............71 4.6.8.1 Multicollinearity............................................................................71 5 RESULTS OF THE STUDY......................................................................................74 5.1 Descriptive Statistics........................................................................................75 5.2 Correlates of Long-term Survival after AIDS Diagnosis................................81 5.3 Results of the Multivariable Nega tive Binomial Regression Model...............84 5.4 Results of the Negative Binomial M odel for All-Diagnosis Utilization..........84 5.5 Results of the Negative Binomial Re gression for Psychiatric Utilization.......88 5.6 Results of the Negative Binomial Regression for Medical Utilization............92 5.7 Chapter Summary............................................................................................94 6 STUDY CONCLUSION............................................................................................96 6.1 Overview of the study......................................................................................96 6.2 Discussion of the Findings and their Implications...........................................98 6.2.1 Long-term Survivorship.........................................................................98 6.2.2 Findings on Other Need Variables.......................................................102 6.2.3 Findings on the Predisposing Variables...............................................104 6.3 Limitations of the Study.................................................................................109 6.4 Suggestions for Future Research...................................................................110 6.5 Conclusion.....................................................................................................112 APPENDIX A AIDS CASE DEFINITION OF TH E CENTERS FOR DISEASE CONTROL AND PREVENTION (CDC)....................................................................................114 B VA PRIORITY GROUPS........................................................................................115 LIST OF REFERENCES.................................................................................................116 BIOGRAPHICAL SKETCH...........................................................................................129

PAGE 10

x LIST OF TABLES Table page 2.1 Sources of Data Extracts and Number of Variables in the National Immunology Case Registry............................................................................................................21 2.2 Selected Patient Characteristics a nd Outcomes from the Immunology Case Registry for the fiscal year 2003..............................................................................22 4.1 Independent Variables Including Operati onal Definitions Arranged According to the Andersen Behavioral Mode l of Health Services Use.........................................62 4.2 Correlation Coefficients and their Leve ls of Significance among the Independent Variables...................................................................................................................72 5.1 Characteristics of Veterans with AI DS by All-Diagnosis Utilization (N=4,717)....76 5.2 Logistic Regression Results of the Corre lates of Long-term Survival after AIDS Diagnosis (N=4716).................................................................................................82 5.3 Negative Binomial Regression Results fo r All-Diagnosis Outpatient Visits, AllDiagnosis Inpatient Visits, and All-Dia gnosis Length of Stay among Veterans with AIDS................................................................................................................86 5.4 Negative Binomial Regression Results for Psychiatric Outpatient Visits, Psychiatric Inpatient Visits, and Psychi atric Length of Stay among Veterans with AIDS................................................................................................................90 5.5 Negative Binomial Regression Results fo r Medical Outpatient Visits, Medical Inpatient Visits, and Medical Length of Stay among Veterans with AIDS Who Used Healthcare Services in the Veterans Health Administration...........................93

PAGE 11

xi LIST OF FIGURES Figure page 1.1 Comorbidities and Their Likely S ources among Patients with Long-Term Survival after AIDS Diagnosis.................................................................................13 3.1 The Andersen Behavioral Model: Utili zation of Health Service among AIDS Patients Experiencing Long-Term Survival.............................................................30 4.1 Histogram of Outpatient Visits among Patients with Known Duration of Survival after AIDS..................................................................................................68 4.2 Histograms of (a) Inpatient Visits and (b) LOS among Patients with Known Duration of Survival after AIDS Diagnosis.............................................................68

PAGE 12

xii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy LONG-TERM SURVIVAL AND THE USE OF HEALTH CARE SERVICES AMONG VETERANS WITH AIDS IN THE ERA OF HIGHLY ACTIVE ANTIRETROVIRAL THERAPY By William Nehemiah Mkanta December 2005 Chair: R. Paul Duncan Major Department: Health Services Research, Management and Policy This dissertation examines th e impact of long-term surv ival after AIDS diagnosis on the rates of outpatient and inpatient vi sits and length of ho spital stay. A crosssectional, retrospective design was used. Th e data were obtained from the Veterans Health Administration database, the Immunol ogy Case Registry, and it consisted of 4,717 veterans who have ever been diagnosed with AIDS and have a history of using highly active antiretroviral therapy (HAART). A nega tive binomial regression model was used to assess the relationship between long-term survival and service ut ilization based on the Andersen Behavioral Model of health services use. Long-term survival was conceptualized as a need variable. Predispos ing variables were race, HIV risk factor, military period of service, and age at AIDS diagnosis. PatientÂ’s region of residence was the enabling variable, while presence of AIDS -defining illness, pr esence of psychiatric comorbidity, number of medical comorbidit ies, death, and CD4 cell count were need

PAGE 13

xiii variables. We hypothesized that long-term survival after AIDS diagnosis would be associated with greater use of the services. We found that long-term surviv al is associated with the use of healthcare services. For all-diagnosis utilization, long-term survival was associ ated with greater use of outpatient and inpatient servic es. Specifically, long-term survival was associated with higher rates of inpatient psychi atric visits, outpatient medical visits, and inpatient medical visits. Being black, having IDU risk factor, older age at AIDS diagnosis, having AIDSdefining illness, and death were associated with higher rates of both outpatient and inpatient visits. However, we found that bl acks and whites differed only in outpatient visits. Blacks had 51% more outpatient psychiatric visits and 14% more outpatient medical visits compared to whites (both with p<.05). Deceased patients had high prevalence of AIDS-defining illnesses and greater use of inpatient care. Because long-term survival is associated with the use of healthcare services, HIV and non-HIV factors leading to greater use of the services among patients with long-term survival need to be delineated. Research has to establish how race, age, and comorbidity influence healthcare needs of the patients who attain long-term su rvival. Evaluation of patient satisfaction with care is needed to identify areas of service improvement.

PAGE 14

1 CHAPTER ONE INTRODUCTION AND THE PROBLEM Since the advent of highly active antiretroviral therapy (HAART) in 1996, the rate of death from AIDS has declined signifi cantly, and more people with AIDS diagnoses experience long-term survival. As a result, mo st AIDS patients now grow older and live well into their advanced age. For instance, Manfredi and Calza (2004) reported that before HAART was introduced, the time between HIV diagnosis and death was estimated to range from 8 to 13 years, and, in general, persons infected with HIV had reduced life expectancy due to high HIV-related death rates. But now, HIV patients live longer regardless of their age at infec tion and have a life expectancy comparable to that of the general population. Correspondingly, the numbe r of middle-aged and older individuals with AIDS has increased (Gebo & Moore, 2004; Tumbarello et al., 2004). Combined with the continued emergence of new HI V diagnoses in the nation (CDC, 2004), the number of AIDS patients will continue to increase because persons with AIDS have the potential to survive longer and contribute to the steady increase of their number. For example, in an investigation of the changes in AIDS incidence, prevalence, and deaths among persons with AIDS, “Update: AIDS – Un ited States” (2002) reported that there was a steady increase in the prevalence of persons living with AIDS during January 1996 and December 2000, that is, in a period that marked the introduction of HAART. Several reports have indicated that veterans with HIV/AI DS have also experienced increased survival rates af ter HAART initiation (Backus et al., 2005; CQM, 2001; Mole et al., 1999). Furthermore, McGinnis et al. (2003) found that although there was no

PAGE 15

2 significant racial difference in the use of HAART, minority veterans with HIV experienced poorer survival ra tes than white veterans. As a result of longer lives, patients now experience HIV/AIDS as a chronic condition and they are more likely to face co mpeting health risks that can potentially generate new healthcare needs. For instance, studies have reported th at comorbidities not related to HIV, for example, liver diseas es, are occurring with increasing frequency among the patients (Brau, 2003; Law et al., 200 4). Further, increasing prevalence of antiretroviral drug resistance (S cott et al., 2004) as well as in creasing levels of toxicity (Piliero, 2003) has been reported among AIDS patients experiencing longer lives. Despite growing evidence that most HI V/AIDS patients experience long-term survival and the consequent health risks, re search has not yet made clear the impact longterm survival has on the use of health services among AIDS patients. Many questions arise as increasing numbers of AIDS patients attain long-term survival. As patients live longer, their care presents new challenges to the health care deliver y: (i) persons with HIV/AIDS live longer and thus require care a nd services for longer pe riods of time; (ii) racial minorities are more likely to have less desirable patterns of health service utilization compared to whites; it is likely that minorities would have a more severe course of disease than whites when they experience longer lives with HIV/AIDS. This course of disease may create new health ca re needs for minorities in particular and patients with long-term survival in general; (iii) with longer surviv al, AIDS patients now are increasingly at the risk of comorbidity not related to HIV; and (iv) increasing numbers of individuals receive HIV diagnosis at an older ag e but survive and live longer, while others receive HIV diagnosis at younger age but survive into older age. For these

PAGE 16

3 reasons, there are increased numbers of olde r patients living with HIV/AIDS now than ever before. Besides HIV/AIDS, these patien ts are likely to confront age-related conditions such as diabetes and hypertension to create greater need of health services. Because these factors may account for increased use of services, we expect that patients experiencing long-term survival after AIDS di agnosis would have greater rates of service utilization. The primary objective of this dissertation was to examine the impact of the length of survival time after an AIDS diagnosis on outpatient visits, inpati ent visits, and length of inpatient stay (LOS) among veterans w ho use HAART. Because long-term survival after AIDS diagnosis can predispose patients to various health risk s (e.g., emergence of new non-HIV conditions), a new set of questions particularly salient for health services research needs to be addressed. For example, we will need to determine how and to what extent the emergence of non AIDS defining illnesses as patien ts live longer has affected both the amount and type of healthcare serv ices consumed. A thor ough understanding of the sociodemographic factors such as age, r ace/ethnicity, and region of residence as well as clinical factors such as comorbidity and CD 4 cell count associated with health services utilization is therefore important for realizing healthcare needs that might be generated as a result of prolonged lives of the patients. Such analysis will also be important in determining the degree of disparities in service utilization among key subgroups of patients. It would also serve to identify areas where certain services may be targeted. 1.1 Long-term Survival with AIDS Use of HAART has emerged as a strong pr edictor of increased survival rates among AIDS patients (Pezzotti et al., 1999; Hoffmann et al., 2001; Skiest & Crosby,

PAGE 17

4 2003). However, the improvements in survival rates have not been uniform across all patient groups. Differential access to medi cal care between racial groups may explain differences in AIDS survival times. Also, person to person variation in both AIDSdefining events at diagnosis and HAART effectiv eness might be part of the differences in survival times among AIDS patients. Additi onally, there is often a delay between the time of diagnosis of AIDS and the time at which AIDS is reported. This delay may differ among different groups of pa tients and thus further complicate measurement and conceptualization of survival lengths. In this dissertation we examined the role of survival time after AIDS diagnosis on service utilization among veterans who had ev er been diagnosed with AIDS and used HAART. Only patients with the use of HAART were included in th is study because of the overwhelming evidence of its associati on with improved survival rates. Most investigators have measured the rates of survival among HIV/AIDS patients by taking into account their use of HAART and the type s of AIDS-defining illnesses (ADIs) they had. Some of the studies that have repor ted survival rates among HIV/AIDS patients using this approach are reviewed. This review is presented to dem onstrate the association between the use of HAART and survival rates. Following the advent of HAART in 1996, there was a remarkable improvement in survival for the most common ADIs (e.g., Pneumocystis carinii pneumonia [PCP], pulmonary tuberculosis, and centr al nervous system [CNS] dis eases). Patients with these illnesses exhibited a significant decrease in the risk of death (Fordyce et al., 2002). In addition, among the ADIs that were characte rized by short survival, for example, progressive multifocal leukoencephalopathy (PML), a three-fold increase in survival rates

PAGE 18

5 was reported (Conti et al., 2000). Although part of the improvement in survival might have been attributed to sp ecific ADI prophylaxis (e.g., PCP pr ophylaxis), proliferation of studies with evidence of prolonged survival among patients who had different types of ADIs and who responded positively to HAART has been a strong indicator of the potency of HAART in the suppression of HIV viral loads. Some of the studies have addressed survival using specific ADIs su ch as AIDS-related malignancies, AIDS dementia complex, etc., and were able to estimate and compare survival times under different HAART status. A review of a selected sample of these studies is given to exhibit long-term survival among AIDS patient s and to justify investigation of health services utilization based on the lengths of their survival times. Conti et al. (2000) investigated survival rates after the diagno sis of AIDS in the period when HAART was introduced for the firs t time in Italy. This investigation used data from 35,318 AIDS cases (age > 12 year s) diagnosed from January 1990 to August 1998. Conti and her colleagues reported that among AIDS patients diagnosed after HAART, the proportion of surv ivors 24 months after diagnosis was more than double and statistically significantly highe r compared to that of patie nts diagnosed before HAART (66% vs. 31%). In addition, 24-month surviv al rates were doubled among patients with the most common ADIs (e.g., PCP and KaposiÂ’s sa rcoma [KS]) as well as in those with ADIs that were characterized by short su rvival before HAART (e.g., PML and brain toxoplasmosis). The investigator s also showed that for AIDS diagnoses that occurred in the last period of the study (1997 1998), the reduction in the ri sk of death occurred for almost all the ADIs, ranging from 55% to 80% compared to the risk before HAART.

PAGE 19

6 However, during this study, primary lymphoma of the brain remained as the ADI with the worst outcome. In another study, Hoffmann et al. (2001) ex amined the survival of AIDS patients with primary central nervous system lympho ma (PCNSL) during the HAART era. In this multicentric study, 29 AIDS patients with pr oven PCNSL were followed over time. Six patients were on HAART, while 12 were given cranial radiation. This study found that survival time of patients receiving radiat ion plus HAART or ra diation alone differed significantly from those receiv ing neither of the two treatments. Median survival times were 1093, 132, and 33 days respectively for the three treatment options. Moreover, among patients on HAART alone, four patien ts (67%) showed a marked immune recovery and survived for more than 1.5 years, and two of them were still alive at the end of the study. Fordyce et al. (2002) used the experience of AIDS patients in New York City to study survival rates in the HAART era. It was found that, ov erall, cumulative survival at 24 months increased from 43% am ong patients diagnosed during 1990 1995 to 76% for those diagnosed during 1996 1998. This study involved 79,878 AIDS patients (age 13 years), and it was reported as the largest si ngle study in the U.S. at a population level by 2002 that evaluated effects of HAART on surviv al. The study also re ported that 56% to 70% twenty-four month survival rates after 1995 were observed among the most common ADIs including PCP, pulmonary tube rculosis, HIV wasting syndrome, and CNS diseases. However, similar to other findings, for example, Conti et al. (2000), primary lymphoma of the brain had only moderate a nd non-significant improvement in survival rates.

PAGE 20

7 A study that examined survival among AIDS patients diagnosed with PCNSL over a seven-year period demonstrated that recei pt of HAART resulted in marked survival benefits (Skiest & Crosby, 2003). Among the 25 patients that were identified with PCNSL during the study period, longer survival was noted for patients who received HAART compared to others who received radiation or none of the two treatments. Further, none of the patients who did not re ceive HAART were alive at 1 year, while six of the patients who received HAART remained alive with follow-up times of 229, 620, 667, 1,424, 2,005, and 2,112 days respectively. In a nother study using patients with AIDS dementia complex (ADC) in Australia, Dore et al. (2003) showed th at survival following AIDS diagnosis significantly increased fr om 19.6 months for those diagnosed in 1993 1995 to 39.6 months for those diagnosed in 1996 2000. The median survival following ADC increased to a greater extent than for all other ADIs, from 11.9 to 48.2 months between the two time periods. In a recent study, use of HAART has been associated with survival improvement among AIDS patients with cancer diagnoses (Bi ggar et al., 2005). Usi ng data from adults with AIDS from New York City (9,015 AI DS-related and 929 non-AIDS-related cancers) the study showed that 24-month survival ra te improved significantly for most cancer types. Specifically for AIDS -related cancers, the 24-month su rvival rate was 58% for KS, 41% for peripheral Non HodgkinÂ’s Lymphoma (NHL), 29% for central nervous system NHL, and 64% for cervical cancer for AIDS diagnoses made between 1996 and 2000. Among non-AIDS-related cancers, survival wa s lowest for lung cancer (10%) but was over 50% for other types of cancer during the same period.

PAGE 21

8 This review has provided evidence of improved survival rates among AIDS patients with rare ADIs (e.g., PCNSL and AIDS dementia complex) as well as among those with the most common ADIs (e.g., PC P, pulmonary tuberculosis and KS). Generally, it has been observed that signi ficant improvement occurred for at least 24month (two-year) survival rates among AIDS pa tients with most ADIs, regardless of the time point in the HAART era in which AIDS diagnosis occurred. Using data from 39,403 patients with AIDS in the United States, CD C (2004) reported that 83% of these patients survived for more that 36 months after AI DS diagnosis. Although the effectiveness of HAART may be compromised by other factors such as smoking (Miguez-Burbano et al., 2005), nutrition (Tang & Kaslow, 2003), or CD4 cell count (Wang et al., 2004), its use in populations of patients with different ADIs has persisten tly and increasingly been associated with increased survival rate s. Furthermore, although HAART does not eradicate HIV-infected cells in a patient, virus replica tion has been shown to rebound when the therapy is terminated, regardless of the duration of th erapy (Yang, 2004). Following these observations, we argue th at in a veteran population of AIDS patients with different types of ADIs and ha ving evidence of HAART use, we can use duration of survival times after AIDS diagnosis to define and measure long-term survival after AIDS diagnosis. Months of survival after AIDS diagnosis through December 31, 2003, for veterans who had outpatient or inpati ent care in the Department of Veterans Affairs (VA) in the year 2003 were computed based on the known date (year) of AIDS diagnosis. In accordance to the median cut-off point of the duration of survival times, we defined long-term survival time for a given patient as a length of survival time after AIDS equal to or greater than the median. The median was taken as the cut-off point

PAGE 22

9 because (i) this classification of AIDS patient s was used for the first time; (ii) it is a central value of the distribu tion and would reflect better the increasing numbers of patients who attain long-term survival; and ( iii) the distribution of survival times was skewed. The median was the best cu t-off point under the circumstances. For the purpose of this dissertation, we defined a patient as having long-term survival if he has ever had an AIDS diagnos is and has attained survival time after AIDS diagnosis equal to or greater th an the median survival time in the sample of all patients included in the study. 1.2 Significance of the Study Health care utilization has advanced as the populationÂ’s need for care has changed over time. Some factors that influence n eed include aging, sociodemographic population shifts, and changes in the prevalence and in cidence of different di seases. Likewise, as AIDS patients experience long-term survival, th eir need for health care services is likely to change. Diverse factors may influence ch anges in the need for healthcare in the population of AIDS patients who experience long-term survival. Hence our point of departure was a basic understanding of a study of health service uti lization of the patients in this population. The following section discus ses some of the basic factors influencing the need for health services and capable of be ing described using hea lth services research. This discussion highlights the importance of this dissertation research and is also used to point out different ways thr ough which the healthcare deliver y system can benefit from service utiliza tion research. 1.2.1 Changing Demographics of the Patients The demographics of the HIV epidemic have shifted because of the advancement in treatment and prophylaxis for some ADIs that were characterized by high mortality rates

PAGE 23

10 among AIDS patients. Major shifts have occu rred in the age and r ace of people living with HIV/AIDS and have brought concerns in the healthcare need s of the HIV/AIDS patients. For instance, due to the shift in th e age of affected persons, HIV and aging has become one of the major topics in HIV research. 1.2.1.1 Age and aging Currently, HIV/AIDS patients are capable of living longer, h ealthier, and more productive lives than ever before in the history of the diseas e. As an increasing number of patients live longer, the average age of th e AIDS patients curren tly reported in the literature is increasing due to longer life expectancy. For instance, Stoff et al. (2004) reported that individuals with HIV/AIDS now live well into their 50s, 60s, and 70s. The shift in the age of AIDS patients is particularly important to health services since aging is associated with increased hea lthcare utilization. Over the past twenty years, studies have increasingly reported that the aging of the population and the subse quent increase in the size of the elderly population are the most important demographi c trends that will affect the future health services in the general population (Vladeck & Firman, 1983; Forti et al., 2000). Further, the aging of the population and AIDS patients in par ticular will increase the total amount of healthcare services de manded, will change the types of services demanded, and will have profound economic impl ications that could affect future coverage policies (e.g., Medicare and retirement programs). Aging AIDS patients possibly have additional health concerns related to their disease. For instance, older AIDS patients ar e likely to have other age-related chronic conditions (e.g., high blood pressure, arthritis, etc.); the coexisten ce of these conditions and AIDS may increase the risk of severe co mplications among the patients. In addition, antiretroviral drugs can interact with medications frequently taken for other age-related

PAGE 24

11 conditions to make older AIDS patients more prone to toxicity or other adverse events compared to an average olde r person in the population. Als o, for an individual patient, these factors may vary substantially as an AIDS patient is aging. Older AIDS patients experiencing long-term survival are thus likel y to encounter a more severe course of the disease compared to younger patients. This c ourse of disease may have a significant and unique influence on service ut ilization among older AIDS pa tients. It is important therefore to assess healthcare service use among patients of various ages to determine how the quantity and types of services vary with aging and with different age-related factors. Other areas of concern, which includ e, but are not limited t o, racial or spatial disparities in service utilization among olde r patients may be addressed through this assessment. 1.2.1.2 Race According to the CDC (2004) African Amer icans and Hispanics represent 64% of males living with AIDS in the U.S. In the hi story of HIV/AIDS, it is reported that the proportion of adults reported with HIV w ho are African American exceeded that of whites for the first time in 1996 (41% vs. 38 %) (“Update: Trends in AIDS incidence,” 1997). A high prevalence of injecting drug use and an increased proportion of black men who have sex with men (CDC, 2004) have been shown to be major factors contributing to the increase in incidence and prevalence of HIV among the racial minorities. It is important to note that this racial shift in the prevalence of HIV occurred during the time of HAART introduction in the country. This implies most of the minorities would have the potential of using HAART and e xperience long-term survival. Previous research has established racial differences in service utilization in both HIV (Shapiro et al., 1999) and in the genera l population (Bertakis et al., 2000); it is

PAGE 25

12 therefore important when examining the role of long-term survival on service utilization to take into account racial/ethni c groups of the AIDS patients as race/ethnicity could alter the use of health services. 1.2.2 Comorbidity Profile The overall profile of comorbidity among AIDS patients is ch anging as increased numbers of patients experience longer live s. Non-AIDS-related comorbidities due to aging, drug toxicity, and complex infections have increased and compete with ADIs (Stoff et al., 2004). Therefore, besides th e commonly known ADIs (e.g., PCP, pulmonary tuberculosis, and KS), non-ADIs such as liver , heart, and mental conditions are becoming increasingly prevalent as patie nts live longer. Presence of co morbidity, its treatment, and the possible interaction between comorb idity and HIV treatments can produce a significant effect on service util ization. Information on comorbidity is therefore vital for a more complete assessment of health serv ice utilization because comorbidity has the potential to become a major determinant of service utilization among AIDS patients who attain long-term survival. Figure 1.1 presents a simplified form of th e multifaceted problem of comorbidity in AIDS patients experiencing long-term surviv al. While describing comorbidity using the diagram, we acknowledge that it is almost im possible to explicitly identify what causes different comorbid conditions we have indicate d because of several reasons, including (i) causes of comorbid conditions could be severa l (e.g., HIV, toxicity from HIV treatment, comorbidity exacerbated by treatment of HIV, a true comorbidity or combination of these causes); (ii) different populat ion groups may have different comorbidity pathways and thus generalization could be misleading; and (iii) high prevalence and negative impact of

PAGE 26

13 mental conditions on treatment for HIV dis ease may promote the occurrence of other medical conditions. We thus describe Figure 1.1 taking into account the above-mentioned problems based on the literature on the general trend of comorbidity after the introduction of HAART and reports describing salient comorb id characteristics associated with longterm survival among AIDS patients. Figure 1.1: Comorbidities and Their Likely Sources among Patients with Long-Term Survival after AIDS Diagnosis First, a diagnosis of AIDS is made whenever a person is HIV-positive and he or she has a CD4 cell count below 200 cells per micro liter, or his or her CD4 cells account for fewer than 14 percent of all lymphocytes, or that person has been diagnosed with one or more of the ADIs (CDC, 1992; Appendix A). Th e ADIs occur due to the weakening of an D Age/Aging Hypertension Diabetes Arthritis Dementia B Long-term Survival Liver disease Heart disease Toxicities related to highly active antiretroviral therapy (HAART) A AIDS-Defining Illnesses Pneumocystis carinii pneumonia KaposiÂ’s sarcoma Bacterial pneumonia G C E F

PAGE 27

14 individualÂ’s immune system following HIV infection, and they could appear during and/or after the first diagnosis of AIDS ( region A). Following the advent of HAART and impr ovements in the therapeutic care of HIV/AIDS in general, patients with a diagnos is of AIDS can expect to survive for a longer period than before. Lee, Karon, Se lik, Neal, and Fleming (2001) reported that beginning in the mid-1990s when HAART was initia ted; the survival rate on the average was at least three years am ong patients with a diagnosis of AIDS. Similarly, the CDC (2004) reported that 83% of persons with HIV survived for more than three years after an AIDS diagnosis. Thus, most AIDS patients have the potential to attain long-term survival and live with HIV as a chronic condition ( region B) . Although ADIs continue to exist among patients at this stage, they are occurring at a much lower prevalence. However, research has increasingly shown a shift in the types and incidence of comorbity among HIV/AIDS patients in the post-HAART era. While the use of HAART has largely transformed HI V disease into a chronic, manageable illness, most antiretroviral agents are metabo lized through the liver and can cause varying degrees of hepatotoxicity. In general, liv er conditions from hepatitis C infection (Benhamou et al., 2001), coronary heart di sease (Glesby, 2003), cance r (Bower et al., 2004), and tuberculosis (Dean et al., 2002) have been reported to occur with increasing incidence in this era. These conditions ha ve become a common cau se of hospitalization for AIDS patients with long-term survival. There is also an increased ra te of liver toxicity caused by long-term use of antiretroviral therapy. Anemia, lipid abnormalities, bone diseases, and glucose intoleran ce are other conditions that ha ve been closely associated with prolonged use of HAART.

