Citation
Methicillin-resistant Staphylococcus aureus Transmission

Material Information

Title:
Methicillin-resistant Staphylococcus aureus Transmission A Genomic and Ecological Approach to Unraveling the Epidemiology of Antibiotic Resistant Pathogens
Creator:
Azarian, Taj Hassan
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (141 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
COOK,ROBERT L
Committee Co-Chair:
SALEMI,MARCO
Committee Members:
MORRIS,JOHN GLENN
JOHNSON,JUDITH A
Graduation Date:
5/2/2015

Subjects

Subjects / Keywords:
Antibiotics ( jstor )
Drug prescriptions ( jstor )
Health care industry ( jstor )
Hospitals ( jstor )
Infants ( jstor )
Infections ( jstor )
Neonatal intensive care units ( jstor )
Phylogenetics ( jstor )
Staphylococcus aureus ( jstor )
Surveillance ( jstor )
Epidemiology -- Dissertations, Academic -- UF
antibiotic -- aureus -- epidemiology -- genome -- mrsa -- nicu -- sequencing -- staphylococcus
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Epidemiology thesis, Ph.D.

Notes

Abstract:
Staphylococcus aureus (SA), an infectious gram-positive cocci bacterium, is responsible for approximately 700,000 hospitalizations annually in the US. Of these hospitalizations, 66% are attributed to a drug resistant and often fatal form of the infection called Methicillin-resistant SA (MRSA). Neonatal and pediatric populations are particularly susceptible to severe infections. Antibiotic use is the most important contributing factor to the development of antibiotic resistance and half of all antibiotics prescribed in the community are unwarranted. Whole-genome sequencing (WGS) and phylogenetic analysis possess an unmatched resolution necessary to investigate the transmission of bacterial pathogens. Here, I apply these methods to understanding the changing epidemiology of MRSA by investigating transmission among patients hospitalized in the neonatal intensive care units (NICU) of two hospitals in northeast Florida. I evaluated risk factors of MRSA colonization, intrahost genetic variation, and changes in demographic history between community-associated and healthcare-associated genotypes. I then assessed the association of outpatient antibiotic prescriptions and SA resistance to explore population-level drivers for the emergence of antibiotic resistance. I identify a shift in the distribution of MRSA causing colonizations in the NICU, a significant proportion of which resulted from intrahospital transmission, despite comprehensive infection control interventions. Furthermore, the bacterial population of most prevalent community-associated MRSA genotype, pulse-field gel electrophoresis (PFGE) type USA300, has been expanding exponentially since 1998 and peaked in 2008. Overall, I demonstrate the utility of phylogenetic analysis for investigating bacterial pathogens epidemics, tracing their emergence, and characterizing epidemiological changes. Last, I demonstrate geographical, temporal, and demographic variations in outpatient antibiotic prescribing. Children aged 1-4 received the greatest proportion of outpatient antibiotics, had the highest proportion of SA that was MRSA, and the highest MRSA incidence rate. Outpatient antibiotic use was significantly associated with the proportion of SA that was MRSA. In conclusion, I elucidate the epidemiology of MRSA in NICUs and explore population level effects of antibiotic use. These finding have implications for both the approach to infection control in healthcare facilities as well as targets for interventions in the community to reduce unwarranted antibiotic use. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2015.
Local:
Adviser: COOK,ROBERT L.
Local:
Co-adviser: SALEMI,MARCO.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2016-05-31
Statement of Responsibility:
by Taj Hassan Azarian.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
5/31/2016
Classification:
LD1780 2015 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

METHICILLIN RESISTANT Staphylococcus aureus TRANSMISSION: A GENOMIC AND ECOLOGICAL APPROACH TO UNRAVELING THE EPIDEMIOLOGY OF ANTIBIOTIC RESISTANT PATHOGENS By TAJ AZARIAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2015

PAGE 2

2015 Taj Azarian

PAGE 3

To my expecting wife and our future son. Continue the dream.

PAGE 4

ACKNOWLEDGMENTS I have been extremely blessed by a life rich with family and friends. I would of course not be the person I am today without my parents , Hossein and Athena, who embody the American dream of every immigrant by working hard for the sole benefit of their children. They have cultivated my curiosity by encouraging my (sometimes reckless) exploration of world around me as well as my never ending thirst for knowledge. Not enough can be said about the educators and mentors who have had such a tremendous impact on my academic and professional endeavors . Foremost, my mother who has been an educator in the Pinellas County school system for over 20 years. Additionally, I owe special thanks to a mentor who changed the way I think about science , epidemiology, and public health. Dr. Richard Hopkins, former state epidemiologist f or Florida (among many other titles) , who I lovingly refer to as Socrates for both his physical and philosophical e mbodiment of that great thinker. He instilled in me a dogma of always asking myself how my research will be translated into practice. Last, r aising a PhD student follows a n “ it takes a village ” mantra ; therefore, I am in debt to my PhD cohort as well as my committee. This research has been greatly influenced and made possible by Drs. Robert Cook, Marco Salemi, Glenn Morris, and Judy Johnson who have provided mentorship, friendship, and financial support. Most importantly, I acknowledge my best friend, my wife Sophia. This would not have been possible without your support, encouragement, and positive influ ence in my life. No person could ask for a better friend or partner. Last, I would especially like to acknowledge the effective population of the Azarian and Morakis lineages; whom, without which, I would literally not be here today. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURE S .......................................................................................................... 9 LIST OF ABBREVIATIONS ........................................................................................... 12 ABSTRACT ................................................................................................................... 1 4 CHAPTER 1 ANTIBIOTIC RESISTANCE AND Staphylococcus aureus ..................................... 16 Introduction ............................................................................................................. 16 Antibiotic Use, Emergence of Antimicrobial Resistance, and MethicillinResistant S. aureus Infections ....................................................................... 17 Methicillin Resistant S. aureus Definitions and Implications of Changing Epidemiology ................................................................................................. 18 Molecular Genotyping and its Role in Differentiating Epidemiologically Related and Unrelated Cases ....................................................................... 20 Community Antibiotic Use as a Driver for Antimicrobial Resistance and the Transmission of CA MRSA Strains ............................................................... 22 Primary Outbreak Investigation and Study Settings ......................................... 24 Methicillin Resistant S. aureus Transmission in Northeast Florida ................... 26 Implications for Infection Control of Phylogenetic Analysis of NICU Isolates .... 26 Studies Assessing the Association between MRSA and Antibiotic Use ........... 28 The Burden of Pediatric Antibiotic Use and Proliferation of Antibiotic Resistance .................................................................................................... 30 Preliminary Data from the Florida Department of Health (FDOH) Highlights Disparities in CA MRSA Incidence ................................................................ 30 Study Goal, Aims, and Research Questions ........................................................... 31 2 WHOLE GENOME SEQUENCING FOR OUTBREAK INVESTIGATIONS OF METHICILLIN RESISTA NT Staphylococcus aureus IN THE NEONATAL INTENSIVE CARE UNIT ......................................................................................... 38 Overview ................................................................................................................. 38 Methods .................................................................................................................. 40 Outbreak Investigati on ..................................................................................... 40 Sample Collection and Epidemiologic Investigation ......................................... 41 Whole Genome Sequencing, Single Nucleotide Polymorphism Detection, and Phylogenetic Analysis ............................................................................ 42 Bayesian Phyl ogenetic Analysis ....................................................................... 43 5

PAGE 6

MLST+ Comparison ......................................................................................... 44 Results .................................................................................................................... 44 Study Population, Risk Factors, and Investigation Timeline ............................. 44 Molecular, Genomic, and Phylogenetic Results ............................................... 45 Putative Transmission Pathways ...................................................................... 46 Discussion .............................................................................................................. 47 3 WHOLE GENOME SEQUENCING DEMONSTRATES MULTIPLE INTRODUCTIONS AND TRANSMISSION OF METHICILLINRESISTANT Staphylococcus aureus IN THE NEONATAL INTENSIVE CARE UNIT .................. 60 Overview ................................................................................................................. 60 Methods .................................................................................................................. 61 MRSA Surveillance and Study Population ....................................................... 61 Sample Selection, spa typing, and Risk Factor Analysis .................................. 62 WGS and Phylogenetic Analysis ...................................................................... 62 Results .................................................................................................................... 65 Clinical and Demographic Comparison of Infants ............................................. 65 Whole Genome Sequencing and Phylogenetic Analysis .................................. 66 Population Dynamics of t008 and t045 ............................................................. 67 Discussion .............................................................................................................. 69 4 ASSOCIATION OF COMMUNITY ANTIBIOTIC USE AND Staphylococcus aureus DRUG RESISTANCE, FLORIDA, 2010 2012 ............................................. 90 Overview ................................................................................................................. 90 Methods .................................................................................................................. 91 Data Sources .................................................................................................... 91 Analysis ............................................................................................................ 93 Descriptive analysis ................................................................................... 93 Correlates of antibiotic prescribing and identification of spatial clusters .... 94 Association of outpatient antibiotic prescribing and S. aureus methicillin resistance ............................................................................................... 95 Results .................................................................................................................... 95 Discussion .............................................................................................................. 98 5 CONCLUSIONS ................................................................................................... 119 Accomplishment s of the Dissertation .................................................................... 119 Genomic Epidemiology of Bacterial Pathogens .............................................. 119 Healthcare Epidemiology of MRSA ................................................................ 121 Community Antibiotic Prescribing and Antibiotic Resistance .......................... 123 Future Directions .................................................................................................. 124 LIST OF REFERENCES ............................................................................................. 127 BIOGRAPHICAL SKETCH .......................................................................................... 141 6

PAGE 7

LIST OF TABLES Table page 1 1 Risk factors for healthcareassociated and community associated MRSA ......... 34 2 1 Pairwise comparison of nucleotide difference between 17 MRSA isolates from patients hospitalized in the NICU ............................................................... 52 2 2 Antibiotic susceptibilities and spa types of 17 MRSA isolates from infants hospitalized in the NICU during the putative outbreak investigation ................... 53 3 1 Frequency of MRSA spa type t008 wholegenome sequences by healthcare facility and hospital unit ...................................................................................... 74 3 2 Characteristics of cases (colonized infants) and controls (uncolonized infants) ................................................................................................................ 74 3 3 Univariate logistic regression of colonization risk factors ordered by statistical significance ......................................................................................................... 75 3 4 Comparison of characteristics between patients with spa typed and nonspa typed isolates ordered by statistical significance ................................................ 75 3 5 Frequency and proportion of spa types identified among colonized infants in Hospital A NICU ................................................................................................. 76 3 6 Comparison of characteristics between patients with community genotype ( spa type t008) and healthcare genotypes (nonspa type t008) orde red by statistical significance ......................................................................................... 76 3 7 Univariate logistic regression of community genotype ( spa type t008) colonization risk factors ordered by statistical significance ................................. 77 3 8 Multivariate logistic regression of community genotype ( spa type t008) colonization risk factors orde red by statistical significance ................................. 77 3 9 Multivariate linear regression of risk factors for increased length of stay among MRSA c olonized infants hospitalized in the Hospital A NICU ................. 78 3 10 Comparison of mean sing le nucleotide polymorphism differences, evolutionary rate ................................................................................................. 78 4 1 Prescriptions per 100 population for select antibiotic groupings, Florida, 2011. ................................................................................................................ 102 4 2 Spatial error regression of healthcare and demographic covariates on the per capita rates of antibiotic prescri bing by ZIP Code, Florida, 2011 ...................... 102 7

PAGE 8

4 3 Spatial error regression of healthcare and demographic covariates on the per capita rates of antibiotic prescribing by county, Flori da, 2011 .......................... 102 4 4 Number of isolates tested and percent resistant to oxacillin (MRSA) by age group, commercial outpati en t laboratory, Florida, 2011 .................................... 103 4 5 Number of isolates tested (n) and percent resistant to oxacillin (MRSA), Commercial Outpatient Laboratory, Florida, 2011 ............................................ 103 4 6 Linear regression of per capita outpatient antibiotic prescriptions on proportion of S. aureus isolates that wer e resistant to oxacillin (MRSA) .......... 103 8

PAGE 9

LIST OF FIGURES Figure page 1 1 Current Centers for Disease Control Active Bacterial Core surveillance case definition for methicillin resistant S. aureus ........................................................ 35 1 2 Maximum likelihood phylogenetic analyses of MRSA t008 in northcentral Florida by healt hcare facility ............................................................................... 36 1 3 Interpr eting phylogenetic dendograms. .............................................................. 36 1 4 Schematic description of the population level effect of outpatient antibiotic use on introduction of MRSA i nto healthcare facilities and subsequent transmission i n the NICU .................................................................................... 37 1 5 Schematic diagram of the relationship between outpatient antibiotic use and S. aureus antibiotic resistance ............................................................................ 37 2 1 Detailed timeline for putative neonatal intensive care unit outbreak. .................. 54 2 2 Pulse field gel electrophoresis of methicillin resistant S. aureus isolates from putative outbreak ................................................................................................ 55 2 3 Unrooted maximum likelihood phylogenetic tree illustrating the relationship between 17 putative outbreak isolates ............................................................... 56 2 4 Integration of epidemiological and phylogenetic data to produce a timeline of putative NICU outbreak ...................................................................................... 57 2 5 Comp arison of neighbor joining trees constructed with genome wide singlenucleotide polymorphism data and MLST+ allelic profiles .................................. 58 2 6 Epidemic curve of putative NICU outbreak incorporating increasing levels of resolution f rom molecular and WGS analysis ..................................................... 59 3 1 Diagram of study population, data sets, and analyses ....................................... 79 3 2 Patient colonizations by MRSA spa type and day .............................................. 80 3 3 Maximum likelihood phylogenetic relationship among spa type t045 and t008 MRSA isolates and patient length of stay with date of positive culture ............... 81 3 4 Minimum spanning tree of isolates obtained from infant s hospitalized from 20082010 .......................................................................................................... 82 3 5 Global phylogeny of common healthcareassociated MRSA genotypes and t045 (S T 225) isolates from Hospital A ............................................................... 83 9

PAGE 10

3 6 DensiTree visualizations of posterior distributions of trees obtained from Bayesian phylogenetic analysis of t008 and t045 datasets ................................ 84 3 7 Bayesian maximum clade credibility phylogeny of 97 spa type t008 from multiple healthcare facilities ................................................................................ 85 3 8 Maximum likelihood phylogeny of 97 spa type t008 isolates from five healthcare facilities in northeast Florida ............................................................. 86 3 9 Comparison of evolutionary rates and 95% highest posterior density of t045 and t008 lineages ............................................................................................... 87 3 10 Comparison of effective population sizes ( Ne ) of t045 and t008 lineages .......... 88 3 11 Bayesian maximum clade credibility phylogenies and genotypic anti biotic resistance and virulence ..................................................................................... 89 4 1 Per capita outpatient antibiotic prescriptions by antibiotic groupi ng and age group, Florida, 2011 ......................................................................................... 104 4 2 Per capita outpatient antibiotic prescriptions by provider type, Florida 2011 .... 104 4 3 Timeseries of the cumulative monthly per capita rates of antibiotic prescribing, Florida, 20102012 ........................................................................ 105 4 4 Timeseries of the cumulative monthly per capita rates of antibiotic prescribing by antibiotic grouping, Florida, 20102012 ........................................................ 105 4 5 Timeseries of the cumulative monthly per capita rates of antibiotic prescribing by antibiotic groupings not displaying seasonality, Florida, 20102012 ............ 106 4 6 Average yearly per capita antibiotic prescriptions by antibiotic gr ouping and county, Florida 2011 ......................................................................................... 107 4 7 ZIP code level analysis of per capita outp atient antibiotic prescriptions ........... 108 4 8 Average yearly per capita rate of antibiotic presc riptions ................................. 109 4 9 County level healthcare correlates of antibiotic prescribing and S. aureus resistance ......................................................................................................... 110 4 10 Local test for spatial autocorrelation of residuals from spatial error regression of demographic correlates and per capita outpatient antibiotic prescription r ates by ZIP Code, Florida, 2011 ...................................................................... 111 4 11 Spatial error residuals from regression of healthcare and demographic covariates by county on per capita antibiot ic prescriptions, Florida, 2011 ........ 112 10

PAGE 11

4 12 Local test for spatial autocorrelation of residuals from spatial error regression of demogr aphic and healthcare correlates and per capita outpatient antibiotic prescription rates by county, Florida, 2011 ....................................................... 113 4 13 Percent of S. aureus Isolates that were oxacillinresistant (MRSA) by age group, commercial outpatient laboratory, Florida, 2011 .................................... 114 4 14 Percent of S. aureus Isolates that were oxacillinresistant (MRSA) by gender, commercial outpatient laboratory, Florida, 2011 ............................................... 114 4 15 Frequency of S. aureus tests and proportion of isolates that were oxacillinresistant (MRSA) by year and report week, commercial outpatient laborat ory, Florida, 20102012 ........................................................................................... 115 4 16 S. aureus cases by Florida County, 2011 ......................................................... 116 4 17 Comparison of outpatient antibiotic prescriptions per capita and proportion of S. aureus isolates resistant to oxacillin ............................................................. 1 17 4 18 Antibiotic prescriptions per capita for 2011 and threeyear average percent of S. aureus resistant to oxacillin by county, Florida ............................................. 118 11

PAGE 12

LIST OF ABBREVIATIONS BEAST Bayesian Evolutionary Analysis Sampling Trees BSP Bayesian skyline plot CDC Centers for Disease Control and Prevention ESS Effective sample size gDNA Genomic deoxyribonucleic a cid HAI Healthcare associate infection GMRF Gaussian Markov random field ICU Intensive care u nit MCC Maximum clade credibility MCMC Markov c hain Monte C arlo ML Maximum likelihood MLE Marginal likelihood estimate MLST Multi locus sequence typing MRSA Methicillin r esistant Staphylococcus aureus MST Minimum spanning tree Ne Effective population size NGS Next generation s equencing NHSN National Healthcare Safety Network NICU Neonatal intensive care u nit NJ Neighbor joining PICU Pediatric i ntensive c are Unit PFGE Pulse field gel e lectrophoresis PVL Panton Valentine leukocidine SNP Single nucleotide polymorphism 12

PAGE 13

SSTI Skin and soft tissue infections TMRCA The most recent common ancestor WGS Whole genome s equencing 13

PAGE 14

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy METHICILLIN RESISTANT Staphylococcus aureus TRANSMISSION: A GENOMIC AND ECOLOGICAL APPROACH TO UNRAVELING THE EPIDEMIOLOGY OF ANTIBIOTIC RESISTANT PATHOGENS By Taj Azarian May 2015 Chair: Robert Cook Cochair: Marco Salemi Major: Epidemiology Staphylococcus aureus (SA ), an infectious gram positive cocci bacterium, is responsible for approximately 700,0 00 hospitalizations annually in the US . Of these hospitalizations, 66% are attributed to a drug resistant and often fatal form of the infection called Methicillin resistant SA (MRSA). Neonatal and pediatric populations are particularly susceptible to severe infections. Antibiotic use is the most important contributing factor to the development of antibiotic resistance and half of all antibiotics prescribed in the community are unwarranted. Whole genome seq uencing (WGS) and phylogenetic analysis possess an unmatched resolution necessary to investigate the transmission o f bacterial pathogens. Here, I apply these methods to understanding the changing epidemiology of MRSA by investigating transmission among patients hospitalized in the neonatal intensive care unit s (NICU) of two hospitals in northeast Florida. I evaluated risk factors of MRSA colonization, intrahost genetic variation, and changes i n demographic history between community associated and healthcare associated genotypes . I then assessed the association of outpatient antibiotic prescriptions and SA resistance to explore populationlevel drivers for the emergence of 14

PAGE 15

antibiotic resistance. I identify a shift in the distribution of MRSA causing colonizations in the NICU, a significant proportion of which result ed from intrahospital transmission, despite comprehensive infection control interventions. Furthermore, the bacterial population of most prevalent community associated MRSA genotype, pulsefiel d gel electrophoresis (PFGE) type USA300, has been expanding exponentially since 1998 and peaked in 2008 . Overall, I demonstrate the utility of phylogenetic analysis for investigating bacterial pathogens epidemics , tr acing their emergence, and characterizing epidemiological changes . Last, I demonstrate geographical, temporal, and demographic variations in outpat ient antibiotic prescribing. Children aged 14 received the greatest propor tion of outpatient antibiotics, had the highest proportion of SA that was MRSA, and the highest MRSA incidence rate. O utpatient antibiotic use was significantly associated with the proportion of SA that was MRSA . In conclusion, I elucidate the epidemiology of MRSA in NICUs and explore population level effe cts of antibiotic use. These finding have implications for both the approach to infection control in healthcare facilities as well as targets for interventions in the community to reduce unwarranted antibiotic use. 15

PAGE 16

CHAPTER 1 ANTIBIOTIC RESISTANCE AND Staphylococcus aureus Introduction Staphylococcus aureus (SA), an infectious gram positive cocci bacterium, is responsible for approximately 700,000 hospitalizations in the US alone, with an approximate annual healthcare cost of $3.2 – $4.2 billion [1] . Of these hospitalizations, 66% are attributed to a severe, drug resistant, and often fatal form called Methicillinresistant SA (MRSA) [2] . Neonatal and pediatric populations are particularly susceptible to severe MRSA infections leading to disproportionate morbidity and mortality. Antibiotic use is an important contributing factor to the emergence and spread of antibiotic resistance and half of all antibiotics are unwarranted or improperly prescribed. Furthermore, neonatal and pediatric populations receive a greater proportion of the antibiotics, potentially increasing their likelihood of contracting an antibiotic resistant pathogen. While rates of MRSA infections acquired in hospitals have decreased, little to no reduction has been obser ved in community associated MRSA (CA MRSA) disease, suggesting a critical need for community based interventions targeting MRSA transmission, or the possible underlying cause, antibiotic use. The Centers for Disease Control and Prevention (CDC) in 2012 est imated an incidence of 5.21 per 100,000 for severe CA MRSA infections [3] . Unfortunately, there are still several gaps in effective control measures for MRSA infections in the community. In the healthcare setting , hospitals have implemented antibiotic stewardship programs to curb unwarranted antibiotic use, reducing the prevalence of antibiotic resistant pathogens , and improved infection control strategies prevent transmission . However, similar control measures have not been translated to the 16