PAGE 28

15 Figure 1.1 also presents the problem of co-occurring ADIs and non-ADIs whose incidence has increased in the post-HAART era ( region C) . Prior to and in the early HAART era the most prevalent ADIs were PCP, KS, and Mycobacterium Avium Complex (MAC). However, reports have s hown more recently that, while many patients experience long-term survival, multiple groups of patients have reported lower incidence of these conditions and higher rates of ne urologic complications, malignancies, and bacterial pneumonia (Gebo & Moore, 2004). This implies that at the same time that HIV is increasingly managed as a chronic diseas e in the post-HAART era, a shift has occurred in the prevalence of ADIs. This shift is co-o ccurring with the increas ed incidence of nonHIV comorbidity in the post-HAART era and could create an influence on service utilization. The prevalence of AIDS in older people is growing because of both long-term survival and growing numbers of older pers ons who become newly infected with HIV. Similar to any other aging popul ation, rapid increases in the prevalence of age-related comorbidities, such as hypertension, stroke, di abetes, etc., are also likely to occur among older AIDS patients (region D). Considering patients who are diagnosed with AIDS at older ages, some of the age-related medical complications may have significant impact on their immune system, and thus further co mplicate the course of the disease and its treatment. Treatment of the age-related condi tions may interact w ith the treatment of ADIs to create clinical adverse events (region E) . AIDS patients who survive to older ages are likely to have a unique experience with comorbidity (region F) . First, these patients may have high prevalence of ADIs compared to other groups of AIDS patients because of the weakened immunity that

PAGE 29

16 occurs with aging. Second, due to the co -occurring of non-HIV comorbidity and agerelated conditions, older patient s are more likely to be ch aracterized by a higher burden of chronic conditions, more effects of th e toxicities relate d to HAART, and drug resistance. For these reasons, as patients e xperience long-term surviv al and live to older age, non-HIV comorbidity (as described from region B of the diagram) would combine with age-related chronic conditions to create an influence on utilization. Lastly, the treatment of comorbidities as well as combinations of antiretroviral therapy is likely to cause drug-drug interact ions for AIDS patients experiencing longterm survival (region C) ; or older, newly diagnosed AIDS patients (region E) ; or those AIDS patients living into their advan ced ages due to long-term survival (region G) . These interactions can drop drug levels too low to be effective or can lead to serious or fatal events. The interactions are t hus an added threat to the burden of comorbidity, and may have major implications in the trea tment plans and serv ice utilization. Since the coexistence of comorbid cond itions and HIV is likely to complicate treatment plans, considerable effects on serv ice utilization may be generated. Knowledge of how utilization varies with comorbiditie s based on important patient characteristics (e.g., sociodemographics) will thus provide va luable input in the management of the patients. 1.2.3 Disparities in Service Utilization Disparities in service utilization have been reported in population studies (Hargraves et al., 2001) as well as HIV studies (Shapiro et al., 1999). Most of the studies on the subject have shown that minority popula tions do not fare as well as whites in the access to and utilization of h ealthcare services. Particular ly in HIV populations, it has been shown that, even when treatment had improved and patients were mainly managed

PAGE 30

17 on outpatient basis, minorities were increasingl y associated with costly inpatient care (Hellinger & Fleishman, 2001; Kass et al., 1999). One of the major reasons for the poor trend of service utiliz ation among minorities is delayed entry into the health system. The delayed entry might be due to fewer res ources, poor access to services, or other unexplored factors. As a result, patients fr om minority populations are more likely to access medical care while at a relatively more severe stage of the disease, which would require more intense and costly inpatient car e. Although disparities in service utilization have been documented in the early postHAART period (Hellinger, 2004), the experience of patients with long-term su rvival has not been investig ated to determine types and extents of disparities that may be occurring as the number of these patient increases. Service utilization data among patients with long-term survival are important for determining health disparities. The data co uld be used to determine whether types of disparities observed in the previous phases of the epidemic persist. Racial disparities in service utilization were an ar ea of major concern in the earlier phases of the epidemic. For example, it has been reported that hospita lization rates have b een disproportionately high for racial minorities in the early hist ory of HIV, and, in addition, members of minority communities have been less likely to use outpatient services compared to whites (Uphold & Mkanta, 2005). The U.S. healthcare system may con tinue with the adoption of existing policies on disparities (e.g., assuring th at all individuals ha ve access to basic health care at affordable cost) if the same t ypes of disparities persist when AIDS patients attain long-term survival. Otherwise, new in tervening policies may be formulated if new trends in disparities are detected.

PAGE 31

18 1.3 Chapter Summary In conclusion, this dissertation research aims to examine the impact of long-term survival after AIDS diagnosis on the use of healthcare services among veterans who use HAART and receive care in the VA system. A pa tient is said to have long-term survival if he has lived at least si x years (median cut-off point) after AIDS diagnosis. Data concerning duration of AIDS survival and utiliz ation of health servi ces are obtained from the VA. The findings of this study are expected to provide a valuable addition in building knowledge and entice further research on health care issues pertaini ng to the population of AIDS patients with long-term survival. The remaining chapters are organized as follows. Chapter two describes the source of data for this research. Chapter three revi ews the relevant litera ture for this study and provides an overview of the underlying theo retical framework th at specifies the relationship between long-term survival and se rvice utilization. Chapte r four describes in detail the data and analytical methods used in the research. This ch apter is divided into several major sections in cluding objectives and hypot heses, methods for finding correlates of long-term survival, and methods for examining the impact of long-term survival on the use of healthcare serv ices. Also, measurements and operational definitions of dependent variables outpatient vi sits, inpatient visits, and length of stay as well as independent variables such as age, comorbidity, and AIDS-defining illnesses are described within chapter four. Chapter five pr esents the resu lts of the study and, lastly, chapter six presents the conclusion of the study that includes disc ussion of the findings, study limitations, and areas for future research.

PAGE 32

19 CHAPTER TWO THE IMMUNOLOGY CASE REGISTRY The Veterans Health Administration (VHA) is the principal agency of the VA and the largest integrated healthca re system in the United States , with a budget of $25 billion for 2003. The VA has 1,300 care facilities that include 163 hospitals, 850 ambulatory care and community-based outpa tient clinics, 206 counselin g centers, 137 nursing homes and, 43 domiciliary facilities (“Facts about the Department of Veterans Affairs,” 2002). These VA facilities are organized into 22 regional VA Integrated Service Networks (VISNs). Regarding HIV/AIDS, the VA is also th e nation’s largest public integrated healthcare system and the largest single provid er of health care to HIV-infected persons in the United States. In the fiscal y ear 2003, for instance, the VA saw about 20,111 HIV infected patients and had almost 9,000 inpati ent discharges for 4,600 patients who had a mean length of stay of 15.7 days (“VA HI V Report,” 2004). The information on the HIV infected patients treated in the VA system is managed through the Immunology Case Registry (ICR), a database that supports the maintenance of local and national registries for the tracking of HIV dis ease and health care among VA patients. The ICR was the source of information for this study. This chap ter describes the establishment of the ICR, its purpose, management, and othe r important char acteristics. 2.1 Establishment and the Purpose of the ICR The VA established the ICR in 1992 with the primary purpose of building a national-level administrative database to c ount the number of HIV infected individuals

PAGE 33

20 receiving care in the VA system. However, over time, following advancements in the technology of transferring and storing medical data, the cap acity of the ICR has been expanded to include quantity of healthcare services received by the patients. Currently, the ICR includes data on dental, mental healt h, and substance abuse se rvices. In the year 2000, the VA established the Center for Quality Management in HIV Care (CQM) and placed the management of the ICR under the cen ter. The main goal of the center is to maintain and assure appropriate, cost-effec tive, high quality HIV care for veterans. In recognition of the health challenges that ar e faced by HIV infected patients in the postHAART period, especially those exhibiting prolonged lives, the mission of CQM was expanded in 2001 to address issues of Hepatitis C in HIV patients. 2.2 Data Input and Extraction from the ICR Each local VA station maintains a local ICR within their electronic medical record system. At these stations, the local staff adds patients manually into the registries. For security and patient privacy purposes, the pa tientsÂ’ identifiers ar e de-identified (i.e., social security numbers and names entered at each local station are scrambled). At the national ICR, located at the VA Palo Alto Heal th Care System California, data extracts from the local ICRs are automatically conducted by the VAÂ’s mainframe system for inclusion in the national registr y. Information for a particular patient can thus be accessed at two levels: at the reporting (local ) station and at the national level. The ICR collects data fields from about 15 different software packages within the VAÂ’s electronic medical record system known as the Veterans Health Information Systems & Technology Architecture (VistA). Fo r example, outpatient records are drawn from the Ambulatory Care Reporting Software, while laboratory data , including type of test, dates, and results, are extracted from the Laboratory Package. The data extracted

PAGE 34

21 from these software applications are updated nightly. We adapted Table 1 from Backus et al. (2001) to show sources of da ta extracts for some ICR data fields. Each data field has several variables associated with the field. Fo r instance, patient file contains demographic data and has the largest number of variables (137) among all da ta fields. Examples of key variables contained in each data fi eld are also presented in Table 2.1. Table 2.1. Sources of Data Extracts and Nu mber of Variables in the National Immunology Case Registry Data Field Number of Variables Key Variables Bed section file Inpatient file Inpatient procedure file Inpatient surgical procedure file IV pharmacy file Laboratory test history file Outpatient file Outpatient prescription file Patient file Radiology file 17 19 11 10 5 8 8 11 137 11 Long-term care bed Discharge diagnosis Admissions, CPT codes Length of stay, CPT codes National drug code (NDC) Test name, value, units Clinic type, ICD-9 code NDC, amount dispensed Demographic information Results, CPT codes CPT: Current Procedural Terminology; ICD-9: International Classification of Diseases, 9th Revision 2.3 Sample Data and Patient Factors from the ICR for the Fiscal Year 2003 People with HIV who use medical servi ces within the VA system are added continually into the ICR. This means the ICR is an active or living registry that holds medical records for all patients who have ever received VA health services for HIV/AIDS. Since its inception in 1992, over 50,000 patients have received care for HIV from the VA and their medical records are co ntained in the national ICR. Records from the ICR for the fiscal year 2003 showed that 20,111 HIV infected patients received health care from the VA medical system. Overall, the population of VA patients with known HIV is described as overwhelmingly male, ol der, predominantly non-white, and receives care in urban centers (Backus et al., 2001). Part of the ICR data for patients who received care in the fiscal year 2003 is shown in Table 2.2. This information gives an elaboration

PAGE 35

22 of the important aspects of the patient populati on and shows the strength of the registry in capturing relevant information for the assessmen t of service utiliza tion. For instance, of the 20,108 HIV infected veterans with know n gender who received care, 19,617 (98%) were male and 9,294 (46%) were black. Table 2.2. Selected Patient Characteristic s and Outcomes from the Immunology Case Registry for the fiscal year 2003 Characteristics Number Percent Gender Male Female 19,617 491 98.0 2.0 Race Black (Not Hispanic) Hispanic White Unknown Other 9,294 1,463 7,094 1,433 827 46.2 7.3 35.3 7.1 4.1 Antiretroviral Drug Use Any highly active anti retroviral therapy (HAART) use Any nucleoside reverse transcriptase inhibitor (nRTI) use Any non nucleoside reverse transcript ase inhibitor (nnRTI) use Any protease inhibitors (PI) use 14,978 14,831 7,451 8,482 Deaths 1,114N/A Inpatient Care Inpatient discharges Patients with inpatient discharges Mean length of stay in days 8,684 4,597 15.7 N/A N/A N/A New Patients Added in the Immunology Case Registry Male Female Black (Not Hispanic) Hispanic White (Not Hispanic) Other race 1,912 43 786 122 629 21 98.0 2.0 50.4 7.8 40.4 1.4 N/A: Not Applicable Table 2.2 further shows nucleoside reverse transcriptase inhibitor (nRTI) as the HAART regimen used by most of the pa tients (14831, 74%). The 1,114 patients who died in the year 2003 represented about 6% of all the patients who received care in the same year. Regarding inpatient care, the data showed that 4,597 (23%) of the patients had

PAGE 36

23 episodes of hospitalization. Last, of the patients entering the re gistry for the first time in 2003, the majority were men (98%) and blacks (50%) “VA HIV Report,” (2004). 2.4 Strengths and Weaknesses of the ICR Patients are added continually into the ICR. Thus, the large number of patients in the registry is one of the ICR’s strengths. As more AIDS patients experience long-term survival, for instance, the ICR provides a unique opportunity of assessing different aspects of the HIV disease over time. Issues on service utilization, death, and comorbidity among patients with long-term survival may be examined at different phases of the disease. Because the ICR provides local and national tracking of HIV positive veterans, it offers an opportunity to study geographical tren ds for several aspects of the disease, for example, variations in resource use by regi on. Also, the ICR offers the ability to track veterans who move from one area to another, as it is capable of id entifying such veterans as long as they receive their he althcare services within the VA. The ICR has some limitations. First, becau se the national ICR extracts data based on select data elements from the local stations , it is subject to high ra tes of local variation in data reporting (Backus et al., 2001), and thus it is considered less complete compared to the local medical record systems. Howeve r, the CQM has been making efforts aimed at minimizing the number of HIV infected pati ents potentially missed by the current ICR capture mechanism. Secondly, the ICR does not contain information on common measures of socioeconomic status, for example, education and income. The impact of socioeconomic factors may vary across soci al, gender, and age groups and thus may provide an important way of su mmarizing different components of service utilization. For example, individuals with lower income s and less education (usually women and

PAGE 37

24 members of racial minorities) are likely to have higher rates of utilization than better educated, wealthier persons (usually whites).

PAGE 38

25 CHAPTER THREE LITERATURE REVIEW This chapter presents some important background information of the available literature related to the study. First, we review literature related to th e selected theoretical model of healthcare service utilization. This is done to show how the model could be applied in this study, and also to produce evid ence that the model can be validly applied to the population of individuals with HIV/AI DS. Second, a review of literature on the use of HAART and the effects of HAART on servi ce utilization is pres ented. Last, a review of literature on service utilizat ion in the population of individuals with HIV/AIDS is presented. This review is organized accord ing to the theoretical model of service utilization, and is meant to offer an opport unity of understanding important trends of service utilization lead ing into the current era of pr olonged survival among HIV/AIDS patients. 3.1 Theoretical Model 3.1.1 Andersen Behavioral Model of Health Service Utilization This research used the Andersen Behavi oral Model of service utilization (ABM) originally developed by Ande rsen (1968) and revised by Andersen and Newman (1973) and Andersen (1995) as the conceptual framework for the assessment of service utilization among AIDS patients with long-term survival. The ABM was selected as the conceptual framework in this research fo r three basic reasons. First, the ABM has become a widely adopted framework for explaining and predicting health service utilization in different populations (Wolinsky, 1994). Second, the ABM has been

PAGE 39

26 extensively used in service utilization research pertaini ng to HIV/AIDS patients in predicting the use of both HAART (Andersen et al., 2000) and medical services such as inpatient hospitalizations and outpatient visits (Fleishman et al., 1994). Lastly, this research involves secondary analysis, and the ABM has been successfully used in the analysis of secondary data focusing on servic e utilization in a national probability sample of AIDS patients using the AIDS Cost and Services Utilization Survey, ACSUS (Fleishman et al., 1994; Hellinger et al., 1994). The original ABM was developed in the 1960s to assist in studying why families use health services (Andersen, 1968). But over tim e, it was determined that use of family as a unit of analysis was practically difficu lt because variations in various aspects of health care among family members did not allow for meaningful and interpretable measures to be obtained. The model was thus revised to focus on the individual as a unit of analysis, while integrating important fa mily characteristics, e.g., family size or household income. The ABM proposes that observed variation in health service utilization is a func tion of an individualÂ’s predis posing, enabling, and need factors (Andersen, 1968, 1995). The predisposing factors include variables that describe the propensity of the individual to use health services. These variables include demographic characteristics such as age, gender, and marita l status; social structure characteristics such as race/ethnicity, education, and occupation; a nd health beliefs. This dissertation will use age, race/ethnicity, HIV risk factor, and veteranÂ’s period of service as predisposing factors. Enabling factors are those affecting the patientÂ’s ability to gain access to services, including income, access to a regular sour ce of care, and insurance coverage. The patientÂ’s region of residence was used as a community enabling factor in this study. Need

PAGE 40

27 factors can be defined as an i ndividualÂ’s health status as he or she perceives it and/or as evaluated by a healthcare provider. Examini ng the level of illness or establishing the presence of specific condition(s) can measure need-related factors. For example, three stages of HIV disease: asymptomatic HIV, symptomatic HIV, and AIDS, as well as the presence of comorbidity, can be used as n eed variables in the AB M framework. We used length of AIDS survival, presence of ADIs, diagnosis of psychiatric comorbidity, CD4 cell count, death, and number of medical comorbidities as need characteristics. The predisposing factors are considered exogenous or fixed and hypothesized to influence service use both direc tly and indirectly through the enabling and need variables. A premise of the ABM is that equitable acce ss to services occurs when demographic and need characteristics account for most of th e variance in service use, while access is considered inequitable when social characte ristics (e.g., ethnicity) and enabling resources (e.g., income) determine who gets the services (Andersen, 1968, 1995). Since its inception, the ABM has undergone several revisions and updates as a result of utilization research findings (Andersen and Newman, 1973; Andersen, 1995; Aday & Awe, 1997), and is now modified w ith important additions that indicate dynamics of service utilization. The earlies t revision of the ABM posited that it was important to pay attention to both individual a nd societal determinants of the utilization of health services rather than to consider it a type of individua l behavior (Andersen & Newman, 1973). The authors pointed out that the societal determin ants of utilization, such as familyÂ’s racial/ethnic background or av ailability of relevant services may affect the individual determinants both directly a nd through the health se rvice system. Thus it was important to include the characteristics of the external environment and healthcare

PAGE 41

28 system in the framework to provide a mean s of assessing what happens to the patient following his or her entry into the system. A revision that took place in the 1990s recognized changes in personal practices (e.g., appropriate diet, exercise, etc.) and the maintenance and improvement of health status as explicit outcomes of health services delivery (Andersen, 1995; Aday and Awe, 1997) . The feedback loop of the ABM was also recognized; that is, though it is expected that health behavi or (personal health practice and use of services) would modify individualsÂ’ need for services, utilization outcomes such as patient satisfaction and h ealth status may also modify subsequent predisposition, enabling resources, need-related characteristics, and health behavior. The revisions and modifications to the ABM are evidence that research on service use and access has shifted from an individual level focu s to a combination of the individual, the healthcare system, the external environment, and the effects that each have on the others. The Behavioral Model for Vulnerable P opulations (Gelberg, Andersen, & Leake, 2000) was a major revision of the ABM that ha d an important application to individuals with HIV and other groups considered to be vulnerable. This model represents an employment of the ABM that includes factors to consider when studying the use of health services in vulnerable populations. Some of the vulnerable populati ons that have been mentioned in the health services literature include women, racial and ethnic minorities, immigrants, elderly, homeless, and HIV/AI DS patients (Aday, 1993, 1994); in HIV care, previous studies have defined vulnerable popul ations as racial/ethnic minorities, women, injection drug users (IDUs), and the less educ ated (Kilbourne et al ., 2002; Andersen et al., 2000). The model for vulnerabl e populations specifies that apart from the traditional components of predisposition, enabling, and n eed, vulnerable populations have specific

PAGE 42

29 vulnerability factors that may impede thei r ability to use healthcare services. For example, in their study of use of health serv ices in a homeless populat ion, Gelberg et al. (2000) added specific vulnerable domains in each of the original components of the ABM. The predisposition vulnerable domain in cluded social structur e factors, such as acculturation, immigration status, and literacy ; childhood characteristics including foster care and parental sickness; living conditions; mental illness; and substance abuse. The enabling vulnerable domain included personal/fa mily resources such as receipt of public benefits and the availability of social servic es, while the need vulnerable domain included perceptions and evaluated need regarding some conditions such as tuberculosis and AIDS. Gelberg and her colleagues showed that these additions to the ABM were important in profiling serv ice use among the homeless. Researchers have successfully used the ABM to describe and predict health services use among HIV/AIDS patients in diffe rent phases of the HIV disease. The ABM has been used to predict use of medical care, such as hospi tal admissions (Fleishman et al., 1994; Kilbourne et al., 2002) and use of HAART (Andersen et al., 2000). The ABM for vulnerable populations has been applied to HIV/AIDS patients to predict the role of symptoms in service use among the vulne rable groups of women, ethnic minorities, IDUs, and people who are less educated (Kilbour ne et al., 2002) as well as to determine the role of age as a vulnera bility factor on service uti lization among midlife and older adults (Emlet & Farkas, 2002). This dissertation applied the ABM for vulnerable populations to examine the influence of the length of AIDS survival on service use. In our analysis, we have considered AIDS patients with long-term survival as a vu lnerable group for the following

PAGE 43

30 reasons that can potentially be associated with increased use of services: (i) many of the patients have reached the most severe stage of HIV disease; (ii) they are more likely to be vulnerable to drug resistance, drug-drug interact ions, and toxicities; (iii) they have high risk of comorbidity; and (iv) they are more likely to suffer AIDS-related fatal events compared to other categories of HIV patie nts. For this reason, we have described (Chapter Four) specific vulnera bility factors that have been included in the ABM to better predict the utilizati on behavior of this population. Figure 3.1 presents the ABM as applied in this study. It shows the original model and the modifications done to integrate vulnerability factors from AIDS patients with long-term survival. Figure 3.1. The Andersen Behavioral Model: Utilization of Health Service among AIDS Patients Experiencing Long-Term Survival. This study used four measures of predispos ing characteristics. These included three demographic variables: veteranÂ’s age, period or era of service, and risk factor, and race as Predisposing Factors Age Period of service ---------------Race Risk factor Enabling Factors Region of residence Need Factors Presence of AIDSdefining illness Non-AIDS-defining illness Psychiatric conditions ---------------Long-term survival Death Service Utilization Hospitalization and Outpatient care

PAGE 44

31 a social structure variable. Of these, race a nd risk factor were included as vulnerability factors based on previous investigations th at regarded racial minorities and IDUs as vulnerable populations among persons with HI V/AIDS (Andersen et al., 2000; Kilbourne et al., 2002). United States region of residence was incl uded as a community enabling variable. We included five measures of n eed characteristics. These included general need-related variables indicating diagnosis of AIDS-defining illnesses, non-AIDSdefining illnesses (medical comorbidity), a nd psychiatric conditions during the year under study. We have defined AIDS patients with long-term survival as a vulnerable population partly because of the health pr oblems caused by their prolonged lives after AIDS diagnosis and high risk of death. For th ese reasons, two variable s were included in the model as vulnerability factors. These we re duration of survival since AIDS that indicated attainment of long-term survival and death, which was included as a measure of severity of HIV/AIDS in the study peri od. Lastly, the model shows how service utilization occurs. Use of a given type and am ount of services occur as a function of the predisposing, enabling, and need factors. 3.2 Review of HIV/AIDS-Related Literature In order to contribute meani ngfully to the discussions that lead into this study, we divided the following section of the review in to three parts: (i) Use of HAART; (ii) Effects of HAART on service utilization; and (iii) The role of predisposing, enabling, and need characteristics on service utilization among HIV/AIDS pa tients. Use and Effects of HAART are emphasized because this study involv es only those patients with evidence of HAART use in the VA system. Also, while keeping other factors constant, the use of HAART is considered as an important link into patientsÂ’ attainment of long-term survival. Furthermore, although this study us es information from AIDS patients, our

PAGE 45

32 review is based on the experience of HIV/AIDS patients in general as a way of showing important trends that lead into the emergen ce of AIDS patients with long-term survival. This review, using important patterns of service utilization among HIV/AIDS from previous inquiries, is expected to provid e the reader with an important link in understanding the context under which questions and hypotheses were formulated. In line with our theoretical model, the review was conducted in the framework of the ABM: that is, we examined the role of predisposing, en abling and need characteristics on service utilization among HIV/AIDS patients. Recalling that this study is based on the in formation from AIDS patients receiving care from the VA system, different parts of th e review have included relevant service use data from studies based on VA patients. Th is information was used to demonstrate utilization experience of veterans with HIV infection, and, where applicable, to compare utilization experi ence between VA and non-VA patients. 3.3 Use of HAART Previous research in the HAART era ha s extensively described the patterns of HAART use among different HIV populations. Mo st people with HIV initiated the use of HAART when it was introduced, and HAART use prevalence grew high with time (Cunningham et al., 2000; Katz et al., 2001). Among healthcare providers, studies showed that almost equal proportions of patients treated by expert generalists and infectious disease specialists were on appr opriate HAART regimens within one year of its introduction (La ndon et al., 2003). HIV infected persons who received care in the VA had high rate s of antiretroviral drug use. Among 18,572 patients who received care in the fiscal year 2001, about 14,076 (76%) patients received at least one prescr iption for an FDA-approved antiretroviral drug,

PAGE 46

33 and the rate of use remained high for two successive years, 2000 and 2001, among the VA users (CQM, 2001). Also, “VA HIV Report” (2004) reported that about 74% of patients who received care for HIV in th e VA in 2003 were on some HAART regimen. Other studies, e.g., Mole, Ockrim, and Holodni y (1999), Keiser et al. (1999), and Keiser et al. (2001), have also in dicated increased used of HAART among VA patients. Some important findings from these studies have been discussed under the effects of HAART on service utilization. Following the high rates of HAART use in the VA, the inclusion of only patients with history of HAART in th is study is still ex pected to produce a representative sample for AIDS patients receiving care in the VA system. 3.3.1 Predisposing Variables and the Use of HAART Despite the reported increases in the ove rall use prevalence, several studies reported persistent disparities in access to HAART in the nation. For example, in a study that applied the framework of ABM for vulnerable populatio ns after the advent of HAART, Andersen et al. (2000) showed that women, IDUs, African Americans, and the least educated were less likely to gain ear ly access to HAART. This study was based on a national probability sample of 2,776 adu lts. Similarly, Cunningham et al. (2000) indicated that by 1998 the proportion of HIV patie nts under care in the U.S. had grown to over 70%, but blacks, drug users, and female heterosexuals were less likely to use HAART. The largest and first national HIV u tilization study, the HIV Cost and Service Utilization (HCSUS), showed that only a few women and racial minorities had started using HAART (Shapiro et al., 1999). Th e HCSUS included 2,864 patients who were selected from lists from a variety of provide r sites including clinic s, outpatient hospital facilities, and physician practi ces. In particular, Shapiro and colleagues found that 22 percent of the women, compared to 13 per cent of the men, and 20 percent of African

PAGE 47

34 American and 16 percent of the Latino, comp ared to 12 percent of the white, in the sample had not received HAART. Gebo, Diener-West, and Moore (2001) coll ected data from 10 U.S. HIV primary care sites under HIV Research Network (HIV RN), and showed that African Americans (odds ratio = .84; p<.05) and IDUs (odds rati o = .86; p<.05) were le ss likely to receive HAART compared to their respective counterpa rts. Female sex was also an independent predictor of not receiving HAART (G iordano et al., 2003). McNaghten, Hanson, Dworkin, and Jones (2003) examined 9,530 patie nts eligible for HAART in 10 U.S. cities and found that differences in HAART use by gender, race, exposure mode, and alcoholism existed among the patients. M oore Stanton, Gopalan, and Chaisson (1994) reported a study involving 838 patients (79% blacks and 20% non-Hispanic whites). Moore and colleagues showed that although ther e were no racial disparities in the stage of HIV at the time of presentation, there were significant racial disp arities in receipt of HAART. Sixty-three percent of el igible whites, but only 48% of eligible blacks, received the therapy. Based on the ACSUS study, Smith and Kirking (1999) showed that female and those aged between 15 and 24 years were associated with lower odds of using HAART. But unlike other studies, it was found that African American race was associated with higher od ds of using the therapy. Our review has demonstrated pe rsistent differences that have occurred in the use of HAART. Most of these differences were e xplained by predisposing factors of race, gender, and IDU among different populati ons of HIV/AIDS patients, suggesting inequitable access to HAART. However, it is wo rth mentioning that most of the analyses did not adjust for indicators of socioeconomic or enabling variables, e.g., income and