PAGE 17

community. Moreover, while it is known that antibiotic contributes to the emergence of antibiotic resistance at both an individual and population level, antibiotic prescribing data depicting the “who”, “how much”, and “where” are seemingly limited . Here, I hypothesize that healthcareonset MRSA cases stem from community MRSA infections and that these epidemics within communities persist due to antibiotic use. Indeed, our preliminary studies suggest a heterogeneous epidemic of MRSA within a healthcare network in a large metropolitan city. My thesis illustrates 1) the transmission dynamics of MRSA within the hospital and between the community and the hospital utilizing whole genome sequencing (WGS) and phylogenetic analysis , and 2) the contribution of antibiotic use in the community to the emergence of antibiotic resistance. Antibiotic Use, Emergence of Ant imicrobial Resistance, and Methicillin Resistant S . a ureus Infections Antibiotic use is recognized as a significant driver of antibiotic resistance and may explain, in part, MRSA infections rates in the community setting [4 – 9] . At an individual level, antibiotic use disrupts normal bacterial flora protecting the body from infection and allows antibiotic resistant bacteria to supplant “healthy” bacteria [10] . At a population level, antibiotic resistant bacteria, including MRSA, can then spread throughout the community, potentially sped by incr easing antibiotic use. Furthermore, w hile antibiotic use is recognized as a population level driver for antibiotic resistance, less is known about antibiotic prescribing patterns in the community, data which is needed to identify potential targets for public health interventions aimed at reducing unwarranted or inappropriate antibiotic use. A s prevalence of MRSA increases in the community, the likelihood of a n admitted patient being colonized increases, r aising the MRSA burden on healthcare facilities and increasing the potential for subsequent 17

PAGE 18

transmission. However, the changing epidemiology of MRSA is not well understood, including MRSA transmission between the community and healthcare. Methicillin Resi stant S. aureus Definitions and Implications of Changing Epidemiology MRSA is a leading cause of healthcareassociated infections (HAIs) [11,12] and a significant contributor to increased healthcare costs [1,13] . MRSA can cause a variety of infectio ns ranging from noninvasive skin and soft tissue infections (SSTIs) (i.e. , furuncles, abscesses, cellulitis, and folliculitis) to invasive infections such a bacteremia, meningitis, and osteomyelitis. Historically, MRSA infections have been categorized into two specific types: healthcareassociated (HA MRSA) and community associated (CA MRSA) based on clinical, epidemiological , and genetic criteria ( Table 1 1 ). CA MRSA traditionally referred to infections among individuals with onset in the community who lacked healthcareassociated risk factors (e.g., rec ent hospitalization) . This definition also described genotypic characteristics frequent among strains causing CA MRSA infections. As these delineations have blurred surveillance definitions have evolv ed and at least three MRSA classifications now classify the epidemiological etiology of MRSA infections (Figure 1 1 ). First, a case is defined as a healthcareonset MRSA (HO MRSA) infection if the culture was obtained on or after the fourth calendar day o f hospitalization. HO MRSA infections are acquired as a result of transmission within the healthcare setting (i.e. , HAI). Second, healthcareassociatedcommunity onset MRSA (HACO MRSA) infections are defined as MRSA infections among people with previous h ealthcareassociated risk factors, which have onset in the community. These predisposing risk factors include history of hospitalization, surgery, dialysis, residence in long term care or assisted living facilities in the previous year, or the a central v ascular 18

PAGE 19

catheter [14] . Third, community associated MRSA (CA MRSA) infections are recognized among healthy individuals who lack HA risk factors and who have onset of infection in the community [15,16] . HO MRSA and HACO MRSA can largely be grouped under healthcareassociated (HA MRSA) infections . Surveillance case definitions allow epidemiologist to quantify the magnitude of a health problem, monitor changes in epidemiology of a pathogen, and detect epidemics. H owever, these surveillance data do not fully elucidate the complex epidemiology of S. aureus , as recent studies have demonstrated the extensive genetic diversity among S. aureus populations . Genetic variations between MRSA strains have historically been used to cluster epidemiologically related strains and differentiate between epide miologically unrelated isolates . For exam ple, MRSA PFGE Type USA300 (ST8 IV) emerged from a methicillinsusceptible SA of clonal complex 8 and was first reported in the USA as a cause of SSTIs among college football players in Pennsylvania and among prisoners in Missouri in 2000 [17] . In the US, USA300 is the predomin ate cause of CA MRSA infections, a strikingly different trend when compared to other countries. CA MRSA strains of the CC8 lineage also have a smaller staphylococcal chromosomal cassette mec element (SCC mec IV ) as well as genes that code for PantonValentine leukocidine (PVL) toxin, which significantly increase the virulence and transmissibility of these infections. Additionally, CA MRSA strains tend to be susceptible to a greater number of antibiotics compared to HA MRSA strains that share a unique genetic lineage [18] . Most recently, CA MRSA strains have been identified as a significant contributor to HAIs [19 – 21] . In 2008, almost 27% of epidemiologically defined HO MRSA infections 19

PAGE 20

were due to strains genetically consistent with community associated infections (e.g., PFGE type USA300) [22] . In many healthcare facilities in the US, strains recognized genetically as CA MRSA are now replacing HA MRSA strains as the primary cause of HAIs [23] . This occurrence has putatively resulted from increased MRSA prevalence in the community leading to increased colonization pressure on healthcare facilities (i.e. , the proportion of patients admitted to the hospital with MRSA colonization or infection). Yet, surveillance data on MRSA in the community are lacking. Among healthcare facilities, interventions including the implementation of a national surveillance system, development and enforcement of infection control practices, and integration of antibiotic stewardship into routine clinical practice have seemingly reduced nosocomial transmission of MRSA and other pathogens [3,24] . These interventions have not been translated for the community healthcare setting, and as a result, there is an absence of surveillance data, deficiency in “community” infection control practices, and need for antibiotic stewardship. Through the following studies , I sought to underscore the need for such undertakings. Ultimately, as CA and HA strains intermix, the exposure of more virulent CA MRSA strains to antibiotic pressure in healthcare facilities is increasing the possibility for development of more virulent and difficult to treat infections leading to greater morbidity and mortality [19] . Molecular Genot yping and its Role in Differentiating Epide miologically Related a nd Unrelated Cases Molecular characterization i.e. , the ability to identify genetic and therefore epidemiological relatedness , is an essential component of epidemic i dentification and investigation, demonstrated most recently by well publicized outbreaks of Escherichia coli in Europe [25] and Salmonella serotype Saintpaul in the US [26] . W ithin healthcare 20

PAGE 21

facilities, passive laboratory surveill ance of epidemiologically important organism s, including MRSA, is the cornerstone for infection prevention [27,28] . The hallmar k of these molecular methods is their ability to discriminate between genetically related and unrelated isolates, essentially identifying outbreaks of epidemiologically linked cases. Over time, these tests have evolved to provide increasing levels of resol ution, higher throughput, and decreased turnaroundtime, which have in turn enhanced epidemi c detection and investigation [29] . A variety of molecular typing methods are currently utilized to characterize MRSA strains including pulsedfield gel electrophoresis (PFGE), multi locus sequence typing (MLST), and spa typing of the highly polymorphic Staphylococcus protein A ( spa ) gene [30,31] . Spa typing, PFGE, and MLST typing have all been used for the investigation of SA epidemics [30,32 – 35] , yet they lack the resolution to discri minate between genetically similar isolates [31,36,37] . I t is therefore difficult to interpret molecular typing results during an outbreak if the s patype or PFGE pattern of the pathogen has a disproportionately high prevalence in that setting (e.g. , S. aureus PFGE USA300 or MLST ST 08). Additionally, these methods provide little information regarding the temporal ity of transmission events beyond the date they were collected as clinical samples and the onset date of symptoms. Establishing this timeline allow s investigators to piece together the series of events, which in an outbreak scenario facilitates the identification of the mode of tr ansmission. Most recently, the application of next generation WGS and phylogenetic analysis to investigating bacterial pathogens has provided the resolution required to discriminate between genetically similar isolates, enhancing outbreak investigation as well as the 21

PAGE 22

study of pathogen emergence and spread [38] . Phylogenetic and population genetic inference (phylodynamics) based on genomewide SNP data can resolve putative outbreaks, investigate their etiology, and provide spatiotemporal context during investigations [39 – 41] . These methods may also be utilized to elucidate macrolevel (population) transmission dynamics and understand pathogen success [42] . The uti lity of phylogenetics in understanding the emergence of pathogens is also well recognized [42 – 44] . While the field of genomic epidemiology is in its infancy, the application of these methods to investigate bacterial pathogens are increasing in frequency [45 – 50] . Gray et al. (2011) and Harris et al. (2010) initially demonstrated that genomewide SNP data from MRSA strains sampled over relatively short period of time (12 years) contain enough phylogenetic information to test epidemiological hypothesis (i.e. , are a measurably evolving population) [39,40] . Moreover, recent advances in sequencing technology have made genomewide phylogenetic analysis plausible for real time epidemic investigation [38,50– 52] . In Prosperi et al. (2013), we applied WGS and phylogenetic analysis to the investigation of MRSA in seven northe ast Florida hospitals [49] . In the present study, we build upon this wor k, applying these methods to the investigation of a putative MRSA out break in a N ortheast FL NICU. Community Antibiotic Use as a Driver for Antimicrobial Resistance and the Transmission of CA MRSA S trains Antimicrobial resistance is a worldwide problem posing a serious threat to human health. As a result of diminished effectiveness of antibiotics, patients with resistant infections have longer hospital stays, take longer to recover, suffer long term sequel la, and have poorer overall health outcomes [9] . In 2013, the CDC conservatively estimated that in the US alone two million illnesses and at least 23,000 deaths were 22

PAGE 23

directly the result of antibiotic resistant infections [3,9] . Each year many more die from complications due to antibiotic resistant organisms and are sickened by infections such as Clostridium difficile for which antibiotic use is a contributing factor. In addition to the cost in human life, annual financial impact has been estimated at $20 billion direct healthcar e costs and $35 billion for lost productivity [53] . S. aureus alone, 66% of which was classified as MRSA, was responsible for an estimated 700,000 hospitalizations in the US, with an approximate healthcare cost of 3.2 4.2 billion dollars a year [1]. Reducing the burden of MRSA would greatly reduce morbidly and mortality caused by this pathogen . Antibiotic use is an important contributing factor to the development of antibiotic resistance and studies have show n that half of all antibiotics are unwarranted or improperly prescribed [9,54,55] . Clinicians in the community frequently succumb to patient requests for antibiotics, even when they may not be indicated [56] . Additionally, patients oftentimes misuse antibiotics by failing to complete therapy, skipping dos es, or reusing leftover medication [57] . This trend is significant in pediatric patients treated for otitis media, pharyngitis, and upper respiratory infections [55,58,59] . The e ffects of anti biotic use on the emergence of drug resistance has previously been demonstrated [4 – 8,60– 62] . Furthermore, populationlevel selective antibiotic pressure has the greatest influence antibiotic resistance in a community by increasing the prevalence of resistance genes able to be acquired by other bacteria through horizontal gene t ransfer [10] . At an individual level, antibiotic use can select for resistant organisms by disrupting normal flora and decreasing microbiota diversity, leading to replacement by resistant strains [10,63,64] . 23

PAGE 24

In recent years, antibiotic stewardship has reduced unwarranted antibi otic use in U S hospitals. I n combination with other prevention measures, we have observed a gradual reduction in the incidence of some antibiotic resistant pathogens, including MRSA, in that setting [5,65,66] . However, despite CDC’s “Get Smart: Know When Antibiotics Work” campaign to promote judicious use of antibiotics in the community, rates of antibiotic resistant infections have remained largely unchanged [65,67,68] . CDC’s 2013 Antibiotic Threat Report identified that the lack of data on antibiotic use in human healthcare and outpatient programs is a current gap in knowledge hindering the improvement in antibiotic prescribing to reduce the prevalence of antibiotic resistance [9]. C ommunity antibiotic prescribing data are needed to identify highrisk populations with high antibiotic consumption and set benchmarks for evaluations of future public health programs. Addressing these highrisk groups is more difficult and will require a multifaceted approach involving community healthcare providers, state and local public health agency, and academic institutions. One possible approach would be for the community retail pharmacist to mirror the antibiotic stewardship role of a hospital pharmacist by monitoring antibiotic use, optimizing treatments, and reducing unwarranted use. However, unlike the hospital setting where specific providers (e.g. , hospitalist vs. infectious disease doctors), antibiotics (e.g. , the broad spectrum antibiotic vancomycin), and wards (e.g. , the intensive care unit) have all been targeted for monitoring based on results from previous studies, similar studies in the community have not been undertak en. Primary O utbrea k Investigation and Study Settings The Hospital B NICU provides medical care for premature and critically ill newborns in a 43bed Level III unit. It also serves as a regional referral center providing 24

PAGE 25

intensive care for infants around northeast Florida and South Georgia. In December of 2010, a cluster of three MRSA infections was identified among patients in the NICU, leading to the implementation of active surveillance for MRSA colonizations. Subsequently, the hospital implemented several infection control interventions as active surveillance begun to identify several colonized neonates. During the investigation, hospital staff identified 33 MRSA isolates including 17 MRSA PFGE type USA300 from unique patients . A detailed epidemiological investigation was conducted, including a review of medical records for pertinent clinical and epidemiological data. Based on the laboratory surveillance data, it was suspected that the cluster of 17 MRSA PFGE type USA300 isolates were epidemiologi cally related. However, the mode of transmission and etiology of the outbreak could not be determined. The limitation of PFGE analysis to discriminate between USA300 isolates, one of the most common CA MRSA strains in the US, contributed to the difficult y in identifying an etiology or mode of transmission. A genomic epidemiological approach utilizing WGS and phylogenetics could have elucidated transmission within the NICU by identifying the cases that were epidemiology linked by genetically similar strain s and ruling out sporadic cases unrelated to transmission events in the NICU. Ultimately, these results would have lead to targeted interventions. The Hospital A ’s NICU is a 48 bed level III unit located in the heart of Jacksonville, FL, four miles away f rom Hospital B . While geographically similar, Hospitals A and B patient demographics vary by socioeconomic status, race, and acuity. Since 2009, Hospital A has been conducting active surveillance for MRSA 25

PAGE 26

colonization in the NICU. Active surveillance detects colonized neonates, allowing for early interventions to prevent transmission. M ethicillin Resistant S. aureus T ra nsmission in Northeast Florida Our research team previously published the results of a 2010 study of MRSA phylog enetics in northcentral/northeast Florida [49] . Ninety seven clinical M RSA isolates were obtained from seven hospitals in northeast Florida, USA during a oneweek period in 2010. Six hospitals participated in Jacksonville, FL including Hospitals A and B. We conducted spa typing of all isolates and phylogenetic analysis of n ext generation WGS (Illumina GAIIX) data of spa type t008 (USA300) strains. Twenty six healthcareassociated (t002) strains, 48 community associated (t008) strains, and 23 strains of other/unknown type were identified. Phylodynamic analysis of SNP data i ncluding 30 t008 strains provided evidence of an ongoing exponential growth of the MRSA effective population size, signifying a growing epidemic. Additionally, we observed the epidemic wa s largely community based, supported back the lack of hospital speci fic clades (clustering of genetically related strains within a specific hospital) or directional geneflow from/to hospitals (Figure 12 ) . This study represented the first phylodynamic characterization of MRSA transmission at the hospitals community interf ace. The findings indicated a complex dynamic of MRSA transmission, possibly driven by a growing epidemic at the community level in hidden reservoirs. The results also indicate that introduction of MRSA from the community is the likely driving forc e for subsequent transmission. I mplications for Infection Control of Phylogenetic Analysis of NICU I solates Phylogenetic dendograms, also referred to as trees or simply phylogenies, depict the relatedness of organisms (e.g. , people, animals, bacteria, viruses etc.) based on 26

PAGE 27

genetic data ranging from a gene, set of genes, or an entire genome. Variations between sequence data, referred to as single nucleotide polymorphisms (SNPs), are used in conjunction with evolutionary models to reconstruct the ancestral relat edness between the individual genetic sequences being compared (Figure 1 3 ). If personto person transmission has occured , isolates would appear tightly clustered in a single clade of a phylogenetic tree (monophyletic) compared to other USA300/t008 isolat es sampled from the community and the comparison NICU. A monophyletic topology would suggest a possible point source introduction of a single CA MRSA clone into the NICU followed by subsequent personto person transmission among secondary cases. Conversely, a paraphyletic topology (i.e. , isolates do not cluster on a single phylogenetic clade and/or were more related to other USA300/t008 strains sampled outside the primary NICU) would suggest multiple introductions into the NICU either from healthcare work ers, parents, or colonized neonates prior to admission. These contrasting scenarios would have seemingly different implications for infection control interventions. The first scenario in which the point source exposure and personto person transmission w as suspected would warrant isolation precautions, cohorting, review of patient services, environmental cleaning, and handhygiene surveillance. If a common intermittent exposure source was suspected, then healthcare worker screening could be implemented and a detailed investigation of procedures and devices would ensue. For the second scenario in which the isolates did not cluster and multiple introductions from the community were suspected, patient screening on admission, decolonization, a strict protoco l on visitors, nursing, and kangaroocare, and screening of expecting mothers may be warranted. E ither of these scenarios could indicate a 27

PAGE 28

larger epidemic within the community population the hospital serves, or even subepidemics within the other wards of the hospital. Second, phylogenetic analysis would enable assessment of the temporal relationship of isolates. Overall, this welldocumented cluster of epidemiologically and molecularly related isolates provided a unique opportunity to retrospectively apply phylogenetic methods to investigate MRSA trans mission within and across healthcare facilities . Studies Assessing the Association b etween MRSA a nd Antibiotic Use Few studies have assessed the association between antibiotic use and drug resistant S. aure us . Among those in the literature, researchers have found mixed results. One of the largest studies, a systematic review and metaanalysis by Tacconelli et al . published in 2004, identified 76 publications between 1976 and 2000, representing 24,230 patie nts, which assessed the role of antibiotic exposure on the risk of MRSA [69] . The metaanalysis showed that recent antibiotic use increased the odds of acquiring MRSA 1.8 times (95% CI, 1.71.9). A separate 2004 study found that antibiot ic usage in the previous si x months was associated with MRSA infection in a rural community (OR 3.1, 95% CI 1.1 to 8.6) [60] . A similar study also found that antibiotic use in the previous 12 months increased children’s odds of CA MRSA colonization (OR 16.1, 95% CI 6.4 to 40.8) [70] . Contrastingly, some studies have demonstrated negative results. Paganini and colleagues assessed CA MRSA in children in Argentina [71] and found that previous antibiotic usage was not an important risk factor for MRSA acquisition (OR=0.98 (95% CI 0.67 to 1.42) [72] . A 2004 study of MRSA skin infections among military trainees found that previous antibiotic use was not associated with MRSA infection (OR 0.7, 95% CI 0.2 to 1.9) [73] . However, it is 28

PAGE 29

possible that studies finding negative results are largely underreported due to publication bias. There have also been attempts to characterize this relationship of antibiotic use and the development of resistance through temporospatial analyses. In an ecological time series analysis spanning four years, Monnet and colleagues found a positive associat ion between antibiotic use and the incidence of MRSA [74] . They delineated seasonal peaks in MRSA infection during spring, but no such variation for methicillinsusceptible S. aureus (MSSA). They repeated their analysis on a sample of inpatient hospital data from 20002004 [75] . They again demonstrated a temporal correlation between antibiotic usage and MRSA; however, neither of these study examined MRSA in the community setting, leav ing a large gap in the literature. In a more recent study, Klein and colleagues utilized national data on MRSA hospitalizations and antibiotic resistance to assess trends in S. aureus from 2005 2009 [1] . Through timeseries a nalysis they revealed seasonality with HA MRSA infections peaking in the winter months among older individuals and CA MRSA hospitalizations peaking in summer months among children. They deduced that these trends are potentially explained by increased anti biotic usage in the winter, which subsequently would reduce HA MRSA infections during the summer. Surprisingly, these finding are contrary to other studies including one coauthored by Klein in 2012, which found that population level increases in antibiot ic prescribing of fluoroquinolones preceded increases in antibiotic resistant S. aureus by one month (0.23, p =0.03) [76] . Taken together, these contradictory results emphasize the current gaps in knowledge related to this association. Additional studies are required to validate previous findings and to assess other community level factors 29

PAGE 30

contributing to the prevalence of drug resistant S. aureus. These results would quantify the magnitude of the effect of antibiotic use on the spread of antibiotic resistance. The Burden of Pediatric Antibiotic Use a nd Prolife ration of Antibiotic Resistance A 1999 survey of pediatricians revealed that 40% had been requested by a parent 10 or more times in the last month to prescribe an antibiotic when they did not feel it was clinically indicated [77] . One third of physicians responded they frequent ly submitted to these requests. P reliminary data indicate that children less than five receive a disproportionate number of antibiotic prescriptions com pared to adults . Additionally, a l arge proportion of antibiotics are misprescribed to pediatric patients treated for viral otitis media, pharyngitis, and upper respiratory infections [55,58,59] . R ates of antibiotic resistance also peak in this age group, subsequently leveling off and then peaking again in adolescents and older adults [78] . Therefore, the misuse of antibiotics in this age group may be driving resistance at the population level, reflecting the epidemiology of other infectious diseases. For instance, seasonal influenza epidemics are largely driven by transmission in pediatric populations [79] . I posit that high community antibiotic use (i.e., per c apita rates of antibiotic prescribing) is associated with a greater level of antibiotic resistance (i.e. , proportion of SA that are MRSA positive) (Figure 14) . Preliminary D ata from the Flori da Department of Health (FDOH) Highlights Disparities in CA MRSA I ncidence Antibiotic susceptibility testing results for SA isolates became reportable in 2008 for all laboratories participating in electronic laboratory reporting. In the Florida Department of Health 2012 Annual Morbidity and Mortality Report, a surveil lance update for S. aureus was provided. From 2008 to 2012, the susceptibility of SA isolates 30

PAGE 31

to penicillin decreased from 7% to 1%. In the 2011, the most recent year for which data was available, resistance to oxacillin (i.e., MRSA) ranged from 4852% a mong Florida counties . North Florida had the highest proportion of SA isolates that were MRSA while south Florida co unties had the lowest . Most importantly, resistance to oxacillin was highest in the 1 to 4 year olds (61% resistant) . In addition, resist ance of SA to erythromycin, a macrolide, was highest among this age group (76% resistant) . This is particularly significant since macrolide usage may induce clindamycin resistance, as crossresistance to both antibiotics has been observed, playing a larger population level role in antibiotic resistance [80,81] . Overall, these data illustrated geographic and demographic variation s in SA antibiotic resistance. Study Goal, Aims, and Research Questions My specific aims of this dissertation are: (Study 1) Determine the epidemiological and genomic relatedness of MRSA infections and colonizations among 17 patients belonging to a putative outbreak in a NE FL neonat al intensive care unit . Using high resolution phylogenetic analysis of genomewide data (i.e. , by comparing similar single nucleotide polymorphisms between MRSA WGS) in conjunction with clinical and epidemiological data, I will discern the interrelatedness of MRSA strains, with emphasis on elucidating the distinctive markers of epidemic and endemic transmissions. I hypothesize that MRSA isolates from the putative NICU outbreak, when assessed at a higher resolution provided by WGS, will not be related by direct transmission event s and will be indicative of multiple introductions of MRSA strains from community sources. (Study 2) Investigate the epidemiology of MRSA colonizations among patients hospitalized in the NICU by comparing clinical, epidemiological, and MRSA whole31