PAGE 48

35 education, which may have strong correlation with gender or race. Racial or gender variations in HAART use may be cau sed by the confounding effects of these socioeconomic factors. For example, in a study based on nationally representative data that showed lower odds of HAART use among African-Americans and IDUs in bivariate analyses, it was found that the effect becam e nonsignificant when education, region, and insurance were controlled (Cunningham et al ., 2000). Some investigators have pointed out that use of medical records, in whic h the socioeconomic indicators may not be reported, is one of the reasons for the lack of adjusted analysis (Gebo et al., 2001). 3.3.2 Enabling Variables and the Use of HAART Previous research has indicated a le ss favorable pattern of HAART use in uninsured and publicly insured (Medicaid) indi viduals. Examples of studies that showed less favorable use of HAART among unins ured or Medicaid enrollees include Cunningham et al. (2000), Shapiro et al . (1999) who used the HCSUS study, Hsu, Vittinghoff, Katz, and Schwarcz (2001), and Keruly, Conviser, and Moore (2002). In addition, Keruly et al. (2002) and Gebo et al . (2001) found that Me dicaid and Medicare patients were significantly more likely to use HAART than uninsured patients. These findings indicated that the di fferences in access to HAART caused by enabling factors such as lack of health insurance might decrease through government intervention through public health programs. 3.3.3 Conclusion Despite the observed disparities in the use of HAART based on predisposing and enabling characteristics of the patients, th e advent of HAART completely changed the face of HIV. In addition to the evidence of the survival benefits of HAART (Chapter One), it is equally important to understand how and to what extent the advent of HAART

PAGE 49

36 changed patterns of service uti lization. The following part of th e review is devoted to the effect of HAART on service utilization among HIV/AIDS patients, taking into account different phases of the HIV disease. 3.4 Effects of HAART on Service Utilization 3.4.1 Effects of HAART on Service Ut ilization within the VA System Use of HAART has been associated with a decrease in the inpatient care after the introduction of HAART. Some important studies using ve teran populations have used costs of care to measure changes in type s and amount of health services utilized. Although we do not use cost of car e as a covariate in this stud y, we have reviewed studies based on cost of care to reflect decreases in the volume of outpatient and inpatient services utilized by HIV patients after the introduction of HAART. Mole et al. (1999) showed in a study of HIV-infected persons seen at a VA health center that HAART produced up to a 70% decline in mortality, a nd costs for HIV-related outpatient resource utilization in 1997 fell by 25% and 59% as compared to 1993 and 1995 respectively. The greatest fall in utilization was for inpatient bed-days of care, where the average cost per patient fell by $2,782 between 1993 and 1997. In another report, investigation of the effect of HAART on outcomes in VA medical centers found that a marked increase in per-patient HAART costs was accompanied by a su stained decrease in the days of acute inpatient hospitalization (Goetz et al., 2001). It was shown that hospitalizations per 1, 000 patients decreased from 5,159 in fiscal year 1996 to 2,354 in fiscal year 1999. This investigation covered four years beginning October 1995, and it exhibited an association between use of HAART with a reduction in the rate and co sts of acute hos pitalizations. Keiser et al. (1999) examined the rate of service use and cost among veterans who received care at the Dallas VAMC between January 1995 and July 1997. It was shown

PAGE 50

37 that a decrease in hospital days, infectious disease clinic visits, emergency department visits, as well as inpatient and outpatient costs had occurr ed. Specifically, per patient costs of HIV care significantly decreased fr om a monthly average of $1,905 in the first interval to $1,122 in the last interval. In another investigation, Keiser et al. (2001) reported increased cost of HA ART per patient from $79 to $518 per patient per month in a study that covered 44 months of HAART use. However, this cost was offset by a decrease in hospitalizations where it was reported that i npatient costs decreased from $1,275 to less than $500 per patien t per month among the veterans. 3.4.2 Effects of HAART on Service Utiliz ation in Other Healthcare Settings The association between HAART use and h ealthcare service utilization has been extensively investigated in the general population (Perez & Moore, 2003; Shapiro et al., 1999; Giordano et al., 2003). Mo st of the early investig ations compared service utilization among HIV/AIDS patients between pre-HAART and HAART eras (e.g., Fleishman & Hellinger, 2001) and found asso ciation between intr oduction of HAART and rates of health service use. It was ev idently shown that compared to the pre-HAART era, patients in the HAART era were associated with decreased rate s of hospitalizations (Perez & Moore, 2003; “Hospital and outpatient services,” 2 002). For instance, “Hospital and outpatient services” (2002) used inform ation from a sample of 5,255 patients from nine U.S. HIV primary and specialty care s ites for 1999 and showed that use of HAART was associated with significantly lower hospi tal utilization compared to non-users (265 vs. 320 days per 100 persons, p<.05). Other reports used costs of care to show changes in the intensity/volume of service utilization. Bozzette and Helli nger (2001) showed that expenditures for medications increased due to the widespread use of HA ART, but reductions in the use of hospital

PAGE 51

38 services were still apparent. Also, using the experience of HIV-infected veterans in VA care, Keiser et al. (2001) s howed that mean monthly hos pitalization and associated inpatient costs decreased and remained low 2 years after the introduc tion of antiretroviral therapy. The authors also showed that hosp italization costs decreas ed from $1,275 to less than $500 per patient per month, within six months of the introduction of the therapy, reflecting a reduction in the rates of inpatient care. Other studies have focused exclusively on the trends of service utilization that occurred in the post-HAART era. Evidence of increased survival grew and outpatient services became the dominant type of care re ceived by the patients (Sherer et al., 2002). Several studies reported furt her decreases in hospitalizati ons and established that the corresponding increase in the use of outpatien t care was associated with the receipt of HAART (Garattini et al., 2001; Gebo et al., 2005). A study that used HCSUS data employed cost analysis to measure the cha nge in the volume of inpatient care after initiation of HAART (Bozzette et al., 1998) . Bozzette and his colleagues showed that while overall expenditures declined by 16 perc ent, the expenditure for hospital care, the largest category of expenditure due to extended LOS, declined by 43 percent. Previous research has also indicated that declining trends of hospital use were not uniform. Predisposing and enabling characteristic s were shown to be important in service use variation after in troduction of HAART. For instance, studies increasingly showed that African Americans, male, and IDUs conti nued to be associated with increased rates of costly hospitaliza tions (“Hospital and outpatient se rvices,” 2002; Pulvirenti et al., 2003). In another study of comprehensive hospita l discharge data from seven states for 1996 to 2000, Fleishman and Hellinger (2003) showed that there were racial and health

PAGE 52

39 insurance disparities in the re duction of LOS. They showed that across all seven states: California, Colorado, Kansas, Maryland, New York, New Jersey, and South Carolina, mean LOS was lowest for white men (8.28 da ys, median = 5) and highest for AfricanAmerican men (9.92 days, median = 6). Furthermore, mean LOS was higher for Medicaid admissions (9.73 days) than for pr ivate (9.09 days) or Medicare (8.88 days) admissions. Similarly, Bozzette and Hellinge r (2001) pointed out that despite high decreases in hospital use among HIV/AIDS pa tients, not all groups benefited equally from HAART. It was shown that by the e nd of 1998, hospital care was still the largest component of expenditures for disadvantage d groups such as women, blacks, patients with public health insurance, and th ose who had not completed high school. Thus, despite the overwhelming evidence of reduced hospital inpatient care and increased use of outpatient car e after HAART, the cited studie s have indicated persistent disparities in service utilization. Racial minorities, women, and persons with public health insurance coverage we re associated with poor ut ilization outcomes. A better understanding of these disparit ies is provided in the next section of the review. This section examined the predictors of serv ice utilization among HIV/AIDS patients. 3.4.3 The Role of Predisposing Va riables in Service Utilization 3.4.3.1 Age and service utilization The effect of age on service utilization is expected to impact veterans with HIV because they are aging and tend to be older than infected individuals in the general population. A study using a national sample (H CSUS) showed that 12% of the population (Bozzette et al., 1998), but about 38% of the veterans in VA care (CQM, 2001), were aged 50 years and older. In addition, the CQ M pointed out that the proportion of older patients receiving care in the VA is increasing. Long-term surv ival of the patients as well

PAGE 53

40 as aging of the cohort of patients receivi ng care in the VA system was suggested as possible explanations for the increase of older patients. Because of the common perception that HI V/AIDS afflicts only young persons, the effect of age on service utiliz ation did not draw much atten tion from researchers in the early years of the disease. However, cha nges in the demographic composition of the patients in the 1990s drew interest in the subject, and mixed results were reported. A study conducted a few months before the in troduction of HAART between July 1995 and June 1996 reported that age did not emerge as a predictor of service utilization (Emlet & Farkas, 2002). In this st udy, medical service was meas ured by hospital admissions, hospital days, and physician visits and util ization was compared in three groups of patients: those 30 49 years, 50 59 years, and 60 years and older. Similarly, in the HAART era, “Hospital and out patient services” ( 2002) found that HIV patients aged 40 years or older did not differ significantly in total admissions, total hospital days, and median length of stay from their younger counterparts. However, other studies, especially thos e conducted in the HAART era, showed age differences in service utilization. Shapiro et al. (1999) showed that older patients used less emergency care, but had increased use of ot her healthcare services , such as inpatient hospitalizations. This finding was supported by other investigations. For example, it was found that compared to younger patients, older patients had increased use of primary care services (Palacio et al., 1999) , inpatient care (Menke et al ., 2000), and had longer lengths of hospital stays (Crystal et al., 1999). Also, a study involving 2,647 HIV infected patients who were seen in a public hospita l in Chicago between 1997 and 1998 found that older age predicted hospitaliz ations (Sherer et al., 2002) . In this study, older age

PAGE 54

41 significantly predicted both increased admissi ons and increased hospital days. Bozzette et al. (1998) used total cost of clinic care a nd hospital days to meas ure volume of service use. The authors showed that patients aged 50 years and older had a higher cost of care compared to younger patients ($1,469 vs. $1,380). In another study using information from the HCSUS, Kilbourne et al. (2002) reported that age was not associated with service use for most bothersome HIV symptoms. The high rates and volume of service uti lization among older patients cited above could be indicative of the comorbidity burde n on these patients due to a combination of HIV and age-related conditions. Although there are some studies (e.g., Grim es et al., 2002) that did not find age differences in service utilization among HIV/ AIDS patients in the post-HAART era, it is likely that as more patients attain long-term survival age differences in service use might occur. For instance, in a recent study that ex amined patients for a longer period of time, it was reported that the mean ag e of hospitalized persons with HIV increased from 38 years when HAART was introduced in 1996 to 41 years in 2000 (Hellinger, 2004). The increase in the age of patients is likely to influence the rates and costs of service utilization, with the influence expected to gr ow more evident as more patients attain longterm survival. 3.4.3.2 HIV risk factor and service utilization HIV/AIDS literature has indicated IDU, men who have sex with men (MSM), and heterosexual contact as important HIV risk factors in predicti ng service utilization. Compared to other risk fact ors, IDU has been found to be highly associated with increased rates and costs of service utili zation. Fleishman, Hsia, and Hellinger (1994) conducted a study of correlates of medical services utiliz ation among people with HIV

PAGE 55

42 from 25 U.S. cities with the highest repor ted number of AIDS cases in 1989. In this study, risk factor was coded as “IDU” or “ non-IDU” and it was shown that IDU history was significantly associated with increased rate of inpatient admi ssions. The IDUs were also shown to have longer inpatient stays, but the difference was nonsignificant when the analysis was controlled for other variables, in cluding insurance status and site of care. Cost of care among patients with different risk factors was also assessed during the HAART era. Gebo, Chaisson, Folkemer, Bar tlett, and Moore (1999) reported that injecting drug use was associated with higher inpatient hospital payments among Medicaid enrollees in Baltimore metropolitan area. However, similar to the study of Fleishman et al. (1994), in an adjusted anal ysis IDU was no longer associated with high healthcare payments. Bozzette and Hellinger (20 01) showed that monthly costs for clinic and hospital days were comparable betw een MSM and heterosexuals ($1,320 vs. $1,291), but was less compared to that of IDUs ( $1,564). Emlet and Farkas (2002) showed that IDUs were 4 and 1.6 times more likely to have excessive ho spital stays than heterosexuals and MSM, respectively. Major studies in the HAART era also repor ted differences in service utilization according to patients’ risk factor. Using HCSUS, Shapiro et al. (1999) found that compared to MSM, heterosexuals had lower odds (OR = 0.82, p<.05) but IDUs had higher odds (OR = 1.12, p<.05) of having emerge ncy department visits. Furthermore, IDUs were 1.2 and 1.41 times more likely to be hospitalized compared to heterosexuals and MSM, respectively. These findings support another major research that used data from 5,255 patients and found that injecting drug use risk was associated with increased number of hospital days (“ Hospital and outpatient servi ces,” 2002). In a study using VA

PAGE 56

43 patients, it was shown that IDUs were more likely to use emergency care and inpatient services, but were less likely to obtain outpatient services ex cept for mental health care (Menke et al., 2000). Our review has consistently shown that in jection drug use is associated with poor pattern of health care utilization among HIV/ AIDS patients with diffe rent risk factors. Some explanations have been suggested. Uphold and Mkanta (2005) pointed out that factors that have been histor ically associated with drug use, such as low socioeconomic status, increased disease severity, and compli cations from life-threatening comorbidities, could partially explain the poor pattern of care among IDUs. These factors are known to be negatively associated with optimal use of healthcare services, so their impact on IDUs may cause the poor pattern of service utilizati on. Also, it is possible that providers have been prescribing less intense care and anti retroviral therapies to IDUs because they believe that these disenfranchi sed patients may not adhere to the expensive treatments (Menke et al., 2000). 3.4.3.3 Gender and service utilization Typically, the overwhelming majority of ve terans with HIV disease in the VA care are men. Due to the small proportion of wo men in this population, our study involved only men in the assessment of the impact of long-term survival after AIDS on service utilization at the multivariabl e stage. Nevertheless, the group of women with HIV under the VA care offers a unique opportunity for re search on issues pertaining to HIV positive women in the country. There are about 400 wome n in the ICR as a whole, and this group is regarded as one of the largest groups of known HIV-positive women in the United States (CQM, 2001).

PAGE 57

44 3.4.3.4 Race/ethnicity and service utilization Racial differences in service utilization have been increasingly reported in all phases of the HIV epidemic. People with HI V from racial and ethnic minorities had increased rates of hospitaliz ation in the pre-HAART era (R osenblum et al., 1992), early HAART period (Sherer et al ., 2002), and post-HAART era (Shapiro et al., 1999; Fleishman & Hellinger, 2003). 3.4.3.5 Race/ethnicity and service use in the VA population In a study of racial disparit ies in survival and service use among veterans using a national administrative database , McGinnis et al. (2003) showed that white veterans were more likely to have had one or more non-VA visits (33%) than were black (27%) and Hispanic (25%) veterans; however, this diffe rence was not statistically significant. Of those on HAART, a significantly higher propor tion of black (86%) and Hispanic (90%) veterans than of white (80%) veterans had at least two visits at the VA clinic six months prior to the survey. Regardi ng survival rates, the study showed that HIV-infected minorities experienced poorer survival rates th an white veterans. The authors suggested that variation in surviv al benefits might be due to varia tions in comorbidities and severity of ADIs. A study using a unique data set that combined 462 male VA patients and 1,309 male patients from the ACSUS was used to inve stigate the association of race and service utilization (Menke et al., 2003). It was f ound that African-Americans and Hispanics were no more likely to be hospitalized than white s, but they had signi ficantly more total inpatient nights than white patients (1.4 and 1.3 more days respectively). In addition, African Americans had more emergency departme nt visits, but less mental health visits compared to whites. Hispanics were less likely than whites to have a mental health visit.

PAGE 58

45 3.4.3.6 Race/ethnicity and service us e in other population studies Most of the population studies consistent ly found that racial minorities had a less desirable pattern of service ut ilization in the HIV epidemic. For example, compared to whites, minorities had higher rates of emerge ncy care use (Shapiro et al., 1999), longer inpatient stays (Hellinger & Fleishman, 2001), a nd fewer outpatient visits (Kass et al., 1999; Sherer et al., 2002). In a study of hosp italization rates in Ch icago, HIV-infected persons were examined in two different years, 1997 and 1998 (Sherer et al., 2002). The authors reported that for each year, whites ha d significantly fewer admissions and shorter hospital stays than Hispanics or Africa n-Americans; also, African-American and Hispanic patients had significantly more vis its than whites in both years (mean visits 8.5 and 8.9 vs. 7.5 respectively). Utilization levels over time have also indicated racial differences in hospitalizations. Hellinger (2004) showed that while the rate of hospitalization in whites with HIV droppe d from 31% to 26% between 1996 and 2000, the rates among blacks increased from 46% to 51%. Behavioral and cultural factors may explain some of the cited racial differences in our review. For instance, HIV-associated stigma among ethnic minority communities is more likely to interfere with treatment programs compared to non-Hispanic whites. Studies have shown that compared to whites, African-American MSM were less likely to disclose their sexuality or associate with ot her gays, bisexuals, or lesbians (Kennamer et al., 2000). Because of this behavior, African Americans are less integrated into HIVrelated networks compared to whites, and thus less likel y to receive optimal health services. In addition, confounding effects of some factors, e.g., socioeconomic factors, such as education, income, or risk beha vior, may account for the racial and ethnic differences in service utilization. For exampl e, after controlling for injection drug use, it

PAGE 59

46 was shown in the analysis of health services use by urban women that the strong effect of race on service utilization was reduced (Solomon et al., 1998). There were no differences between African American and white women in use indicators such as having a usual doctor, recent emergency department visit, or recent hospitalization. Some studies found no racial differences in utilization. These included Kilbourne et al. (2002) who examined the role of sympto m in service utilization; Emlet & Farkas (2002) who investigated age differences in me dical service, psychosoc ial service, and inhome service; and Gebo et al . (1999) who evaluated cost of medical care in the HAART period. This part of the review has shown that racial differences persist in service utilization among HIV patients. Combined with reports of increasi ng new infections in minority women (Levine, 2002) and MSM (Har awa et al., 2004; Jimenez, 2003), it is likely that minorities will have increased need of healthcare services. This need has to be determined and appropriately addressed amid growing disparities in utilization outcomes. Also, because of the previously establishe d tendency of poor treatment adherence and delays in getting health care, people from racial minorities are likely to have increased use of inpatient care and e xperience longer lengths of hosp ital stays compared to whites even when experiencing long-term survival. 3.4.3.7 Summary on the predisposing variables Our review focused on predisposing factors a ssociated with use of health services among HIV/AIDS patients. The review can be summarized as follows: (i) race/ethnicity has played an important role in explaining the use of health services. Persistently, in all phases of the HIV epidemic, racial minorities had less favorable pattern of service use. They had excessive use of i npatient and emergency care compared to whites; (ii) HIV

PAGE 60

47 risk factor has also been a strong predic tor of service utiliz ation. Among HIV patients with known risk factors, IDUs had the poorest pattern of service use. These patients had increased use of costly inpatient care compar ed to patients reporti ng other risk factors; (iii) with time, patientÂ’s age has become an important factor of utilization. The effect of age has been mixed depending on the sample characteristics. However, as increasing numbers of HIV/AIDS patients attain long-term survival, combined with the growing interest among researchers on age-related issues , the role of age on service utilization will hopefully become clearer. 3.4.4 The Role of Enabling Variab les on Service Utilization 3.4.4.1 Insurance coverage, region of residence and service utilization Health insurance is one of the major enab ling factors in studyi ng health services utilization. However, veterans are regarded to have essentia lly the same level of health insurance coverage because mo st of them receive their health care under the same, equal access VA system. For this reason, McGinnis et al. (2003) have argued that HIV patients receiving care in the VA system represent a un ique and scientifically important group for the study of racial disparities in health care (McGinnis et al., 2003). The authors pointed out that one of the reasons for the uniquene ss of these patients is the possibility of studying the disparities without having to cont rol for health insurance status. Since the VA is regarded as an equal access system (M enke et al., 2000), othe r factors, e.g., region of residence, could be an important enabling factor. For example, any regional variation in service utilization associat ed with the process of care (e.g., providerÂ’s experience in HIV/AIDS) may be regarded as variation in servi ce use coming from an enabling factor. Shapiro et al. (1999) reported regional differences serv ice utilization based on the data from the HCSUS, a nationally representati ve sample of adults receiving HIV care in

PAGE 61

48 the U.S. Shapiro and his colleagues found th at patients in the Northeast region had increased use of emergency departments and hospitalizations compared to those in the West, South, and Midwest in that order. Thes e regional differences were similar to those reported by Bozzette et al. (1998) based on the cost of clinic and hospital days. It was found that monthly cost of care for patient s in the Northeast region was significantly higher than for patients in other regi ons ($1,739 vs. $1,315, Midwest; $1,296, South; and $1,242, West). In their study of the role of sympto ms on service use based on a national probability sample of 2,864 patients, Kilbourne et al. (2002) showed that patients in the Midwest had a significantly lower rate of hosp ital use compared to patients in the West. Patients in the South region had an increased ra te of use compared to those in the West, but the difference was not significant. Hell inger (2004) used Healthcare Cost and Utilization Project (HCUP) da ta to examine hospital utilization for 1996 and 2000 among HIV patients in eight states California, Colorado, Florida, Kansas, New Jersey, New York, Pennsylvania, and South Ca rolina. The data for both years showed that patients in New York and New Jersey had the longest hospital length of stay, while those in Colorado had the shortest stay. At the multivariate stage, data from both years were combined and it was shown that being a pati ent in New York predicted longer hospital stay compared to all other states. Among the cited studies, Shapiro et al . (1999) and Kilbour ne et al. (2002) controlled their studies for clin ical and demographic factors, but the regional differences in service use persisted. This observation ma y suggest that factors other than patient

PAGE 62

49 characteristics, such as process of care, may be the source of regional variation in service use. 3.4.4.2 Summary on the enabling variables We discussed only one enabling factor, i.e., region of residence, for the purpose of our study. The review consisten tly showed that patients resi ding in the Northeast region had increased use of emergency department a nd inpatient care. Patients in New York, an important state in the Northeast region, had in creased lengths of hospital stay compared to other patients from states in the No rtheast and other regions of the country. 3.4.5 The Role of Need-related Variables on Service Utilization 3.4.5.1 AIDS-defining illnesses and other comorbidities Improvement in the management of ADIs si nce the onset of the epidemic has been an important factor in reducing rates of health service use, mortality, and cost of care in HIV/AIDS patients. In the early days, tran smission of HIV infection was related to homosexuality and injection drug use. During this period, the leading cause of morbidity was infection involving the central nervous system, the respiratory system, and the lymphatic system; correspondingly, use of nur sing care was considered important for infection prevention and control (Ungvarski, 1984). In the late 1980s and early 1990s the number of AIDS patients continued to ri se and predominant care for HIV-related illnesses took place in the emergency departme nt and intensive care units (Markson et al., 1998; De Palo et al., 1995). PCP and respirator y failure continued to occur as the most common illnesses (Talan & Kennedy, 1991; Wachter et al., 1991). Reports of favorable outcomes in the management of AIDS-relate d illnesses from the Intensive Care Unit (ICU) resulted in the increased rates of using the ICU for the management of the illnesses. Some of these reports included that of Montaner et al. (1991) who found that a

PAGE 63

50 major reduction in mortality had occurred among patients undergoing mechanical ventilation for PCP in the ICU. In another study, Rogers et al. (1989) asserted that ICU admission was beneficial to AIDS patients becau se of the increased proportion of patients who had prolonged survival after discharge. In the pre-HAART era, PCP continued to be the most preval ent illness among HIV patients, and research focused on how it infl uenced service utilization. Bastian et al. (1993) examined gender differences among patie nts with hospital admission for their first episode of PCP. The authors showed that, compared to men, women with PCP were less likely to be white or have private health in surance and more likely to be admitted through the emergency department. Skiest, Rubinstie n, Carley, Gioiella, and Lyons (1996) used the interaction effect of como rbidity and age to examine prevalence of ADIs and rates of service use among 129 HIV patients in two age cohorts between 1986 and 1993. It was found that there were no significant differe nces in ADIs between the two cohorts. Furthermore, PCP was the most common illn ess in both cohorts, while esophageal candidiasis and MAC were th e second most prevalent fo r older and younger cohorts respectively. Regarding service use, the ol der cohort had both significantly more nonADIs hospitalizations (12.9 vs. 8.1 per 100 patient months) and ADIs (13.4 vs. 9.2 per 100 patient months). Bennett et al. (1995) examined racial di fferences in hospitalizations for PCP between VA and non-VA patients. There were no racial differences among VA patients in the use of inpatient services (bronc hoscopy and anti-PCP medications), but among non-VA patients it was shown that minorities were less likely to us e the services. The

PAGE 64

51 authors reported that the racial differences in service use for PCP among non-VA patients were accounted for by the differences in hea lth insurance and hospital characteristics. 3.4.5.2 The role of non-AIDS-defining comorbidity In the HAART era comorbidities became mo re complex because of the increasing occurrence of medical comorbidities or non-ADIs among AIDS patients. The medical comorbidities increased with appearance of long-term survival. During this period, researchers started to view HIV as a ch ronic medical condition. Previously reported illnesses, e.g., PCP (Uphold et al., 2004) and pneumonia (Sureka et al., 2004) still occurred in the post-HAART era, but in lowe r proportions. Literature showed that there was an increase in the prevalence of me dical comorbidities among HIV/AIDS patients, and these comorbidities were associated with service utilization. For example, Paul et al. (2002) reported that despite la rge decreases in opportunistic infections, hospitalizations among HIV patients have been stable due to in creases in conditions such as hepatitis, cirrhosis, and cellulitis. Furtherm ore, Kaplan et al. (2000) sh owed that decreases in the most common ADIs, including PCP, esophage al candidiasis, and MAC, were more pronounced when HAART was introduced in th e medical care. However, Kaplan and colleagues also pointed out th at due to HAART resistance and coinfections, it is possible that hospitalizations may increase. Gebo et al. (2003) conducted a prospec tive cohort study of 3,730 HIV patients who were longitudinally followed between 1995 and 2000 to determine the effect of hepatitis C virus (HCV) infection on hos pitalization. The authors show ed that hospital rates for patients who were HCV-positive decreased between 1995 and 1997 from 55.4 to 43.5 per 100 patient-years, but increased between 1997 and 2000 from 43.5 to 62.9 per 100 patient-years. In addition, the data showed that HCV infection was a significant predictor