PAGE 32

genome sequences among patients hospitalized in NE FL NICUs from 20082011 and a large healthcare sample. Building on Aim 1, I will utilize genomic epidemiology and phylodynamics to a.) distinguish between personto person transmission of MRSA among NICU patients a nd repeated introductions from the community leading to sporadic cases , b.) determine risk factors for colonization, c.) compare colonizations and infections between neonates t o determine if variations exist, d.) assess the demographic history of CA MRSA a nd HA MRSA lineages, and e.) assess intrahost diversity of MRSA. I hypo thesize that a large proportion of MRSA cases in the NICU will be sporadic, marked by few transmission events (i.e. , sub epidemics), indicating a communi ty etiology, there be variations in risk between noncolonized and colonized and colonized and infected infants , and that changes in effective population size of CA MRSA strains will reflect the community epidemic . (Study 3) Assess population level factors associated with CA MRSA infect ions by conducting a temporospatial analysis of outpatient antibiotic prescriptions and community prevalence of SA drug resistance in Florida. I plan to investigate the association between per capita rates of outpatient antibiotic prescriptions (main predictor) and the county level prevalence of MRSA (main outcome) (Figure 1 5) . I hypothesize that the high county level per capita rates of pediatric antibiotic prescriptions will be associated with greater populationlevel prevalence of MRSA. Taken together, these data will contribute new knowledge regarding the MRSA emergence and transmission, characterizing its epidemiology through use of novel technology, methods and data sets. These findings will underscore the importance of translating co ntrol measures for antibiotic resistant pathogens from the healthcare 32

PAGE 33

setting to the community by delineating the pathway from populationlevel antibiotic use to a hospitalized infant in the NICU. Specifically, aim one will demonstrate how MRSA WGS data a nd phylogenetic analysis can provide actionable information for clinicians during epidemic investigations. A im two will track MRSA from the hospital to community in an effort to understand the etiology of these infections. Last, a im three will test the as sociation of community antibiotic use and antibiotic resistance while highlighting healthcare providers and patient populations that may potentially contribute disproportionately to antibiotic prescribing and consumption. 33

PAGE 34

Table 11. Risk factors for hea lthcareassociated and community associated methicillin resistant S. aureus MRSA in Healthcare (HA MRSA) MRSA in the Community (CA MRSA) Prevalent genotypes * (US ) USA100, USA200 USA300, USA400 Antimicrobial resistance Multiple agents Few agents SCCmec ( mecA resistance gene) Types I III Types IV, V PVL toxin gene Rare Common Risk Factors Hospitalization Surgery Long term care Dialysis Indwelling devices Crowding Contact Compromised skin Contamination Cleanliness HA MRSA = he althcare associated methicillin resistant S. aureus CA MRSA = c ommunity associated methicillinresistant S. aureus *As defined by pulsefield gel electrophoresis (PFGE) 34

PAGE 35

Figure 1 1 . Current Centers for Disease Control (CDC) Active Bacterial Core surveillance case definition for methicillin resistant S. aureus . 35

PAGE 36

Figure 1 2 . Maximum likelihood phylogenetic analyses of MRSA t008 in northcentral Florida by healthcare facility (colored by tip) . The results demonstrate an intermixing of strains with no evidence of endemic transmission. Figure 1 3 . Interpreting phylogenetic dendograms. Two separate scenarios depicting different relatedness patterns of isolates collected from three different hospitals. 36

PAGE 37

Figure 14 . Schematic description of the population level effect of outpatient antibiotic use on introduction of methicillinresistant S. aureus (MRSA) into healthcare facilities (HCF) and subsequent transmission in the neonatal intensive care unit (NICU). Figure 1 5 . Schematic diagram of the relationship between outpatient antibiotic use and S. aureus antibiotic resistance in consideration of healthcare and demographic correlates. 37

PAGE 38

CHAPTER 2 WHOLE GENOME SEQUENCING FOR OUTBREAK INVESTIGATIONS OF METHICILLIN RESIS TANT Staphylococcus aureus IN THE NEONATAL INTENSIVE CARE UNIT Overview Methicillin resistant Staphylococcus aureus (MRSA) is a leading cause of healthcareassociated infections (HAI), significantly contributing to morbidity and mortality of hospitalized patients. Infants in the neonatal intensive care unit (NICU) are at increased risk for infection and colonization wi th MRSA, often resulting in poor outcomes and long term sequelae [82] . MRSA in the NICU may be acquired from colonized parents, healthcare workers, and other neonates [83,84] . The CDC estimates that ~50% of MRSA infections for patients 389 days old are hospital onset [85] . Community reservoirs have been implicated in the introduction of MRSA into NICUs by increasing coloni zation prevalence among patients and visitors [86] . However, identifying reservoirs and tracking the source of implicated strains has proven difficult, resulting in the persistence of transmission despite aggressive control measures [87 – 89] . Limitations in current genotyping techniques available in clinical practice may hinder the investigation of MRSA outbreaks in the healthcare setting [89] . Genotyping, is increasingly becoming a necessary component of epidemic detection and investigation, provides discrimination among genetically similar strains and identification of epidemiologically important isolates. Pulsedfield gel electrophoresis (PFGE), spa typing, antibiograms, and multilocus sequence typing ( MLST) are commonly employed to investigate MRSA transmission. However, these methods may not be optimal, as the unit of categorization (e.g., PFGE type, spa type, MLST profile) can encompass broad genetic and epidemiological diversity [87,89] , making it difficult to differentiate sporadic 38

PAGE 39

from epidemic cases [19 – 21,23,49,90] , particularly when a prevalent strain type is commonly identified. MRSA PFGE type USA300 is an important pathogen in community and healthcare settings. In the U.S., these s trains were historically associated with “community associated” (CA) infections acquired outside of hospitals. However, in many healthcare facilities, including those in our study area, CA MRSA strains are displacing healthcareassociated (HA) strains as a cause of HAIs. The increasing prevalence of USA300 emphasizes the need for advanced typing methods in clinical practice. Recently, phylogenetic analysis of wholegenome sequencing (WGS) data has provided the resolution to discriminate between closely r elated isolates of bacterial pathogens through comparison of single nucleotide polymorphisms (SNP) [51,91] . As a result, epidemiologically important isolates may be identified among a population sample that appears homogenous when analyzed using conventional genotyping methods. WGS technology is often not readily available to investigators of putative outbreaks in the healthcare setting [38,50 – 52] . Epidemiological linkages between patients may then be spuriously attributed and transmission sources obscured, leading to ineffective interventions to curb transmission and the possibility for uninterrupted transmission. We sought to determine whether phylogenetic analysis of SNP would facilitate identification of the source of MRSA transmission amidst a putative NICU outbreak, compared to the initial investigation that utilized traditional genotyping. We considered multiple typing methods including WGS SNP data assessed using maximum likelihood (ML) and Bayesian phylogenetics. Epidemiological and phylogenetic data were co39

PAGE 40

visualized to illustrate the temporal and genetic relationship among cases, allowing for assessment of patient to patient transmission events. We demonstrate how this approach would have enhanced the investigation, ruling out several sporadic cases of MRSA and potentially augmenting infection control interventions. Methods Outbreak I nvestigation The 43 bed Level III NICU in Hospital B provides medical care for premature and critically ill newborns, serving as a regional referral center for Northeast Florida and South Georgia. In December 2010, a temporal cluster of four MRSA infections was identified among neonates. The county health department was notified and a joint investigation was initiated together with hospital infection control and prevention department. In January 2011, active MRSA surveillance was implemented through collection of nares swabs from neonates on admission and weekly thereafter. Surveillance swabs were analyzed using the GeneXpert molecular diagnostic system Xpert MRSA assay (Cepheid, Sunnyvale, CA). Positive surveillance swabs were cultured on Columbia nalidixic acid agar (Becton Dickinson, Sparks, MD). Cli nical samples were obtained from infants demonstrating signs of infection as part of routine clinical evaluation as determined by the clinical team caring for the patient. S. aureus isolates were identified using conventional biochemical methods and MRSA isolates were confirmed using cefoxitin disk diffusion testing. MRSA isolates were sent to the Florida Department of Health, Bureau of Laboratories for PFGE typing. County health department and hospital investigators conducted a detailed investigation involving the review of medical records for pertinent clinical and epidemiological data (e.g., dates of admission and discharge, laboratory results, patient demographics, bed assignments, 40

PAGE 41

and procedures). While the investigation was underway, a number of in fection control interventions were implemented. These included limited patient visitation, discontinuation of “kangaroocare” (the process by which an adult coddles the infant through skinto skin contact), patient cohorting, contact precautions, and enhanced environmental cleaning. A review of infection control practices was also conducted, focusing on cleaning, disinfection of equipment, and medication delivery. From December 2010 to October 2011, 34 MRSA isolates were identified, including 17 MRSA PFG E type USA300 from unique patients (7 clinical and 10 surveillance isolates). Based on these results, it was suspected that neonates with PFGE type USA300 were epidemiologically linked through recent transmission events in the NICU (Figure 1). Non USA300 strains were suspected to represent sporadic “background” cases. Despite the intensive epidemiological investigation, neither the origin of the epidemic strain nor route of transmission could be identified. Sample Collection and Epidemiologic Investigation To determine whether phylogenetic analysis of genomewide SNPs could elucidate putative transmission events, the 17 USA300 isolates were further evaluated in our laboratory. The first positive surveillance or clinical isolate for each patient was selected for spa typing, WGS, and phylogenetic analysis. Dates of admission and discharge, positive MRSA laboratory culture reports, and previously negative clinical and surveillance tests were used to reconstruct the epidemic timeline and assess ov erlapping lengths of stay. To illustrate how increasing levels of genetic resolution could be used to investigate transmission events, we created three epidemic curves including: 1) Laboratory collection date of positive clinical and surveillance tests and 41

PAGE 42

PFGE analysis performed by the Florida Department of Health, 2) spa typing results of PFGE type USA300 isolates, and finally 3) WGS and phylogenetic analysis. Whole Genome Sequencing, Single Nucleotide Polymorphism Detection, and Phylogenetic Analysis Sample preparation, gDNA isolation, and spatyping was performed as previously described [49] . Isolated gDNA were sequenced using Illumina HiSeq 2000 sequencing system. FASTQ files with raw 2x100 pairedend reads were trimmed and then filtered by Phred quality score of 20 and a minimum length of 30 base pairs using Sickle v. 1 .2. Filtered FASTQ files were then mapped to MRSA USA300 reference genome FPR3757 (GenBank accession no. 87125858) with Bowtie 2 v. 2.2.3 short read aligner software using default settings [92] . Local realignment around insertions and deletions was performed using the Genome Analysis Toolkit v. 3.1.1, a framework for analyz ing next generation sequence data [93] . SNPs were called using FreeBayes Bayesian v. 0.9.14 genetic variant detector for haploid organisms using the following settings ( --ploidy 1 --left align indels --min basequality 20 --min alternatefraction 0.75) [94] . Putative SNPs were initially filtered by depth of coverage (<10) and quality (<20) using VCFtools v. 0.1.10. SNP locations not conserved across all samples (ambiguous sites) were rem oved. Unmapped reads and mobile genetic elements (e.g. plasmids) were not included in the alignment for phylogenetic analysis as previously described [40] . A FASTA multiple alignment of SNPs was then generated for phylogenetic analysis. The final SNP alignment for the 17 USA300 isolates contained 981 SNPs in the core genome. The bioinformatics pipeline was constructed using the University of Florid a High Performance Computing Canter's local instance of Galaxy [95] . 42

PAGE 43

The frequencies of SNP differences were compared between isolates. We selected a cutoff of < 30 SNP differences to indicate a putative transm ission event as studies have suggested differences ranging from 2340 are indicative of epidemiological linkages between individuals [96,97] . Furthermore, we previously identified a diversity of 138 SNP differences between unrelated spa type t008 MRS A strains among healthcare facilities within our study area [49] . A prel iminary neighbor joining (NJ) phylogenetic tree was inferred. The best nucleotide substitution model was then selected using the base tree and a hierarchical likelihood ratio test. Calculations were carried out with MEGA v6.06 [98] . A maximum likelihood (ML) tree was then inferred using RAxML v0.7.4, using the GTR+ nucleotide substitution model and ascertainment bias correction, with branching patterns evaluated by bootstrapping (1000 replicates) [99] . Bayesian Phyl ogenetic Analysis A molecular clock was calibrated to assess the timescale of MRSA spread utilizing the Bayesian framework implemented in the BEAST v1.7.5 package [100] . A Markov Chain Monte Carlo (MCMC) was run for 1 billion generations with sampling every 100,000 generations using the HKY substitution model and the Bayesian skyline plot demographic prior. MRSA evolutionary timescale was es timated by enforcing a lognormal uncorrelated (relaxed) molecular clock where the age for each tip represented by the sampling date of the first positive MRSA isolate from the hospitalized infants. Effective sample size > 200 for each estimated parameter was used as the cutoff to assess proper mi xing of the MCMC using Tracer v 1.5. A maximum clade credibility (MCC) tree was selected from the trees posterior distribution using treeAnnotator v1.7.5. 43

PAGE 44

MLST+ Comparison To explore the feasibility and reproducibil ity of available software packages with the potential for implementation in the clinical or publi c health laboratory setting, we analyzed the WGS data using SeqSphere+ software version 2.0 (Ridom, Muenster, Germany). This software conducts a geneby gene comparison (MLST+) to produce genomewide allelic profiles. This software is one example of a Windows based graphical user interface that provides an alternative to complex bioinformatic pipelines and advanced phylogenetic analysis. MLST+ allelic profile s for S. aureus were downloaded from the SeqSphere+ server. Assembled genomes in BAM format from our bioinformatics pipeline were imported into SeqSphere+ and allelic profiles were used to create a NJ tree of the 17 NICU isolates. NJ phylogenies from the SNP based and MLST+ phylogenetic analyses were compared to determine concordance. Results Study Population, Risk Factors, and Investigation Timeline The identification of four positive clinical cultures within ten days of admission (patients 14) alerted hospital infection control staff of a putative outbreak, and they initiated an investigation and implemented active surveillance. Over the 10month investigation, 17 patients with the suspected outbreak strain, PFGE type USA300, were identified. The 17 patients included eight males (47%) and nine females (53%). A timeline of the putative outbreak was obtained by review of admission dates, discharge dates, and report dates of relevant laboratory results (Figure 2 1). Mapping of the patient lengths of stay showed two discrete periods of the outbreak including patients 111 and 1217 and separated by 37 days, leading investigators to posit that transmission was persisting despite interventions. All neonates in the study were born in the hospital 44

PAGE 45

and direct ly admitted to the NICU, reducing the likelihood of MRSA introduction from an outside referring hospital. Seven clinical isolates from normally sterile sites were obtained from incident MRSA cases (from patients 14, 8, 13, and 14) and ten isolates were obtained from surveillance cultures (from patients 57, 9 12, and 15 17). Six infants had a MRSA negative clinical culture and five infants a negative surveillance culture prior to a subsequent positive result (Figure 2 1). The average incubation period fr om admission to positive surveillance or clinical culture result was 12.2 days (range 1 33), 14.4 days and 10.7 days for clinical isolates and surveillance cultures, respectively. Molecular, Genomic, and Phylogenetic R esults The primary investigation foc used on 17 patients with PFGE type USA300 isolates (Figure 2 2 ). Among these 17 isolates, spa typing identified five isolates that were discordant from the expected t008 genotype (Figure 2 1). SNP differences between WGS ranged from four to 188 (mean = 143), and patients 1, 2, and 6 differed by < 10 nucleotides, meeting our criteria for putative epidemiological linkage ( Table 2 1 ). Patients 13 and 14, who were siblings admitted to the NICU on the same day, differed by only four nucleotides suggesting vertical transmission through birthing or other common source; however, the parents were not screened to confirm this association. The 12 remaining patients’ MRSA genome sequences differed by a minimum of 77 nucleotides. The timescaled Bayesian Maximum Clade Credibility phylogeny utilizes an estimated rate of molecular evolution (molecular clock) to identify divergence dates of common ancestors. In the context of our outbreak investigations, this allowed the temporal assessment of transmission events. Patients 1, 2, and 6 were clustered withi n a monophyletic clade (Figures 2 3 and 2 4 ) and shared a recent common ancestor with patients 3, 9, 7, and 15. The remaining isolates clustered on a distinct clade, including 45

PAGE 46

the closely related sibling isolates fr om patients 13 and 14. The most recent common ancestor of all 17 MRSA isolates is dated just seven weeks [95% HPD 1 35] prior to the admission date of the earliest patient. When the genomic data are considered in the context of the epidemic curve, several transmission events were ruled out (Figure 2 4 ). Antibiotic susceptibility patterns demonstrated the ambiguity of compar ing antibiograms (Table 2 2 ). Antibiotic susceptibilities were consistent with CA MRSA strains possessing resistance to oxacillin and erythromycin, with the exception of two isolates (patients 4 and 17) susceptible to erythromycin and two isolates (patients 10 and 11) intermediately resistant to levofloxacin. Phylogenetic analysis did not epidemiologically link these patients throug h a recent transmission event (Figures 2 3 and 2 4 ). SeqSphere+ typed all isolates as MLST ST8, consistent with PFGE type USA300 genotype. Comparison of SeqSphere+ MLST+ phylogenies to the genomewide SNP based analysis produced analogous results in regard to phylogeny topology. While branch lengths varied by orders of magnitude, the topology (i.e., clustering of sequence pairs) differed by three variations (Figure 2 5 ); however, MLST+ phylogenetic analysis did not provide a method for assessing statisti cal robustness. Overall, topological differences between the phylogenies would not have affected the interpretation of the results in assessment of transmission events . Putative Transmission P athways The integration of epidemiological and genomewide SNP data disproved several transmission events that were suspected during the initial outbreak investigation. Patients 1, 2, and 6 possessed nearly identical genomes and overlapping lengths of stay, providing strong evidence of recent transmission and epidemi ological linkage (Figure 2 6 B). The laboratory collection dates for isolates from patients 1 and 2 were 46

PAGE 47

separated by just two days. Phylogenetic analysis suggests that patient 6’s colonization resulted from a common exposure, as they shared a common genetic ancestor basal to patients 1 and 2. However, while patient 6 was known to have a negative blood culture on admission, a surveillance swab was not collected; therefore, it is unknown whether the patient was colonized. Twin patients 13 and 14 may have acquired the strain through vertical transmission or from a common parent, visitor, or healthcare worker; although, transmission via a common source within the NICU cannot be ruled out. The remainder of the patients were not linked by recent transmission events. Among 17 isolates assessed by phylogenetic analysis, 12 (70.5%) represented genetically unique isolates, whereas five patients were grouped into two clusters (Figure 2 2 and 2 4). Also of note, spa typing spuriously indicated a putative transmiss ion event between patients 7 and 8, which was disprov en by phylogenetic analysis. Discussion MRSA PFGE Type USA300, the prevalent cause of community associated MRSA infections, was responsible for 31.6% of healthcareassociated invasive MRSA infections in 2011 [3,19,101] . Models have proposed the replacement of HA MRSA by CA MRSA strains in the healthcare setting and recent epidemiological studies have documented this occurrence [102,103] . With the increasing prevalence of USA300 MRSA, advanced molecular typing methods are necessary for surveillance, outbreak detection, and epidemic investigation. These activities rely on the ability to discriminate between genetically related and unrelated isolates, identifying putative transmission events that may then be investigated further . As we demonstrate, conventional typing methods (e.g., MLST, PFGE, and spa typing) used to investigate putative outbreaks of S. aureus provide macro level discriminatory resolution and are unable to resolve the 47

PAGE 48

epidemiological relationships among genetic ally similar isolates [31,34 – 37] . More generally, when the spa type or PFGE pattern of the putative outbreak strain has a high prevalence within a setting (e.g., hospital, healthcare system, or community) it becomes increasingly difficult for investigators to interpret molecular typing results, rule out transmission events, and identify epidemiologically linked patients among closely related isolates. At the onset of the NICU investigation, a cluster of MRSA infections among neonates alerted hospital infection control staff to a suspected outbreak. The subsequent implementation of active surveillance identified several neonates colonized with prevalent CA MRSA strains, further obscuring the relationship between epidemic and sporadic (i.e., background or endemic) cases. In the absence of a clear mode of transmission, nonspecific interventions were employed, yet new cases continued to emerge [82,104] . Outbreak investigation and response relies heavily on laboratory data to inform case investigation and determine appropriate interventions. Since certain infection control interventions such as contact precautions and isolation may decrease staff’s in teraction with patients to provide the necessary care, targeted inventions are required to stop transmission while mitigating unwanted outcomes [87,105,106] . Two possible MRSA transmission scenarios may have emerged from the real time, prospective utilization of WGS and phylogenet ic analysis during the investigation. Each scenario would have then dictated distinct infection control interventions. First, a close phylogenetic relationship among most or all isolates would have indicated recent transmission events, focusing the inves tigation on transmission pathways within the NICU. This may have warranted review of healthcareworker patient assignments, 48

PAGE 49

medical procedures, devices, medication administration, and bassinet locations [82] . General interventions may have also involved screening of healthcare workers, enh anced environmental cleaning, and hand hygiene compliance surveillance. Identification of colonized neonates through active surveillance or universal decolonization, as has been recently proposed, may have also been implemented [87] . Second, and consistent with our findings, distant phylogenic relationships among isolates ruled out recent transmission events for a majority of cases. This finding also indicated multiple introductions of diverse MRSA strains into the NICU, focusing interventions on colonization prevalence among parents, visit ors, and newly admitted patients. Interventions may then have involved parent/caregiver screening and decolonization and a review of infection control protocol for patient visitation and patient parent interaction (e.g., kangaroocare) [82,107] . Ultimately, the increased cost, diminished effectiveness, and potential adverse outcomes of non targeted interventions may have been averted with the added findings from phylogenetic analysis guiding the investigation. Phylogenetic analysis is continually improving our knowledge of MRSA epidemiology in the NICU [48,108,109] . Utilizing WGS data, Harris and colleagues successfully tracked MRSA between the community and hospital, documenting transmission between infants and parents in the postnatal ward and eventually identifying a colonized healthcare worker as the source [109] . Both Koser et al and Nubel et al identified several epidemiological linkages among neonates based on WGS data [48,108] . In contrast, we found that a large proportion of the cases were unrelated. Our findings are more consistent with those of Price et al who demonstrated that only 49

PAGE 50

18.9% of S. aureus transmission events in the ICU were linked to other colonized patients, concluding that patient to patient transmission was rare in that setting [96] . The diversity of isolates in our study may reflect a larger epidemic within the hospital or community, which would require more extensive sampling of healthcare workers, parents in the peri and post natal periods, and visitors. As with similar studies, surveillance of MRSA in the NICU may not capture the true prevalence of colonization or infection since neonates may be discharged before surveillance cultures are collected. Additionally, some patients may become colonized or develop infection after discharge and would be mi ssed in the absence of follow up or hospital readmission. As a result, some transmission events between patients may have been missed. In conclusion, phylogenetic analysis is rapidly emerging as the next stage in the evolution in molecular, now termed genomic, epidemiology [51] . The application of WGS additionally extends to the determination of genotypic antibiotic resistance and virulence [110,111] . WGS may provide improved turnaroundtime for diagnostic purposes with the added benefit of enhanced surveillance, outbreak detection, and investigation. With the availability of benchtop sequencers (e.g., Illumina’s MiSeq), real time analysis delivering actionable results is increasingly feasible [52,91] . These technologies are further facilitated by the development of streamline software packages such as Ridom’s SeqSphere that reduce computational demands, simplifying bi oinformatic and phylogenetic analysis. While this technology may not be readily available in all clinical laboratories, the transition will likely follow that of previous technology such as automated PCR (e.g., GeneXpert). Meanwhile, public health refere nce laboratories in the U.S. Centers for Disease Control and Prevention’s 50

PAGE 51

laboratory response network already provide the infrastructure for widespread implementation of WGS. Last, as with all laboratory methods, WGS alone cannot replace the need for surv eillance and epidemiological investigation[112] . WGS should be employed in concert with traditional epidemic investigation techniques. 51

PAGE 52

Table 21 . Pairwise comparison of nucleotide difference between 17 methicillin resistant S . aureus isolates from patients hospitalized in the neonatal intensive care unit . There were an average of 143 nucleotide differences (range 4188) am ong isolates. We d efined nucleotide differences < 3 0 in the core genome as epidemiological linked. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 6 * 3 105 102 4 133 128 160 5 137 137 163 129 6 10 * 7 105 123 142 7 124 122 147 131 143 118 8 151 147 173 151 158 146 147 9 129 128 152 170 173 131 159 181 10 153 148 174 128 135 143 147 165 184 11 154 150 176 132 141 145 148 165 182 144 12 152 154 179 153 147 158 158 173 187 167 162 13 137 133 159 134 145 128 131 150 169 146 144 163 14 139 137 162 137 147 131 133 154 172 149 148 165 4 * 15 84 80 106 132 141 77 120 145 130 146 146 158 129 133 16 146 142 168 144 153 139 140 151 178 158 158 170 143 147 138 17 150 154 181 153 149 158 160 177 188 167 170 120 165 167 160 172 *Denotes isolates with pairwise nucleotide differences of < 30. 52

PAGE 53

Table 22. Antibiotic susceptibilities and spa types of 17 methicillin resistant Staphylococcus aureus isolates from infants hospitalized in the neonatal intensive care unit during the putat ive outbreak investigation. Patient s pa Type Ox Cc E TMP/ SMX Le Gn Rf Va Tet Ln Dp 1 t008 R S R S S S S S S S S 2 t008 R S R S S S S S S S S 3 t008 R S R S S S S S S S S 4 t5593 R S S S S S S S S S S 5 t008 R S R S S S S S S S S 6 t008 R S R S S S S S S S S 7 t118 R S R S S S S S S S S 8 t118 R S R S S S S S S S S 9 t008 R S R S S S S S S S S 10 t008 R S R S I S S S S S S 11 t008 R S R S I S S S S S S 12 t008 R S R S S S S S S S S 13 t211 R S R S S S S S S S S 14 t211 R S R S S S S S S S S 15 t008 R S R S S S S S S S S 16 t008 R S R S S S S S S S S 17 t008 R S S S S S S S S S S Abbreviations: Ox, oxacillin; Cc, clindamycin; E, erythromycin; TMP/SMX, trimethoprim/sulfamethoxazole; Le, levofloxacin; Gn, gentamicin; Rf, rifampicin; Va, vancomycin; Tet, tetracycline; Ln, Linezolid; Dp, Daptomycin; S, susceptible; I, intermediate; R, resistant 53

PAGE 54

Figure 2 1 . Detailed timeline for putative neonatal intensive care unit outbreak. Lengths of stay are indicated in grey for the 17 pulsefield gel electrophoresis type USA300 isolates. Spa types are indicated next to the patient number. Positive (red) and previously negative (green) surveillance (S) and clinical (C) isolates are illustrated. A 37 day gap between the two discrete outbreak periods is designated with a double vertical line shaded in gray. 54

PAGE 55

Figure 22 . Pulse field gel electrophoresis (PFGE) of m ethicillin resistant S . aureus (MRSA) isolates from putative outbreak. PFGE of 17 of 34 MRSA isolates from infants hospitalized in the neonatal intensive care unit belonging to the putative outbreak. PFGE pattern was identified as USA300. References USA300 a and USA3000114 are included. While PFGE patterns demonstrate variations, dissimilar patterns could n ot be excluded from consideration of recent transmission. 55

PAGE 56

Figure 23. Unrooted maximum likelihood phylogenetic tree illustrating the relationship between 17 putative outbreak isolates. Percentages on branches represent bootstrap support. Tip labels include spa type and patient identification number. Dashed branches depict nont008 spa types. Longer branch lengths represent more distant phylogenetic relationships and decrease the likelihood of epidemiological linkages during the incident hospitalizati on. Isolates from patients 1, 2, and 6 differed by less than 10 nucleotides, while isolates 13 and 14, later identified as siblings, different only by four nucleotides. 56

PAGE 57

Figure 24 . Integration of epidemiological and phylogenetic data to produce a timeline of putative neonatal intensive care unit (NICU) outbreak. The Bayesian Maximum Clade (MCC) Credibility tree represents the phylogenetic relationship between PFGE type USA300 isolat es from 17 infants hospitalized in the NICU. The MCC phylogeny is scaled in time with tip dates, indicated by colored squares (red = clinical isolate, green = surveillance isolate), corresponding to the date of incident laboratory result. The lengths of stay of each infant are represented as lines extending from the phylogeny tip dates. The first blue diamond represents the day of admission scaled by the earliest admission date among all cases. The last blue diamond represents the date of discharge. An asterisk on the branch marks subtending clades supported by posterior probability > 0.75. The Bayesian MCC tree of seventeen NICU isolates was constructed using HKY nucleotide substitution model, Bayesian Skyline demographic model, and lognormal uncorrelated (relaxed) molecular clock. 57

PAGE 58

Figure 25 . Comparison of neighbor j oining (NJ) trees constructed with genome wide singl e nucleotide polymorphism data and MLST+ allelic profiles . Scale bars represent genetic distances based on nucleotide substitutions per SNP site (A) and allelic distance (B). A) MEGA v6.0 was used to infer phylogeny A using Kimura 2parameter nucleotide substitution model, and branching patterns were evaluated by bootstrapping (1000 replicates). B) Ridom’s SeqSphere software package was used to create phylogeny B using MLST+ data from assembled genomes. Differences in the clustering of isolates are indicated with an asterisk. While branch lengths and genetic distance scale vary between phylogenies, the overall topology and interpreta tion remain largely unchanged. 58

PAGE 59

Figure 26 . Epidemic curve of putativ e neonatal intensive care unit ( NICU) outbreak incorporating increasing levels of resolution from molecular and WGS analysis. A) Epidemic curve of 34 cases using dates of incident clinical or surveillance MRSA positive laboratory results. Cases identified as USA300 pulsedfield gel electrophoresis (PFGE) type (n=17) during the primary outbreak investigation are indicated in blue. B) Epidemic curve of 17 PFGE type USA300 cases stratified by spatype conducted retrospectively to identify five nont008 spa types among the 17 PFGE type USA300 isolates. C) Epidemic curve of 10 remaining PFGE type USA300 and spatype t008 isolates further stratified by results from phylogenetic analyses. Cluster 1 (patients 1, 2, and 6) represent epidemiological linkages based on phylogenetic data (e.g., SNP distances) and epidemiological assessment (e.g. , overlapping lengths of stay). 59

PAGE 60

CHAPTER 3 WHOLE GENOME SEQUENCING DEMONSTRATES MULTIPLE INTRODUCTIONS AND TRANSMISSION OF METHICILLINRESISTANT Staphylococcus aureus IN THE NEONATAL INTENSIVE CARE UNIT Overview Amo ng pediatric populations, the risk of methicillinresistant Staphylococcus aureus (MRSA) colonization and infection is greatest among infants hospitalized in the neonatal intensive care unit (NICU) [86,113] . MRSA colonization is a well recognized risk factor leading to infection, and colonized infants are sources for subsequent transmission to other NICU patients [48,86,1 08] . Despite interventions, NICU outbreaks continue to occur, and substantive questions remain about management of MRSA in this setting [114] . Recent studies have sought to determine the relative contribution of repeated introductions (i.e., sporadic cases) and intrahospital transmission (i.e., endemic cases) to rates of S. aureus colo nization and infection. Results have been variable, depending on outcome (e.g., colonization vs. infection), patient population (e.g., adult vs. pediatric), and setting (e.g., ICU vs. nonICU) studied [96,115] . Throughout the last decade, surveillance data have demonstrated a shift in prevalent MRSA genotypes within healthcare facilities, illustrating the evolving epidemiology of this pathogen. Particularly, the most common community associated (CA ) MRSA genotype, identified as USA300 by pulsedfield gel elect rophoresis (PFGE) and ST 8 by multi locus sequence typing (MLST), has increased in prevalence among hospitalized patients, including those in the NICU [102,116] . Until recently, this shift has largely been characterized using molecular typing methods including PFGE, MLST, and spa typing. Ho wever, there has been increasing interest in phylogenetic analysis of MRSA whole genome sequences, with earlier data from our group and others 60

PAGE 61

demonstrating substantial diversity within S. aureus populations previously thought to be clonal [49,97,117] . In the current study, we utilized wholegenome sequencing (WGS) coupled with phylogenetic analysis to explore the epidemiology and population level epidemic dynamics of MRSA in a NICU and surrounding healthcare network, in an effo rt to determine the primary points of origin and risk factors for MRSA colonization [40,43,49,108] . Methods MRSA Surveillance and Study Population Since 2004, patients admitted to the 48 open bed level III NICU in Florida (Hospital A) underwent weekly MRSA screening of the nares until detection of colonization or discharge using a standardized protocol [90] . We assessed colonization and infection among NICU patients from 2008 through 2010. Clinical and demographic information for all hospitalized infants were abstracted from electronic medical records. Colonization was defined as a positive surveillance culture, and infection was defined as MRSA isolation from a clinical specimen collected during routine clinical care. Results of weekly MRSA surveillance were made available to the NICU team; however, clinical management of all infants was le ft to the discretion of the attending neonatologist. Infection prevention and treatment practices followed current guidelines [104,118] . Colonized infants were placed on contact precautions , cohorted, and assigned dedicated clinical staff. Visitors were educated on hand hygiene and contact precautions. Decolonization was attempted using nasal mupirocin, though infants were not rescreened to determine success. Hand hygiene and contact precaution adherence remained high during the study period. 61

PAGE 62

Sam ple Selection, spa typing, and Risk Factor Analysis Clinical and demographic covariates were compared between colonized and noncolonized infants hospitalized in the Hospital A NICU (Figure 3 1A). All available colonization isolates (n=100) were spa typed as previously described [49] . Univariate and multivariate logistic regression assessed the association between risk factors and colonization and compared risk factors between spa type t008 and nont008 colonized infants (Figure 3 1B). Multivariate linear regression assessed the association between t008 and nont008 colonizati on and lengths of stay (LOS). Admission, culture, and discharge dates were used to assess overlapping LOS as well as the daily colonization prevalence. All analyses were performed using R v 3.1.1 . MRSA isolates identified as spa type t008 (n=40), represe nting the most prevalent CA MRSA genotype, and spa type t045 (n=16), representing the most prevalent HA MRSA genotype in the hospital, were selected for WGS (Figure 3 1C). Additionally, to provide an evolutionary context to the study sample, a convenience sample of “historical” t008 (n=6) and t045 (n=22) isolates from infants hospitalized in the NICU from 2003 2007 and 2011 was selected for WGS (Figure 3 1D). These latter infants were excluded from statistical risk factor comparison of the study populat ion. WGS and Phylogenetic A nalysis WGS was performed with the Illumina MiSeq, and filtered 2x250 pairedend reads were mapped to MRSA USA300 genome FPR3757 using a bioinformatic pipeline previously described [49,119] . Local realignment around insertions and deletions was performed and SNPs were called separately for t045 and t008 samples. Putative SNPs were initiall y filtered by depth of coverage and quality. Highly clustered SNPs and SNP loca tions not conserved across all samples were removed. Unmapped reads and 62

PAGE 63

mobile genetic elements (e.g. plasmids) were not included in the alignment for phylogenetic analysis . A FASTA multiple alignment of SNPs for spa type t008 and t045 isolates were then generated for phylogenetic analysis using a custom python script . The final SNP alignment for t008 isolates (n=46) and t045 isolates (n=38) contained 486 and 218 SNPs in the core genome, respectively. The bioinformatics pipeline was run using the Univer sity of Florida High Performance Computing Canter's local Galaxy instance [95] . Phylogenetic analyses were carried out with MEGA v6.06 [98] . The best fitting nucleotide substitution models were selected using the Akaike information criterion . ML phylogenies were then inferred for t008 and t045 isolates using HKY nucleotide substitution model with branching patterns support evaluated by bootstrapping (1000 replicates) . Single nucleotide polymorphism ( SNP ) frequencies were compared among t045 and t008 isolates. Maximum likelihood (ML) phylogenetic trees were inferred from SNP alignments (supplementary methods). Eight genom es representing t002 (ST 5), t003 (ST 225), and t1003 (ST 228) lineages were phylogenetically compared to de novo assemblies of t045 isolates. Minimum spanning trees (MST) were constructed with R to visualize the inferred transmission networks from the parsimonious relationship between MRSA genomes. A putative recent transmission event was defined as two or more patients whose pairwise nucleotide differences were less than the median pairwise of the entire population. To estimate intrahost diversity we sequenced three additional MRSA genomes from three infants with multiple available isolates. We compared SNP differences between two t008 isolates from the incident surveillance culture of the first infant, two t008 isolates from surveillanc e cultures collected seven days apart from the second 63

PAGE 64

infant, and two t045 isolates from a surveillance and clinical culture collected on the same day of the third infants. Similarly, to investigate the diversity of spa type t008, we expanded our Hospital A NICU sample to include 42 previously published genomes concurrently collected from five hospitals in the local healthcare network (Hospitals A, B, C, D, and E), as well as nine random pediatric intensive care (PICU) s urveillance isolates from Hospital A (Figure 3 1E and Table 31) [49] . Th e demographic histories of spa types t008 and t045 from the Hospital A NICU and the demographic history of spa type t008 in the local healthcarenetwork were explored utilizing the Bayesian framework implemented in BEAST v1.8.0 [119] . The t045 and t008 datasets included colonization isolates from 20032011 to provide a more accurate estim ation of demographic histories. The 20032011 t008 data set (n=97) including the healthcarenetwork sample enabled us to track the emergence of spa type t008 among hospitals and the surrounding community. The age for each tip were represented by the sampling date of positive MRSA surveillance screen from the hospitalized infants. For each MRSA dataset, a posterior distribution of phylogenies was obtained. Evolutionary rates (molecular clock), dating of the most recent common ancestor (TMRCA), as well as changes in effective population sizes ( Ne ) – a measure of genetic diversity representing the number of genomes effectively contributing to the next generation – were estimated [120,121] . The GMRF Skygrid model has been sh own to outperform other demographic models (e.g. Bayesian skyline plot and Bayesian Skyride) by parameterizing Ne and smoothing the trajectory. This model also provides a better estimate of TMRCA since the prior is independent of the genealogy. For each MRSA dataset, a Markov Chain Monte Carlo (MCMC) was run for 750 million 64

PAGE 65

generations with sampling every 75,000. Model convergence was assessed by examination of the effective sampling size (ESS) using Tracer v1.6. Parameter estimates with ESS values of > 200 were accepted. Marginal likelihoods estimates for each model were obtained using path sampling and stepping stone analyses and the best fitting models were selected by comparison of Bayes Factors [121 – 124] . The GMRF Skygrid model enforcing a relaxed molecular clock was selected as the most appropriate representation of the demographic history of t045 and t008 lineages [122,125,126] . DensiTrees and maximum clade credi bility (MCC) phylogenies were constructed to compare tree topologies between spa types t008 and t045. Genotypic determinants of antibiotic resistance and virulence were compared between t008 and t045 strains to explain variations in population dynamics. De novo genome assemblies were processed through MLST v1.7, ResFinder v2.1, and VirulenceFinder v1.2, to assign MLST profiles and detect antibiotic resistance and virulence genes [127] . Results Clinical and Demographic C omparis on of I nfants From 2008 through 2010, 1940 infants were hospitalized in the NICU with an average LOS of 25.9 days ( median = 14; range : 1 295) (Figure 3 2) . NICU occupancy remained constant with an average daily census of 45 patients. Surveillance identified 177 colonized patients, translating to a colonization rate of 9.1% (Table 32 ). The mean duratio n from admission to colonization was 21.0 days . Among the colonized infants, 96 (54.2%) were male and 137 (77.4%) were inborn (i.e. not transferred from another hospital). Infection was identified in 44 infants, including 33 with a prior colonization. T here was an average of 7.5 days (range: 0 43 days) from colonization to infection . The 11 infants with infection and no prior colonization were excluded from the risk factor 65

PAGE 66

analysis. Colonized infants had a significantly lower weight at birth, lower g estational age, and longer LOS compared to the noncolonized (Table 32 ). They were also more likely to be born by cesarean delivery and to be inborn. Higher birth weight and gestational age reduced odds of colonization, while caesarian birth and black race in creased the odds (Table 33 ). Of the 177 colonized infants, 100 (56.5%) isolates were available for spa typing. Infants with available isolates were not significantly different from those that were not in terms of gender, race, birth weight, gestational age, LOS, and days to colonization ( Table 3 4 ). Colonization isolates were spa type t008 (n=54), t045 (n=22), t002 (n=7), and other (n=17) ( Table 35 ). In univariate and multivariate models, infants colonized with spa type t008 had a signific antly lower gestational age, as compared with infants colonized with nont008 isolates (Tables 3 6, 3 7, and 38 ). The gestational age, in turn, was a primary driver for increased LOS, so that infants who had t008 colonization had a significantly longer LOS than infants colonized with other spa types (Table 39 ). Of t008 and t045 isolates, 40 (81.5%) and 16 (72.7%), respectively, were sequenced. The remaining isolates were unable to be recovered. Whole Genome Sequencing and Phylogenetic Analysis Sequencing of m ultiple isolates from three patients estimated intrahost diversity to range from 23 SNPs. Two t008 surveillance isolates collected on the same day differed by four SNPs while two t008 surveillance isolates collected seven days apart from another patient differed by only two SNPs. Last ly , two spa type t045 isolates obtained from endotracheal tube cultures collected on the same day differed by three SNPs. This provided an estimate of MRSA intrahost diversity, contextualizing the epidemiological r elationship among strains colonizing infants. 66

PAGE 67

Two datasets including Hospital A NICU infants with spa type t008 (ST 8) (n=40) and t045 (ST 225) (n=16) were phylogenetically analyzed. The t008 and t045 datasets included 345 and 147 SNPs with mean pairwise nucleotide differences of 32.0 (SE=1.86) and 36.4 (SE=1.11), respectively. ML phylogenies of t008 and t045 isolates were covisualized with patient LOS to assess temporal relationships of genetically similar isolates (Figure 3 3 ). This demonstrated sever al temporal clusters of phylogenetically related isolates corresponding to the MST transmission network ( Figures 34 ). Pairwise nucleotide differences suggested recent transmission events among 33 t008 isolates (82.5%) and 14 t045 isolates (87.5%). When isolates with nucleotide differences greater than the median cutoff were removed, the mean pairwise distances were reduced to 9.2 (SE 1.43) and 4.4 (SE 0.77) for the t008 and t045 datasets respectively. WGS and epidemiological data suggested seven cluster s of t008 isolates and two clusters of t045 isolates (Figure 33 ). It was also evident two t045 lineages were present, representing two introductions of divergent ST 225 lineages. A global phylogeny illustrated the close relatedness to other prevalent HA MRSA strains ( Figure 35 ). In conjunction with spa type level data identifying unique strain types, at least 28 separate introductions or reintroductions of MRSA were likely, including 10 subsequently leading to putative transmission events. Population D ynamics of t008 and t045 The demographic histories of MRSA t045 and t008 spa types were explored through phylodynamic analysis using the Bayesian coalescent framework. Visualization of DensiTree phylogenies illustrated several well supported introductions of diverse t008 lineages into the Hospital A NICU as well as two divergent lineages of t045 (Figure 3 6 ). When compared to the healthcarenetwork sample, the majority of Hospital A NICU 67

PAGE 68

isolates (77.4%) formed a distinct monophyletic clade, with the remaining isolates interspersed among a panmictic population of t008 concurrently circulating among regional hospitals (Figure 37 ). The Bayesian and ML analyses also demonstrated the range of community level interhost diversity of MRSA among individuals seek ing care within the immediate healthcare region ( Figure 38 ). Variations in evolutionary rates, TMRCA and Ne were evident between t045 and t008 isolates. The evolutionary rate for the community sample of t008 isolates was estimated at ~3.22 (95% HPD, 2.37 to 3.99) SNP year1, marginally faster than the t045 rate; however, the difference did not reach significance ( Table 310 and Figure 39 ). TMRCA for t045 isolates was dated significantly older than t008 isolates, likely representing divergence of two t045 lineages. Interestingly, TMRCA for the community wide sample of t008 isolates was dated at 1997 (95% HPD, 1994 to 2000), immediately pre ceding the increased incidence of community onset MRSA infections observed in national surveillance data. Population dynamics of t045 and t008 spa types were further investigated by assessing changes in bacterial effective population size ( Ne ), a measure of genomic diversity and population growth over time. While the t045 Ne between 1998 and 2012 remained unchanged (i.e., no population increase or decline), the population of NICU and community samples of t008 isola tes grew exponentially (Figure 3 10). Th e beginning of this increase was observed significantly earlier in the healthcarenetwork sample compared to Hospital A NICU isolates, supporting the hypothesis that community level changes in population dynamics drove similar increases among individual hospitals. 68