PAGE 65

52 of hospitalization. Some other recent studies ha ve not directly discu ssed health services but have reported increases in non-ADIs th at might be associated with quantity of hospital services. In a French study of hospitalized HIV-in fected patients during the first six months of 2000, it was shown that 535 of 62,000 patients died and the cause of death was related to complications from HIV in 51% of deaths (Lowden et al., 2002). Furthermore, the most frequent non-AIDS rela ted causes of death were hepatitis C virus infection (10%) and non -AIDS defining cancer (9%). Also, in two small studies from Boston (n = 22) and Madrid (n = 20), relativ e mortality from liver disease reached fifty percent (Bica et al., 2001; Martin-Carbonero et al., 2001). Despite the growing evidence that AIDS pa tients now live longer, there is still no known cure yet for HIV. It is likely that the longer the patients liv e the higher would be the chances of experiencing psychiatric cond itions because of both prolonged use of medications and despair. Conditions such as depression and anxiety may be highly prevalent in AIDS patients and could create an influence on servi ce utilization. Hoover, Sambamoorthi, Walkup, and Crystal (2004) s uggested that psychi atric comorbidities might increase the time required to treat a nd discharge HIV/AIDS patients. Hoover and his colleagues reported that from 1990 to 1992 nearly 30% of AIDS inpatient hospital admissions involved patients with psychiatri c comorbid conditions. In addition, Cheng, Mijch, Hoy, Wesselingh, and Fairley (2001) found that psychiatric diagnoses were associated with increased risk of more than three months hospitaliza tion for patients with HIV/AIDS. We have shown that comorbidity is a majo r need factor in service utilization among AIDS patients. As AIDS patients experience long-term survival, comorbidity will be

PAGE 66

53 even more important need factor. Assessmen t of how it influences utilization while taking into account important patient factors such as age and race would be useful. For example, Stoff, Khalsa, Monjan, and Port egies (2004) argued that among older patients with prolonged survival, non-HIV related co morbidities now constitute at least an equivalent disease burden as HIV-related cond itions. It is therefore important to establish the prevalence of comorbidity using the most recent data and assess how it may influence both quantity and type of hea lthcare service provided to AIDS patients with long-term survival. 3.4.5.3 Influence of immune status and vira l status on service utilization CD4 cell count is the most commonly used marker to determine the progression of HIV. The purpose of determining the CD4 count is to measure the strength of a personÂ’s immune system if he has been diagnosed w ith HIV. For these reasons, levels of CD4 count are used to guide clin ical and therapeutic manageme nt of HIV-infected persons (Fernandez-Cruz et al., 1990). One of the criter ia for an AIDS diagnosis is a CD4 count of below 200 cells per cubic milliliter of blood (o r cells per mL). At this point there is a great risk for developing oppor tunistic infections like PCP. CD4 cell count has been extensively used to examine service use am ong HIV/AIDS patients. Viral load measures amount of the HIV RNA virus in blood, and it is usually reported as copies of HIV in one milliliter of blood (or copies per mL). If the viral load measurement is high, it indicates that HIV is reproducing and that the disease will likely progre ss faster than if the viral load is low. A high viral load can be anywhere from 50,000 to 100,000 copies per mL and can range as high as one million or more. Research has found that a decrease in viral load is associated with an increase in CD 4 cells (Montaner & Harris, 2002). This study

PAGE 67

54 used CD4 cell count as a need variable. The role of CD4 cell count on service utilization is reviewed next. A study of determinants of hospital admission among HIV-positive patients in British Columbia reported that hospital admission was associated with high viral load, low CD4 cell count, and AIDS diagnosis (W eber et al., 2000). However, at the multivariate stage, this study of 947 patients showed that only having high viral loads predicted hospital admission. In other studies based on the cost s of service Bozzette et al. (2001) showed that having AIDS and low CD4 cell count was associat ed with increased clinic and hospital costs, Gebo et al. (1999) showed that nonuse of HAART was associated with increased inpatient costs at all levels of CD 4 cell count (< 50, 50-200, and 200-500 cells/L), and Bavry and Mauldin (1 996) reported that the total cost of AIDS care (inpatient, hospitalization, and physician co st) was highest for the sickest group of patients in the study (those with CD4 ce ll count<200, having AIDS or deceased). Investigations in the HAART era have also reported on the role of immune and viral indicators on service use. The HIV Res earch Network showed that declining CD4 level and increasing viral load were both a ssociated with higher hospital and outpatient utilization (“Hospital and outpatient services ,” 2002). Specifically, it was reported that outpatient visit rates ranged from 9.7 to 13.2 per person annually as the CD4 declined from higher than 500 to less than 50 cells/L . Kilbourne, Herndon, Andersen, Wenzel, and Gelberg (2002) used a national probabil ity sample of 2,864 from the HCSUS, and reported that symptom intensity score, cough, and low CD4 level predicted increased use of health services. Emlet and Farkas (2002) found that having AIDS predicted increased physician visits, while death was associat ed with increased hospital admissions.

PAGE 68

55 3.4.5.4 Summary on need variables Overall, most studies examining service ut ilization have found need-related factors such as symptoms, presence of ADIs, medical comorbidities, viral load, disease stage, and CD4 cell count to be the majority of pr edictors of the use of health services. Investigators have extensively explored th e influence of HIV and non-HIV comorbidities on service utilization. In additi on, studies that used different sample characteristics and methodology have consistently found that CD4 cell count was an impor tant predictor of service utilization. Because HIV stage reflects the severity of the disease, disease stage was an important predictor of the types of se rvices used by HIV/AIDS patients. Presence of ADIs, number of medical comorbidities, diagnosis of psychiatric comorbidity, and CD4 cell count have been used as need variables in this study.

PAGE 69

56 CHAPTER FOUR RESEARCH OBJECTIVES AND METHODS 4.1 Introduction This dissertation examines th e role of long-term survival after AIDS diagnosis on service utilization among AIDS patients who use antiretrovira l therapy. The study population consisted of AIDS patients receiv ing health services in the VA facilities nationwide. Although these patients may have received care from non-VA facilities, our study is based on the information gathered solely from the VA healthcare system. Veterans seeking care in the VA system have to be enrolled to receive care each year based on the priority groups. The prio rity groups (Appendix B) are based on serviceconnected disabilities (if any), income, and ot her factors. For example, priority groups 1 6 include veterans with service-connected disa bilities, low-income veterans, and veterans in special categories (e.g., former prisoners of war). Priority group 7 is comprised of veterans without compensable service-connect ed disabilities and with incomes above the statutory threshold for free care and who agr ee to pay specified co-payments. Under the priority system, the Secretary of Veterans Affairs decides each year whether the VAÂ’s medical budget is adequate to serve veterans in all priority groups w ho seek care. If not, those in the lowest priority groups (P7 and P8) would be the first to lose access to care. Although the priority groups indicate the ab ility of veterans to receive care, once enrolled to receive care, all veterans have equal access to all th e services in the VA system. Thus, AIDS patients involved in th is study are a part of patient population

PAGE 70

57 enrolled under the VA system having equal access to healthcare services in the VA facilities nationwide. We hypothesized that patients with long-term survival would have greater use of healthcare services. We conceptu alized long-term survival afte r AIDS diagnosis as a need factor in this dissertati on research, and based on the ABM framework, we hypothesized that AIDS patients with long-term survival would have higher rates of outpatient and inpatient visits and longer hospital stays. We examined the impact of long-term survival on service utilization controlli ng for predisposing factors (race , HIV risk factor, age, and military service period); enabling factor (regi on of residence); and other relevant need factors (i.e., presence of ADI s, diagnosis of psychiatric comorbidity, number of medical comorbidities, and CD4 cell count). 4.2 Objectives The objectives of this study were: To identify correlates of long-term surviv al after AIDS diagnosis among veterans who received healthcare servic es in the VA system; and To examine the impact of long-term surv ival after AIDS diagnosis on the quantity and types of healthcare servic es used in the VA system by veterans with history of HAART. 4.3 Research Questions and Hypotheses The long-term objective is to promote effi cient use of healthcare services and provision of suitable treatment for a growing population of AIDS patients with long-term survival. We explored the following se t of research questions in addressing the objectives of our study. For each research question, a corresponding set of relevant hypotheses being tested is presented.

PAGE 71

58 4.3.1 Research Question I and Hypotheses 4.3.1.1 Research question I What are the sociodemographic and clinical correlates of long-te rm survival after AIDS diagnosis among patients using HAART? 4.3.1.2 Hypotheses White race, younger age, non-IDU risk factor s, and living in the West region are associated with better long-term survival rates following the diagnosis of AIDS; and Absence of ADIs, absence of psychiat ric comorbidity, and having less medical comorbidity are associated with better long-term survival rates following the diagnosis of AIDS. 4.3.2 Research Question II and Hypotheses 4.3.2.1 Research question II Among patients with psychiatri c and medical comorbidities, are long-term survival, race, HIV risk factor, age, and ADIs associat ed with their use of healthcare services? 4.3.2.2 Hypotheses Long-term survival following AIDS diagnos is, black race, IDU risk factor, older age, and presence of ADIs are associated with greater rates of outpatient alldiagnosis visits, inpatient all-diagnos is visits, and all-diagnosis LOS; Long-term survival following AIDS diagnos is, black race, IDU risk factor, older age, and presence of ADIs are associated w ith greater rates of psychiatric outpatient visits, psychiatric inpatient visits , and longer psychiatric LOS; and Long-term survival following AIDS diagnos is, black race, IDU risk factor, older age, and presence of ADIs are associated with greater rates of medical outpatient visits, medical inpatient visi ts, and longer medical LOS. The hypotheses formulated under the two re search questions are based on two rationales. First, increased incidence of non-ADIs occurring in the post-HAART era indicates that most patients with long-term survival would have more psychiatric and medical comorbidities that are likely to cr eate more use of psychiatric and medical

PAGE 72

59 services. In addition, a patientÂ’s sociodemograp hic and clinical factors may contribute to greater use of the services. For instance, highe r rates of severe comorbidities that might occur to black patients who traditionally ha ve added factors of poverty and poor access to health services may result in a poor pattern of service util ization (e.g., higher rates of inpatient visits and longer LOS). Thus, in general, comorbidity has the potential to increase the rates of both the us e of health services and deat hs, leading to shorter lengths of survival among subgroups of patients w ho had ever received an AIDS diagnosis. Second , research has shown in the history of the HIV epidemic that racial minorities have poor patterns of using HAART and HIV medical care. For this reason, we expect that even when the majority of HIV/AIDS patients in th e country experience long term survival, racial minorit ies would still be facing a seve re course of the disease as a result of a history of poor management of the disease. Minorities with AIDS are therefore more likely to be increasingly vul nerable to health risk s occurring with longterm survival such as comorbidity and mort ality, and thus (i) ma y have poor survival rates compared to whites; and (ii) those who attain long-term survival may have greater use of services compared to whites. And follo wing racial differences, we expect that HIV risk factor and age would also be associated with se rvice utilization. 4.4 Data and Methods The primary purpose of this research is to examine the impact of long-term survival time after diagnosis of AIDS on th e use of healthcare se rvices among veterans receiving care in the VA system. Long-term su rvival after AIDS was conceptualized as a need factor for the purpose of this dissertation. To accomplish this research, the most current and adequately available data repres enting this veteran population was sought to

PAGE 73

60 examine the relationship between outpatient visits, inpatient visits, and LOS, and veteransÂ’ predisposing, enabling, and need factors. 4.4.1 Source of Data and Availability The source of data for this research wa s the national ICR, which is owned by the Veterans Health Administration (detailed desc ription of ICR is given in Chapter Two). We determined that information available in the ICR was adequate to meet the objectives of this study and was used to address the research questions. The ICR contains national utilization and patient information for HIV-inf ected veterans who rece ive health services in the VA facilities. The information is amenab le to investigation of service utilization with the application of the A ndersen Behavioral Model. The ICR database is publicly held, and is accessible through PHSHG data access policy. We followed all the necessary pro cedures for the acquisition of the data, including: (i) application for the data th rough the local VA Medical Center, (ii) study approval by the Institutional Review Board (IRB), University of Florida, (iii) study approval by the VA Subcommittee for Clinical Investigation (SCI), (iv) approval by the VA Research and Development Committee (R&D), and (v) collaboration between the principal investigator and officials at the national ICR located at Palo Alto VA Medical Center, California, for the release of the data . A completely de-identified data (free of all patient identifying information in accordan ce to Health Insurance Portability and Accountability Act [HIPAA] privacy policy) was obtained from the ICR through these steps. 4.4.2 Study Design A cross-sectional study of patients experiencing long-te rm survival after AIDS diagnosis was conducted. For the purpose of captu ring the effects of l ong-term survival in

PAGE 74

61 the study, we used the ICR data for the cale ndar year 2003. We select ed to use the data from this year because it had the latest and complete information on healthcare service utilization among veterans with AIDS recei ving care from the VA system. We analyzed the data in accordance with the ABM fram ework to meet the objectives of the study. 4.4.3 Subjects The study participants (units of analysis) in this study were veterans who had ever received an AIDS diagnosis and who used th e VA system for their health care in the calendar year 2003. The specific inclusion crite ria formulated to obtain patients with long-term survival were: Patient having a known date (year) of AIDS diagnosis; Had evidence of using HAART; and Adult, aged 18 and above. 4.5 Variables in the Study 4.5.1 Dependent Variables We examined three dependent variables. One variable, number of outpatient visits, measured the use of outpatient medical care. Two variables, hospital admissions and LOS, measured the use of inpatient medical care. Outpatient Visits were operationalized as the number of distinct visits that include both HIV primary and specialty care visits a nd psychiatric visits during 2003. Outpatient visits for psychiatric care we re distinguished from outpatie nt medical (non-psychiatric) visits on the basis of ICD-9 codes for out patient care used by the patients. Inpatient Visits were defined as the total number of hospital admissions for medical care and psychiatric care a patient with AIDS made during 2003. Admissions to psychiatric services were distinguished fr om admissions to medical (non-psychiatric) services on the basis of ICD-9 codes for inpatient care provided to the patients.

PAGE 75

62 LOS: Defined as the total number of days a patient with AIDS was admitted in a hospital all year for medical or psychiatric care. LOS was distinguished for psychiatric and non-psychiatric care. In the ICR databa se, LOS was computed as discharge date minus admission date + 1. All dependent variables were assessed as count variables. To obtain meaningful and interpretable results reflecting use of medi cal and psychiatric care for AIDS patients, our analysis excluded visits and hospitalizati ons for surgical and radiological procedures. 4.5.2 Independent Variables Table 4.1 provides a list and operational definitions of all independent variables used in the study. The variables are arra nged according to the framework of ABM. Table 4.1. Independent Variables Including Oper ational Definitions Arranged According to the Andersen Behavioral Model of Health Services Use Predisposing variables Operational definitions RACE\ETHNICITY Categorical Variable: Race\ ethnicity as reported in patient record: 1=White (Not Hispanic), 2=Black (Not Hispanic), 3=Hispanic, and 4=Other Races (American Indian s and Alaska Natives, Asian Americans and Pacific Islanders). RISK FACTOR* Categorical Vari able: Risk factor reported in patient records: 1=IDU, 2=MSM, 3=Heterosexual, 4=Other Risks (Transfusion, Clotting factor, Healthcare exposure), 5=Un known Risk (Starting 1999, risk factor was not recorded in the ICR). PERIOD OF SERVICE Categorical Variable: Ve teranÂ’s first recorded period of military service: 1=Vietnam (8/5/64 to 5/7/ 75), 2=Post-Vietnam/Persian Gulf War (5/8/75 to 8/1/90 and 8/2/90 to -), 3=Korean/Post-Korean War (6/27/50 to 1/31/55), and 4=Others. AGE AT AIDS DIAGNOSIS Continuous Variable: Age of patient at the year of AIDS diagnosis, defined as the difference between the year of AIDS diagnosis and year of birth (both reported in the ICR), measured in years. Enabling variables REGION Categorical Variable: Region of residence in U.S. based on the location of the VA facility obtained from the 5-digit zip code of the VA facility where care was receiv ed in the year under study, 1=West, 2=Midwest, 3=South, and 4=Northeast.

PAGE 76

63 Table 4.1 continued Need variables PRESENCE OF ADIs Dichotomous Variable: (Had ADIs=1; Had No ADIs=0) indicating diagnosis of any AIDS-defining illness during the year under study, ICD-9 codes were used to identify patients with a diagnosis of ADIs as listed in Appendix A. PSYCHIATRIC COMORBIDITY Dichotomous Variable: (YES=1; NO=0) indicating diagnosis of comorbid psychiatric conditions during the year under study based on the search of the following ICD-9 codes for mental disorders: 295-Schizophrenic Disorders; 296-Affective Psychoses; 297Paranoid States; 301-Personality Disorders; 302-Sexual Disorders; 306-Psychophysiologic Disorders; 307-Special Mental Symptoms Not Classified Elsewhere; 308Acute Reaction to Stress; 309Adjustment Reaction; 311-Depressi ve Disorder; and 780 Alterations of Consciousness. NUMBER OF MEDICAL COMORBIDITIES Continuous Variable: Number of distinct medical comorbidities diagnosed per AIDS patient during the year under study, as determined through the ICD-9 codes (excluding codes for ADIs and psychiatric comorbidities). LONG-TERM SURVIVAL Categorical Variable: Months of survival after AIDS diagnosis through December 31, 2003, for vete rans who had outpatient or inpatient care in the VA system in the year 2003. Categorized as long-term survival attained =1 and not attained=0. Long-term survival is defined as attainment of survival time equal to or longer than the median survival time (median was 72 months or 6 years). DEATH Dichotomous Variable: (Deceased =1; Alive =0) death that occurred in the year 2003 to AIDS patients who also used health services during the same year. Death was used as an indicator of severity of AIDS for patients with long-term survival. CD4 CELL COUNT Continuous Variable: A meas ure of CD4 cell count in cells/mm3 as reported based on the latest laborat ory test results for patients who used outpatient or inpatient care in 2003. * Many HIV patients have multiple risk factors, but on ly one risk factor per patient was indicated in the ICR database. 4.6 Methods of Analysis Analysis was conducted in three stages. Stage one consisted of the descriptive statistics to describe the sample characteris tics of the patients included in the study. Stage two consisted of the development of a multivariable logistic model of the correlates of long-term survival after AIDS diagnosis. Finally , stage three consisted of the fitting of a multivariable negative binomial regression m odel to determine the impact of long-term

PAGE 77

64 survival on the use of outpatient visits, inpa tient visits, and LOS for specific diagnoses among AIDS patients. 4.6.1 Stage One: Methods for the Description of the Characteristics of the Sample of Veterans with AIDS who participated in the Study We aimed to describe the important charac teristics of the patients with AIDS who were included in the study base d on the inclusion criteria. First, descriptive statistics were computed for all patients. The description of the sample was based on the mean values of outpatient visits, inpatient vi sits, and LOS based on the cate gories of the categorical independent variables (race, risk factor, milita ry period of service, region, duration of survival, presence of ADIs, and presence of psychiatric comorbidity). For continuous independent variables, age, number of medi cal comorbidities, and CD4 cell count, the following procedures were followed: (i) range , mean, and standard errors of the mean were computed for each variable; then, base d on clinically meaningful cut-off points, CD4 levels were categorized, a nd (ii) mean values of outpati ent visits, inpatient visits, and LOS were assessed based on these categorie s to complete the sample description. Differences in the number of outpatient vi sits, inpatient vis its, and LOS among the groups of independent variables were assess ed by Analysis of Variance (ANOVA) and ttest procedures using version 13.0 of SPSS for Windows Statistical Package (SPSS Inc, Chicago, IL). 4.6.2 Stage Two: Methods for Examining th e Correlates of Long-Term Survival After AIDS Diagnosis In this stage of analysis we investigat ed correlates of long-term survival among veterans with AIDS and use of HAART. Th e dependent variable was an indicator (dichotomous) variable that was coded ‘1’ if patients attained long-term survival according to the median cut-off definition and was coded ‘0’ if patients did not attain

PAGE 78

65 long-term survival. The independent variables were race, HIV risk factor, age at AIDS diagnosis, region, diagnosis of ADIs, number of medical como rbidities, and diagnosis of psychiatric comorbidities. All these vari ables were used as defined on Table 4.1. To determine the correlates of long-te rm survival, we f itted a multivariable logistic regression (Hosmer & Lemeshow , 2000) model by regressing the dependent variable on the above-listed independent variab les: race, HIV risk factor, age at AIDS diagnosis, region (sociodemographic factors); and diagnosis of ADIs, number of medical comorbidities, and diagnosis of psychiatric comorbidities (clinical factors). Referent categories were appropriately selected for each categorical variable before running the analysis. The estimated logistic regression co efficients were assessed to determine the significance of the variables. Sta tistical significance was set at p 0.05. The sign of the coefficient was assessed to determine the eff ect of the variable on long-term survival. Finally, the exponent of the estimated l ogistic regression co efficients, exp { }, was assessed to determine the odds of the variable predicting patient attaining long-term survival and to address Research Question I. We used version 13.0 of SPSS for Window s Statistical Package (SPSS Inc, Chicago, IL) to fit the logistic regression m odel for the assessment of the correlates of long-term survival after AIDS diagnosis. 4.6.3 Stage Three: Methods for Examining the Impact of Long-Term Survival After AIDS Diagnosis on The Use of Heal thcare Services Among Veterans Using HAART After examining the relationship between sociodemographic and clinical factors and long-term survival, this part of analys is was conducted to address Objective 2 and Research Question II of the study. That is, we assessed the impact of long-term survival after AIDS diagnosis (a need factor) on all-diagnosis utili zation, psychiatri c utilization,

PAGE 79

66 and medical utilization (all measured by outpati ent visits, inpatient visits, and LOS) for patients with AIDS in the year 2003. All three measures of utilizati on, outpatient visits, inpatient visits, and LOS were assessed as count variables. We applied a multivariable negative binomial regression model to examine the relationship between outpatient visits, inpatient visits, and LOS and the predisposing, enabling, and need variables listed and descri bed in Table 4.1. This analytical procedure was conducted to address Research Question II , that is, “Among patie nts with psychiatric and medical comorbidities, are long-term survival after AIDS, race, age, and ADIs associated with their use of healthcare se rvices?” All three hypothe ses formulated under this question were tested using the negative binomial regression model. We assessed each dependent variable using three models based on all-diagnosis util ization, psychiatric utilization, and medical utilization during 2003, as defined below. Model 1: All-diagnosis utilization combining psychiatric a nd medical utilization among AIDS patients for th e year under study; Model 2: Psychiatric utilization consisting of patients who had psychiatric utilization, i.e., psychiatri c outpatient visits and psychiatric hospital admissions; and Model 3: Medical utilization consisting of patients w ho had medical utilization, i.e., utilization for non-ADIs and non-psychiatric comorbidities. The negative binomial regression was select ed in this analysis based on both the nature of the dependent variab les and the purpose of the study. 4.6.4 Use of the Negative Binomial Model Based on the Nature of the Dependent Variables Service utilization was assessed as counts of outpatient visits, inpatient visits, and LOS for different diagnoses among AIDS patie nts with long-term survival. For this reason, traditional linear regression models were inappropr iate for our analyses because

PAGE 80

67 count data are typically highly skewed. The high degree of skewness of the count data prevents meaningful transformations that w ould allow application of linear regression analysis. Figures 4.1 and 4.2 show the histograms representing the three dependent variables used in this research. All histograms depict ch aracteristics of the count data distributions. The distributions had high frequencies on the va lue of zero, indicating that there were many counts of zero in the data. Al so, the frequencies for the counts decreased steadily to provide highly positively skewed di stributions. As a result, for meaningful and interpretable results from our analysis , we assumed that the data for all three dependent variables followed a negative binom ial distribution. This is a distribution of discrete variables derived by Greenwood and Yule (1920) that falls in a broader class of Poisson distributions. The negative binomial model is a nonlinea r regression model with the capability to adequately describe and fit discrete, skewed data. Also, the model predicts values that are restricted to non-negative numbers, i.e., it allows for our analysis to predict only values that are theoretically possible (non-negative) for outpati ent visits, inpatient visits, and LOS.

PAGE 81

68 250 200 150 100 50 0 Outpatient visits in 2003 1,400 1,200 1,000 800 600 400 200 0 Fre qu en cy Figure 4.1: Histogram of Outp atient Visits among Patients with Known Duration of Survival after AIDS 500 400 300 200 100 0 LOS in 2003 70 0 60 0 50 0 40 0 30 0 20 0 10 0 0 Fre qu en c y 25 20 15 10 5 0 Inpatient visits in 2003 60 0 50 0 40 0 30 0 20 0 10 0 0 Fr eq ue nc Figure 4.2. Histograms of (a) Inpatient Visi ts and (b) LOS among Patients with Known Duration of Survival after AIDS Diagnosis Moreover, in the event of di sproportionately occurrence of zeros such as indicated in our data, or occurrence of over-dispersi on in the count data (variance greater than mean, violation of the basic assumption of th e mean-variance equality under the Poisson distribution) the negative binomial model woul d still be capable of correctly fitting the data. That is, the negative binom ial model allows for the natura l spread of the count data

PAGE 82

69 and it fits the data without imposing any upper boundaries on the counts (Cameron and Trivedi, 1998). Accordingly, we used a negative binomial model according to the ABM framework to examine the relationship of the three measures of se rvice utilization and predisposing, enabling, and need factors. 4.6.5 Use of the Negative Binomial Model to Address the Purpose of the Study The negative binomial model falls within the framework of generalized linear models and has the following general form: ) ...... (3 3 2 2 1 1 0 k kX X X X f n where n represents the number of times a particular type of medical service was used (rate of utilization) a nd ƒ represents the negative binomial distribution. The vector represents parameters predicting amount of services utilized base d on the predisposing, enabling, and need factors repres ented by X. Based on this form of the model, our aim of determining predictors of serv ice utilization was satisfied. 4.6.6 Interpretation of the Regression Coefficients The regression coefficients ( s) produced by the negative binomial model represent changes in the incide nt rates (or expected counts). These are coefficients to be estimated. Because of the nonlinearity of the negative binomial distribution, the s are not directly interpretable. The interpretation of the coefficients is based on an exponential scale that transforms the s into incident rate ratios (IRR). Specifically, the IRRs are obtained by exponentiation of the individual regression coefficien ts, that is, exp ( ). By definition, the IRRs describe the change in th e rate of healthcare use (outpatient visits, inpatient visits, and LOS) associated with a one-unit increment in co ntinuous independent variable. Among categorical variables, a refe rent category was selected and the IRR was

PAGE 83

70 interpreted as a relative change in the rate of healthcare use for each category, compared with the selected referent category. For ease of interpretati on, the expression 100*{exp ( )-1} indicated the percentage change in th e rate of service u tilization for each unit increase in the independent variable. Corre sponding standard errors and significance levels for the IRRs were reported for each stag e of analysis. We used version 8.0 of Stata (Stata Corporation, College Station, TX) for running multivariable negative binomial regression models. 4.6.7 Assessment of the Regression Models An assessment of the regression mode ls was conducted to determine how effective the models described the correlate s long-term survival and the impact of longterm survival on the rates of outpatient visits, inpatient vis its, and LOS. For the logistic regression model of the corre lates of long-term survival the assessment included the Hosmer and Lemeshow goodness-of-fit test, the Cox & Snell R-square, and overall percent correct prediction. The Hosmer and Le meshow test was performed to determine whether the model provided the best fit for the data. The C ox & Snell R-square was used to determine the proportion of the variance in the model at tributed to the dependent variable (long-term survival after AIDS di agnosis). The overall percent correct prediction was used to generate the propor tion of patients classified as attaining long-term survival in the model. For the negative binomial models, the assessment included the likelihood ratio tests for overdispersion that compared the adequacy of the negative binomial model compared to the Poisson model for the data and Wald tests that were performed to test for the global significance of the IRRs in the model.