PAGE 69

As virulence and antibiotic resistance genes may affect pathogen success, we assessed variations among t008 and t045 accessory genomes (Figure 311 ). A minoglycoside resistance genes differed between spa types, with t045 strains possessing genes c onferring resistance to a wider range of a minoglycoside antibiotics [128,129] . Macrolide resistance in t045 strains was mediated by ermA , while t008 strains possessed macrolide phosphotransferase C ( mphC ) and efflux pump msrA [130] . Notably, four t008 strains possessed a 2.4 kb ermC carrying plasmid , conferring macrolide and lincosamide resistance [131] . Leukotoxin and P antonValentine leucocidin, prototypical of CA MRSA strains, were common among t008 and t045 isolates. Nine t008 isolates possessed the SaPI5 pathogenicity island carrying sek2 and seq2 genes coding for enterotoxins K and Q. Adhesion genes involved in host cell binding and invasion were largely present only in t008 isolates. Discussion We sought to investigate the genomic epidemiology of MRSA colonization among infants in the level III NICU, specifically focusing on the two most prevalent spa types, t008 ( CA MRSA) and t045 (HA MRSA). While the colonization rate in our NICU population was consistent with previous studies [86] , phylogenetic analysis suggested 84% of colonizations with t008 and t045, accounting for 76% of all typed isolates, were acquired intrahospital . Active surveillance, decolonization, and other infection prevention interventions likely contributed to low rates; however, multiple introductions of MRSA and subsequent transmission were identified. These results contrast with recent studies by Price et al. and by Long et al. that found colonization and infection, respectively, were not the result of recent intrahospital transmission events [96,115] . Differences in patient populations and setting likely explain contrasting results, 69

PAGE 70

considering a large proportion of infections in the study by Long et al. were present on admission, and only 17% were ICU patients. In our study, MRSA colonization was p resent on admission among only nine infants. Taken together, these findings suggest the epidemiology of MRSA in the NICU is unique compared to other hospital settings, requiring tailored prevention interventions. Overall, while spa types t008 and t045 wer e both implicated in recent transmission events, significant epidemiological differences existed. CA MRSA spa type t008 was the most prevalent, and phylogenetic analysis illustrated multiple introductions, putatively resulting in frequent intrahospital tr ansmission events. Whereas colonization with spa type t045 (ST 225), a common HA MRSA strain in Central Europe [132] , appeared to result more often from horizontal transmission. Of note, our analysis illustrated two lineages of the spa type t045 , which were independently introduced to the local healthcare community. The predominance of t008 and ev idence of multiple introductions are consistent with previous findings suggesting CA MRSA genotypes are often acquired post partum, potentially from parents and caregivers, while HA MRSA genotypes are acquired intrahospital through horizontal transmission [133] . The repeated introduction of highly related t008 strains also suggests maintenance of external reservoirs continually seeding the unit. These reservoirs may exist within the hospital, healthcare network, or community where parents and caregivers may have contact during the perinatal period. More recently, the household has been highlighted as a likely reservoir [97,134] . High resolution phylogenetic analysis has been applied to understanding the evolution and global spread of MRSA, with applications to investigating hospital 70

PAGE 71

transmission [39,40,135] . Previous studies have shown variations in Ne among successful clones of EM RSA 15 in Europe [136] ; however, similar assessments of the demographic histor y of USA300/t008 in the healthcare setting have not been conducted. We found that spa type t008, the most prevalent CA MRSA genotype, has experienced rapid population expansion since 1998, while HA genotype t045 remained constant. This likely reflects th e population level shift between genotypes resulting in the purported displacement of HA MRSA strains in healthcare settings including the NICU [19,116,137,138] . This displacement may have resulted from increased prevalence of MRSA in the community, impacting the proportion of patients, visitors, and healthcare workers with unrecognized MRSA carriag e (i.e., colonization pressure) [49] . The beginning of the t008 populati on size increase coincides with both an increase in national outpatient fluoroquinolone prescriptions and the emergence of widespread community onset MRSA infections in the United States [97] . Overall, this indicates a larger community epidemic may be associated with populationlevel evolutionary drivers such as antibiotic use [10] . CA and HA MRSA strains are known to affect different populations and to vary genetically in terms of antibiotic resistance and virulence [15] . We identified several risk factors for MRSA colonization, consistent with previous reports [90,133] . Among colonized patients, t008 colonization was associated specifically with low gestational age. The association between gestational age and t008 colonization may result f rom underlying epidemiological or genetic differences among MRSA genotypes. Low gestational age may be associated with variations in patient care and the frequency or route of MRSA exposure (e.g., more exposure from community reservoirs). 71

PAGE 72

Additionally, infants with low gestational age may have increased comorbidity and decreased immunocompetence, resulting in greater susceptibility to spa type t008 strains harboring additional antibiotic resistance and virulence genes. Both t008 and t045 strains possessed genes conferring resistance to aminoglycosides, which have a wide range of clinical uses. However, t008 isolates belonging to the most prevalent lineage in the NICU often possessed adherence genes not found in t045 strains including ebh, fnbA, and fnbB (Figure 3 11). These virulence genes play a role in host immune evasion, cell binding, and, possibly in this case, evolutionary success [139] . Overall, additional research is needed to explore these associations. The current approach to MRSA prevention in the NICU includes identification and elimination of sources for transmission through isolation, cohorting, and/or decolonization of colonized patients [104] . This approach may be ineffective due to lag time between MRSA acquisition and identification of new carrier, when MRSA may be spread [87] . Additionally, only an estimated 41.5% of National Healthcare Surveillance Network hospitals conduct active MRSA surveillance in the NICU [140] . We show that despite a comprehensive prevention program, multiple introductions of t008 strains contributed significantly to colonization. Therefore, while infection prev ention should continue to target horizontal transmission, interventions should be developed to mitigate introductions of MRSA into the NICU. This could potentially include routine screening or universal decolonization of parents/caregivers in the prenatal period and a focus on postnatal skin to skin contact and parental hygiene [106] . Furthermore, the association between gestational age and t008 colonization may highlight highrisk groups to target these interventions. Clinical trials should further be considered to determine the 72

PAGE 73

effectiveness of interventions in preventing the introduction of MRSA into the NICU and reducing infant colonization. 73

PAGE 74

Table 31. Frequency of MRSA spa type t008 wholegenome sequences by healthcare facility and hospital unit. Healthcare Facility Unit 2003 2006 2007 2008 2009 2010 2011 Total Hospital A NICU 1 3 1 8 14 18 1 46 Hospital A PICU 2 7 9 Hospital A General 8 8 Hospital B NICU 3 9 12 Hospital B General 9 9 Hospital C General 7 7 Hospital D General 2 2 Hospital E General 4 4 Total 1 3 1 8 16 58 10 97 Table 32 . Characteristics of cases (colonized infants) and controls (uncolonized infants). Cases (n=177) Controls (n=1763) Level of significance (p value) Birth weight (median and range) 1.59 kg (0.46 4.38 kg) 2.42 kg (0.35 5.28 kg) <0.001 Gestational age (median and range) 31 weeks (23 42 weeks) 35 (22 42 weeks) <0.001 Length of stay (median and range) 49 days (1 295 days) 13 days (1 248 days) <0.001 Birth by caesarean section 73.4% (130/177) 61.8% (1090/1763) 0.003 White race 57.6% (102/177) 69.7% (1229/1763) 0.004 Born off site 22.6% (40/177) 33.0% (581/1763) 0.006 Multiple births 24.3% (43/177) 20.1% (355/1763) 0.23 Male gender 54.2% (96/177) 57.1% (1006/1763) 0.52 MRSA infection 18.6% (33/177) n/a Days to positive MRSA (median and range) 12 (0 167) n/a * As determined by Kruskal Wallis and Chi2 tests 74

PAGE 75

Table 33 . Univariate logistic regression of colonization risk factors ordered by statistical significance. Variable (reference) Odds Ratio and 95% CI p value Birth weight by 1 kg 0.44 (0.36 0.53) <0.001 Gestational age by 1 week 0.84 (0.81 0.87) <0.001 Birth by caesarean section (vaginal) 1.71 (1.22 2.44) 0.002 Black race (white) 1.72 (1.22 2.39) 0.002 Born off site 0.59 (0.41 0.85) 0.005 Other race (white) 1.58 (0.82 2.82) 0.14 Multiple births 1.27 (0.88 1.81) 0.19 Gender (male) 1.12 (0.82 1.53) 0.47 T able 34 . Comparison of characteristics between patients with spa typed and nonspa typed isolates ordered by statistical significance. spa typed (n=100) Not spa typed (n=77) Level of significance (p value) * White race 53.0% (53/100) 63.6% (49/77) 0.11 Length of stay (median and range) 56 days (6 268 days) 36 days (1 295 days) 0.11 Days to positive MRSA (median and range) 14.5 (2 150) 12 (0 167) 0.13 Born off site 19.0% (19/100) 27.3% (21/77) 0.20 Gestational age (median and range) 31 weeks (23 42 weeks) 32 weeks (23 40 weeks) 0.23 MRSA infection 16% (16/100) 22.0% (17/77) 0.40 Birth weight (median and range) 1.58 kg (0.54 4 kg) 1.6 kg (0.46 4.38 kg) 0.49 Male gender 52.0% (52/100) 57.1% (44/77) 0.59 Birth by caesarean section 75.0% (75/100) 71.4% (55/77) 0.72 Multiple births 24.0% (24/100) 24.7% (19/77) 1.00 *As determined by Kruskal Wallis and Chi2 tests 75

PAGE 76

Ta ble 35 . Frequency and proportion of spa types identified among colonized infants in Hospital A NICU. spa type Frequency (percent) t008 54 (54.0) t045 22 (22.0) t002 7 (7.0) t5160 3 (3.0) t019 2 (2.0) t214 2 (2.0) t4554 2 (2.0) t922 2 (2.0) t148 1 (1.0) t3263 1 (1.0) t330 1 (1.0) t586 1 (1.0) t711 1 (1.0) t14545 1 (1.0) Total 100 Table 36 . Comparison of characteristics between patients with community genotype ( spa type t008 ) and healthcare genotypes (nonspa type t008) ordered by statistical significance. spa type t008 (n=54) Non t008 spa types (n=46) Level of significance (p value) * Length of stay (median and range) 60 days (7 268) 44.5 days (6 182 days) 0.04 Gestational age (median and range) 28.5 weeks (23 42 weeks) 32 weeks (23 41 weeks) 0.08 Birth weight (median and range) 1.33 kg (0.54 3.64 kg) 1.7 kg (0.69 4.0 kg) 0.20 MRSA infection 20.4% (11/54) 10.9% (5/46) 0.31 White race 53.7% (29/54) 52.2% (24/46) 0.48 Multiple births 20.4% (11/54) 28.3% (13/46) 0.49 Birth by caesarean section 77.8% (42/54) 71.7% (33/46) 0.64 Days to positive MRSA (median and range) 15 (2 125) 13.5 (2 150) 0.72 Male gender 50.0% (27/54) 45.7% (21/46) 0.82 Born off site 20.4% (11/54) 17.4% (8/46) 0.90 *As determined by Kruskal Wallis and Chi2 tests 76

PAGE 77

Table 37 . Univariate logistic regression of community genotype ( spa type t008 ) colonization risk factors ordered by statistical significance. Variable (reference) Odds Ratio and 95% CI p value Length of Stay 1.01 (1.00 1.02) 0.06 Gestational age by 1 week 0.92 (0.85 1.01) 0.08 Birth weight by 1 kg 0.75 (0.47 1.16) 0.20 Multiple births 0.65 (0.25 1.63) 0.36 Birth by caesarean section (vaginal) 1.38 (0.55 3.46) 0.49 Black race (white) 0.78 (0.34 1.82) 0.57 Other race (white) 1.93 (0.48 9.71) 0.58 Gender (male) 1.19 (0.54 2.63) 0.67 Born off site 1.22 (0.45 3.44) 0.70 Days to positive MRSA 0.99 (0.98 1.02) 0.82 Table 38 . Multivariate logistic regression of community genotype ( spa type t008 ) colonization risk factors ordered by statistical significance. Variable (reference) Odds Ratio and 95% CI p value Gestational age by 1 week 0.79 (0.62 0.99) 0.05 Multiple births 0.38 (0.12 1.10) 0.08 Black race (white) 0.55 (0.20 1.46) 0.24 Other race (white) 2.26 (0.51 12.38) 0.30 Days to positive MRSA 0.99 (0.97 1.01) 0.32 Born off site 1.75 (0.56 5.82) 0.35 Birth by caesarean section (vaginal) 1.56 (0.55 4.49) 0.41 Birth weight by 1 kg 1.63 (0.498 5.66) 0.42 Gender (male) 1.02 (0.43 2.41) 0.96 77

PAGE 78

Table 39 . Multivariate linear regression of risk factors for increased length of stay among MRSA colonized infants hospitalized in the Hospital A neonatal intensive care uni t. Variable (reference) Coefficient (95% CI) p value Days to positive MRSA 0.84 (0.56 1.12) <0.001 MRSA Infection (No infection) 30.00 (13.03 46.97) <0.001 Gestational age by 1 week 3.85 ( 7.25 -0.45) 0.03 spa type (non t008) 8.89 ( 3.63 21.17) 0.16 Birth weight by 1 kg 1.74 ( 15.42 18.89) 0.84 Table 310. Comparison of mean single nucleotide polymorphism (SNP) differences, evolutionary rate, the most recent common ancestor (TMRCA) of spa type t045 isolates from Hospital A neonatal intensive care unit (N ICU) and spa type t008 isolates from Hospital A NICU and community sampl e. Sample Count s pa type SNPs Mean nucleotide differences Evolutionary Rate TMRCA WCH NICU 46 t008 482 39.5 1.58E 06 [95% HPD: 2.30E 06, 8.56E 07] 2000 [95% HPD: 1995, 2004] Community 97 t008 1928 64 1.12E 06 [95% HPD: 1.39E 06, 8.26E 07] 1997 [95% HPD: 1994, 2000] WCH NICU 2 0 t045 214 46.6 9.24E 07 [95% HPD: 1.52E 06, 3.93E 07] 1979 [95% HPD: 1953, 2000] 78

PAGE 79

Figure 31 . Diagram of study population, data sets, and analyses. Data sets are labeled A E and correspond to specified analyses. 79

PAGE 80

Figure 32 . Patient colonizations by MRSA spa type and day. The graph represents the daily prevalence of colonized patients in Hospital A NICU, assuming that patients remained colonized from the date of positive surveillance culture until discharge. 80

PAGE 81

Figure 3 3. Maximum likelihood (ML) phylogenetic relationship among spa type t045 and t008 MRSA isolates and patient length of stay with date of positive culture. The left side of the figure displays the ML phylogeny scaled in SNPs per site and ordered by decreasing genetic distance (i.e., closer phylogenetic relationships appear on top). Asterisks represent clades with bootstrap support values above 80% (i.e., well supported). The right side of the figure displays the corresponding clinical data for each patient. Grey bar spans the day of admission to day of discharge, and black point represents the collection date of positive surveillance c ulture. Infants closely related by p hylogenetic analysis and lengtho f stay are shaded similarly. A) 40 spa type t008 isolates obtained from infants hospitalized from 20082010 and co rresponding length of stays. B ) 16 spa type t045 isolates obtained from infants hospitalized from 20082010 and corresponding length of stays. 81

PAGE 82

Figure 34 . Minimum spanning tree (MST) of isolates obtained from infants hospitalized from 20082010. Nodes represent individual infants. Branches are labeled by the number of nucleotides separating the nearest genetic neighbor. A) MST of 40 spa type t008. Thirty three isolates are separated by < 30 nucleotides from one or more isolates signifying a recent transmission event. B) MST of 16 spa type t045 isolates. Fourteen isolates are separated by < 30 nucleotides from one or more isolates signifying a recent transm ission event. 82

PAGE 83

Figure 35 . Global phylogeny of common healthcareassociated MRSA genotypes and t045 (ST 225) isolates from Hospital A. NICU groups 1 and 2 correspond to Figure 33 B. Genomes in this phylogeny include 0402981 (NC_017340), ECT R 2 (FR714927), N315 (NC_002745), 18583 (HE579073), Mu50 (NC_002758), Mu3 (NC_009782), ED98 (NC_013450), CBD 635 (ASHS00000000). De novo assemblies of t045 (ST 225) isolates from Hospital A NICU were aligned to comparison genomes using ProgressiveMau ve. Single nucleotide polymorphisms (SNPs) were extracted from homologous regions of the genome. A maximum likelihood phylogeny was inferred using Mega v6.0.6 using GTR nucleotide substitution model with 100 bootstrap replicates. Bootstrap support was 100% for all branches with the exception of the labeled polytomy. 83

PAGE 84

Figure 36 . DensiTree visualizations of posterior distributions of trees obtained from Bayesian phylogenetic analysis of t008 and t045 datasets using the GMRF skygrid model and relaxed mol ecular clock as implemented in BEAST v1.8.0. Tip dates are assigned to each node based on the date of collection of positive MRSA surveillance swab, allowing the phylogeny to be scaled in time. The frequency of node clustering is used to assess statistic al support for clades, and well supported branches are indicated by solid colors. A) DensiTree of 46 spa type t008 isolates from colonized patients hospi talized in the NICU of HospitalA from 20032010. B) DensiTree of 40 spa type t045 isolates from colonized patients hospi talized in the NICU of Hospital A from 20052010. 84

PAGE 85

Figure 37 . Bayesian maximum clade credibility phylogeny of 97 spa type t008 from multiple healthcare facilities including 46 from colonized patients hospitalized in Hospital A’s NICU (blue branches and diamond tips). The phylogeny is scaled in time with tip dates corresponding to collection dates of positive MRSA cultures. The shaded area represents a monophyletic clade comprised of 31/48 (64.6%) of Hospital A NICU isolates. 85

PAGE 86

Figure 38 . Maximum likelihood phylogeny of 97 spa type t008 isolates from five healthcare facilities in northeast Florida, including 46 isolates from Hospital A neonatal intensive care unit (NICU), nine isolates from Hospital A pediatric intensive care unit (PICU), eight from adult hospital associated with Hospital A NICU, 12 isolates from Hospital A NICU, and 9, 7, 2, and 4 from Hospitals A E respectively. Tip labels are colored corresponding to healthcare facilities and branches are scaled in SNP s per site. Asterisks represent clades with bootstrap support values above 80%. The shaded area represents a monophyletic clade comprised of 31/48 (64. 6%) of Hospital A NICU isolates 86

PAGE 87

Figure 39 . Comparison of evolutionary rates and 95% highest posterior density (HPD) (credibility intervals) of t045 and t008 lineages estimated from Bayesian phylogenetic analysis of 46 spa type t008 isolates and 40 spa type t045 isolates from colonized patients hospitalized in Hospital A NICU from 2003 2010 as well as 97 spatype t008 isolates (Community) from multiple healthcare facilities including 46 from colonized patients hospitalized in Hospita A NICU. 87

PAGE 88

Figur e 310. Comparison of effective population sizes ( Ne ) of t045 (black) and t008 (dark grey) lineages estimated from Bayesian phylogenetic analysis of 46 spa type t008 isolates and 40 spa type t045 isolates from colonized patients hospitalized in H ospi tal A NICU from 2003 2010 as well as 97 spa type t008 from multiple healthcare facilities in northeast Florida. 88

PAGE 89

Figure 311. Bayesian maximum clade credibility phylogenies and genotypic antibiotic resistance and virulence. Phylogenies are scaled in time wit h tip dates corresponding to collection dates of positive MRSA cultures. A) MRSA spa type t008. B) MRSA spa type t045. Spa type t008 strains possessed ant(6) la (previously referred to as aadE ) and aph(3') III (previously referred to as aphA 3), which confer streptomycin and kanamycin resistance, while t045 strains possessed aadD (previously referred to as ant (4')Ia) and spc ( transposon Tn554). M acrolide resistance in t045 strains was mediated by ermA , while t008 strains possessed macrolide phosphotransferase C ( mphC ) and efflux pump msrA . 89

PAGE 90

CHAPTER 4 ASSOCIATION OF COMMUNITY ANTIBIOTIC USE AND Staphylococcus aureus DRUG RESISTANCE , FLORIDA, 2010 2012 Overview Antibiotic use is the most important contributing factor to the development of anti biotic resistance and studies show that half of all antibiotics prescribed are unwarranted or improperly prescribed [9]. The introduction of penicillin in the 1940s greatly reduced mortality from bacterial infections , parti cularly Staphylococcus aureus [141] . However, almost immediately, the first antibiotic resistant bacteria were identified [7,142] . S everal studies have documented the association between antibiotic use and the development of antibiotic resistance [4 – 8] , and p opulationlevel antibiotic pressure is recognized as a significant driver of community level antibiotic resistance [10] . At an individual level, antibiotic use can increase risk of acquisition of a drug resistant organism by disrupting normal flora, decreasing microbiota diversity, and leading to replacement by resistant strains [10,63,64] . Methicillin re sistant S . aureus (MRSA) is one of the most well recognized antibiotic resistant pathogens. A common cause of skin and soft tissue infections (SSTI) in the community, MRSA is estimated to result in ~75,000 cases of invasive infections in year in the US. The incidence of community onset infections is increasing, and rates of MRSA in the community vary both seasonally and geographically, with some evidence that pediatric populations and racial minorities may be disproportionately affected [23,101,143,144] . Interestingly, seasonal fluctuations is S. aureus antibiotic resistance appear to be correlated wi th outpatient antibiotic use [76,145 – 147] . H owever, data on MRSA cases in the outpatient setting are sparse . 90

PAGE 91

D espite the Centers for Disease Control ’s report identifying antimicrobial resistance as a critical problem facing healthcare, surveillance data depicting trends in community antibiotic use are largely limited [5,65,66] . Furthermore, while acute care hospitals have made significant progress in recent years to implement antibiotic stewardship programs to regulate and optimize antimicrobial use, these efforts have not pervaded into community healthcare settings [67,68] . One of first steps in community stewardship efforts is to understand trends in outpatient antimicrobial use so that communities and community healthcare provider groups can be directly targeted with public health messaging. Additionally, studies are needed to assess the association between community antibiotic use and antibiotic resistance. To address the current gap in knowledge of community prescribing practices we assessed 24 months of outpatient antibiotic prescriptions in Florida. We sought to elucidate differences in antibiotic prescribing among counties with varying demog raphic and community healthcareproviders compositions. We additionally assessed the association between outpatient antibiotic prescribing and populationlevel S. aureus anti biotic resistance in the community. These data highlight communities as well as provider groups in which direct public health messaging should be aimed and to establish a baseline of antibiotic usage for future evaluations. Methods Data Sources Outpatient antibiotic prescription data were obtained from IMS Health Solutions Xponent dat a system [148,149] . This surveillance system is a unique source of data, capturing local and regional prescription drug dispensing and containing information on all pharmaceutical products dispensed in retail and mail order channels. Outpatient 91

PAGE 92

antibiotic prescri ption data were obtained for Florida, encompassing prescriptions from May 2010 through April 2012, including patient 5digit ZIP Code, patient age, patient gender, and provider type for each antimicrobial class. Antibiotics were aggregated into 11 main groupings: a lactams (Other), cephalosporins, macrolides, metronidazole, mupirocin, oxazolidinones, fluoroquinolones, trimethoprim/sulfamethoxazole (SMZ TMP), tetracycline, and v ancomycin. These groupings were chosen based on the clinical indications and associations with epidemiologically important organisms. All groupings included oral antibiotics with the exception of the topical antibiotic mupirocin. P opulation data was used to calculate per capita prescriptions at the ZIP code and count y level as a proxy for antibiotic use in Florida. Additionally, Florida counties were dichotomized as “low” or “high” prescribing based on whether they fell below or above the median per capita prescription rate of all counties. S. aureus antibiotic resistance testing data from 2010 through 2012 was obtained from the Florida Department of Health through collaboration with Quest Diagnostics Laboratory. These data encompass S. aureus testing from outpatient healthcare providers throughout Florida. Data included patient county , specimen collection date, specimen site (e. g. , blood, wound, urine etc.), patient age, patient gender, and Oxacillin susceptibility results. The proportion of S. aureus iso lates resistant to Oxacillin ( i.e., MRSA) were calculated for each county by report year. The threeyear average (20102012) was also calculated. Florida Census data were obtained from the US Census Bureau TIGER Line shapefiles containing geographic and c artographic information from the Census 92

PAGE 93

Bureau's MAF/TIGER (Master Address File/Topologically Integrated Geographic Encoding and Referencing) database. C ounty level demographic factors including population, median age, median income, and percent urban we re abstracted. Population at the ZIP Code level was utilized for the ZIP Code level analysis. For ZIP Codes with no population values, the ZIP Codes and corresponding antibiotic prescriptions were merged with neighboring ZIP Codes. Florida acute care and healthcare facility data were obtained from the Florida Geographic Data Library ( http://www.fgdl.org/). D ata contains selected fields specifying the name, physical address, and other facility information, includi ng t he number of hospital beds. Florida healthcare facility data was obtained from the Florida Department of Health (FDOH) and includes information on clinics, medical doctor, nursing home, osteopath, State Laboratory/Clinic, and surgicenter/walk in clinics. T he number of licensed healthcare providers and the total number of acute care hospital beds was obtained from the FDOH and Agency f or Health Care Administration. Analysis Descriptive analysis The 2011 per capita rates of outpatient antibiotic prescribing for all 11 major antibiotic groupings were stratified by patient age group, gender, and prescribing provider type. Rates were also mapped by county using ESRI ArcGIS v10.1. The average yearly per capita rate of antibiotic prescriptions was mapped by ZIP Code and co visualized with the distribution of healthcare facilities/providers and acute care hospitals in Florida. T he 3 year average proportion of S. aureus isolates resistant oxacillin (MRSA) were mapped by Florida County. The proportion of S. aureus that was 93

PAGE 94

MRSA was stratified by age group, gender, and collection site. R ates of S. aureus and MRSA were also assessed. C orrelates of antibiotic prescribing and identification of spatial clusters We conducted a spatial analysis of outpatient ant ibiotic prescription data for 2011. ArcGIS v10.1 was used to test the global spatial autocorrelation of per capi ta antibiotic rates by ZIP Code. SatScan was then utilized to identify spatial clust ers of high prescribing rates. To assess the healthcare c orrelates at the ZIP Code level, the Euclidian distance from the ZIP Code centroid to the nearest acute care hospital was calculated. We then identified the number of healthcare facilities, which fell within a 25mile buffer around the ZIP Code centroid. Additionally, the number of acute care beds for the nearest acute care hospital was obtained. OpenGeoDa v1.2.0 was used to fit a log linear spatial error regression model of the healthcare correlates and the per capita rate of antibiotic prescribing by Z IP Code using a Queen Contiguity of two. The residuals of this model were then exported and mapped and a global test for spatial autocorrelation was performed . We then assessed the association of demographic and healthcare correlates with outpatient antibiotic prescribing at the county level. For this analysis, the median household income, median age, and percent rural were included as demographic correlates. The number of licensed physicians per 10,000 population and the number of acute care hospital be ds per 10,000 population were included as healthcare correlates . OpenGeoDa v1.2.0 was used to conduct a log linear spatial error regression models using a Queen Contiguity of one. The residuals were again exported and mapped, and a global test of spatial autocorrelation was performed. 94

PAGE 95

Association of outpatient antibiotic prescribing and S. aureus methicillin resistance The populationlevel association of antibiotic prescribing and S. aureus resistance was assessed by analyzing per capita antibiotic prescr iption rates and the 3 year average proportion of S. aureus isolates resistant oxacillin (MRSA). Single linear regression models for demographic and healthcare correlates of S. aureus resis tance were first assessed. Col inearity of predictors, specifically percent urban, healthcare providers per capita, and median income, were evaluated, and significant covariates were moved to a final multivariate regression model. All analysis was performed using R v 3.1.1 . Results We analyzed a total of 15,196,983 antibiotic prescriptions for 2011. Overall, the average yearly per capita antibiotic prescription rate during the study period was 0.81 prescriptions per person (Rx/person). The highest prescription rates were observed for the combined lactams ( including cephalosporins and amoxicillin clavulanic acid) (0.31 Rx/person), macrolides (0.20 Rx/person), and fluoroquinolones (0.11 Rx/person) (Table 41 and Figure 41 ). Among provider types, the general practitioner provider category, which includes inte rnal medicine, family medicine, nurse practitioner and physician assistant, prescribed 42.6% of the total per capita antibiotics (Figure 42 ). Notably , dermatologists were responsible for 31% of the tetracycline prescriptions, and pediatricians and dermat ologists were responsible for 28% and 17% of mupirocin pres criptions respectively , significantly higher than other provider types (p <0.001) (Figure 4 2 ). Within age groups, children 14 years of age were prescribed 33.3% of per capita 95

PAGE 96

antibiotic prescript ions, and were the highest prescribed age group among the lactam , macrolide, SMZ TMP, and mupirocin antibiotic groupings. For fluoroquinolones, the highest prescribed age group was among individuals 65 years and older. The monthly per capita rate of to tal antibiotic prescriptions demonstrated seasonality with peaks during the winter months and tro ughs during the summer (Figure 43 ). Macrolides, lactams, and fluoroquinolones demonstrated similar seasonality, consistent with previous findings (Figure 4 4) [76,148,150] . The remaining antibiotic groupings did not demonstrate seasonality (Figure 4 5) . County level per capita rates of antibiotic prescribing for 2011 were mapped for each antibiotic grouping (Figure 46). Per capita rates of antibiotic prescribing ranged f rom 0 to 10.89 within ZIP Codes; however, per capita prescribing rates were not sig nificantly spatially autocorrelated (Z score = 0.81, Moran’s I = 0.007) (Figure 4 7A ) . Fifty seven statistically significant clusters were identified of high per capita antibiotic prescribing rates were identified using SatScan . Significant cluster s of h igh rates were identified in Duval, Flagler, Volusia, Orange, Pinellas, Hillsborough, Lee, Palm Beach, and Miami Dade counties, among several others (Fig ure 47 B) . W e assessed access to care as a correlate to antibiotic prescribing using the number of ac ute care hospitals, acute care hospital beds, and healthcare facilities within ZIP Codes (Figure 4 8 A B ). We then preformed a spatial error regression of healthcare covariates on the per capita rate of antibiotic prescribing at the ZIP Code level (Table 4 2 ) and county (Table 4 3 ) . The county level analysis included additional demographic and healthcare covariates including the median household income, median age, percent rural, hospital beds per 10,000 population, and licensed physicians per 10,000 96

PAGE 97

popul ation (Figure 9A B) . The distance to the nearest healthcare facility was significant in the ZIP Code level analysis ( p <0.01) ; however, the spatial coefficient ( p =0.17) was not significant and ZIP code level covariates were poor predictors of per capita antibiotic prescriptions (R2 = 0.07). The county level analysis performed significantly better (R2 = 0.55). At the county level, the percent rural ( p =0.03) and hospital beds per 10,000 population ( p <0.01) were significant ly associate d with outpatient ant ibiotic prescribing. The residuals from the ZIP Code and county level analysis were mapped to visualize the unexplained variation in outpatient antibiotic prescribing rates (Figures 4 10A ) , and a local test for spatial autocorrelation was conducted. The analysis showed some residual autocorrelation but low Moran’s I values for the ZIP Code (Moran’s I = 0.13) (Figure 4 10B ) and county (Moran’s I = 0.02) (Figure 4 11A B ) analyses. The 3 year average proportion of S. aureus that was resistant to oxacillin ( MRSA ) w as 53.08% (range: 41.4% 69.7%). Thirty four counties were above the 3year average, 22 counties were consistently classified as high resistance for all three years , and 23 counties were consistently classified as low prescribing (20102012). In 2011, the proportion MRSA was 49.7% and the incidence of MRSA was 17.96 cases per 10,000 population (range : 5.2 48.5 ) (Table 44). The proportion MRSA and incidence rate remained stable from 2010 through 2012. Children aged 14 had both the highest incidence of MRSA (44.16 cases per 10,000 population) (Table 44) and the highest proportion of S. aureus that were resistant to oxacillin (MRSA) (62.7%) (Figure 4 13). The proportion of S. aureus isolates that were oxacillin resistant (MRSA) did not vary significantly by gender (Figure 414). Temporally, S. aureus testing and the proportion identified as oxacillinresistant (MRSA) did not display seasonality when assess ed 97

PAGE 98

weekly from 2010 through 2012 (Figure 415). Among b ody site categories, 79.9% of isolates w ere categorized as SSTI or SSTI other in 2011 (Table 45). An estimated 8.9% were nasopharyngeal cultures, presumably collected to test for colonization. The average proportion of S. aureus isolates resistant to ox acillin (MRSA) was comparable among body collection sites (Table 45). Assessed geographically, several western coastal counties including Taylor, Lafayette, Dixie, and Levy demonstrated the highest rates of S. aureus (Figure 4 16A) and MRSA (Figure 416B) . Assessment of county level per capita rates of antibiotic prescribing (Figure 417A) and the proportion of S. aureus identified as oxacillin resistant (MRSA) by county (Figure 417B) demonstrated several Florida panhandle counties with high rates of ant ibiotic prescribing and a high proportion of S. aureus that was MRSA. H owever, analytically, the relationship was not as clear (Figure 419). Since the spatial coefficient was not significant in the Z IP Code and County level analyses of per capita antibi otic prescribing, we determined the association between antibiotic prescribing and S. aureus resistance could be assessed with linear regression. The final regression model included median age, percent urban, and per capita antibiotic prescriptions (Table 4 6). A one unit increase in per capita outpatient prescribing was associated with a ~6% increase in the proportion of S. aureus resistant to oxacillin (MRSA) ( p = 0.04). Discussion Our study sought to assess spatial and temporal variations in outpatient antibiotic prescribing and to test the association between prescribing rates and S. aureus antibiotic resistance. To do this, we obtained 24 months of prescribing data and 3years of outpatient S. aureus testing from a large commercial laboratory. We ar e the first to 98

PAGE 99

assess antibiotic prescribing at this detailed level of spatial resolution , while taking into consideration healthcare and demographic factors potentially associated with variations in prescribing. Similar studies conducted in the US and in ternationally have focused on state, regional, or country level data and have not accounted for healthcare or demographic factors , which may confound interpretation of the results . Additionally, these studies did not assess spatial autocorrelation to expl ore geographic heterogeneity of antibiotic prescribing. We identified that lactams , macrolides , and fluoroquinolones were among the highest prescribed antibiotics , consistent with national data [149] . Additionally, we observe seasonal increases in antibiotic prescriptions in winter months likely correlating to cold and flu season [148] . Pediatric patients age 14 had the highest antibiotic prescription rates, specifically for lactam and macrolide antibiotics , which is alarming since previous studies hav e identified a significant proportion of these antibiotics are prescribed for viral infections [151] . Similarly, this population has the highest inc idence rates of S. aureus and MRSA. Interestingly, prescriptions for mupirocin , which is used in the treatment of S. aureus infections, were disproportionately higher among the 14 age group. In consideration of the implications of mupirocin resistance, this finding may require further investigation [152,153] . At the county level, both the percent rural and number of hospital beds per 10,000 population were associated outpatient prescribing. This reinforces the need to include demographic and healthcare factors in analyses of the antibiotic use and drug resistance. Taken together, we demonstrate spatial and temporal variation in outpatient antibiotic prescribing in Florida and a clear association with S. aureus antibiotic 99

PAGE 100

resistance . This findings echo the results of other ecologi cal studies . For example, Monnet and colleagues found a positive association between antibiotic use and the incidence of MRSA using time series analysis spanning four years of data [74] . P opulation level increases in fluoroquinolone prescriptions have also been shown to precede increases in antibiotic resistant S. aureus [76] . We further explored the association of demographic and healthcare correlates with outpatient antibiotic prescriptions and documented the need to include these f actors when assessing the association between antibiotic use and drug resistant organisms. Furthermore, w hile antibiotic prescribing rates did not demonstrate spatial autocorrelation, subsequent studies should test spatial structure. Our study should be interpreted in light of certain limitations. Primarily, our analysis was limited to outpatient antibiotic prescription which did not capture antibiotic prescriptions in the acute care setting as well as long term care and assisted livin g facilities, all of which contribute significantly to global antibiotic prescribing. Additionally , we utilized the distance from ZIP Code centroid to acute care hospital as a proxy measure for healthcare access. Last , in approximately 1% of antibiotic prescriptions obtained from IMS Healthcare Solutions, the patient ZIP Code was unavailable in which case the prescribing provider’s ZIP Code was assigned to the prescription. It is evident that efforts must be made to combat the continued emergence of drug resistant organisms . One approach may focus on optimizing outpatient antibiotic prescribing to reduce unwarranted antibiotic use, as s everal mechanisms for the development antimicrobial resistance hinge on antibiotic “pressure” . Until now, information on outpatient antibiotic prescribing was largely unavailable. The results 100

PAGE 101

from this study elucidate the trends in antibiotic prescribing in Florida. These data also establish a benchmark for current use that may be used to evaluate targeted interventions aimed at reducing unwarranted antibiotic use. Education of the public on the appropriate uses of antibiotics, provider education, and reports of susceptibility trends in the community are all areas that can be strengthened through the use of these data. S pecifically, it is evident that pediatric populations in the outpatient setting are receiving a large proportion of lactam antibiotics . Combined with the seasonality of prescriptions, which corresponds to seasonal increases in viral upper respiratory in fections, it brings into question whether t hese antibiotics are warranted as i t is well documented that clinicians in the outpatient setting often succumb to patient reques ts for antibiotics, even when it may not be clinically indicated [56] . When prioritizing public health interventions to reduce unwarranted antibiotic use, these data could be used to target specific provider groups or patient populations. Ultimately, these data are highly valuable for policy makers and public health officials in the fight against the growing prevalence of antibiotic resistant organism s. 101

PAGE 102

Table 41 . Prescriptions per 100 population for select antibiotic groupings , Florida, 2011. Age Group <1 1 4 5 14 15 24 25 64 65+ Total p value Antibiotic prescriptions per capita (zip code) 0.071 Healthcare Facility Count 0.42 0.42 0.42 0.47 Hospital Beds 1.00 1.00 1.00 0.29 Distance to Nearest Acute Care Hospital 1.00 1.00 1.00 <0.01 Spatial Coefficient 1.10 0.96 1.26 0.17 Table 43 . Spatial error regression of healthcare and demographic covariates on the per capita rates of antibiotic prescribing by county, Florida, 2011. County Level Analysis R squared Coefficient 95% CI p value Antibiotic prescriptions per capita (county) 0.546 Median Age 0.99 0.97 1.00 0.02 Median Income 0.99 0.99 0.99 0.14 Percent Rural 1.00 0.99 1.00 0.03 Hospital Beds per 10,000 population 1.01 1.00 1.02 <0.01 Licensed Phscians per 10,000 population 1.00 0.99 1.01 0.67 Spatial Coefficient 0.85 0.59 1.22 0.39 102

PAGE 103

Table 44 . Number of isolates tested (n) and percent resistant to oxacillin (MRSA) by age group, commercial o utp atient l aboratory, Florida, 2011 and three year average (20102012) . 2011 3 year Average Age Group n Percent resistant Cases Per 10,000 Population n Percent resistant Cases Per 10,000 Population <1 1,413 51.6% 32.66 1,404 48.9% 30.73 1 4 5,897 62.7% 44.16 5,767 62.0% 42.73 5 14 7,745 42.9% 15.01 7,849 44.0% 15.60 15 24 7,537 46.2% 14.26 7,510 46.5% 14.29 25 40 9,530 49.9% 12.80 9,597 50.8% 13.11 41 64 19,731 48.9% 15.76 19,700 49.2% 15.82 65+ 16,516 50.3% 24.60 16,328 50.7% 24.55 missing 129 52.7% 172 49.1% Total 68,498 49.7% 17.96 68,327 49.9% 18.03 Table 45 . Number of isolates tested (n) and percent resistant to oxacillin (MRSA), Commercial Outp atient Laboratory, Florida, 2011 and three year average (20102012) . 2011 3 year Average Site n Percent Resistant n Percent Resistant Blood 84 50.0% 88 54.4% Invasive Other 243 41.6% 245 41.2% Urine 3,844 44.1% 4,085 44.8% Medical device 263 53.8% 266 56.0% Nasopharyngeal 6,020 33.3% 6,460 32.3% Respiratory 576 48.5% 548 51.2% SSTI 12,919 59.1% 12,999 59.2% SSTI Other 41,352 49.7% 41,136 50.1% Unknown 2,570 52.7% 2,465 53.7% None Available 37 33.3% 35 43.3% Total 67,908 50.1% 68,327 50.0% SSTI = skin and soft tissue infection Table 46 . Linear regression of per capita outpatient antibiotic prescriptions on proportion of S. aureus isolates that were resistant to oxacillin (MRSA) . Estimate 95% CI p value Total antibiotics per capita, 2011 5.96 (3 .25 11.7) 0.04 Percent urban 0.15 ( 0.19 0.11) <0.001 Median age 0.18 ( 0.37 0.02) 0.08 103

PAGE 104

Figure 41. Per capita outpatient antibiotic prescriptions by antibiotic g rouping and age group, Florida, 2011. Figure 42. Per capita outpatient antibiotic prescriptions by provider t ype, Florida 2011. 104

PAGE 105

Figure 43 . Timeseries of the cumulative monthly per capita rates of antibiotic prescribing , Florida, 20102012. Figure 44 . Timeseries of the cumulative monthly per capita rates of antibiotic prescribing by antibiotic grouping , Florida, 20102012 lactams include prescriptions for cephalosporins and amoxicillin clavulanic acid. 105

PAGE 106

Figure 45 . Timeseries of the cumulative m onthly per capita rates of antibiotic prescribing by antibiotic grouping s not displaying seasonality , Florida, 20102012. 106

PAGE 107

Figure 46. A verage yearly per capita antibiotic prescriptions by ant ibiotic grouping and county , F lorida 2011 . 107

PAGE 108

F igure 47 . ZIP code level analysis of per capita outpatient antibiotic prescriptions. A) Average yearly per capita antibiotic prescription r ates by ZIP Code, Florida, 20102012. B) Statistically significant clusters of high per capita antibiotic prescription r ates by ZIP Code, Florida, 20102012 108

PAGE 109

Figure 4 8 . Average yearly per c apita rate of antibiotic prescriptions . A ) Acute care h ospital (red) by ZIP Code, Florida 20102012. B) Healthcare f acilities (purple) by ZIP Code, Florida 20102012 . 109

PAGE 110

Figure 49. County level healthcare correlates of antibiotic prescribing and S. aureus resistance. A ) P er capita licensed physicians per 10,000 populat ion by county, Florida, 2011. B ) T otal acute care hospital beds by county, F lorida, 2011. 110

PAGE 111

Figur e 410. Local test for spatial autocorrelation of residuals from spatial error regression of demographic correlates and p er capita outpatient antibiotic prescription r ates by ZIP Code, Florida, 2011 . A) Residuals from spatial error regression by ZIP Code, Florida, 2011. B) Local test for spatial autocorrelation of residuals. 111

PAGE 112

Figure 411. S patial error residuals from regression of healthcare and demographic covariates by c ounty on per capita antibiotic prescriptions, F lorida, 2011 . A) H ealt hcare c ovariates by zip code, F lorida, 2011 . B) H ealthcare and demographic covariates by county , F lorida, 2011. 112

PAGE 113

Figure 4 12. L ocal test for spatial autocorrelation of residuals from spatial error regression of demographic and healthcare correlates and per capita outpatient antibiotic prescription rates by county, F lorida, 2011. A) Significant clusters of residuals from spatial error regression. B) The association between high and low prescribing counties. B A 113

PAGE 114

Figure 413. Percent of S. aureus Isolates that were oxacillinr esistant (MRSA) by age group, commercial o utp atient l aboratory, Florida, 2011 and three year average (20102012) . Figure 414. Percent of S. aureus Isolates that were oxacillinr esistant (MRSA) by g ender, commercial o utp atient l aboratory, Florida, 2011 and threeyear average (20102012) . 114

PAGE 115

Figure 415. Frequency of S. aureus tests and proportion of i solates that were oxacillin r esistant (MRSA) by year and report week, commercial o utp ati ent l aboratory, Florida, 20102012. 115

PAGE 116

Figure 416. S. aureus cases by Florida County, 2011. A) Rate of S. aureus cases per 10,000 population by county, Florida, 2011. B) Rate of MRSA cases per 10,000 population by county, Florida, 2011. 116