PAGE 84

71 4.6.8 Data Processing Prior to the Fi tting of the Regression Models Prior to fitting the regression models as described in the prec eding section, some procedures were undertaken to prepare the data for adequa te analysis. We conducted a procedure to evaluate the problem of multic ollinearity among the independent variables. 4.6.8.1 Multicollinearity The problem of multicollinearity arises when independent variables in a regression model are highly co rrelated. When running a regression analysis in the presence of multicollinearity, individual regression parameters will be biased, i.e., multicollinearity (i) leads to wrong magnitudes of a coefficient or making the magnitude meaningless; (ii) leads to an incorrect theoretical sign, a high F-value, and low t-values for some or all of the coefficients; (iii) aff ects the stability of the coefficients so that different parts of the data will give substantia lly different results; and (iv) leads to a very high R2 value while some or all of the coeffici ents are insignifican t. In addition, these problems become more serious when using sma ll samples with already limited statistical power. Therefore, multicollinearity may affect the data to the extent that it would be difficult to know the separate influence of each independent (or explanatory) variable. Different degrees of multicollinearity exist with all multivariate regression analysis. Thus, resolving the problem of multicollinearity is necessary only if the independent variables are hi ghly correlated. A rule of thumb mentioned by Farrar and Glauber (1967) states that if the correlat ion between two independe nt variables exceeds 0.8 or 0.9, then one should have multicollin earity problems. Bryman and Cramer (1997, p.257) suggest 0.80 instead of 0.90 as the thresh old: “The Pearson’s r between each pair of independent variables shoul d not exceed 0.80; otherwise the independent variables that

PAGE 85

72 show a relationship at or in excess of 0.80 may be suspected of exhibiting multicollinearity.” An examination of the correlation matrix is used to identify multicollinearity. The presence of high correlation, those of 0.80 and above, may be an indication of substantial multicollinearity in the data. We used 11 i ndependent variables in this study. Table 4.2 lists correlation coefficients between these va riables. The correlation coefficient indicates the strength of the association between these variables, and it is considered significant if p 0.05. Table 4.2. Correlation Coefficients and th eir Levels of Significance among the Independent Variables Military Period Race Risk Factor Age Region ADIs MC PC LTS Death CD4 Military Period 1.000 Race .025 .077 1.000 Risk Factor .141 .000 -.095 .000 1.000 Age .006 .684 -.019 .193 -.018 .306 1.000 Region .026 .073 -.065 .000 .080 .000 -.006 .697 1.000 ADIs .003 .809 .007 .608 .000 .993 .001 .928 .034 .019 1.000 MC -.031 .033 .036 .012 -.015 .395 .127 .000 -.010 .509 .199 .000 1.000 PC -.104 .000 .009 .520 -.076 .000 -.027 .062 -.038 .009 .136 .000 .198 .000 1.000 LTS -.041 .005 -.045 .002 -.075 .000 -.290 .000 -.029 .043 -.093 .000 .025 .084 -.004 .781 1.000 Death .009 .540 .029 .041 -.015 .405 .121 .000 -.015 .286 .157 .000 .079 .000 .047 .001 -.084 .000 1.000 CD4 .006 .693 -.031 .029 .017 .344 .004 .790 -.007 .623 -.030 .035 -.009 .522 -.020 .161 .035 .015 -.020 .171 1.000 Age: Age at AIDS diagnosis; ADIs: AIDS-defining Illnesses; PC: Psychiatric comorbidity; LTS: Longterm survival; MC: Medical comorbidity. Coefficients ar e significant if level of significance is equal to or less than .05

PAGE 86

73 There is significant correlation between al l the independent variables as listed in Table 4.2. Of the 55 coefficients, 29 were si gnificant but none had a value greater than 0.2. The highest coefficient was 0.199 betw een the number medical comorbidity and presence of AIDS-defining illness, and th e lowest was 0.003, between military period of service and presence of AIDS-defining illness. Thus, the highest coefficient of correlation is well below the cut-off point of 0.80 for the multicollinearity problem. Thus, we concluded that these correlati ons were not strong, and they were not expected to cause problems with multicollinearity in the regression models.

PAGE 87

74 CHAPTER FIVE RESULTS OF THE STUDY This dissertation research has two major objectives: to determine the correlates of long-term survival, and to examine the impact of long-term survival after AIDS diagnosis on outpatient visits, inpatient visits, and LOS among ve terans who use HAART and receive their healthcare services in the VA system. We assessed the relationship between long-term survival and outpatient and inpa tient visits and LOS based on the Andersen Behavioral Model of health services use. Th e predisposing variables were race, HIV risk factor, military service period, and age at AI DS diagnosis, while region of residence was the enabling factor. Need factors included pres ence of AIDS-defining il lness, presence of psychiatric comorbidity, number of medical comorbidities, duration of AIDS survival, death, and CD4 cell count. We assessed servi ce utilization in three ways: all-diagnosis utilization, psychiatric utili zation, and medical ut ilization. The dependent variables (or measures of utilization) we re outpatient visits, inpatient visits, and LOS. All three dependent variables were assess ed as count variables. There were 20,296 unique patients in the ICR during 2003. Of these, 16,043 (79%) were on HAART. Being on HAART indicates that a pa tient had at least one outpatient prescription fill for a Food and Drug Administration (FDA) approved antiretroviral therapy during 2003. Out of the 16,043 patients on HAART, 4,808 met our inclusion criteria, i.e., they ha d ever received an AIDS diagnos is and they were at least 18 years old in 2003. These patients represented about 30% of all veterans using HAART.

PAGE 88

75 Of the 4,808 patients with AIDS who met th e inclusion criteria of this study, 4,717 (98.1%) were men. This gender distribution is typical among patients who are treated in the VA system. The VA treats mostly men w ho have served in the U.S. military in different service periods. Thus, due to thei r small number, women were not included in the study sample. Only the 4,717 male vetera ns with AIDS who met the study criteria were included in all stages of analysis. 5.1 Descriptive Statistics The predisposing, enabling, and need char acteristics of the 4,717 AIDS patients who were included in the study were examined. Forty-four percent of these patients were whites, 45.9% blacks, 9.1% Hispanics, and less than one percent other races (Table 5.1). About 31% of the patients were MSM a nd 21% were classified as IDU. The distribution of risk factors s howed that 7.4% of the patients had other HIV transmission modes that included clotting factor a nd transfusion. Starting April 1999 the ICR discontinued collecting informa tion on patientsÂ’ risk factor. As a result, the ICR does not contain information on HIV risk factors for a large proportion of patients. For example, the database had no information on risk f actors for about 57% of the patients who received HIV care in the VA during fiscal year 2004. In our sample, about 22% of the patients had unknown HIV risk factor as a result of the omission of this data element in the ICR. Based on the distribution of the first milita ry service period, the data showed that half of the patients (50.7%) st arted their service in the Vietnam era (August 1964 to May 1975), while 41.4% started during the post-Vi etnam or Gulf war era (May 1975 to August 1990 and beyond). The mean age of the patients was 50 years (range, 26 83;

PAGE 89

76 SE=8.61), while the mean age at AIDS diagnosis was 45 years (range, 21 81; SE=9.15). A majority of the patients (45.3%) were from the South Region; others were from West (25.8%), Northeast (16.6%), and Midwest (12.3%). Slightly over half of the patients (51.6%) attained long-term survival following AIDS. By our definition, these are patients with survival time equal to or greater than 72 months, the median survival time after AIDS. Table 5.1. Characteristics of Veterans with AIDS by All-Diagnosis Utilization (N=4,717) Variables Outpatient Visits Inpatient Visits Length of Stay Predisposing N (%) Mean SE Mean SE Mean SE Race † White 2096 (44.4)29.54.763.09 .17 4.83.36 Black 2163 (45.9)34.03*.944.42* .20 8.44*.50 Hispanic 427 (9.1)36.39*2.003.87.40 6.85.91 Others 31 (.7)28.675.343.00*1.21 2.801.20 Min Max Mean SE Age, years 26 83 50.13 8.61 Age at AIDS, years 21 81 44.53 9.15 Risk Factor † Injection Drug Use (IDU) 992 (21.0)38.861.595.72 .38 10.66.85 Men who have Sex with Men (MSM) 1455 (30.8)28.63*.782.70* .16 3.61*.29 Heterosexual 863 (18.3)28.48*1.102.95* .24 6.04*.67 Others 348 (7.4)30.83*2.603.68*.47 6.86*1.03 Unknown 1059 (22.5)34.401.304.10* .28 7.53*.69 Military Service Period † Post Vietnam/ Gulf War 1951 (41.4)29.79.903.39 .18 6.20.44 Vietnam 2392 (50.7)34.66*.834.12* .19 7.26.44 Korean War 283 (6.0)31.081.993.70.45 5.63.91 Others 91 (1.9)23.35*3.722.82.74 3.801.08

PAGE 90

77 Table 5.1 continued Variables Outpatient Visits Inpatient Visits Length of Stay N (%) Mean SE Mean SE Mean SE Enabling Region † Midwest 580 (12.3)39.532.474.43* .396.94.80 Northeast 784 (16.6)34.461.424.58 .389.40.95 South 2136* (45.3) 30.96803.49 .176.72.44 West 1217*(25.8)29.48.923.42 .234.64.42 Need AIDS-Defining Illness (ADI) Had ADIs 969 (20.5)42.841.458.36 .3815.41.86 No ADIs 3748 (79.5)29.47*.622.58* .124.3*.28 Psychiatric Comorbidity ‡ Had psychiatric comorbidity 2305 (48.9)44.91.986.00 .2310.63.53 Had no psychiatric comorbidity 2412 (51.1)20.08*.531.63* .092.86*.24 Long-Term Survival ‡ Long-term survival attained 2435 (51.6)33.34.813.24 .165.23.36 Long-term not attained 2282 (48.4)31.01*.834.33* .198.17*.46 Death ‡ Living by the end of 2003 4236 (89.2)31.57.592.85 .114.99.26 Death in 2003 481 (10.2)37.90*2.3111.86* .6821.33 * 1.56 Min Max Mean SE Number of Medical Comorbidities 0.0 9.0 1.82 1.43 CD4 cell count 58.01431401.74 112.9 8 CD4 cell count categories † 50 to < 200 186 (3.9)35.663.398.02 1.2212.712.06 200 to < 350 626 (13.3)37.661.4613.11* .5424.51 * 1.40 350 to < 500 3628 (76.9)29.52*.561.99* .093.37*.22 500 and above 276 (5.9)52.80*4.992.68* .405.11*1.16† Based on ANOVA; ‡ Based on t-tests; * Different (p<0.05) from the first category

PAGE 91

78 Approximately 21% of the patients were diagnosed with at least one AIDSdefining illness during the y ear under study. A further anal ysis of the data on ADIs showed that the three leading diagnoses of ADIs reported for outpatient care were psychosis (encephalopathy) (ICD-9 code 298.9; 30% of all ADIs diagnoses received), herpetic disease (ICD-9 code 54.9; 10%); and Cachexia or wasting syndrome (ICD-9 code 799.4; 9%). The most fr equent ADIs that resulted in hospitalizations were PCP (ICD-9 code 136.3; 33%); psychosis (ICD-9 code 298.9; 11%); and cryptococcosis, a fungal infection (ICD-9 code 117.50; 9%). Almost half of the patients (48.9%) were diagnosed with psychiatric comorbidity and had psychiatric utilization during 2003. The most common psychiatric comorbidity was drug dependence (Opioid type depende nce), which was reported by 18% of all patients with psychiatric comorbidities. In addition, patients with psychiatric comorbidities consumed a mean number of 44.91 outpatient all-diagnosis visits, 6.00 alldiagnosis inpatient visits, and had a mean all-diagnosis LOS of 10.63 days. Overall, patients had a mean number of 1.82 (range, 0 9; SE=1.43) medical comorbid conditions (excluding ADIs and psychiatric comorbidities). A separate analysis showed that the top three medical comorbidities that were pres ented during the outpat ient visits were hypertension (ICD-9 code 401.9), hepatitis C in fection (ICD-9 code 70.51), and chronic renal failure (ICD-9 code 585.0), while data fo r inpatient visits for medical comorbidities showed that the most frequent medical condi tions that resulted in hospitalizations were hypertension (ICD-9 code 401.9), chronic hepa titis C infection (IC D-9 code 70.54), and hepatitis C infection (ICD-9 code 70.51).

PAGE 92

79 The mean number of CD4 cell count fo r the patients was 401.74 cells/mm3 (range, 58 1431; SE=112.98). The CD4 cell counts re ported in this study were based on the maximum counts for each patient, obtai ned from several tests made in 2003. After grouping the patients according to their CD4 ce ll count, we found that a majority of the patients (76.9%) had CD4 cell count between 350 and 500 cells/mm3, while only small proportions had either the lowe st CD4 cell count of 50 to 200 (3.9%) or the highest CD4 cell count of 500 or more (5.9%). Table 5.1 also shows that there were di fferences in predisposing, enabling, and need variables in all-diagnosis outpatient visits, inpatient visits and LOS among the patients. Hispanic patients had significantl y greater number of out patient all-diagnosis visits than patients of other races (Indian/Alaskan Natives), but had a comparable mean number of outpatient visits to blacks (36.39 vs. 34.03). Black patients had greater number of inpatient all-diagnosis visi ts and longer LOS compared to all other races. Patients with IDU risk factor had greater use of both outpa tient and inpatient al l-diagnosis services compared to patients with unknown or with other risk factors in the sample. With the exception of the difference between IDUs a nd patients with unknown risk factors in outpatient visits (p>.05), all ot her differences between IDUs a nd patients with other risk factors for all-diagnosis outpatient visits, i npatient visits, and LO S were statistically significant. While there were no differenc es in all-diagnosis LOS among patients according to their military service periods, the re sults showed that veterans who started to serve in the military during the Vietnam Er a had more all diagnosis outpatient and inpatient visits. Specifically, the mean num ber of outpatient visits for Vietnam Era

PAGE 93

80 patients was 34.66 vs. 29.79 for post-Vietnam/Gu lf war veterans, 31.08 for Korean War veterans, and 23.35 for veterans from other periods. The distribution of patientsÂ’ region and se rvice use indicated that patients from the Midwest Region had the highest rates of out patient all-diagnosis visits compared to patients from other regions. But patients fr om the Northeast Region had greater number of inpatient all-diagnosis visits and longer LOS. The data also revealed that patients who were diagnosed with ADIs had greater use of both outpatient and i npatient all-diagnosis care compared to patients who had no ADI s. These patients had 42.84 vs. 29.47 mean outpatient visits; 8.36 vs. 2.58 mean inpatient visits, and 15.41 vs. 4.39 mean number of inpatient days. The ADIs are displayed in detail in Appendix A together with their corresponding ICD-9 codes. Similarly, patie nts who had diagnoses of psychiatric comorbidities had significantly greater use of all-diagnosis care: 44.91 vs. 20.08 mean outpatient visits, 6.00 vs. 1.63 mean inpatient visits, and 10.63 vs. 2.86 mean LOS in days. Differences in the use of services were also observed according to the duration of AIDS survival. While patients who attained long-term survival had higher rates of alldiagnosis outpatient visits (33.34 vs. 31.01), patients who have not yet attained long-term survival had greater number of all-diagnosis inpatient visits ( 4.33 vs. 3.24) as well as LOS (8.17 vs. 5.23 days). Although these differen ces were statistically significant, they were not as large as those f ound in other factors such as pr esence of ADIs or presence of psychiatric comorbidity. Death in the y ear under study was included in our study as a measure of severity of HIV/AIDS. We found th at compared to patients who were alive at the end of 2003, patients who died had greater use of inpatient all-diagnosis care. They

PAGE 94

81 had more inpatient visits (11.86 vs. 2.85), a nd longer LOS (21.33 vs. 4.99) than patients who were alive by the end of 2003. Finally, the distribution of service use according to the levels of CD4 cell count showed that pati ents with CD4 cell count equal to or greater than 500 had greater use of outpatient all-di agnosis services, while those with CD4 cell count less than 350 had greater use of inpatient all-diagnosis care. 5.2 Correlates of Long-term Sur vival after AIDS Diagnosis Prior to running the multivariable mode l to assess the impact of long-term survival on outpatient visits, inpatient visi ts, and LOS, we conducted an analysis to identify correlates of long-term survival. This analysis was done using the multivariable logistic regression model with duration of survival after AIDS as the dependent variable, coded ‘1’ for patients with long-term surv ival and ‘0’ for patients who have not yet attained long-term survival. Long-term surviv al was defined as attainment of survival time after AIDS that is greater than or equa l to 72 months, the me dian duration of AIDS survival among the patients in the sample. Th e independent variables in the logistic model were divided into two groups of sociodemographic and clinical factors. Sociodemographic variables were race, age, HIV risk factor, milita ry service period, and region. Clinical variables were presence of ADIs, presence of psychiatric comorbidity, number of medical comorbidities, and CD4 cell count. Table 5.2 displays the results of logistic regression of the correlates of long-term survival after AIDS. Odds ratio (OR), 95% confidence intervals (95% CI), and significance levels are reported. The Hosmer and Lemeshow chi-square statistic was 34.46 (p<.001), indicating that the model provided a good fit for the data. In addition, Cox & Snell R-square was .276, indicating that the model increases its prediction power on long-term survival by at least 28% by using the fitted data, and last, the model

PAGE 95

82 correctly classified 74.6% of the patients according to whet her they had attained longterm survival or not. This score indicates a high predictive power of the fitted model. Table 5.2. Logistic Regression Results of the Correlates of Long-term Survival after AIDS Diagnosis (N=4716) Variable Odds Ratio 95% Confidence Interval p Race White (Referent category) Black Hispanic Other races 1.00 .85 .91 .85 .735 .992 .706 1.183 .385 1.879 .038 .494 .689 Risk Factor Heterosexual exposure (Referent category) Injection Drug Use (IDU) Men who have Sex with Men (MSM) Other risk factors Unknown risk factor 1.00 1.17 1.47 .59 1.22 .944 1.465 1.204 1.799 .441 .815 .992 1.516 .148 <.001 .001 .059 Military Service Period Post-Vietnam/Gulf War (Referent category) Vietnam Era Korean/Post Korean Other Era 1.00 8.10 109.37 5.64 6.614 9.941 69.712 170.207 2.833 11.239 <.001 <.001 <.001 Age at AIDS diagnosis .82 .814 .837 <.001 Region West (Referent category) Midwest Northeast South 1.00 1.11 1.46 .85 .880 1.407 1.168 1.827 .723 1.015 .371 .001 .074 Presence of AIDS-defining illness (ADI) Had no ADIs (Referent category) Had ADIs 1.00 .68 .576 .815 <.001 Presence of psychiatric comorbidity No psychiatric comorbidity (Referent category) Had psychiatric comorbidity 1.00 .89 .779 1.031 .125 Number of medical comorbidities. 1.18 1.122 1.240 <.001 CD4 cell count 1.01 1.004 1.005 <.001 The results in Table 5.2 also show that ra ce is associated with long-term survival after AIDS diagnosis. Our hypothesis that whit e patients have better long-term survival

PAGE 96

83 rates than black patients was supported. Black patients were found to have significantly lower odds than white patients of attaining long-term survival after AIDS diagnosis (OR=.85; 95% CI .735 to .992; p=.038). Patients who were classified as MSM (OR=1.47; 95% CI 1.204 to 1.799; p <.001) had higher odds of attaining long-term survival, while those patients with other risk factors such as transfusion and clotting factor had 41% lower odds (OR=.59; 95% CI .441 to 815; p< .01) of attaining long-term survival compared to patients with heterosexual exposur e. Compared to patients who served in the post-Vietnam and Gulf war eras, patients who served in the Vietnam era, Korean era and other eras had higher likelihood of achieving long-term surv ival after receiving AIDS diagnosis. Among these groups, the highest odds of a better long-term survival were found among veterans who served duri ng the Korean War (OR=109.37; 95% CI 69.712 to 170.207; p<.001). Logistic regression results al so showed that age at AIDS diagnosis was associated with long-term survival. Specifically, the mode l predicted that the odds of attaining longterm survival decrease by 18% for each one ye ar increase in the age at AIDS diagnosis (OR=.82; 95% CI .814 to 837; p<.001). Pa tients living in the Northeast Region (OR=1.46; 95% CI 1.168 to 1.827; p<.01) had hi gher odds of having long-term survival after diagnosis of AIDS compared to those living in the West Region. Considering the clinical f actors, our results showed that presence of ADIs was associated with decreased odds of long-te rm survival (OR=.68; 95% CI .576 to .815; p<.001), but diagnosis of psychiatric comorb idity was not associated with long-term survival. Also, patients who attained long-term survival had higher odds of having greater number of medical comorbidities (OR=1.18; 95% CI 1.122 to 1.240; p<.001). Finally, the

PAGE 97

84 odds of attaining long-term survival were shown to increase by 1% for each additional unit of CD4 cells (OR=1.01; 95% CI 1.004 to 1.005; p<.001). 5.3 Results of the Multivariable Negat ive Binomial Regression Model Results of the multivariable negative bi nomial regression on the impact of longterm survival on outpatient visits, inpatient visits, and LOS are presented within this section. The results are presented in three part s covering different type s of services; these are (i) all-diagnosis utilizati on (i.e., utilization for psychiat ric and medical comorbidities combined); (ii) psychiatric u tilization; and (iii) medical u tilization (i.e., utilization for non-ADIs, non-psychiatric comorbidities). ICD-9 codes were used to identify psychiatric and medical comorbidities presented at outpa tient and inpatient encounters. The three parts of analysis were used to test th e four hypotheses under Research Question II. 5.4 Results of the Negative Binomial Model for All-Diagnosis Utilization We fitted the negative binomial model using AIDS patients with information on outpatient visits, inpatient visi ts and LOS for all diagnoses in the year 2003. The aim of this analysis was to test the hypothesis that “Long-term survival following AIDS diagnosis is associated with greater rates of all-diagnosis outpatient visits, all-diagnosis inpatient visits, and all-dia gnosis LOS.” We examined the association after controlling for predisposing variables (race, HIV risk, age, and military period of service); enabling variable (region of residen ce); and need variables (pre sence of ADIs, presence of psychiatric comorbidity, number of medical comorbidities, death, and CD4 cell count). IRRs and p-values are reported for all pred isposing, enabling, and need variables. The Wald chi-square tests and likelihood ratio tests for overd ispersion for all models (outpatient visits, inpatient vi sits, and LOS) were significan t (p<.05), indicating that the

PAGE 98

85 IRRs were globally significant and that the negative binomial model provided the best fit for the data. Results presented in Table 5.3 show that black patients with AIDS had higher rates of outpatient all-diagnosis visits comp ared to white patients with AIDS (IRR=1.13; p<.001). The IRR value of 1.13 implies that th e model predicted a 13% increase in the number of outpatient visits among black pa tients with AIDS as compared to white patients with AIDS. Although there was no diffe rence in the number of inpatient alldiagnosis visits between blacks and whites, we found that being black was associated with longer all-diagnosis LOS (IRR=1.20; p<.01) compared to whites. In addition, Hispanics had higher rates of outpatient vi sits than whites (IRR=1.11; p<.01), but there were no other racial differences in the rates of healthcare use. Patients having IDU risk factor had higher rates of all-diagnosis outpa tient visits, inpatient visits and longer LOS compared to those reported with heterosexua l exposure. The largest difference in service use between these two groups was found in inpa tient all-diagnosis visits. IDUs had 40% more inpatient visits than heterosexuals. Patients reported with MSM risk factor had higher rates of all-diagnosis outpatient and inpatient visits compared to heterosexuals, but had shorter LOS (IRR=.80; p<.05). Similarly, patients with other risk f actors as well as those with unknown risk had higher rates of healthcare use compared to heterosexuals. In general, considering HIV risk factor, these results show that heterosexuals had the lo west rates of all-diagnosis outpatient and inpatient car e among all risk groups.

PAGE 99

86 Table 5.3. Negative Binomial Regression Result s for All-Diagnosis Outpatient Visits, All-Diagnosis Inpatient Visits, and All-Diagnosis Length of Stay among Veterans with AIDS Who Used Healthcar e Services in the Veterans Health Administration Variable Outpatient Visits (N =4711) Inpatient Visits (N=1397) Length of Stay (N=1375) RC = Referent Category IRR SE p IRR SE p IRR SE p Race White (RC) Black Hispanic Other races 1.00 1.13 1.11 1.19 .027 .046 .164 <.001 .007 .195 1.00 0.95 1.08 1.41 .044 .070 .292 .285 .554 .769 1.00 1.20 0.99 0.69 .074 .105 .274 .002 .981 .351 Risk Factor Heterosexual (RC) Injection Drug Use (IDU) Men who have sex with Men (MSM) Other risk factors Unknown risk factor 1.00 1.21 1.14 1.26 1.18 .042 .035 .060 .039 <.001 <.001 <.001 <.001 1.00 1.41 1.14 1.33 1.29 .083 .067 .116 .077 <.001 .020 .001 <.001 1.00 1.34 0.80 1.21 1.23 .113 .068 .151 .105 <.001 .012 .122 .012 Military Service Period Post-Vietnam/Gulf (RC) Vietnam Korean Era Other Era 1.00 0.79 0.52 0.70 .021 .030 .058 <.001 <.001 <.001 1.00 0.82 0.47 0.64 .038 .047 .101 <.001 <.001 .005 1.00 0.73 0.31 0.54 .049 .045 .124 <.001 <.001 .008 Age at AIDS diagnosis 1.03.001<.0011.03.002<.001 1.04.003<.001 Region West (RC) Midwest Northeast South 1.00 1.35 1.10 1.14 .051 .039 .030 <.001 .005 <.001 1.00 1.14 1.04 1.04 .074 .064 .051 .040 .526 .318 1.00 1.17 1.49 1.30 .112 .133 .094 .089 <.001 <.001 Presence of ADIs Had no ADIs (RC) Had ADIs 1.00 1.26.035<.001 1.00 1.25.048 <.001 1.00 1.51.086<.001 Had psychiatric comorbidity No (RC) Yes 1.00 2.02.045<.001 1.00 1.70.070 <.001 1.00 1.77.107<.001 Medical comorbidities 1.41.011<.0011.16.014<.001 1.12.021<.001 Long-Term Survival Not Attained (RC) Attained 1.00 1.24.029<.001 1.00 1.32.053 <.001 1.00 1.18.070.004 Death Alive end of 2003 (RC) Died in 2003 1.00 0.95.035.188 1.00 1.52.069 <.001 1.00 1.66.111<.001 CD4 cell count 1.01.000<.0010.99.000<.001 0.99.000.111

PAGE 100

87 The examination of patientsÂ’ use of hea lthcare services according to their military service periods showed that veterans with AIDS who served during the post-Vietnam Era or Gulf war (May 1975 to September 2003) had higher rates of all-diagnosis utilization compared to those who served during the Vietnam war or Korean conflict. The largest margin of difference in serv ice use was between the post-Vietnam /Gulf war veterans and those who served in the Ko rean conflict: IRR=.52 (outpatient visits), IRR=.47 (inpatient visits), and IRR=.31 (LOS) all with p<.001. The results also showed that each one-y ear increase in age at AIDS diagnosis increases the rates of both outpatient and inpatient all-diagnosi s visits by 3% and increases all-diagnosis LOS by 4%. Furtherm ore, patients who live in the Midwest, Northeast, and South regions were found to ha ve greater use of out patient all-diagnosis services compared to patients living in the West Region. However, patients in the Midwest Region (IRR=1.14; p<.05) had more inpatient all-diagnosis visits, while those in the Northeast (IRR=1.49; p<.001) and South (IRR=1.30; p<.001) regions had longer alldiagnosis LOS compared to patients in the West. Presence of ADIs among the patients was a ssociated with greater use of inpatient and outpatient all-diagnosis services. For ex ample, those who were diagnosed with ADIs had 50% longer hospital stays compared to those with no ADIs diagnoses during the year. Similarly, patients who had psychiatric comorbidities had overall greater use of services compared to those without psychiatric diagnoses . Patients with psychiatric comorbidities had more than double the number of outpatient all-diagnosis visits compared to those without psychiatric co morbidities (IRR=2.02; p<.001) and at least 70% more all-diagnosis inpatient visits or longer LOS. Moreover, number of medical

PAGE 101

88 comorbidities was also associated with all-di agnosis utilization of healthcare services. Each additional medical comorbidity a pati ent had increased outpatient visits by 41% (p<.001), inpatient visits by 16% (p<.001), and LOS by 12% (p<.001). Long-term survival impacted both outpatient care and inpatient all-diagnosis care. Specifically, our hypothesis that long-term su rvival after AIDS diagnoses would be associated with greater rates of all-diagnosis outpatient visits, inpatie nt visits, and longer LOS was supported. Patients who attained long-term survival had higher rates of outpatient visits (IRR=1.24; p< .001), higher rates of inpatient visits (IRR=1.32; p<.001), and longer LOS (IRR=1.18; p=.004) compared to patients who had not yet attained longterm survival. We also considered the experi ence of AIDS patients who used services in 2003 but died in the same year. The results s howed that patients who died had greater use of inpatient all-diagnosis se rvices; they had 52% more i npatient visits and 66% longer LOS compared to those who were alive at the end of 2003. These re sults indicate that patients experience severe conditions of their disease that involves intensive and longer inpatient care towards the end of their lives. La st in this section, th e results showed that higher CD4 cell count is associated with more outpatient all-diagnosis visits but fewer inpatient all-diagnosis visits and shorter all-diagnosis LO S. This results show that patients who are healthy, i.e., those with hi gher CD4 cell count, w ould have greater use of outpatient than intense a nd costly inpatient services. 5.5 Results of the Negative Binomial Reg ression for Psychiatric Utilization To test the hypothesis that long-term survival after AIDS diagnosis would be associated with greater rates of psychiatric outpatient visits, psychi atric inpatient visits, and psychiatric LOS, we fitted the multivariable negative binomial regression model considering only AIDS patients who received diagnosis of psychiat ric comorbidities and

PAGE 102

89 used healthcare services for those comorbid ities. We examined the association after controlling for predisposing variables (race, HIV risk, age, and military period of service); enabling variable (region of residen ce); and need variable s (presence of ADIs, presence of psychiatric como rbidity, number of medical co morbidities, death, and CD4 cell count). As in the previous section, we report IRR and p-values for each predisposing, enabling, and need variable used to adjust the analysis. Results fo r this analysis are shown in Table 5.4. Similar to the results of previous section, the Wald chi-square tests and likelihood ratio tests for overdispersion fo r all models (outpatient visits, inpatient visits, and LOS) were signifi cant indicating that the IRRs were globally significant and that the negative binomial model provided th e best fit for the data on psychiatric outpatient and inpatient care among AIDS patients. Among the racial groups, we found that black patients had higher rates of outpatients psychiatric visits than white patients (IRR=1.51; p< .001), while Hispanics had shorter LOS for psychiatric conditions compared to whites (IRR=.42; p<.001). There were no racial differences in the number of psychiatric hospitalizations. Of the patients with known risk factors, the re sults showed that patients re ported with IDU risk were associated with greater use of inpa tient and outpatient psychiatric care. For instance, compared to heterosexuals, IDUs had more out patient psychiatric visits (IRR=1.81; p<.001) and inpatient ps ychiatric visits (IRR=1.57; p<.001).This finding supports that of Marimoutou et al. (2003) who found that IDUs had higher rates of psychiatric hospitalizations in a study of the causes of hospitalizations and mortality among HIV patients with history of injection drug use. Use of psychiatric services was also associated with military period of servi ce. Patients who served in the Korean conflict

PAGE 103

90 (relatively older in the sample) had lower ra tes of both psychiatric outpatient (IRR=.46; p<.001) and inpatient visits (I RR=.33; p<.001) compared to those who served in the postVietnam/Gulf war era. Table 5.4. Negative Binomial Regression Resu lts for Psychiatric Outpatient Visits, Psychiatric Inpatient Visits, and Psychi atric Length of Stay among Veterans with AIDS Who Used Healthcare Se rvices in the Veterans Health Administration Variable Outpatient Visits (N =2147) Inpatient Visits (N=715 ) Length of Stay (N=208) RC = Referent Category IRR SE p IRR SE p IRR SE p Race White (RC) Black Hispanic Other races 1.00 1.51 0.93 1.18 .094 .095 .440 <.01 .531 .652 1.00 1.08 0.82 1.53 .085 .113 .816 .288 .152 .419 1.00 0.85 0.42 1.01 .123 .104 .930 .288 <.01 .983 Risk Factor Heterosexual (RC) Injection Drug Use (IDU) Men who have sex with Men (MSM) Other risk factors Unknown risk factor 1.00 1.81 0.96 1.03 1.62 .161 .079 .129 .139 <.01 .681 .787 <.01 1.00 1.57 0.99 1.34 1.48 .168 .115 .217 .161 <.01 .957 .071 <.01 1.00 1.24 1.15 1.14 1.52 .237 .251 .345 .277 .241 .509 .663 .019 Military Service Period Post-Vietnam/Gulf (RC) Vietnam Korean Era Other Era 1.00 0.96 0.46 0.99 .064 .073 .245 .613 <.01 .999 1.00 0.77 0.33 0.58 .065 .073 .207 .003 <.01 .130 1.00 0.83 0.52 0.38 .120 .350 .258 .216 .333 .155 Age at AIDS diagnosis 1.02.002<.011.02.003<.01 1.05 .006<.01 Region West (RC) Midwest Northeast South 1.00 1.74 1.25 1.03 .160 .109 .071 <.01 .010 .666 1.00 1.13 1.49 1.25 .136 .169 .116 .299 <.01 .015 1.00 1.31 2.35 1.29 .296 .505 .229 .223 <.01 .142 AIDS-Defining Illness (ADI) Had no ADIs (RC) Had ADIs 1.00 1.43.095<.01 1.00 1.12.081 .090 1.00 1.17 .157.227 Long-Term Survival Not Attained (RC) Attained 1.00 1.11.063.066 1.00 1.25.092 .002 1.00 1.16 .149.240 Death Alive end of 2003 (RC) Died in 2003 1.00 0.80.076.023 1.00 0.85.074 .074 1.00 1.30 .257.184 CD4 cell count 1.01.000<.010.99.000.038 1.00 .000.499

PAGE 104

91 Also, patients who served in the Vietnam era had less inpatient psychiatric visits than those who served in the post-Vietnam/ Gulf war era (IRR=.77; p=.003). As the age at AIDS diagnosis increases the use of both ps ychiatric outpatient a nd inpatient care was shown to grow higher. For each additional year of age at AIDS diagnosis, both outpatient and inpatient psychiatric visi ts increased by 2%, while psyc hiatric LOS increased by 5%. Patients who live in the Northeast Region were shown to have greater use of psychiatric services compared to those from the West Region: IRR=1.25, p=.010 (psychiatric outpatient visits); IRR=1.49, p<.001 (psychia tric inpatient vis its); and IRR=2.35, p<.001 (psychiatric LOS). Additionally, living in the Midwest was associated with more outpatient psychiatric visits, while living in the South was associated with greater number of inpatient psychiatric visits . Generally, living in the West Region was associated with less use of both outpatient and i npatient psychiat ric services. Presence of ADIs was associated with greater rates of outpatient psychiatric visits (IRR=1.43; p<.001), but not with inpatient psyc hiatric visits or ps ychiatric LOS. Longterm survival was associated with greater inpatient psychiatric vi sits (IRR=1.25; p=.002). Neither frequency of outpatient psychiatric visits nor psychiatric LOS was associated with long-term survival. Also patients who died in 2003 had fewer outpatient psychiatric visits (IRR= .80; p<.001) compared to thos e who were still alive by the end of 2003. When we compare these results with those obtained under all-diagnosis utilization, we can say that patients who die may have seve re course of illne sses but they are not characterized by having greater psychiatric utilization. Las tly, increasing CD4 cell count was associated with greater number of outpa tient psychiatric visi ts, while decreasing levels of CD4 cells were associated with gr eater number of inpatie nt psychiatric visits.

PAGE 105

92 5.6 Results of the Negative Binomial Regression for Medical Utilization This section presents results for the test of the hypothesis that long-term survival following AIDS diagnosis is associated with greater rates of outpatient medical visits, inpatient medical visits, and longer medical LOS. Similar to the previous sections, the association was controlled for predisposing va riables (race, HIV risk, age, and military period of service); enabling vari able (region of residence); and need variables (presence of ADIs, presence of psychiatric comorbidit y, number of medical comorbidities, death, and CD4 cell count). These results presented wi thin this section were obtained from the negative binomial regression model and invol ved patients who had medical utilization (i.e., utilization for non-ADIs, non-psychiatri c comorbidities) during 2003. We present IRR and significance values for each predispos ing, enabling, and need variable that was used to adjust the analysis. The Wald chi-square te sts and likelihood ratio tests for overdispersion for all medical utilization mode ls (outpatient visits, inpatient visits, and LOS) were significant, indicating that the IRRs were globally signifi cant and the negative binomial model provided the best fit for the data on medical utilization. Results presented in Table 5.5 show that the negative binomial model predicted greater numbers of outpatient medical visits for black and Hispanic patients compared to whites (IRR=1.14; p=.001 and IRR=1.26; p=.0 01 respectively). In addition, we found that there were no racial differences in the ra tes of medical inpatient care (inpatient visits and LOS). Based on the HIV risk factor, the results showed that being IDU, MSM, and having other risk factors was associated with more outpatient medical visits compared to heterosexuals. However, considering inpati ent care, only IDUs had higher rates of inpatient medical visits (IRR=1.41; p=.025) and longer medical LOS (IRR=1.59; p=.046) compared to heterosexuals.

PAGE 106

93 Table 5.5. Negative Binomial Regression Result s for Medical Outpatient Visits, Medical Inpatient Visits, and Medical Length of Stay among Veterans with AIDS Who Used Healthcare Services in the Veterans Health Administration Variable Outpatient Visits (N =1459) Inpatient Visits (N=222) Length of Stay (N= 218) RC = Referent Category IRR SE p IRR SE p IRR SE p Race White (RC) Black Hispanic Other races 1.00 1.14 1.26 1.41 .046 .092 .275 .001 .001 .078 1.00 1.13 1.21 0.75 .110 .209 .530 .177 .247 .689 1.00 1.07 0.95 0.73 .156 .235 .731 .624 .843 .753 Risk Factor Heterosexual (RC) Injection Drug Use (IDU) Men who have Sex with Men (MSM) Other risk factors Unknown risk factor 1.00 1.12 1.19 1.49 1.18 .064 .061 .114 .064 .040 .001 <.001 .002 1.00 1.41 1.30 1.06 1.12 .221 .194 .210 .160 .025 .069 .757 .416 1.00 1.59 1.03 1.14 1.33 .377 .232 .335 .281 .046 .886 .649 .172 Military Service Period Post-Vietnam/Gulf (RC) Vietnam Korean Era Other Era 1.00 0.71 0.50 0.49 .033 .044 .063 <.001 <.001 <.001 1.00 0.78 0.48 0.47 .087 .099 .142 .031 <.001 .013 1.00 0.53 0.19 0.40 .088 .062 .169 <.001 <.001 .030 Age at AIDS diagnosis 1.03.002<.0011.03.005<.001 1.05.008<.001 Region West (RC) Midwest Northeast South 1.00 1.31 1.07 1.16 .083 .063 .052 <.001 .221 .001 1.00 1.26 0.78 0.97 .172 .116 .110 .085 .099 .845 1.00 0.90 0.96 1.04 .194 .206 .179 .638 .867 .809 Medical comorbidities 1.44.023<.0011.20.036<.001 1.12.054.020 Long-Term Survival Not Attained (RC) Attained 1.00 1.24 .050<.001 1.00 1.23.121 .033 1.00 1.00.150.955 Death Alive end of 2003 (RC) Died in 2003 1.00 1.01 .073.831 1.00 1.66.177 <.001 1.00 2.42.405<.001 CD4 cell count 1.01.000<.0010.99.000.301 0.99.000.101 There was an overall significant higher ra te of medical health care services used by veterans who served in the post-Vietnam/Gu lf war era compared to those who served in other eras. Similar to the findings in the previous sections, the negative binomial model predicted greater use of services as age at AI DS diagnosis increases. Specifically in this

PAGE 107

94 section, the model predicted an increase of 3% in the rates of both medical inpatient and outpatient visits and an increase of 5% in me dical LOS for each one year increase in age at AIDS diagnosis. Patients from Midwest and South regions were found to have greater use of medical outpatient care compared to thos e from the West Region (IRR=1.31; p<.001 and IRR=1.16; p=.001 respectively for medical outpa tient visits). No regional differences were found in the consumption of medical i npatient care. As expected, the number of medical comorbidities was associated with both medical inpatient and outpatient care for AIDS patients. For each unit increase in the number of medical comorbidities, outpatient medical visits were pr edicted to increase by 44%, inpatient medical visits by 20%, and medical LOS by 12%. This finding implies that a majority of the medical comorbidities is more likely to be treated in the outpatient care setting. Long-term survival after AIDS diagnosis was found to be associated with greater number of outpatient medical visits (I RR=1.24; p<.001) and inpatient medical visits (IRR=1.23; p=.033), but not with medical LOS. Patients who died had greater use of medical inpatient care. They had more inpa tient medical visits (IRR=1.66; p<.001) and longer medical LOS (IRR=2.42; p<.001). Finally, the results showed that higher CD4 cell count predicted greater use of medical outpatient care (IRR=1.01; p<.001 for outpatient visits). 5.7 Chapter Summary We presented results of all statistical analyses within this chapter. Descriptively, the data showed that there were differences in outpatient visits, inpatient visits, and LOS according to patientsÂ’ predisposing, enabling, and need characteristics. Racial minorities,

PAGE 108

95 patients with ADIs, and those with diagnoses of psychiatric conditions were shown to have higher levels of using the services. By using the lo gistic regression model, we showed that being black, havi ng ADIs, and having a diagnosis of psychiatric comorbidity was associated with poor long-term survival rates after AIDS diagnosis, while being MSM and living in the Northeast Region was as sociated with better long-term survival rates after AIDS. The multivariable negative binomial regr ession model assessed the impact of long-term survival on the rates of service use. The model was used for all-diagnosis utilization, psychiatric utili zation, and medical utilization. We found that long-term survival was associated with greater use of se rvices under all utiliza tion scenarios, but its effect varied according to the types of serv ice used. When all-dia gnosis utilization was considered, long-term survival was shown to be associated with more outpatient and inpatient visits and LOS. However, neith er psychiatric LOS nor medical LOS was associated with long-term survival. Potent ial confounding factors su ch as race, age at AIDS diagnosis, and presence of ADIs were shown to be important in adjusting the analyses. Our results have consistently show n that black patients with AIDS had greater rates of service utiliz ation than white patients. Also, ag e at AIDS diagnosis was shown to be associated with greater use of the both ps ychiatric and medical services. Presence of ADIs, which was conceptualized as a measure of disease severity, was also associated with greater rates of outpatie nt and inpatient visits for both psychiatric and medical services.

PAGE 109

96 CHAPTER SIX STUDY CONCLUSION This dissertation has examined the correla tes of long-term survival following AIDS diagnosis and the impact of long-term survival after AIDS diagnosis on outpatient visits, inpatient visits, and LOS among veterans receiving healthcare services from the VA system. The five preceding chapters presented the introduction and background to the study, discussion on the source of data, a revi ew of literature, objectives and methods, and the results of the study. Within this last chapter of the disser tation we present four sections covering the conclusion of the study. Fi rst, an overview of the study is presented to give a summary of the main ideas of the preceding chapters. Second, a discussion of the findings generated by the analytical met hods used in this study is presented. Policy and clinical practice implications of thes e findings are discussed accordingly. Third, a section discussing important limitations of th is study will follow; finally, suggestions for directions of future research will be presented. 6.1 Overview of the study The purpose of this dissertation was to ex amine correlates of long-term survival after AIDS diagnosis and the impact of longterm survival after diagnosis of AIDS on outpatient visits, inpatient vi sits, and LOS among veterans r eceiving healthcare in the VA system. The ABM for health services use was used to model outpatient visits, inpatient, visits and LOS (all as count variables) fo r the assessment of th e relationship. The independent variables were divided into three groups in accordance to the ABM framework. Predisposing variables were race, HIV risk factor, military period of service,

PAGE 110

97 and age at AIDS diagnosis. Region of residen ce formed the only enabling variable, while presence of ADI, diagnosis of psychiatric comorbidity, number of medical comorbidities, death, and CD4 cell count were the need variables. Long-term survival was conceptualized as a need vari able in the model. Negative binomial regression model, a regression model for count data , was used to fit the data. The data for this dissertation research were obtained from the National ICR, a database owned by the Department of Ve terans Affairs containing information on veterans with HIV who use healthcare serv ices within the VA system. This study used information on service utilization for th e calendar year 2003 using a study sample consisting of 4,717 veterans who had ever been diagnosed with AIDS . These patients had also been using HAART within the VA system. Data analysis was conducted in three stag es. Stage one involved computation of the descriptive statistics for the sample. Mean leve ls of outpatient and i npatient visits as well as LOS were compared among categories of independent variables of race, HIV risk factor, military period of service, region of residence, presence of ADIs, presence of psychiatric comorbidity, death, and CD4 cell c ount. Stage two involved determination of the sociodemographic and clinical determin ants of long-term survival after AIDS diagnosis. A logistic regression model was used in this analysis to determine the odds of attaining long-term survival after AIDS di agnosis. Last, a negative binomial regression was used to determine the impact of long-term survival on the rates of outpatient visits, inpatient visits, and LOS. Three models feat uring all-diagnosis util ization, psychiatric utilization, and medical util ization were analyzed. IRRs were used to assess the significance of the predisposing factors (race, HIV risk factor, age, military service

PAGE 111

98 period), enabling factor (regi on of residence), and need factors (presence of ADIs, presence of psychiatric como rbidity, number of medical co morbidities, death, and CD4 cell count) in the models. The logistic a nd negative binomial regression models were assessed and found to provide the best fit fo r the data in all modules of analysis. 6.2 Discussion of the Finding s and their Implications 6.2.1 Long-term Survivorship Although the advent of HAART in the mi d-1990s has resulted in prolonged lives among HIV/AIDS patients, the association be tween long-term survival and the use of healthcare services is a topic that has so fa r received little attention in HIV research. However, the increasing number of AIDS patie nts who attained and with the possibility of attaining long-term survival, combined with the costs associated with caring for AIDS patients, warrants public polic y interest in the issue. Our findings suggested that longterm survival after AIDS diagnosis is im portant in predicting patterns of service utilization. A considerable amount of evidence has demonstrated that long-term survival after AIDS diagnosis is associated with incr eased rates of outpatient and inpatient visits and LOS. For example, when service utiliz ation was investigated for all-diagnosis utilization, we found that long-term survival was associated with increased rates of outpatient visits, inpati ent visits, and LOS. In this investigation, the a ssociation between long-term su rvival and increased rates of all-diagnosis service use persisted despite controlling for predisposing and enabling factors as well as other need factors. For example, three n eed factors, presence of ADIs, presence of psychiatric comorbidity, and in creased number of medical comorbidities, were independently associated with increased rates of a ll-diagnosis outpatient visits,

PAGE 112

99 inpatient visits, and LOS. This means long-term survival by itself is an important need factor of service utilization. A detailed analysis examining the nature of the association between long-term survival and use of services was conducted based on psychiatric utilization and medical utilization (for non-ADIs, non-psychiatric como rbidities). The association between longterm survival and increased use of services was still detectable in both psychiatric utilization and medical utiliza tion, although it occurred in differ ent ways for each type of service utilization. For instance, while long -term survival was associated with only increased psychiatric inpatient visits, it was found to be associ ated with incr eased rates of both outpatient and inpatient medical visits . However, neither psychiatric LOS nor medical LOS was associated with long-term survival. Our finding that long-term surv ival is associated with increased rates of service utilization reveals that the patterns of service use among patients with long-term survival may be unique. However, the interpretation of this finding need to be done carefully in light of multiple factors that might lead to increased rates of service utilization. These factors need to be assessed to determine in what ways th e independent impact of longterm survival on service utiliz ation could be demonstrated. Fo r instance, increased use of the services among patients with long-term su rvival after AIDS diagnosis could be a function of either HIV-related treatment, ex isting risk factors pr ior to HIV or AIDS diagnosis, or aging cohort effect. Many patients undergoing HIV-treatment have problems with overlapping drug toxicity, drug interactions, and difficulty a dhering to drug regimens (Antonious & Tseng, 2002). These problems may be more seriou s among AIDS patients with co-existing

PAGE 113

100 conditions such as cancer (Antoniou & Tse ng, 2005). The treatment of these problems can potentially increase the rate of service utilization among th e patients. For instance, in this study, hepatitis C infection was highly prevalent among the AIDS patients, and it contributed to a large number of medical outpatient and inpatient visits. At the same time, previous research found that risk of liver toxicity is higher among HIV/AIDS patients with hepatitis C (Sulkowski, Mast, Seeff, & Thomas, 2000; Sulkowski, Thomas, Chaisson, & Moore, 2000) implying that the risk of liver toxicity was relatively high in our sample. For this reason, clinicians have to monitor closely surviving AIDS patients for conditions such as hepatitis C that may increase their risk of acquiring toxicities. Identification of patients with, or at risk of, hepatitis C or other conditions known to increase the risk of toxicity or drug interactions would allo w for a better management of AIDS patients, and consequently reduce the ra tes of service utilization associated with long-term survival. Another risk factor for in creased rates of se rvice utilization among AIDS patients could be a high incidence of chronic conditi ons occurring independent of HIV. It is important to acknowledge that as AIDS patients live longer, thei r risk of chronic diseases such as diabetes, cardiovascular disease, and cancer may increase because of their life style and other factors prior to HIV (or AI DS). Such factors, including smoking, heavy use of alcohol, elevated chol esterol level, hypertension, etc ., may have their cumulative or long-term impact (occurrence of chronic c onditions) that may coin cide with long-term survival after AIDS diagnosis. In this way, patients with long-term survival after AIDS might have increased rates of services due to the conditions not dir ectly related to HIV. On the other hand, however, as a result of th e immune suppression due to HIV/AIDS, the

PAGE 114

101 onset of the chronic conditions might occur ear lier and with a more severe course among AIDS patients compared to other patients in th e population. As a result, an increase in the rate of service use could be observed among AIDS patients with long-term survival. Development of database containing in formation on risk factors for commonly known chronic conditions such as cancer, diabet es, and cardiovascular disease may assist in conducting studies in utilization with the ability of separating the effect of diseases occurring independently of HIV/AIDS. Als o, comparative studies between individuals with AIDS and those without AIDS on their experience with specifi c chronic conditions could reveal the independent effect of longterm survival on these conditions. This could lead to a timely implementation of clinical interventions to reduce occurrence and to improve prognosis of the conditions. Such an intervention would eventually lead to reduced rates of medical utilization among patients w ith long-term survival. Of the 2,435 patients who had attained l ong-term survival in the sample, 1,327 (55%) were aged 50 years or older. The comp arison of the rates of service use between older and younger patients with long-term survival showed that there were no differences in the rates of all-diagnosis inpatients visits (m ean visits 12.39 vs. 12.48; p=. 533) or alldiagnosis LOS (19.32 vs. 21.51; p=. 811), but olde r patients had signifi cantly higher rates of all-diagnosis outpatient visits (35.86 vs. 30.33; p<. 005). The association between long-term survival and increase in service utilizat ion may thus be attributable in part to an aging cohort of AIDS patients. The higher medi cal comorbidity seen in the older patients with long-term survival was reflected in th eir overall higher rates of outpatient visits. This finding has demonstrated that medical comorbidity is an important determinant of service utilization among older pa tients with long-term survival, and it suggests a need to

PAGE 115

102 increase the knowledge of the biological and physi cal risk factors for comorbidity among long-term survivors of AIDS. These factors might be unique to HIV/AIDS patients and their knowledge would enable appropriate and timely clinical intervention to reduce the rates of both incidence of comorb idity and service utilization. 6.2.2 Findings on Other Need Variables Other need variables that were assessed in our analysis included presence of ADIs, presence of psychiatric comorbidity, death, a nd CD4 cell count. All these variables were found to be associated with service utilization. For example, although only 20% of the patients in the sample had ADIs, our results showed that having ADIs was associated with increased rates of all-diagnosis outpati ent and outpatient visits as well as longer LOS. Though the prevalence of ADIs is lo w when AIDS patients attain long-term survival, our results suggest that ADIs incr ease use of the services. In addition, ADIs have also been associated with poor long-term survival rates, that is, patients who had ADIs following their first AIDS diagnosis were less likely to attain long-term survival. These findings suggest a need for clinical intervention designed for patients with longterm survival. Plans should be developed fo r continued monitoring of incidence of ADIs among the surviving patients for the improveme nt of both their survival and service use rates. Almost half of the patients in the whol e sample (48.9%) as well as among those with long-term survival (48.7%) had psychiat ric comorbidities. In addition, we found that psychosis, a psychiatric condition, was among the leading ADIs. Unlike ADIs, presence of psychiatric comorbidity was not associated with long-term survival in the logistic regression model of th e correlates of long-term surviv al, but it was associated with increased rates of service utilization. Most not ably, we found that pres ence of psychiatric