PAGE 117

Figure 417. Comparison of outpatient antibiotic prescriptions per capita and proportion of S. aureus isolat es resistant to oxacillin. A) Antibiotic prescriptions per capita by Florida County, 2011. B) Three year average proportion of S. aureus isolates resistant to oxacillin (MRSA) by county, Florida, 20102012. 117

PAGE 118

Figure 418. Antibiotic prescriptions per capita for 2011 and threeyear average percent of S. aureus resistant to oxacillin by county, Florida. 118

PAGE 119

CHAPTER 5 CONCLUSIONS Accomplishments of the Dissertation Taken together, these studies address ed several gaps in S. aureus epidemiology. Specifically, we elucidate the role of community antibiotic use as a driver for population level antibiotic resistance, address unresolved measurement issues in bacterial phylogenetics, and subsequently apply these methods to the understanding CA MRSA genotype USA300 emergence as a healthcareassociated pathogen. Antibiotic resistance is a worldwide problem. Bacterial resistance to antibiotics is growing and treatment options are becoming limited; meanwhile, the development of new antibiotics to replace those no longer effective is disproportionately small [154] . Inappropriate use of antibiotics promotes resistanc e , and recent studies have show n that current interventions to curb unwarranted antibiotic use in the community have not been successful [155] . Antibiotic misuse is a societal issue, as decisions at the individual level affect the population. It is studies like these that are necessary to illustrate the problem, and as a result of my work, a large, rich dataset will be available for future studies to delineate the issue and identify target s for future interventions. Genomic Epidemiology of Bacterial P athogens WGS and phylogenetic analysis have applications for the real time investigation of pathogen transmission during outbreaks. The feasibility of applying these methods in a multitude of settings (e.g., intrahospital, interhospital, nationally, and international) is increasing as benchtop sequencing platforms are becoming more common among teaching hospitals, academic institutions, and reference laboratories. Computationally, streamlined 119

PAGE 120

bioinformatics pipelines and commercial software are reducing the need for specialized personnel to run these analyses. In addition, sequencing advances will soon allow for direct sample analysis without the need for pathogen isolation (i.e., direct from blood) [156,15 7] . This will provide rapid pathogen identification and genomic “ metadata ” including phenotypic antibiotic resistance, pathogenicity, and virulence predictions. From a public health perspective, WGS and phylogenetic analysis are poised to replace other typing methods (e.g . , PFGE ) for the routine investigation of putative diseas e outbreaks [111] . In our first study, we demonstrated how PFGE typing of MRSA isolates from a putative outbreak spuriously indicated transmission events among patients in the NICU. Subsequently, in study two we show that intrahospital transmission of both CA and HA MRSA strains was occurring over a threeyear period, during a time where cases of colonization were not considered an epidemic. In both of these scenarios , WGS revealed aspects of the epidemic that were not elucidated with the genotyping methods currently available to hospital and public health investigators. The result of these analyses would have provided actionable information, which would have augment ed infection control measures. As these methods evolve, researchers will be able t o address long standing questions regarding the epidemiology of bacterial pathogens. Specifically, we can now track transmission chains, allowing for identification of modes of transmission and risk factors, which have remained elusive. Methodologically, as WGS data are explored in greater detail, new methods will have to be developed to define epidemiological linkages and the timescale of transmission. In study two, we explore three aspects contextualizing 120

PAGE 121

the relationship between the genomes of pathogens colonizing and infecting patients hospitalized in the NICU . Frist, we glimpse the range of intrahost diversity of MRSA, comparing genomes of isolates collected on the same day as well as several days apart. We find that genomes from the same patient may differ with a range of 24 SNPs. Second, we assess the population level diversity of the most prevalent CA MRSA strain, spa type t008/PFGE type USA300, among a network of healthcare facilities in a large metropolitan city. W e show that unrelated isolat es of MRSA differ on average by 72.9 (SE=1.7) SNPs. Last, by calibrating a molecular clock, we estimate an accumulation of 35 SNPs per year. With these three measures, we can begin to understand the diversity expected between related and unrelated strai ns as well as the timescale of transmission event s. This has direct application to the investigation of bacterial disease outbreaks by providing a guide for interpretation of WGS results. Analytical methods are currently being developed to provide a likelihood of transmission events given pairwise genetic differences among wholegenome sequences [158] . Together, investigators will have a new tool for understanding the transmission dynamics of bacterial populations. Healthcare Epidemiology of MRSA MRSA remains a serious health threat to hospitalized patients, particularly in the NICU. Hospital infection control interventions in the NICU target the introduction and subsequent transmission of MRSA. This is often accom plished by passive surveillance of laboratory results from routine clinical samples and active surveillance for colonized infants, as it has been shown that the presence of an unknown colonized infant is the greatest risk factor for other patients 121

PAGE 122

acquirin g MRSA [108] . Patients who are identified as MRSA colonized are cohorted or placed in contact isolation, and precautions are taken (e.g., mask, gowns, gloves) to prevent transmission while the patient is decolonized. Yet, despite these int erventions, MRSA outbreaks in the NICU still occur [86] . Indeed, it is evident from our analyses that new MRSA strains are introduced to the unit , often frequently, and subsequently spread to other hospitalized neonates . Studies one and two (see chapters 3 and 4) focus on understanding the mode by which MRSA strains are introduced into hospitals and subsequently spread by leveraging recent advances in bacterial WGS . Until the inception of WGS, investigators were unable to discern between genetically similar strains of MRSA, especially in high prevalence settings. We are now able to discriminate between isolates at the nucleotide level, significantly reducing false positive rate of suspected transmission events. Through our studies, it is evident that the epidemiology of MRSA strains varies among hospitals and within hospital wards . We demonstrate that s patype t008 is the most prevalent MRSA genotype within the NICU, and while we observed several introductions of diverse strains, we also found several instances of putative transmission events between neonates. For both spa type t008 and t045 strains, comparison of phylogenies in conjunction with patient LOS suggests maintenance of external reservoirs continually seeding the unit. This is illustrated by gaps in overlapping patient LOS followed by the reappearance of strains whose genomes vary by more than what would be expected if they belonged to the same transmission chain (Figure 3 2 and 33). These reservoirs may exist within the hospital and immediate healthcare network (e.g., clinics, laboratories) where patients, parents, and 122

PAGE 123

caregiver may have contact during the perinatal period. This is further reflected by the diversity of MRSA spa type t008 isolates observed at a population level between hospitals in the same healthcare region. With the exception of the Hospital A NICU, we see little clustering of isolates in the phylogeny, evidence of an underlying community level epidemic. Based on our phylodynamic analysis, this epidemic dates to at least 1998 and has been experiencing exponential population expansion since that time, likely leading to increased colonization pressure on hospitals . Interestingly, this time scale correlates to a national increase in outpatient antibiotic prescriptions for fluoroquinolones , whi ch increased 50% from 1998 through 2008. Overall, it becomes apparent that to combat the issue of MRSA in the healthcare setting, interventions should in part focus on the community. Therefor e, we sought to further elucidate the association between commu nity antibiotic prescriptions and S. aureus antibiotic resistance. Community Antibiotic Prescribing and Antibiotic R esistance Children less than five years of age are disproportionately affected by S. aureus and MRSA. Similarly, this population receives t he highest annual per capita antibiotic prescriptions. We demonstrate spatial and temporal variations in antibiotic prescribing coinciding to spatial variations in high rates of MRSA. Urbani zation, possibly as a proxy for socio economic status and healthcare infrastructure, appears to play are role in the association between outpatient antibiotic prescribing and antibiotic resistance. While our study was ecological in nature, it adds to the growing abundance to literature highlighting the association bet ween population level antibiotic use and resistance. Taken together, it is 123

PAGE 124

evident that the <5 year old population is a target group for unwarranted antibiotic use. This can be done in three ways. First, better diagnostic tests will allow the rapid diag nosis of pathogens , reducing the number of antibiotics prescribed for viral infections (e.g., viral otitis media and other upper respiratory infections) . Second, diagnostic tests employing WGS may provide rapid identification of genotypic antibiotic resistance, allowing the appropriate prescription of the optimal antibiotic therapy. Third, electronic medical records combined with the expanding role of the community pharmacist may reduce unwarranted antibiotic use as well as optimization of therapy through correct dosing and better patient adherence from monitoring. To date, community based interventions such as CDC’s “Get Smart” campaign have not significantly impacted antibiotic prescribing, as seasonal peaks in prescribing during upper respiratory v irus season are evident. As we demonstrate, one additional antibiotic prescription per person per year increases S. aureus oxacillin res istance by 6%. Therefore, reducing pediatric antibiotic prescriptions from the current estimate of two prescriptions per year to one would seemingly have a significant affect on antibiotic resistance. This would impact not only S. aureus resistance but a number of other epidemiologically important organisms. Future D irections T he use of WGS data to reconstruct phylogenet ic and demographic history of bacterial pathogens is not well recognized. As sequencing costs continue to decline and technology improvements make real time sequencing a reality, these data will become more readily available. However, we will have to 124

PAGE 125

ide ntify the best use of these data in the context of clinical and epidemiological information to understand the transmission dynamics of bacterial pathogens . These studies are a step in the right direction, and while there are barriers that need to be overc ome (e.g. , understanding recombination in the context of coalescent theory), the potential for the use of genomic data to improve and inform epidemiology is endless. Genomewide data also have implications for pathogen diagnosis, virulence detection, and antibiotic resistance profiling [159] . WGS will soon become the new norm and individuals who are able to analyze, interpret, and apply these methods will be in high demand. Moving forward, we must shift the focus from healthcare facilities to the larger community. Faster screening tests may allow us to identify colonized infants sooner, and hospitals may move to screening healthcare workers and p arents/caregivers during the prenatal period. More effective isolation and decolonization methods may also be developed. However, a large number of these interventions remain untested, and clinical trials are warranted to assess effectiveness of current infection prevention interventions. Furthermore, these interventions do not address the overarching issue, which is the unchecked epidemic in the communit y. Future studies need to prospectively investigate transmission in the community and betw een the community and hospitals by assessing community reservoirs. For example, recent evidence has highlighted the role of the household in MRSA transmission servi ng as a bacterial reservoir [97,134] . Subsequent NICU studies could assess household members of infants who become colonized during thei r NICU stay. We also see evidence that the epidemiology of S. aureus varies by strain (e.g., USA300/t008). The success of 125

PAGE 126

this particular lineage may be in part to adhesion and superantigen virulence genes that may play a role in colonization, host immun e evasion, and pathogenicity. However, this needs to be investigated further, potentially using repeated cross sectional sampling of a cohort of individuals transiently and persistently colonized with various MRSA strains. Last , since pediatric populations are disproportionately affected, research should target this population to determine if these rates are do to surveillance bias or to actual variations in disease incidence. Furthermore, interventions to reduce unwarranted antibiotic use must be developed and validated on their ability to reduce populatio n level antibiotic resistance. 126

PAGE 127

LIST OF REFERENCES 1. Klein EY, Sun L, Smith DL, Laxminarayan R. The changing epidemiology of methicillin r esistant Staphylococcus aureus in the United States: A national observational s tudy. Am. J. Epidemiol. 2013 ; 177:666 – 74. 2. Klevens RM, Morrison MA , Nadle J, et al. Invasive m ethicillin resistant Staphylococcus aureus infections in the United States. JAMA 2007; 298:1763– 71. 3. Centers for Disease Control and Prevention. Active Bacterial Core Surveillance Report, Emergi ng Infections Program Network, methicillinr esistant Staphylococcus aureus , 2011. 2012. http://www.cdc.gov/abcs/reports findings/ survreports/ . Accessed 5 August 2013. 4. McGowan JE Jr. Antimicrobial resistance in hospital organisms and its relation to antibiotic use. Rev Infect Dis 1983; 5:1033 – 1048. 5. Muto CA, Jernigan JA , Ostrowsky BE, et al. SHEA guideline for preventing nosocomial transmission of multidrug resistant strains of Staphylococcus aureus and enterococcus. Infect. Control Hosp. Epidemiol. Off. J. Soc. Hosp. Epidemiol. Am. 2003; 24:362 – 86. 6. Shlaes DM, Gerding DN, John JF, et al. Society for Healthcare Epidemiology of America and Infectious Diseases Society of America Joint Committee on the Prevention of Antimicrobial Resistance: guidelines for the prevention of antimicrobial resistance in hospital s. Clin. Infect. Dis. an Off. Publ. Infect. Dis. Soc. Am. 1997; 25:584 – 99. 7. Wood A, Gold H. Antimicrobial drug resistance. N. Engl. J. Med. 1996; 335:1445– 1453. 8. Walsh C. Molecular mechanisms that confer antibacterial drug resistance. Nature 2000; 406:775 – 81. 9. Centers for Disease Control and Prevention. Antibiotic resistance threats in the United States, 2013. Atlanta, GA: 2013. http://www.cdc.gov/ drugresistance/pdf/ar threats 2013508.pdf . Accessed 15 May 2014. 10. Lipsitch M, Samore MH. Anti microbial use and antimicrobial resistance: a population perspective. Emerg. Infect. Dis. 2002 ; 8:347 – 54. 127

PAGE 128

11. Diekema DJ, Pfaller MA , Schmitz FJ, et al. Survey of infections due to Staphylococcus species: frequency of occurrence and antimicrobial susceptibility of isolates collected in the United States, Canada, Latin America, Europe, and the Western Pacific region for the SENTRY Antimicrobial Surveillanc. Clin. Infect. Dis. 2001; 32 Suppl 2:S114 – 32. 12. Jarvis WR, Schlosser J, Chinn RY, Tweeten S, Jackson M. National prevalence of methicillin resistant Staphylococcus aureus in inpatients at US health care facilities, 2006. Am. J. Infect. Control 2007; 35:631 – 7. 13. Hubben G, Bootsma M, Luteijn M, et al. Modelling the costs and effects of selective and universal hospital admission screening for methicillinresistant Staphylococcus aureus . PLoS One 2011; 6:e14783. 14. Johnson JK, Khoie T, Shurland S, Kreisel K, Stine OC, Roghmann M C. Skin and soft tissue infections caused by methicillinresistant Staphylococcus aureus USA300 clone. Emerg. Infect. Dis. 2007 ; 13:1195 – 200. 15. Naimi T, LeDell K, Como Sabetti K. Comparison of community and healthcare – associated methicillinr esistant Staphylococc us aureus i nfection. JAMA J. 2003; 290:2976– 2984. 16. Salgado CD, Farr BM, Calfee DP. Community acquired methicillin resistant Staphylococcus aureus : a metaanalysis of prevalence and risk factors. Clin. Infect. Dis. 2003 ; 36:131 – 9. 17. Tenover FC, Goering R V. Methicillin resistant Staphylococcus aureus strain USA300: origin and epidemiology. J. Antimicrob. Chemother. 2009 ; 64:441– 6. 18. Salangsang JM, H arrison LH, Brooks MM, Shutt KA, Saul MI, Muto CA . Patient associated risk factors for acquisition of methicillin resistant Staphylococcus aureus in a tertiary care hospital. Infect. Control Hosp. Epidemiol. 2010; 31:1139– 47. 19. Otter J a, French GL. Community associated methicillin resistant Staphylococcus aureus strains as a cause of healthcareassociated infection. J. Hosp. Infect. 2011; 79:189 – 93. 20. Klevens RM, Edwards JR, Tenover FC, McDonald LC, Horan T, Gaynes R. Changes in the epidemiology of methicillinresistant Staphylococcus aureus in intensive care units in US hospitals, 19922003. Cl in. Infect. Dis. 2006; 42:389 – 91. 21. Otter JA , French GL. Molecular epidemiology of community associated meticillin resistant Staphylococcus aureus in Europe. Lancet Infect. Dis. 2010; 10:227 – 39. 128

PAGE 129

22. Centers for Disease Control and Prevention. Active Bac terial Core Surveillance (ABCs ) Report Emerg ing Infections Program Network methicillinr esistant Staphylococcus aureus , 2008. 2008. http://www.cdc.gov/abcs/reports findings/ survreports/ . Accessed 5 July 2013 . 23. Fridkin S, Hageman J, Morrison M. Methicillin resistant Staphylococcus aureus disease in three communities. New Engl. J. Med. J. 2005; 352:1436 – 1444. 24. Liu C, Bayer A, Cosgrove SE, et al. Clini cal practice guidelines by the Infectious Diseases Society of A merica for the treatment of methic illin resistant Staphylococcus aureus infections in adults and children. Clin. Infect. Dis. 2011; 52:e18– 55. 25. Bielaszewska M, Mellmann A, Zhang W, et al. Characterisation of the Escherichia coli strain associated with an outbreak of haemolytic uraemic syndrome in Germany, 2011: a microbiological study. Lancet Infect. Dis. 2011; 11:671 – 676. 26. Jungk J, Baumbach J, Landen M, et al. Outbreak of Salmonella serotype Saintpaul infections associated with multiple raw produce items – United States, 2008. Morb. Mortal Wkly. Rep 2008 ; 57:929 – 934. 27. Hidron AI, Edwards JR, Patel J, et al. NHSN annual update: antimicrobial resistant pathogens associated with healthcareassociated infections: annual summary of data reported to the National Healthcare Safety Network at the Centers for Disease Control and Prevention, 20062007. Infect. Control Hosp. Epidemiol. 2008; 29:996– 1011. 28. Duffy J, Sievert D, Rebmann C, et al. Effective statebased surveillance for multidrug resistant organisms related to health careassociated infections. Public Health Rep. 2011 ; 126:176 – 85. 29. Tenover FC, McAllister S, Fosheim G, et al. Characterization of Staphylococcus aureus isolates from nasal cultures collected from individuals in the United States in 2001 to 2004. J. Clin. Microbiol. 2008; 46:2837 – 41. 30. Harmsen D, Claus H, Witte W, Rothgnger J, Turnwald D, Vogel U. Typing of methicillin resistant Staphylococcus aureus in a university hospital setting by using novel software for spa repeat determination and database management. J. Clin. Microbiol. 2003; 4 1:5442– 5448. 31. Palavecino E. Clinical, epidemiological, and laboratory aspects of methicillinresistant Staphylococcus aureus (MRSA) infections. Methods Mol. Biol. 2007; 391:1– 19. 129

PAGE 130

32. Enright M, Day N, Davies C. Multilocus s equence typing for charact erization of methicillin resistant and methicillin susceptible c lones of Staphylococcus aureus . J. Clin. Microbiol. 2000 ; 38:1008– 1015. 33. Deleo FR, Otto M, Kreiswirth BN, Chambers HF. Community associated meticillin resistant Staphylococcus aureus . Lan cet 2010; 375:1557 – 68. 34. Okuma K, Iwakawa K, Turnidge JD, et al. Dissemination of new methicillinresistant Staphylococcus aureus clones in the community. J. Clin. Microbiol. 2002; 40:4289 – 4294. 35. Carey AJ, Della Latta P, Huard R, et al. Changes in the molecular epidemiological characteristics of methicillin resistant Staphylococcus aureus in a neonatal intensive care unit. Infect. Control Hosp. Epidemiol. 2010 ; 31:613 – 9. 36. Vivoni AM, Moreira BM. Application of molecular techniques in the study of Staphylococcus aureus clonal evolutiona review. Mem. Inst. Oswaldo Cruz 2005; 100:693 – 8. 37. David MZ, Taylor A, Lynfield R, et al. Comparing pulsedfield gel electrophoresis with multilocus sequence typing, spa typing, staphylococcal cassette chrom osome mec ( SCCmec ) typing, and PCR for pantonvalentine leukocidin, arcA , and opp3 in methicillin resistant Staphylococcus aureus isolates. J. Clin. Microbiol. 2013; 51:814– 9. 38. Le VT, Diep BA. Selected insights from application of wholegenome sequencing for outbreak investigations. Curr. Opin. Crit. Care 2013; 19:432 – 439. 39. Gray RR, Tatem AJ, Johnson JA , et al. Testing spatiotemporal hypothesis of bacterial evolution using m ethicillin resistant Staphylococcus aureus ST239 genomewide data within a bayesian framework. Mol. Biol. Evol. 2011; 28:1593– 603. 40. Harris SR, Feil EJ, Holden MTG, et al. Evolution of MRSA during hospital transmission and intercontinental spread. Science 2010; 327:469– 74. 41. Lemey P, Salemi M . The phylogenetic handbook: a practical approach to phylogenetic analysis and hypothesis testing. Cambridge University Press, New York, 2009. 42. DeLeo FR, Chambers HF. Reemergence of antibiotic resistant Sta phylococcus aureus in the genomics era. J. Clin. Invest. 2009; 119:2464. 43. Grenfell BT, Pybus OG, Gog JR, et al. Unifying the epidemiological and evolutionary dynamics of pathogens. Science 2004; 303:327 – 32. 130

PAGE 131

44. Kluytmans V andenbergh MFQ, Kluytmans J A . Community acquired methicillin resistant Staphylococcus aureus : current perspectives. Clin. Microbiol. Infect. 2006; 12:9 – 15. 45. Walker TM, Ip CL, Harrell RH, et al. Wholegenome sequencing to delineate Mycobacterium tuberculosis outbreaks: a retrospective observational study. Lancet Infect. Dis. 2012 ; 3099. 46. Kser CU, B ryant JM, Becq J, et al. Wholegenome sequencing for rapid susceptibility t esting of M. tuberculosis . N. Engl. J. Med. 2013 ; 369:290– 292. 47. Snitkin ES, Zelazny AM, Thomas PJ, et al. Tracking a hospital outbreak of carbapenem resistant Klebsiella pneumoniae with wholegenome sequencing. Sci. Transl. Med. 2012 ; 4:148ra116. 48. Kser CU, Holden MTG, Ellington MJ, et al. Rapid wholegenome sequencing for investigation of a neonat al MRSA outbreak. N. Engl. J. Med. 2012 ; 366:2267 – 75. 49. Prosperi M, Veras N, Azarian T, et al. Molecular epidemiology of community associated m ethicillin resistant Staphylococcus aureus in the genomic era: a crosssectional s tudy. Sci. Rep. 2013; 3:190 2. 50. Diep BA. Use of wholegenome sequencing for outbreak investigations. Lancet Infect. Dis. 2012 ; 3099:10– 11. 51. Gardy J. Investigation of disease outbreaks with genome sequencing. Lancet Infect. Dis. 2012 ; 3099:1 – 2. 52. Eyre DW, Golubchik T, Gordon NC, et al. A pilot study of rapid benchtop sequencing of Staphylococcus aureus and Clostridium difficile for outbreak detection and surveillance. BMJ Open 2012 ; 2.3 : e001124. 53. Roberts RR, Hota B, Ahmad I, et al. Hospital and societal costs of antimicrobial resistant infections in a Chicago teaching hospital: implications for antibiotic stewardship. Clin. Infect. Dis. 2009; 49:1175 – 84. 5 4. Pichichero ME. Dynamics of antibiotic prescribing for c hildren. JAMA 2002; 287:3133– 3 135. 55. Centers for Disease Control and Prevention. Officerelated antibiotic prescribing --United States, 19931994 to 20072008. MMWR. Morb. Mortal. Wkly. Rep. 2011 ; 60:1153 – 6. 56. Cals JWL, Boumans D, Lardinois RJM, et al. Public beliefs on antibiotics and respiratory tract infections: an internet based questionnaire study. Br. J. Gen. Pract. 2007 ; 57:942 – 7. 131