PAGE 116

103 comorbidity was highly associated with alldiagnosis outpatient visits. For this reason, these results suggest that le ss intensity of care per patient could be expected in the management of psychiatric conditions compared to the management of ADIs or medical comorbidities because most of these condi tions are addressed in the less intense outpatient care. The finding that AIDS patie nts with psychiatric comorbidity have low rates of inpatient care is contra ry to that of Hoover et al. (2004) who used a sample of AIDS patients with high prevalence of psyc hiatric comorbidity and found that patients had long LOS. However, Hoover and his colleagues used a sample that consisted of patients with both psychiatric cond itions and substance abuse. Additionally, high prevalence of psychiatric comorbidity emphasizes the need for intervention to treat, prevent, and reduce th e frequency of psychiatric comorbidity among AIDS patients with long-term survival. Res earch has shown that psychiatric illnesses could impede delivery of appropriate care to AIDS patients (Uldall et al., 1994), mainly because of deficits in decision-making competence (Monahan et al., 2003) that makes patients unable to work with the treatment team and comply with treatment (Uldall et al., 1998). Thus, psychiatric comorbidities can indire ctly increase costs and rates of medical service utilization because they can potentially cause delaye d or inappropriate treatment of other medical conditions. For example, b ecause of delayed treatment, most of these conditions would be treated while in severe stages that might require intense health care. Our study included AIDS patients who di ed in 2003 but with information on service utilization for the same year. We concep tualized death as a measure of severity of the disease. Of all deaths that occurred, 40% occurred among patients with long-term survival. We found, as expected, that death was associated with increased use of all-

PAGE 117

104 diagnosis inpatient care. Deceased patients ha d greater number of inpatient visits and corresponding longer LOS compared to all other patient subgroups. We could not establish causes of death for th e deceased patients with the da ta used in this study. But a distinction of non-HIV and HIV causes of d eaths among patients surviving after AIDS diagnosis is important for monitoring the pa tterns of morbidity and mortality in the population of patients with long-term survival. For instance, our results showed that death at a visit was a strong predicto r of increased medical LOS. This implies that there are some non-ADIs, non-psychiatric conditions that are strongly associated with increased rates of death among patients with long-term survival. Early identi fication and treatment for these conditions may reduce death rates and shift the type of care used by the patients from more intense and costly inpatient to less intense outpatient services. Patients who experienced improvements in their CD4 cell count were associated with increased use of outpatient care. Improve ments in the CD4 cell count indicate that a person becomes healthier and may encounter less severe or a decreased number of medical problems that would require less in tense outpatient care. Although all patients included in this study had evidence of using HAART, patterns of use could be different, thus affecting their CD4 cell recovery. Efforts must be di rected in improving access and adherence to HAART, especially among vulne rable subgroups such as minorities and IDUs who have traditionally been known to have poor access and adherence to HAART. 6.2.3 Findings on the Predisposing Variables Our study included race, HIV risk factor, military service peri od, and age at AIDS diagnosis as a set of variables for two purpos es in our analyses. First, we used these variables as sociodemographic factors in the examination of the correlates of long-term survival. Second, these variables were us ed as predisposing factors in the ABM

PAGE 118

105 framework to take into acc ount their confounding effect in the association between longterm survival and service utilization. We f ound that these variables were important in both analysis modules. Using the logistic model, we found that white patients and MSM had better longterm survival rates compared with their re spective counterparts. These findings may be interpreted on the basis of the history of the HIV epidemic. For example, it has been reported that compared to white patients with HIV, minorities were less likely to receive HAART (Andersen et al., 2000; Gebo et al., 2001) and prophylaxis for PCP (Piette et al., 1991). PCP is an AIDS-defining illness char acterized by short-term survival among HIV patients. In addition, “Minorit y patients ignored in HIV dr ug trials” (2002) reported that minorities were also less likely to obtain experi mental treatments through participation in clinical research trials. For these reasons, mi norities are more likely than whites to have poor disease prognosis due to, among other things, decrease d CD4 cell count, increased viral load, and comorbidity. Consequently, th ey would have undesirable survival patterns after AIDS diagnosis. Compared to other HIV ri sk groups, studies have found that MSM had greater access to private insurance and obtained care in private hospitals (Kass et al., 1999) and had more favorable patterns of care, for inst ance, regular outpatient visits and decreased emergency department visits (Shapiro et al., 1999). This patter n of care was typical among white MSM who were involved in networking and information exchange between MSM organizations in different parts of th e country. In this study, we found that about 487 (64%) of all MSM who attained longterm survival were whites. This disproportionate number of white MSM with lo ng-term survival could be indicative of

PAGE 119

106 the optimal pattern of care th at they had in the course of their HIV infection. These results suggest that interven tions for reduction of health disparities between whites and minorities have to be extended in the era of increased survival among HIV/AIDS patients. In addition, these interventions have to be culturally sensitive because behavioral and cultural factors may account for some of the disparities observed in the treatment and outcomes among HIV/AIDS patients. Our results showed that patients who receiv e an AIDS diagnosis at an older age had poor long-term survival rates. High level of comorbidity and generally severe course of the disease due to relatively higher immune suppression in older patients compared to younger patients might be some of the reasons for this outcome. In addition, race and age may have a combined effect in the poor su rvivorship of older patients. For instance, McGinnis et al. (2003) found that 28% of white veterans had a 90 100% chance of being alive in 5 years, compared to 22% of blacks and 14% of Hispanics. Poor patterns of HIV treatment and service utilizati on among racial minorities in the course of HIV, as pointed out earlier in this study, might be associated with the gene ral poor long-term survival rates among older patients. More over, regional differences in survival that were found in this study could be due to inherent racial di fferences in survivorship. Patients in the Northeast Region were more likely to attain long-term survival, while those in the South, where the percentage of the population that is black is hi ghest in the country, showed patterns of poor survival ra tes after AIDS diagnosis. Our study also found differences in the ra tes of services utilization based on the predisposing factors. Black pati ents were found to have incr eased rates of all-diagnosis inpatient visits and longer LOS compared to whites. However, this trend was different

PAGE 120

107 when we adjusted the analysis for type of comorbidity. For instance, blacks had increased rates of psychiatric outpatient visits compared to whites, but there were no significant differences in psychiatric inpa tient care. Similar results were obtained when patients with medical utilization was examined. Our fi ndings suggest that blacks did not have increased rates of costly and intense psychiatric and medical inpatient care compared to whites. This is contrary to several previ ous studies (e.g., Menke et al., 2003 and Kass et al., 1999) that found that blacks were significan tly more likely to use inpatient health services compared to whites. Although these findings are indicative of reduced disparity in inpatient care, they warrant further resear ch to determine how they may occur based on the characteristics of the surviving patients. For instance, as more data on patients with prolonged survival becomes available, research is needed to determin e if there are intraracial differences that make some black patien ts more likely to surviv e than others, and if they have desired patterns of service utilization. Patients with IDU risk factor were expect edly found to be major users of overall healthcare services among all other patient subgroups. IDUs have been found to have greater propensity for mental illnesses (Ste in, 1994) and are norma lly associated with high prevalence of medical comorbidities (H oover et al., 2004) that can influence increased use of the services. In this study patients with I DU risk factor had 81% more psychiatric outpatient visits and 59% longer psychiatric LOS than heterosexuals. On the other hand, MSM were found to have a decreas ed use of medical i npatient care and only a moderate increase in medical outpatient car e compared to heterosexuals. These findings suggest that a patientÂ’s risk f actor is an important factor in predicting service utilization. Also, it should be known that is common for HIV patients to have multiple risk factors.

PAGE 121

108 Efforts should be directed in early determinati on of risk factors when patients come into contact with the health care system. This would give providers an opportunity to make appropriate and timely treatment plans accordin g to a patientÂ’s HIV risk predisposition. Patients who served during the post-Vietna m and Gulf war eras (relatively younger in the sample) were shown to have increased rates of services compared to those who served during the Vietnam or Korean wars . In addition, considering age at AIDS diagnosis, we found that there were only minima l increases in service utilization as age at AIDS diagnosis increased. Taken together, th ese results suggest that differences in service utilization occurring ac cording to patientÂ’s age are minor compared to those occurring due to patientÂ’s race and HIV risk factor. Noticeable differences for psychiatric visi ts were found based on patientsÂ’ region of residence. Patients in the Northeast regi on were found to have increased rates of psychiatric outpatient and inpatient care compar ed to those in the West Region. Patients in the South Region had increased number of psychiatric inpatient vi sits. We used region as enabling factor in this st udy, but we could not establish to what extent facility or patient characteristics could have influenced the observed diffe rences. Research that aims at examining specific facility characteristics related to service uti lization would provide a better meaning of our findings. Moreover, by using region of residence as the only enabling factor is likely to have biased our results to some extent because of the unaccounted variability of tradit ional enabling factors such as income in our estimates. Lack of information on veteransÂ’ priority groups that are based on veteransÂ’ serviceconnected disabilities and income (Appendix B) has partly contribute d to this problem.

PAGE 122

109 6.3 Limitations of the Study This study is subject to seve ral limitations that may be at tributed to the database and design used for analytical methods. The majo r limitation of this st udy was the lack of some patient characteristics in the data th at would have been im portant in explaining service utilization behavior among AIDS patient s and for best possible application of the ABM. Variables such as marital status, edu cation level, and income that could have potentially explained some of the variability in the rates of service utilization were not available in the database. Also, information on the veteransÂ’ priority groups, an important enabler of service utilization among veterans, was lacking. However, the large size of the database as well as the availability of info rmation on patientsÂ’ race, age, and HIV risk factor has substantially assisted in explaining the variation in service use for our results to have valid interpretations. Another limitation concerns the use of cross-sectional data for our analyses. Studying the impact of long-term survival on th e outcomes such as service utilization is basically a longitudinal problem that might be difficult to examine by using a crosssectional data. As pointed out in several parts of the discussion of th e results, the rates of service utilization among patients who had ever had an AIDS diagnos is could be a result of a series of events such as changes in th e number or types of comorbidities or having ADIs. These changes occur with time and ca nnot be fully captured in a cross-sectional study that utilizes information from a single poin t in time. However, given that this is the first large-scale study to examine service utilization among patients with long-term survival, the observed differences in util ization may be used reliably in making inferences on use patterns for HIV/AIDS pa tients who attain long-term survival. Also, similar cross-sectional studies might be conduc ted in the future to gather information on

PAGE 123

110 service utilization. Data from these studies could then be assessed to identify any time trends in utilization am ong HIV/AIDS patients w ith long-term survival. Although we used a national database for ve terans with AIDS, our results may not be fully generalizable to the population of ve terans with AIDS or the general population due to several reasons. First, information on the use of non-VA care was not available in the database and could not be readily obtaine d from other sources. Veterans with private health insurance may use services from nonVA sites as a supplement to their VA care, and the extent of care recei ved or the number of recipients of non-VA care may vary from year to year and influe nce the rates of using the VA se rvices. However, for patients with HIV/AIDS who need continued care fo r their condition, use of non-VA care could be minimal due to the high costs of care. Fo r these reasons, the fi ndings might be, to a large extent, generalizable to the population of veterans with HIV/AIDS. Second, because only men were studied, our findings cannot be generali zed to women with AIDS. Previous studies have shown that women w ith HIV/AIDS have a different pattern of healthcare use compared to men (FlorisMoore et al., 2003; Fleishman & Hellinger, 2003; Mauskopf et al., 1994). An independent study on the duration of survival after AIDS diagnosis and use of healthcare services among women in the ICR may be conducted to determine the trend of service use among surviving women. 6.4 Suggestions for Future Research As increasing numbers of patients with HI V/AIDS survive to live longer than ever before, we expect that the topic of this dissertation will entice more research from a wider range of disciplines. Apart from health services utilization, issues such as quality of care, quality of life, social suppor t, and aging would draw the attention of researchers to generate different questions pertaining to su rvivorship of HIV/AIDS patients. As an

PAGE 124

111 addition to the specific recommendations that were made in the discussion of the findings, this section presents suggestions fo r future research on the use of healthcare services among AIDS patients with long-term survival. The present study used a cr oss-sectional design repres enting service utilization information for the year 2003 alone. We suggest that a longitudinal st udy is the best way to examine sequences of serv ice utilization among patients wi th different duration times after AIDS diagnosis. Such a design would iden tify critical points of vulnerability after AIDS diagnosis, where service utilization may indicate drastic changes. Identification of such points would allow for appropriate interv entions to be designed to retain desirable levels of service utilization. The longitudina l design would also be more effective in distinguishing between HIV a nd non-HIV factors of service utilization. In general, a longitudinal design would indicate how and to wh at extent the rates of service utilization change over time as patients atta in longer lives with HIV/AIDS. Future research should find a means of linking the ICR with other VA and non-VA databases to capture the whole array of types and quantities of services used by veterans with HIV/AIDS. As patients live longer, diffe rent types of services, such as home health care, rehabilitation services, and nursing home care, would c ontribute a large proportion of their service utiliz ation. In addition, the linkage of di fferent databases would facilitate completeness of the data by adding informati on on variables such as income, priority groups, and education that is not available in the ICR. Previously, age and aging were not issues of priority in HIV/AIDS research because, since its onset, the disease has been affecting younger persons. In addition, due to the unavailability of ther apy and disease specialists du ring the early days of the

PAGE 125

112 epidemic, the infected persons were expected to live only for a short time. However, in the current era of the epidemic, increasing numbers of older persons become newly infected with HIV and, in addition, due to advancements in treatment; increasing numbers of patients live into older ages. For these reasons, age and the process of aging are likely to impact the use of healthcare se rvices to create an important area of future research. Clinical trials invol ving patients of different ages may be designed to study agerelated factors such as comorbidity and ch ronic conditions that may have potential influence the rates of service utilization as well as the well-being of the patients with long-term survival. Last, HIV/AIDS patients w ith prolonged survival present an emerging population that has specific need s related to their survivorship. Fo r instance, we found evidence of increasing utilization that was related to pr esence of psychiatric comorbidity and number of medical comorbidities. Quality and appropriateness of care received by these patients has to be monitored to ensure optimal outcome s with the healthcare services they receive. For example, patient satisfaction is an importa nt area of research that could be used for identification of areas in need of improvement. 6.5 Conclusion In conclusion, this study has demonstrated that long-term su rvival after AIDS diagnosis is associated with utilization of health care services. We found that rates of inpatient care increased for patients who had increased numbers of medical comorbidities, while those patients who had a diagnosis of psychiatric comorbidity had increased use of outpatient care. Also, we found that being black, having an IDU risk factor, and having an older age at AIDS diagnos is were associated with increased use of the services. However, the differences in se rvice utilization between whites and blacks

PAGE 126

113 were mainly in outpatient care, not in inpatien t care as it has been previously found. It is hoped that these findings would create an understanding of the pattern of service utilization among AIDS patients with long-term survival and help in the development of further research to address different issues pertaining to long-term survival of HIV/AIDS patients.

PAGE 127

114 APPENDIX A AIDS CASE DEFINITION OF THE CE NTERS FOR DISEASE CONTROL AND PREVENTION (CDC) The 1993 AIDS Surveillance Case Definition of the U.S. Centers for Disease Control and Prevention (CDC) A diagnosis of AIDS is made whenever a person is HIV-positive and: He or she has a CD4 cell count be low 200 cells per microliter OR His or her CD4 cells account for fewer th an 14 percent of all lymphocytes OR That person has been diagnosed with one or mo re of the AIDS-defining illnesses below. AIDS-Defining Illnesses and ICD-9 Codes Candidiasis of bronchi, trach ea, or lungs (112.40, 112.80,112.90) Candidiasis, esophageal (112.84) Cervical cancer, invasive (622.1) Coccidioidomycosis, disseminated (114.2) Cryptococcosis, extrapulmonary (117.5) Cryptosporidiosis, chroni c intestinal (007.4) Cytomegalovirus disease (other than liv er, spleen, or lymph nodes) (078.5) Cytomegalovirus retinitis (w ith loss of vision) (078.5) Encephalopathy, HIV-related Dementia (298.90) Herpes simplex: chroni c ulcer(s) (054.7, 054.9) Histoplasmosis, disseminated (115.01, 115.11, 115.91) Isosporiasis, chronic intestinal (007.2) KaposiÂ’s sarcoma (176.0) Lymphoma, BurkittÂ’s (200.2) Lymphoma, immunoblastic (200.8) Lymphoma, primary, of brain (primary central nervous system lymphoma) (200.1) Mycobacterium avium complex or disease caused by M. Kansasii , (031.2) Disease caused by Mycobacterium tuberculosis , any site (010 018.9) Disease caused by Mycobacterium , other or unidentified species (031.0, 031.1, 031.9) Pneumocystis carinii pneumonia (136.3) Pneumonia, recurrent (482.80 482.84, 482.89 482.90) Progressive multifocal leukoencephalopathy (046.3) Salmonella septicemia, recurrent (003.1) Toxoplasmosis of brain (e ncephalitis) (130.0) Wasting syndrome caused by HIV infection (799.40)

PAGE 128

115 APPENDIX B VA PRIORITY GROUPS Priority Group 1. Veterans with service-connected conditions rated 50 percent or more disabling or have been discharged or releas ed from the military less than 12 months (365 days). Priority Group 2. Veterans with service-connected conditions rated 30 40 percent or more disabling. Priority Group 3. > Veterans who are former prisoners of war. > Veterans with serv ice-connected conditions rated 10 20 percent disabling. > Veterans discharged from active duty for a disability incurred or aggravated in the line of duty. > Veterans awarded special eligibili ty classification under 38 U.S.C., Section 1151. Priority Group 4. > Veterans who are receiving Aid and Attendance or Housebound benefits. > Veterans who have been determ ined by VA to be catastrophically disabled. Priority Group 5. Veterans who have an annual income and net worth below the means test threshold. Priority Group 6. All other eligible veterans who are not required to make co-payments for their care including: > World War I and Mexican Border War veterans. > Veterans solely seeking care for disorders associated with exposure to a toxic substance, radiation or for disorder asso ciated with service in the Persian Gulf. > Compensable zero per cent service-connected veterans. Priority Group 7. Non service-connect ed veterans and zero percent non-compensable service-connected veterans w ith income and net worth above the statutory threshold and who agree to pay specified co-payments. At this time there is a co-payment of $2.00 for each prescription. Office visits are $50.80. The VA can collect from your insurance carrier but not from Medicare. The established income dollar thresholds for a nonservice-connected veteran or a serviceconnected veteran with a rated disability of zero percent is $22,352 and above if single, or $26,825 and above with one dependent, plus $1,496 for each additional dependent (or combined income and net worth is greater that $50,000 excluding pe rsonal property). The income levels will be adjusted on January 1st of each year by the percentage that VA pension benefits are increased.

PAGE 129

116 LIST OF REFERENCES Aday, L. A. (1993). Equity, accessibility, and et hical issues: is the U.S. health care reform debate asking the right questions? American Behavioral Scentisti, 36 , 724 740. Aday, L. A. (1994). Health stat us of vulnerable populations. Annual Review of Public Health, 15 , 487 509. Aday, L.A., Awe, W. (1997). H ealth services utilization models. Gochman D.S.(Ed.). Handbook of health behavior research I: Personal and social determinants, I53 I72. Plenum Press, New York. Andersen, R. (1968). A behavioral model of families Â’ use of health services . Chicago: University of Chicago, Center for Health Administration Studies, Research Series No. A25. University of Ch icago Press, Chicago. Andersen, R., Bozzette, S., Shapiro, M., Saint Clair, P., Morton, S., Crystal, S., Goldman D., Wenger, N., Gifford, A., Leibowitz, A., Asch, S., Berry, S., Nakazono, T., Heslin, K., & Cunningham, W. (2000) . Access of vulnerable groups to antiretroviral therapy among persons in car e for HIV disease in the United States. HCSUS Consortium. HIV Cost a nd Services Utilization Study. Health Services Research, 35 , 389 416. Andersen, R., & Newman, J. F. (1973). Societal and Individual Dete rminants of Medical Care Utilization in United States. Milbank Memorial Fund Quarterly-Health and Society, 5 1, 95 124. Andersen, R. M. (1995). Revisiting the beha vioral model and access to medical care: does it matter? Journal of Health and Social Behavior, 36 , 1 10. Antoniou, T., & Tseng, A. L. (2002). Inter actions between recreational drugs and antiretroviral agents. Annals of Pharmacotherapy, 36 , 1598 1613. Antoniou, T., & Tseng, A. L. (2005). Inte ractions between antiretrovirals and antineoplastic drug therapy. Clinical Pharmacokinetics, 44 , 111 145. Backus, L., Mole, L., Chang, S., & Deyton, L. (2001). The Immunology Case Registry. Journal of Clinical Epidemiology, 54 Suppl 1 , S12 15.

PAGE 130

117 Backus, L. I., Phillips, B. R., Boothroyd, D. B., Mole, L. A., Burgess, J., Rigsby, M. O., & Chang, S. (2005). Effects of hepatitis C virus coinfection on survival in veterans with HIV treated with highly active antiretroviral therapy. Journal of Acquired Immune Deficiency Syndromes, 39 , 613 619. Bastian, L., Bennett, C. L., Adams, J., Wask in, H., Divine, G., & Edlin, B. R. (1993). Differences between men and women with HIV-related Pneumocystis carinii pneumonia: experience from 3,070 cases in New York City in 1987. Journal of Acquired Immune Deficiency Syndromes, 6 , 617 623. Bavry, A., & Mauldin, P. D. (1996). Resource util ization and stage of illness: a study of an outpatient AIDS clinic. AIDS Patient Care and STDs, 10 , 288 291. Benhamou, Y., Bochet, M., Thibault, V., Calvez, V., Fievet, M. H., Vig, P., Gibbs, C.S., Brosgart, C., Fry, J., Namini, H., Katla ma, C., Poynard, C. (2001). Safety and efficacy of adefovir dipivoxil in patients co-infected with HIV-1 and lamivudineresistant hepatitis B virus: an open-label pilot study. Lancet, 358 (9283), 718 723. Bennett, C. L., Horner, R. D., Weinstein, R. A., Dickinson, G. M., DeHovitz, J. A., Cohn, S. E., Kessler, H.A., Jacobson, J., Goetz, M. B., Simberkoff, M. et al. (1995). Racial differences in care among hospitalized patients with Pneumocystis carinii pneumonia in Chicago, New York, Los A ngeles, Miami, and Raleigh-Durham. Archives of Internal Medicine, 155 , 1586 1592. Bertakis, K. D., Azari, R., Helms, L. J., Callahan, E. J., & Robbins, J. A. (2000). Gender differences in the utilization of health care services. Journal of Family Practice, 49 , 147 152. Bica, I., McGovern, B., Dhar, R., Stone, D ., McGowan, K., Scheib, R., Scheib, R., & Snydman, D. (2001). Increasing mortality due to end-stage liver disease in patients with human immunodeficiency virus infection. Clinical Infectious Diseases, 32 , 492 497. Biggar, R. J., Engels, E. A., Ly, S., Kahn, A ., Schymura, M. J., Sackoff, J., Virgo, P., & Pfeiffer, R.M. (2005). Survival after can cer diagnosis in persons with AIDS. Journal of Acquired Immune Deficiency Syndromes, 39 , 293 299. Bower, M., Powles, T., Newsom-Davis, T., Thirlwell, C., Stebbing, J., Mandalia, S., Nelson, M., & Gazzard, B. (2004). HIV-Asso ciated Anal Cancer: Has Highly Active Antiretroviral Therapy Reduced th e Incidence or Improved the Outcome? Journal of Acquired Immune Deficiency Syndrome, 37 , 1563 1565.

PAGE 131

118 Bozzette, S.A., Berry, S.H., Duan, N., Franke l, M.R., Leibowitz, A.A., Lefkowitz, D., Emmons, C.A., Senterfitt, J.W., Ber k, M. L., Morton, S.C., Shapiro, M.F., Andersen, R.M., Cunningham, W.E., Marc us, M., Wenger, N.S., Athey, L.A., Smith, S.M., Bing, E.G., Brown, J.A., Bu rnham, M.A., Goldman, D.P., Kanouse, D.E., Keesey, J.W., McCaffrey, D.F., Perl man, J.F., Schuster, M.A., Cleary, P.D., Fleishman, J.A., Hays, R.D., McCutchan, J.A., & Richman, D. (1998). The care of HIV-infected adults in the United States . HIV Cost and Services Utilization Study Consortium. New England Journal of Medicine, 339 , 1897 1904. Bozzette, S. A., & Hellinger, F. (2001). Study shows HAART reduced inpatient costs. AIDS Alert, 16 (8), 98, 103. Brau, N. (2003). Update on chronic hepatitis C in HIV/HCV-coinfect ed patients: viral interactions and therapy. AIDS, 17 , 2279 2290. Bryman, A., & Cramer, D. (1997). Quantitative data analysis with SPSS for Windows: a guide for social scientists . New York; London: Routledge. Cameron, A. C., & Trivedi, P. K. (1998). Regression analysis of count data . Cambridge; New York: Cambridge University Press. Centers for Disease Control and Preventi on (CDC). (1992). 1993 Re vised Classification System for HIV Infection and Expanded Surveillance Case Definition for AIDS Among Adolescents and Adults. Morbidity and Mortality Weekly Report, 1992; 41 (RR-17). Centers for Disease Control and Prevention (CDC). (2004). HIV/AIDS Surveillance Report: HIV Infection and AIDS in the United States , 2003, Vol.15. Center for Quality Management in HIV Ca re (CQM). 2001. Caring for Veterans with HIV Disease: Characteristics of Vete rans in VA Care, Fiscal Years 2000-2001. Retrieved November 15, 2004 from Public Health Strategic Health Care Group Web site: http://www.publichealth.va.gov/cqm/caring_2001.pdf Cheng, A. C., Mijch, A. M., Hoy, J. F., Wesselingh, S. L., & Fairley, C. K. (2001). Psychosocial factors are associated with prolonged hospitaliza tion in a population with advanced HIV. International Journal of STD & AIDS, 12 , 302 306. Conti, S., Masocco, M., Pezzott i, P., Toccaceli, V., Vichi, M., Boros, S., Urciuoli, R., Valdarchi, C., & Rezza, G. (2000). Differen tial impact of combined antiretroviral therapy on the survival of Italian patients with specific AIDS-defining illnesses. Journal of Acquired Immune Deficiency Syndromes, 25 , 451 458. Crystal, S., Lo Sasso, A. T., & Sambamoorthi, U. (1999). Incidence and duration of hospitalizations among persons with AIDS: an event history approach. Health Services Research, 33 , 1611 1638.

PAGE 132

119 Cunningham, W. E., Markson, L. E., Andersen, R. M., Crystal, S. H., Fleishman, J. A., Golin, C., Gifford, A., Liu, H.H., Nakazono, T.T., Morton, S.C., Bozzette, S.A., Shapiro, M.F., & Wenger, N.S. (2000). Preval ence and predictors of highly active antiretroviral therapy use in patients w ith HIV infection in the United States. HCSUS Consortium. HIV Cost and Services Utilization. Journal of Acquired Immune Deficiency Syndrome, 25 , 115 123. De Palo, V. A., Millstein, B. H., Mayo, P. H., Salzman, S. H., & Rosen, M. J. (1995). Outcome of intensive care in patients with HIV infection. Chest, 107 , 506-510. Dean, G.L., Edwards, S.G., Ives, N.J., Matthew s, G., Fox, E.F., Navaratne, L., Fisher, M., Taylor, G.P., Miller, R., Taylor, C.B., de Ruiter, A., & Pozniak, A.L. (2002). Treatment of tuberculosis in HIV-infect ed persons in the era of highly active antiretroviral therapy. AIDS, 16 (1), 75 83. Dore, G. J., McDonald, A., Li, Y., Kaldor , J. M., & Brew, B. J. (2003). Marked improvement in survival following AIDS dementia complex in the era of highly active antiretroviral therapy. AIDS, 17 (10), 1539 1545. Emlet, C. A., & Farkas, K. J. (2002). Correla tes of service utilizat ion among midlife and older adults with HIV/AIDS: the role of age in the equation. Journal of Aging and Health, 14 , 315 335. Facts about the Department of Veterans Affairs. (2002). Retrieved November 2004 from the Department of Veterans Affairs Web site: http://www.va.gov/pressrel/vafacts.htm Farrar, D. E., & Glauber, R. R. (1967). Mu lticollinearity in Regression Analysis Problem Revisited. Review of Economics and Statistics, 49 (1), 92 107. Fernandez-Cruz, E., Desco, M., Garcia Mont es, M., Longo, N., Gonzalez, B., & Zabay, J. M. (1990). Immunological and serological markers predictive of progression to AIDS in a cohort of HIV-infected drug users. AIDS, 4 (10), 987 994. Fleishman, J. A., & Hellinger, F. H. (2003). Recent trends in HI V-related inpatient admissions 1996-2000: a 7-state study. Journal of Acquired Immune Deficiency Syndromes, 34 , 102 110. Fleishman, J. A., & Hellinger, F. J. (2001). Trends in HIV-related inpatient admissions from 1993 to 1997: a seven-state study. Journal of Acquired Immune Deficiency Syndromes, 28 , 73 80. Fleishman, J. A., Hsia, D. C., & Hellinger, F. J. (1994). Correlates of medical service utilization among people with HIV infection. Health Services Research, 29 , 527 548.

PAGE 133

120 Floris-Moore, M., Lo, Y., Klein, R. S., Budne r, N., Gourevitch, M. N., Moskaleva, G., Schoenbaum, E.E. (2003). Gender and hospita lization patterns among HIV-infected drug users before and after the availability of highly active antiretroviral therapy. Journal of Acquired Immune Deficiency Syndromes, 34 , 331 337. Fordyce, E. J., Singh, T. P., Nash, D., Gallaghe r, B., & Forlenza, S. (2002). Survival rates in NYC in the era of combination ART. Journal of Acquired Immune Deficiency Syndromes, 30 , 111 118. Forti, E. M., Johnson, J. A., & Graber, D. R. (2000). Aging in America: challenges and strategies for health care delivery. Journal of Health and Human Services Administration, 23 , 203-213. Garattini, L., Tediosi, F., Di Cintio, E ., Yin, D., & Parazzini, F. (2001). Resource utilization and hospital cost of HIV/AIDS care in Italy in the era of highly active antiretroviral therapy. AIDS Care, 13 , 733 741. Gebo, K. A., Chaisson, R. E., Folkemer, J. G ., Bartlett, J. G., & Moore, R. D. (1999). Costs of HIV medical care in the era of highly active anti retroviral therapy. AIDS, 13 , 963 969. Gebo, K. A., Diener-West, M., & Moore, R. D. (2001). Hospitalization rates in an urban cohort after the introduc tion of highly active antiretroviral therapy. Journal of Acquired Immune Deficiency Syndromes, 27 , 143 152. Gebo, K.A., Diener-West, M., & Moore, R. D. (2003). Hospitalization rates differ by hepatitis C status in an urban HIV cohort. Journal of Acquired Immune Deficiency Syndromes, 34 , 165 173. Gebo, K.A., Fleishman, J.A., Conviser, R., Re illy, E.D., Korthuis, P.T., Moore, R.D., Hellinger, J., Keiser, P., Rubin, H.R., Cr ane, L., Hellinger, F.J., & Mathewset, W.C. (2005). Racial and gender disparities in receipt of highly ac tive antiretroviral therapy persist in a multistate sample of HIV patients in 2001. Journal of Acquired Immune Deficiency Syndromes, 38 , 96 103. Gebo, K.A., & Moore, R.D. (2004). Treatment of HIV infection in the older patient. Expert Review of Anti-Infective Therapy, 2 , 733 743. Gelberg, L., Andersen, R.M., & Leake, B.D. (2000). The Behavioral Model for Vulnerable Populations: application to me dical care use and outcomes for homeless people. Health Services Research, 34 , 1273 1302. Giordano, T. P., White, A. C., Jr., Sajja, P ., Graviss, E. A., Arduino, R. C., Adu-Oppong, A., Lahart, C. J., Visnegarwala, F. (2003). Factors associated with the use of highly active antiretroviral therapy in patients newly entering care in an urban clinic. Journal of Acquired Immune Deficiency Syndromes, 32 , 399 405.

PAGE 134

121 Glesby, M. J. (2003). Coronary heart disease in HIV-infected persons. The AIDS Reader, 13 (4 Suppl), S15 19. Goetz, B. M., Morreale, A. P., Rhew, D. C., Berman, S., Ing, M., Eldridge, D., Justis, J. C., & Lott, E. (2001). Effect of highly act ive antiretroviral th erapy on outcomes in Veterans Affairs Medical Centers. AIDS, 15 , 530 532. Greenwood, M. & Yule, C. V. (1920). An i nquiry into the nature of frequency distributions representative of multiple happenings, with particular reference to the occurrence of multiple attacks of disease or repeated accidents. Journal of the Royal Statistical Society 89, 255 279. Grimes, R. M., Otiniano, M. E., Rodriguez-Ba rradas, M. C., & Lai, D. (2002). Clinical experience with human immunodeficiency virusinfected older patients in the era of effective antiretroviral therapy. Clinical Infectious Diseases, 34 , 1530 1533. Harawa, N. T., Greenland, S., Bingham, T. A., Johnson, D. F., Cochran, S. D., Cunningham, W. E., Celentano, D. D., Koblin, B. A., LaLota, M., MacKellar, D. A., McFarland, W., Shehan, D., Stoyanoff, S., Thiede, H., Torian, L., & Valleroy, L. A. (2004). Associations of race/ethnicity with HIV prevalence and HIV-related behaviors among young men who have sex w ith men in 7 urban centers in the United States. Journal of Acquired Immune Deficiency Syndromes, 35 , 526 536. Hargraves, J. L., Cunningham, P. J., & Hughes, R. G. (2001). Racial and ethnic differences in access to medical care in managed care plans. Health Services Research, 36 , 853 868. Hellinger, F. J. (2004). HIV patients in the HC UP database: a study of hospital utilization and costs. Inquiry, 41 , 95 105. Hellinger, F. J., & Fleishman, J. A. (2001). Location, race, and hospital care for AIDS patients: an analysis of 10 states. Inquiry, 38 , 319 330. Hellinger, F. J., Fleishman, J. A., & Hsia, D. C. (1994). AIDS treatment costs during the last months of life: evidence from the ACSUS. Health Services Research, 29 , 569 581. Hoffmann, C., Chow, K. U., Wolf , E., Faetkenheuer, G., Stellb rink, H. J., van Lunzen, J., Jaeger, H., Stoehr, A., Plettenberg, A., Wa smuth, J.C., Rockstroh, J., Mosthaf, F., Horst, H. A. & Brodt, H. (2004). Strong impact of highly active antiretroviral therapy on survival in patients with hu man immunodeficiency virus-associated Hodgkin's disease. British Journal of Haematology, 125 , 455 462. Hoffmann, C., Tabrizian, S., Wo lf, E., Eggers, C., Stoehr, A., Plettenberg, A., Buhk, T., Stellbrink, H., Horst, H., Jä ger, H., & Rosenkranz, T. (2001). Survival of AIDS patients with primary central nervous syst em lymphoma is dramatically improved by HAART-induced immune recovery. AIDS, 15 , 2119 2127.

PAGE 135

122 Hoover, D. R., Sambamoorthi, U., Walkup, J. T ., & Crystal, S. (2004). Mental illness and length of inpatient stay for medicaid recipients with AIDS. Health Services Research, 39 , 1319 1339. Hosmer, D. W., & Lemeshow, S. (2000). Applied logistic regression (2nd ed.). New York: Wiley. Hospital and outpatient health se rvices utilization among HIV-in fected patients in care in 1999. (2002). Journal of Acquired Immune Deficiency Syndromes, 30 , 21 26. Hsu, L. C., Vittinghoff, E., Katz, M. H., & Schw arcz, S. K. (2001). Predictors of use of highly active antiretroviral therapy (HAART) among persons with AIDS in San Francisco, 1996-1999. Journal of Acquir Immune Deficiency Syndromes, 28 , 345 350. Jimenez, A. D. (2003). Triple jeopardy: targ eting older men of color who have sex with men. Journal of Acquired Immune Deficiency Syndromes, 33 Suppl 2 , S222 225. Kaplan, J. E., Hanson, D., Dworkin, M. S., Fr ederick, T., Bertolli, J., Lindegren, M. L., Holmberg, S., & Jones, J. L. (2000). Epidemiology of human immunodeficiency virus-associated opportunistic infections in the United States in the era of highly active antiretroviral therapy. Clinical Infectious Diseases, 30 Suppl 1 , S5 14. Kass, N., Flynn, C., Jacobson, L., Chmiel, J. S., & Bing, E. G. (1999). Effect of race on insurance coverage and health serv ice use for HIV-infected gay men. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology, 20 , 85 92. Katz, M. H., Cunningham, W. E., Fleishma n, J. A., Andersen, R. M., Kellogg, T., Bozzette, S. A., & Shapiro, M. F. (2001) . Effect of case management on unmet needs and utilization of medical care and medications among HIV-infected persons. Annals of Internal Medicine, 135 (8 Pt 1), 557 565. Keiser, P., Kvanli, M. B., Turner, D., Reisc h, J., Smith, J. W., Nassar, N., Gregg, C., & Skiest, D. (1999). Protease inhibitor-based therapy is associated with decreased HIV-related health care costs in men treated at a Veterans Administration hospital. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology, 20 , 28 33. Keiser, P., Nassar, N., Kvanli, M. B., Turn er, D., Smith, J. W., & Skiest, D. (2001). Long-term impact of highly active antiretroviral therapy on HIV-related health care costs. Journal of Acquired Immune Deficiency Syndromes, 27 , 14 19. Kennamer, J. D., Honnold, J., Bradford, J ., & Hendricks, M. (2000). Differences in disclosure of sexuality among African Am erican and White gay/bisexual men: implications for HIV/AIDS prevention. AIDS Education and Prevention, 12 , 519 531.

PAGE 136

123 Keruly, J. C., Conviser, R., & Moore, R. D. (2002). Association of medical insurance and other factors with receipt of antiretroviral therapy. American Journal of Public Health, 92 , 852 857. Kilbourne, A. M., Herndon, B., Andersen, R. M., Wenzel, S. L., & Gelberg, L. (2002). Psychiatric symptoms, health services , and HIV risk factors among homeless women. Journal of Health Care for the Poor and Underserved, 13 , 49 65. Landon, B. E., Wilson, I. B., Cohn, S. E., Fich tenbaum, C. J., Wong, M. D., Wenger, N. S., Bozzette, S. A., Shapiro, M. F. & Cl eary, P. (2003). Physician specialization and antiretroviral therapy for HIV. Journal of General Internal Medicine, 18 , 233 241. Law, W. P., Duncombe, C. J., Mahanontharit, A., Boyd, M. A., Ruxrungtham, K., Lange, J. M., Phanuphak, P., Cooper, D.A., & Dore, G. J. (2004). Impact of viral hepatitis co-infection on response to an tiretroviral therapy and HI V disease progression in the HIV-NAT cohort. AIDS, 18 , 1169 1177. Lee, L. M., Karon, J. M., Selik, R., Neal, J. J., & Fleming, P. L. (2001). Survival after AIDS diagnosis in adolescents and adults during the treatment era, United States, 1984-1997. JAMA, 285 , 1308 1315. Levine, A. M. (2002). Evaluation and management of HIV-infected women. Annals of Internal Medicine, 136 , 228 242. Lewden C., L. Heripret, F. Bonnet, S. Bevil acqua, E. Jougla, J. Boileau, Costagliola, D., Morlat, P., May, T., Salmon, D., & Chene, D. (2002). Causes of death in HIVinfected adults in the era of highly active antiretroviral therapy (HAART), the French Survey ‘Mortalite 2000.’ Ninth Conference on Retroviruses and Opportunistic Infections. Seattle, February 2002 [Abstract 753 W]. Manfredi, R., & Calza, L. (2004). HIV in fection and AIDS in advanced age. Epidemiological and clinical issues, a nd therapeutic and management problems. Le Infezioni in Medicina, 12 , 152 173. Marimoutou, C., Carrieri, P., Poizot-Marti n, I., Loundou, A., Tremolieres, F., Rey, D., Obadia, Y. (2003). Hospitalization for de pressive syndrome in a cohort of HIVinfected patients contaminated throug h injecting drug use: MANIF 2000 cohort, France, 1995-1999. AIDS Care, 15 , 729 734. Markson, L. E., Houchens, R., Fanning, T. R., & Turner, B. J. (1998). Repeated emergency department use by HIV-infected persons: effect of clinic accessibility and expertise in HIV care. Journal of Acquired Immune Deficiciency Syndromes and Human Retrovirology, 17 , 35 41.

PAGE 137

124 Martin-Carbonero, L., Soriano, V., Valencia , E., Garcia-Samaniego, J., Lopez, M., & Gonzalez-Lahoz, J. (2001). In creasing impact of chronic viral hepatitis on hospital admissions and mortality among HIV-infected patients. AIDS Research and Human Retroviruses, 17 , 1467 1471. Mauskopf, J., Turner, B. J., Markson, L. E., Houchens, R. L., Fanning, T. R., & McKee, L. (1994). Patterns of ambulatory care fo r AIDS patients, and association with emergency room use. Health Services Research, 29 , 489 510. McGinnis, K. A., Fine, M. J., Sharma, R. K., Skanderson, M., Wagner, J. H., RodriguezBarradas, M. C., Rabeneck, L. & Justi ce, A. C. (2003). Understanding racial disparities in HIV using data from the veterans aging cohort 3-site study and VA administrative data. American Jouranl of Public Health, 93 , 1728 1733. McNaghten, A. D., Hanson, D. L., Dworkin, M. S., & Jones, J. L. (2003). Differences in prescription of antiretroviral therapy in a large cohort of HIV-infected patients. Journal of Acquired Immune Deficiency Syndromes, 32 , 499 505. Menke, T. J., Giordano, T. P., & Rabeneck, L. (2003). Utilization of he alth care resources by HIV-infected white, African-American, and Hispanic men in the era before highly active antire troviral therapy. Journal of the National Medical Association, 95 , 853 861. Menke, T. J., Rabeneck, L., Hartigan, P. M ., Simberkoff, M. S., & Wray, N. P. (2000). Clinical and socioeconomic determinants of health care use among HIV-infected patients in the Department of Veterans Affairs. Inquiry, 37 , 61 74. Miguez-Burbano, M. J., Shor-Posner, G., & Hadrigan, S. (2005). Non-tuberculous mycobacteria in HIV-infected patie nts: geographic, behavioural, and immunological factors. Lancet Infectious Diseases, 5 , 394 395; author reply 396. Minority patients ignored in HIV drug tials. (2002). AIDS Alert, 17(7), 86 7. Mole, L., Ockrim, K., & Holodniy, M. (1999) . Decreased medical expenditures for care of HIV-seropositive patients. The impact of highly active antiretroviral therapy at a U.S. Veterans Affairs Medical Center. Pharmacoeconomics, 16 , 307 315. Monahan, J., Swartz, M., & Bonnie, R. J. (2003). Mandated treatment in the community for people with mental disorders. Health Affairs (Millwood), 22 , 28 38. Montaner, J. S., & Harris, M. (2002). Mana gement of HIV-infected Patients with Multidrug-resistant Virus. Current Infectious Disease Report, 4 , 259 265.

PAGE 138

125 Montaner, J. S., Lawson, L. M., Gervais, A., Hyland, R. H., Chan, C. K., Falutz, J. M., Renzi, P. M., MacFadden, D., Rachlis, A. R., & Fong, I. W. (1991). Aerosol pentamidine for secondary prophylaxis of AIDS-related Pneumocystis carinii pneumonia. A randomized, placebo-controlled study. Annals of Internal Medicine, 114 , 948 953. Moore, R. D., Stanton, D., Gopalan, R., & Ch aisson, R. E. (1994). Ra cial differences in the use of drug therapy for HIV disease in an urban community. New England Journal of Medicine, 330 , 763 768. Palacio, H., Shiboski, C. H., Yelin, E. H., Hessol, N. A., & Greenblatt, R. M. (1999). Access to and utilization of primary care services among HIV-infected women. Journal of Acquired Immune Deficiency Syndromes, 21 , 293 300. Paul, S., Gilbert, H. M., Lande, L., Vaamonde , C. M., Jacobs, J., Malak, S., & Sepkowitz, K.A. (2002). Impact of antiretroviral ther apy on decreasing hosp italization rates of HIV-infected patients in 2001. AIDS Research and Human Retroviruses, 18 , 501 506. Perez, J. L., & Moore, R. D. (2003). Greater effect of highly active antiretroviral therapy on survival in people aged > or =50 y ears compared with younger people in an urban observational cohort. Clinical Infectious Diseases, 36 , 212 218. Pezzotti, P., Napoli, P. A., Acciai, S., Boro s, S., Urciuoli, R., Lazzeri, V., & Rezza, G. (1999). Increasing survival time after AIDS in Italy: the role of new combination antiretroviral therapies. Tuscany AIDS Study Group. AIDS, 13 , 249 255. Piette, J., Mor, V., & Fleishman, J. (1991). Patt erns of survival with AIDS in the United States. Health Services Research, 26 , 75 95. Piliero, P. J. (2003). Mechanisms of lipid el evations associated w ith the treatment of patients with HIV infection. Medscape General Medicine, 5 (2), 1. Pulvirenti, J. J., Glowacki, R., Muppiddi, U., Surapaneni, N., Gail, C., Kohl, B., & Jezisek, T. (2003). Hospitalized HIV-infected patients in the HAART era: a view from the inner city. AIDS Patient Care and STDs, 17 , 565 573. Rogers, P. L., Lane, H. C., Henderson, D. K ., Parrillo, J., & Masur, H. (1989). Admission of AIDS patients to a medical in tensive care unit: causes and outcome. Critical Care Medicine, 17 , 113 117. Rosenblum, L. S., Buehler, J. W., Morgan, M., & Moien, M. (1992). Increasing impact of HIV infection on hospitalizati ons in the United States, 1983 1988. Journal Acquired Immune Deficiency Syndromes, 5 , 497 -504.

PAGE 139

126 Scott, P., Arnold, E., Evans, B., Pozniak, A., Moyle, G., Shahmenesh, M., White, D., Shirley, D., Cane, P., & Pillay, D. (2004). Surveillance of HIV antiretroviral drug resistance in treated indi viduals in England: 1998-2000. Journal of Antimicrobiology and Chemotherapy, 53 , 469 473. Shapiro, M. F., Morton, S. C., McCaffrey, D. F., Senterfitt, J. W ., Fleishman, J. A., Perlman, J. F., Athey, L. A., Keesey, J. W., Goldman, D.P., Berry, S. H., & Bozzette, S.A.(1999). Variations in the care of HIV-infected adults in the United States: results from the HIV Cost and Services Utilization Study. JAMA, 281 , 2305 2315. Sherer, R., Pulvirenti, J., Stieglitz, K., Narra, J., Jasek, J., Green, L., Moore, B., Shott, S., & Cohen, M. (2002). Hospitalization in HIV in Chicago. Journal of the International Association of Physic ians in AIDS Care (Chic Ill), 1, 26 33. Skiest, D. J., & Crosby, C. (2003). Survival is prolonged by highly active antiretroviral therapy in AIDS patients with primar y central nervous system lymphoma. AIDS, 17 , 1787 1793. Skiest, D. J., Rubinstien, E., Carley, N., Gioiella, L., & L yons, R. (1996). The importance of comorbidity in HIV-infected patients over 55: a retrospective case-control study. The American Journal of Medicine, 101 , 605 611. Smith, S. R., & Kirking, D. M. (1999). Access and use of medications in HIV disease. Health Services Research, 34 (1 Pt 1), 123 144. Solomon, L., Stein, M., Flynn, C., Schuman, P ., Schoenbaum, E., Moore, J., Holmberg, S., & Neil, N. M. H. (1998). Health services use by urban women with or at risk for HIV-1 infection: the HIV Epid emiology Research Study (HERS). Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology, 17 , 253 261. Stein, M. D. (1994). Injected-dr ug use: complications and costs in the care of hospitalized HIV-infected patients. Journal of Acquired Immune Deficiency Syndromes, 7 , 469 473. Stein, M. D., Cunningham, W. E., Nakazono, T., Turner, B. J., Andersen, R. M., Bozzette, S. A., Shapiro, M.F. (2001). Sc reening for cervical cancer in HIVinfected women receiving care in the United States. Journal of Acquired Immune Deficiency Syndromes, 27 , 463 466. Stoff, D. M., Khalsa, J. H., Monjan, A., & Portegies, P. (2004). Introduction: HIV/AIDS and Aging. AIDS, 18 Suppl 1 , S1 2. Sulkowski, M. S., Mast, E. E., Seeff, L. B., & Thomas, D. L. (2000). Hepatitis C virus infection as an opportunistic diseas e in persons infected with human immunodeficiency virus. Clinical Infectious Diseases, 30 Suppl 1 , S77 84.

PAGE 140

127 Sulkowski, M. S., Thomas, D. L., Ch aisson, R. E., & Moore, R. D. (2000). Hepatotoxicity associated with antiretrovira l therapy in adults infected with human immunodeficiency virus and the role of hepatitis C or B virus infection. JAMA, 283 , 74 80. Sureka, A., Parada, J. P., Deloria-Knoll, M., Ch miel, J. S., Phan, L., Lyons, T. M., Ali, S., Yarnold, P. R., Weinstein, R.A., DeHovitz, J. A., Jacobson, J. M., Goetz, M. B., Campo, R. E., Berland, D., Bennett, C. L., Uphold, C. R. (2004). HIV-related pneumonia care in older patients hosp italized in the early HAART era. AIDS Patient Care and STDs, 18 , 99 107. Talan, D. A., & Kennedy, C. A. (1991). The management of HIV-related illness in the emergency department. Annals of Emergency Medicine, 20 , 1355 1365. Tang, J., & Kaslow, R. A. (2003). The imp act of host genetics on HIV infection and disease progression in the era of hi ghly active antire troviral therapy. AIDS, 17 Suppl 4 , S51 60. Tumbarello, M., Rabagliati, R., de Gaetano Donati, K., Bertagnolio, S., Montuori, E., Tamburrini, E., Tacconelli, E., & Cauda, R. (2004). Older age does not influence CD4 cell recovery in HIV-1 infected patie nts receiving highly active antiretroviral therapy. BMC Infectious Diseases, 4 , 46. Uldall, K. K., Koutsky, L. A., Bradshaw, D. H., Hopkins, S. G., Katon, W., & Lafferty, W. E. (1994). Psychiatric comorbidity a nd length of stay in hospitalized AIDS patients. American Journal of Psychiatry, 151 , 1475 1478. Uldall, K. K., Koutsky, L. A., Bradshaw, D. H., & Krone, M. (1998). Use of hospital services by AIDS patients with psychiatric illness. General Hospital Psychiatry, 20 , 292 301. Ungvarski, P. J. (1984). Infection co ntrol in the patient with AIDS. Journal of Hospital Infection, 5 Suppl A , 111 113. Update: Trends in AIDS inci dence, deaths, and prevalence United States, 1996. (1997). Morbidity and Mortal Weekly Report, 46: 165 173. Update: AIDS United States, 2000. (2002). Morbidity and Mortal Weekly Report, 51 (27), 592 595. Uphold, C. R., Deloria-Knoll, M., Palella, F. J ., Jr., Parada, J. P., Ch miel, J. S., Phan, L., & Bennett, C. (2004). U.S. hospital care for patients with HIV infection and pneumonia: the role of public, private, and Veterans Affairs hospitals in the early highly active antiretr oviral therapy era. Chest, 125, 548 556.

PAGE 141

128 Uphold, C. R., & Mkanta, W. N. (2005). Re view: use of health care services among persons living with HIV infection: state of the science and future directions. AIDS Patient Care and STDs, 19 , 473 485. VA HIV Report. (2004). Veterans Affairs HIV Report: Volume 1, Issue 2. Howe, J. M. & Rigsby, M.(Eds.). Retrieved November 15, 2004 from the Department of Veterans Affairs Web site: http://www.public health.va.gov/Report/1-2-June.pdf Vladeck, B. C., & Firman, J. P. (1983). The aging of the population and health services. Annals of the American Academy of Political and Social Science, 468, 132 148. Wachter, R. M., Russi, M. B., Bloch, D. A ., Hopewell, P. C., & Luce, J. M. (1991). Pneumocystis carinii pneumonia and respiratory failure in AIDS. Improved outcomes and increased use of intensive care units. American Review of Respiriratory Disease, 143 , 251 256. Wang, C., Vlahov, D., Galai, N., Bareta, J., Strathdee, S. A., Nelson, K. E., Sterling, T.R.(2004). Mortality in HIV-seropositive versus -seronegative persons in the era of highly active antiretroviral therapy: im plications for when to initiate therapy. Journal of Infectious Diseases, 190, 1046 1054. Weber, A. E., Yip, B., OÂ’Shaughnessy, M. V., Montaner, J. S., & Hogg, R. S. (2000). Determinants of hospital admission among HIV-positive people in British Columbia. Canadian Medical Association Journal, 162 , 783 786. Wolinsky, F. D. (1994). Health services utilization among older adults: conceptual, measurement, and modeling issu es in secondary analysis. Gerontologist, 34 , 470 475. Yang, Q. E. (2004). Human immunodeficien cy virus reservoir might be actively eradicated as residual malignant cells by cytotoxic chemotherapy. Medical Hypotheses, 62 , 358 363.

PAGE 142

129 BIOGRAPHICAL SKETCH William Mkanta received his BSc. degree in mathematics and statistics from the University of Dar es Salaam in 1991; and hi s Master of Statistics (biostatistics) from Makerere University, Kampala, Uganda, in 1997. Mr. Mkanta began his career at the University of Dar es Salaam in April 1991 where he was a lecturer in the Department of Statistics. In th is position, Mr. Mkanta taught different statistical courses and also advised students from the Departments of Economics and Commerce. He also served as a statistical consultant for other departments at the university as well as for governmental and non-governmental organizations. During the time between 1991 and 2001 he participated in various research studies with different capacities such as statistical consulta nt, collaborating investigator, and principal investigator. Mr. Mkanta started his Ph.D. program in hea lth services research at the University of Florida in August 2001. His research intere sts include health services in HIV/AIDS, access and utilization of health care services, chronic diseas e, and health informatics.