PAGE 132

57. Kardas P, Devine S, Golembesky A, Roberts C. A systematic review and meta analysis of misuse of antibiotic therapies in the community. Int. J. Antimicrob. Agents 2005 ; 26:106 – 13. 58. Grijalva CG, Nuorti JP, Griffin MR. Antibiotic prescription rates for acute respiratory tract infections in US ambulatory settings. JAMA 200 9 ; 302:758 – 66. 59. Shapiro DJ, Gonzales R, Cabana MD, Hersh AL. National trends in visit rates and antibiotic prescribing for children with acute sinusitis. Pediatrics 2011 ; 127:28– 34. 60. Baggett HC, Hennessy TW, Rudolph K, et al. Community onset methicillin resistant Staphylococcus aureus associated with antibiotic use and the cytotoxin PantonValentine leukocidin during a furunculosis outbreak in rural Alaska. J. Infect. Dis. 2004 ; 189:1565– 73. 61. Prosperi M , De Luca A, Di Giambenedetto S, et al. The threshold bootstrap clustering: a new approach to find families or transmission clusters within molecular quasispecies. PLoS One 2010; 5:e13619. 62. Carleton H a, Diep BA, Charlebois ED, Sensabaugh GF, PerdreauRemington F. Community adapted methicil lin resistant Staphylococcus aureus (MRSA): population dynamics of an expanding community reservoir of MRSA. J. Infect. Dis. 2004; 190:1730– 8. 63. Monnet DL. Methicillin resistant Staphylococcus aureus and its relationship to antimicrobial use: possible implications for control. Infect. Control Hosp. Epidemiol. 1998; 19:552– 9. 64. Jernberg C, Lfmark S, Edlund C, Jansson JK. Long term ecological impacts of antibiotic administration on the human intestinal microbiota. ISME J. 2007; 1:56 – 66. 65. Huttner B, Samore M. Outpatient antibiotic use in the United States: time to ‘ get smarter’. Clin. Infect. Dis . 2011; 53:640 – 3. 66. Dellit TH, Owens RC, McGowan JE, et al. Infectious Diseases Society of America and the Society for Healthcare Epidemiology of Amer ica guidelines for developing an institutional program to enhance antimicrobial stewardship. Clin. Infect. Dis. 2007; 44:159 – 77. 67. Kaki R, Elligsen M, Walker S, Simor A, Palmay L, Daneman N. Impact of antimicrobial stewardship in critical care: a systematic review. J. Antimicrob. Chemother. 2011; 66:1223 – 30. 132

PAGE 133

68. Hurford A, Morris AM, Fisman DN, Wu J. Linking antimicrobial prescribing to antimicrobial resistance in the ICU: before and after an antimicrobial stewardship program. Epidemics 2012; 4:203 – 10 . 69. Tacconelli E, Venkataraman L, De Girolami PC, D 'Agata E . Methicillin resistant Staphylococcus aureus bacteraemia diagnosed at hospital admission: distinguishing between community acquired versus healthcareassociated strains. J. Antimicrob. Chemother. 2004 ; 53:474 – 9. 70. Lo W T, Lin W J, Tseng M H, et al. Nasal carriage of a single clone of community acquired methicillin resistant Staphylococcus aureus among kindergarten attendees in northern Taiwan. BMC Infect. Dis. 2007; 7:51. 71. Paganini H, Della Latta MP, Muller Opet B, et al. Community acquired methicillin resistant Staphylococcus aureus infections in children: multicenter trial. Arch. Argent. Pediatr. 2008; 106:397 – 403. 72. Cost elloe C, Metcalfe C, Lovering A, Mant D, Hay AD . Effect of a ntibiotic prescribing in primary care on antimicrobial resistance in individual patients: systematic review and metaanalysis. Bmj 2010; 340:c2096– c2096. 73. Campbell KM, Vaughn AF, Russell KL, et al. Risk factors for community associated methicillinres istant Staphylococcus aureus infections in an outbreak of disease among military trainees in San Diego, California, in 2002. J. Clin. Microbiol. 2004; 42:4050– 3. 74. Monnet DL, MacKenzie FM, Lpez Lozano JM, et al. Antimicrobial drug use and methicillin resistant Staphylococcus aureus , Aberdeen, 19962000. Emerg. Infect. Dis. 2004; 10:1432 – 41. 75. Aldeyab MA , Monnet DL, Lpez Lozano JM, et al. Modelling the impact of antibiotic use and infection control practices on the incidence of hospital acquired me thicillin resistant Staphylococcus aureus : a time series analysis. J. Antimicrob. Chemother. 2008; 62:593 – 600. 76. Sun L, Klein EY, L axminarayan R. Seasonality and temporal correlation between community antibiotic use and r esistance in the United States. Clin. Infect. Dis. 2012; 55:687 – 94. 77. Bauchner H, Pelton SI, Klein JO. Parents, physicians, and antibiotic use. Pediatrics 1999 ; 103:395– 401. 78. Centers for Disease Control and Prevention. Activ e Bacterial Core Surveillance ( ABCs ) Report Emerging Infections Program Network MethicillinResistant Staphylococcus aureus , 2010. 2010. http://www.cdc.gov/abcs/reports findings/ survreports/ . Accessed 5 July 2013 . 133

PAGE 134

79. Reichert TA, Sugaya N, Fedson DS, Glezen WP, Simonsen L, Tashiro M. The Japanese experience with vaccinating s choolchildren against Influenza. N. Engl. J. Med. 2001; 344:889 – 896. 80. Barber M, Waterworth PM. Antibacterial activity of lincomycin and pristinamycin: A comparison with Erythromycin. Br. Med. J. 1964; 2:603 – 6. 81. Leclercq R. Mechanisms of resistance to macrolides and lincosamides: nature of the resistance elements and their clinical implications. Clin. Infect. Dis. an Off. Publ. Infect. Dis. Soc. Am. 2002; 34:482 – 92. 82. Gerber SI, Jones RC, Scott M V, e t al. Management of outbreaks of methicillinresistant Staphylococcus aureus infection in the neonatal intensive care unit: a consensus statement. Infect. Control Hosp. Epidemiol. 2006 ; 27:139– 45. 83. Saiman L, Cronquist A, Wu F. An outbreak of methicillinresistant Staphylococcus aureus in a neonatal intensive care unit. Infect. Control Hosp. Epidemiol. 2003 ; 24:317– 321. 84. Geva A, Wright SB, Baldini LM, Smallcomb JA, Safran C, Gray JE. Spread of me thicillin resistant Staphylococcus aureus in a large tertiary NICU: network analysis. Pediatrics 2011; 128:e1173 – 80. 85. Iwamoto M, Mu Y , Lynfield R, et al. Trends in invasive methicillinr esistant Staphylococcus aureus i nfections. Pediatrics 2013; 132:e817 – e824. 86. Zervou FN, Zacharioudakis IM, Ziakas PD, Mylonakis E. MRSA colonization and risk of infection in the neonatal and pediatric ICU: a metaanalysis. Pediatrics 2014; 133:e1015 – 23. 87. Popoola VO, Budd A, W ittig SM, et al. Methicillin r esista nt Staphylococcus aureus transmission and infections in a neonatal intensive care unit despite active surveillance cultures and decolonization: Challenges for infection prevention. Infect. Control Hosp. Epidemiol. 2014 ; 35:412– 8. 88. Popoola VO, Carroll KC, Ross T, Reich NG, Perl TM, Milstone AM. Impact of colonization pressure and strain type on methicillinresistant Staphylococcus aureus transmission in children. Clin. Infect. Dis. 2013; 57:1458 – 60. 89. Mangini E, Srinivasan P, Bur ns J, et al. Unrelated strain methicillinresistant Staphylococcus aureus colonization of health care workers in a neonatal intensive care unit: Findings of an outbreak investigation. Am. J. Infect. Control 2013; 41.11:11021104. 134

PAGE 135

90. Maraqa NF, Aigbivbalu L, MasnitaIusan C, et al. Prevalence of and risk factors for methicillin resistant Staphylococcus aureus colonization and infection among infants at a level III neonatal intensive care unit. Am. J. Infect. Control 2011; 39:35– 41. 91. Price J, Gordon NC, Crook D, Llewelyn M, Paul J. The usefulness of whole genome sequencing in the management of Staphylococcus aureus infections. Clin. Microbiol. Infect. 2013; 19:784 – 9. 92. Li H, Durbin R. Fast and accurate short read alignment with Burrows Wheeler tra nsform. Bioinformatics 2009; 25:1754– 60. 93. McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next generation DNA sequencing data. Genome Res. 2010; 20:1297– 303. 94. Garrison E, Marth G. Haplotypebas ed variant detection from short read sequencing. 2012 ; arXiv preprint arXiv:1207.3907. 95. Goecks J, Nekrutenko A, Taylor J, Team TG. Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol 2010; 11:R86. 96. Price JR, Golubchik T, Cole K, et al. Wholegenome sequencing shows that patient to patient transmission rarely accounts for acquisition of Staphylococcus aureus in an intensive care unit. Clin. Infect. Dis. 2 014 ; 58:609 – 618. 97. Uhlemann A C, Dordel J, Knox JR, et al. Molecular tracing of the emergence, diversification, and transmission of S. aureus sequence type 8 in a New York community. Proc. Natl. Acad. Sci. 2014; 111:6738– 43. 98. Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S. MEGA5: Molecular Evolut ionary Genetics Analysis using maximum likelihood, evolutionary distance, and maximum parsimony m ethods. Mol. Biol. Evol. 2011 ; 28:2731 – 2739. 99. Stamatakis A. Phylogenetic models of rate het erogeneity: a high performance computing perspective. Proc. 20th IEEE Int. Parallel Distrib. Process. Symp. 2006; pp. 8 pp. 100. Drummond AJ, Suchard MA , Xie D, Rambaut A. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. 2012; 29: 19 691973. 135

PAGE 136

101. Como Sabetti KJ, Harriman KH, Fridkin SK, Jawahir SL, Lynfield R. Risk factors for community associated Staphylococcus aureus infections: results from parallel studies including methicillinresistant and methicillinsensitive S. aureus com pared to uninfected controls. Epidemiol. Infect. 2011; 139:419– 29. 102. David MZ, Cadilla A, Boyle Vavra S, Daum RS. Replacement of HA MRSA by CA MRSA infections at an academi c medical center in the midwestern United States, 2004 to 2008. PLoS One 2014; 9:e92760. 103. D’Agata EM , Webb GF, Horn MA, Moellering , Jr. RC, Ruan S. Modeling the invasion of community acquired methicillin r esistant Staphylococcus aureus into h ospitals. Clin. Infect. Dis. 2009; 48:274 – 284. 104. Si egel JD, Rhinehart E, Jackson M, Bre nnan PJ, Bell M. Management of mult drug resistant organisms in healthcare settings , 2006. Am. J. Infect. C ontrol . 2006; 35.10: S165 S193. 105. Morgan DJ, Diekema DJ, Sepkowitz K, Perencevich EN. Adverse outcomes associated with contact pre cautions: a review of the literature. Am. J. Infect. Control 2009 ; 37:85 – 93. 106. Deeny SR, Cooper BS, Cookson B, Hopkins S, Robotham J V. Targeted versus universal screening and decolonization to reduce healthcareassociated meticillin resistant Staphyl ococcus aureus infection. J. Hosp. Infect. 2013 ; 85:44– 33. 107. Heinrich N, Mueller a, Bartmann P, Simon A , Bierbaum G, Engelhart S. Successful management of an MRSA outbreak in a neonatal intensive care unit. Eur. J. Clin. Microbiol. Infect. Dis. 2011 ; 30:909 – 13. 108. Nbel U, Nachtnebel M, Falkenhorst G, et al. MRSA transmission on a neonatal intensive care unit: epidemiological and genomebased phylogenetic analyses. PLoS One 2013; 8:e54898. 109. Harris SR, Cartwright EJ, Trk ME, et al. Wholeg enome sequencing for analysis of an outbreak of meticillin resistant Staphylococcus aureus : a descriptive study. Lancet Infect. Dis. 2012 ; 3099:1 – 7. 110. Gordon NC, Price JR, Cole K, et al. Prediction of Staphylococcus aureus antimicrobial resistance fro m whole genome sequencing. J. Clin. Microbiol. 2014 ; 52:1182 – 1191. 111. Reuter S, Ellington MJ, Cartwright EJP, et al. Rapid bacterial wholegenome sequencing to enhance diagnostic and public health microbiology. JAMA Intern. Med. 2013 ; 173:1397– 404. 136

PAGE 137

112. Senn L, Zanetti G, Bally F, et al. Investigation of classical epidemiological links between patients harbouring identical, nonpredominant meticillin resistant Staphylococcus aureus genotypes and lessons for epidemiological tracking. J. Hosp. Infect. 2 011 ; 79:202 – 5. 113. Nelson MU, Gallagher PG. Methicillin resistant Staphylococcus aureus in the neonatal intensive care unit. Semin. Perinatol. 2012; 36:424 – 30. 114. Milstone AM, Carroll KC, Ross T, Shangraw KA, Perl TM. Community associated methicillin resistant Staphylococcus aureus strains in pediatric intensive care unit. Emerg. Infect. Dis. 2010 ; 16:647 – 55. 115. Long SW, Beres SB, Olsen RJ, Musser JM. Absence of patient to patient intrahospital transmission of Staphylococcus aureus as determined by wholegenome sequencing. MBio 2014; 5:e01692 – 14. 116. Gregory ML, Eichenwald EC, Puopolo KM. Sevenyear experience with a surveillance program to reduce methicillinresistant Staphylococcus aureus colonization in a neonatal intensive care unit. Pediatrics 2009 ; 123:e790 – 6. 117. Carpaij N, Willems RJL, Rice TW, et al. Genetic variation in spatio temporal confined USA300 community associated MRSA isolates: a shift fro m clonal dispersion to genetic evolution? PLoS One 2011; 6:e16419. 118. Liu C, Bayer A, Cosgrove SE, et al. Clini cal practice guidelines by the Infectious Diseases Society of A merica for the treatment of methicillinresistant Staphylococcus aureus infect ions in adults and children. Clin. Infect. Dis. 2011; 52:e18– 55. 119. Azarian T, Ali A, J ohnson JA, et al. Phylodynamic analysis of clinical and e nvironmental Vibrio cholerae isolates from Haiti reveals diversification driven by positive s election. MBio 2 014 ; 5:e01824 – 14 . 120. Gill MS, Lemey P, Faria NR, Rambaut A, Sha piro B, Suchard M a. Improving B ayesian population dynamics inference: a coalescent based model for multiple Loci. Mol. Biol. Evol. 2013; 30:713 – 24. 121. Baele G, Lemey P, Vansteelandt S. Make the most of your samples: Bayes factor estimators for highdimensional models of sequence evolution. BMC Bioinformatics 2013; 14:85. 122. Kass R. Bayes factors. J. Am. Stat. Assoc. 1995; 90:773 – 795. 123. Baele G, Li WLS, Drummond AJ, Suchard MA, Lemey P. Accurate model selection of relaxed molecular clocks in bayesian phylogenetics. Mol. Biol. Evol. 2013; 30:239 – 43. 137

PAGE 138

124. Baele G, Lemey P. Bayesian evolutionary model testing in the phylogenomics era: matching model complexity with computational e fficiency. Bioinformatics 2013; 29:1970 – 9. 125. Newton MA, Raftery AE. Approximate Bayesi an inference with the weighted likelihood bootstrap. J. R. Stat. Soc. Ser. B 1994; 56:3– 48. 126. Suchard MA , Weiss RE, Sinsheimer JS. Bayesian selection of continuous time Markov chain evolutionary models. Mol. Biol. Evol. 2001 ; 18:1001 – 13. 127. Zankari E, Hasman H, Cosentino S, et al. Identification of acquired antimicrobial resistance genes. J. Antimicrob. Chemother. 2012; 67:2640 – 4. 128. Werner G, Hildebrandt B, Witte W. Aminoglycosidestreptothricin resistance gene cluster aadE sat4 aphA 3 disseminated among multiresistant isolates of Enterococcus faecium . Antimicrob. Agents Chemother. 2001; 45:3267– 9. 129. Choi SM, Kim S H, Kim H J, et al. Multiplex PCR for the detection of genes encoding aminoglycoside modifying enzymes and methicillin resistance among Staphylococcus species. J. Korean Med. Sci. 2003; 18:631 – 6. 130. Lina G, Quaglia A, Reverdy ME, Leclercq R, Vandenesch F, Etienne J. Distribution of genes encoding resistance to macrolides, lincosamides, and streptogramins among staphylococci . Antimicrob. Agents Chemother. 1999; 43:1062 – 6. 131. Mayford M, Weisblum B. Conformational alterations in the ermC transcript in vivo during induction. EMBO J. 1989 ; 8:4307– 14. 132. Nbel U, Roumagnac P, Feldkamp M, et al. Frequent emergence and limited geographic dispersal of methicillinresistant Staphylococcus aureus. Proc. Natl. Acad. Sci. US A. 2008 ; 105:14130– 5. 133. Seybold U, Halvosa JS, White N, Voris V, Ray SM, Blumberg HM. Emergence of and risk factors for methicillinresistant Staphylococcus aureus of community origin in intensive care nurseries. Pediatrics 2008; 122:1039 – 46. 134. Alam MT, Read TD , Pet it RA, et al. Transmission and m icroevolution of USA 300 MRSA in US households: Evidence from wholegenome s equencing. MBio 2015 ; 6:e00054– 15. 135. Nbel U, Dordel J, Kurt K, et al. A timescale for evolution, population expansion, and spatial spread of an emerging clone of methicillinresistant Staphylococcus aureus . PLoS Pathog. 2010; 6:e1000855. 138

PAGE 139

136. Holden MT , Hsu L Y, Kurt K, et al. A genomic portrait of the emergence, evolution, and global spread of a methicillinresistant Staphylococcus aureus pandemic. Genome Res. 2013 ; 23:653 – 64. 137. Popovich KJ, Weinstein RA, Hota B. Are community associated methicillin resistant Staphylococcus aureus (MRSA) strains replacing traditional nosocomial MRSA strains? Clin. Infect. Dis. 2008; 46:787 – 94. 138. Healy CM, Hulten KG, Palazzi DL, Campbell JR, Baker CJ. Emergence of new strains of methicillin resistant Staphylococcus aureus in a neonatal intensive care unit. Clin. Infect. Dis. 2004; 39:1460– 6. 139. Gordon RJ, Lowy FD. Pathogenesis of methicillinresistant Staphylococcus aureus infection. Clin. Infect. Dis. 2008; 46 Suppl 5:S350– 9. 140. Hocevar SN, Lessa FC, Gallagher L, Conover C, Gorwitz R, Iw amoto M. Infection prevention practices in neonatal intensive care units reporting to the national healthcare safety network. Infect. Control Hosp. Epidemiol. 2014 ; 35:1126– 32. 141. Chambers HF, Deleo FR. Waves of resistance: Staphylococcus aureus in the antibiotic era. Nat. Rev. Microbiol. 2009; 7:629 – 41. 142. Arias CA , Murray BE. Antibiotic resistant bugs in the 21st century --a clinical super challenge. N. Engl. J. Med. 2009 ; 360:439– 43. 143. Merme l LA , Machan JT, Parenteau S. Seasonality of MRSA in fections. PLoS One 2011; 6:e17925. 144. Na imi T, LeDell K. Comparison of community and healthcare – associated methicillin r esistant Staphylococcus aureus i nfection. JAMA 2003; 290:2976– 2984. 145. Klein E, Smith DL, Laxminarayan R. Community associated methicillin resistant Staphylococcus aureus in outpatients, United States, 19992006. Emerg. Infect. Dis. 2009; 15:1925 – 30. 146. Otter JA, Klein JL, Watts TL, Kearns AM, French GL. Identification and control of an outbreak of ciprofloxacinsusceptible EM RSA 15 on a neonatal unit. J. Hosp. Infect. 2007 ; 67:232 – 9. 147. Tacconelli E, De Angelis G, Cataldo M a, Pozzi E, Cauda R. Does antibiotic exposure increase the risk of methicillinresistant Staphylococcus aureus (MRSA) isolation? A systematic review and metaanalysis. J. Antimicrob. Chemother. 2008; 61:26 – 38. 139

PAGE 140

148. Hicks LA , Chien Y W, Taylor TH, H aber M, Klugman KP. Outpatient antibiotic p rescribing and nonsusceptible Streptococcus pneumoniae in the United States, 19962003. Clin. Infect. Dis. 2011; 53:631 – 639. 149. Hicks LA, Suda KJ, Pharm D, et al . Antimicrobial prescription data reveal wide geographic variability in antimicrobial use in the United States , 2009 National Center for Immunization and Respiratory Diseases. Inectious Dis. Soc. Am. 20 10 ; Poster Pre:217416. 150. Coenen S, Muller A, Adriaenssens N, Vankerckhoven V, Hendrickx E, Goossens H. European Surveillance of Antimicrobial Consumption (ESAC): outpatient parenteral antibiotic treatment in Europe. J. Antimicrob. Chemother. 2009; 64: 200 – 5. 151. Neuzil KM, Mellen BG, Wright PF, Mitchel EF, Griffin MR. The effect of influenza on hospitalizations, outpatient visits, and courses of antibiotics in c hildren. N. Engl. J. Med. 2000; 342:225– 231. 152. Cookson BD. The emergence of mupirocin resistance: a challenge to infection control and antibiotic prescribing practice. J. Antimicrob. Chemother. 1998; 41:11– 18. 153. Annigeri R, Conly J, Vas S, et al. Emergence of mupirocinresistant Staphylococcus aureus in chronic peritoneal dialysis pat ients using mupirocin prophylaxis to prevent exit site infection. Perit. Dial. Int. 2001 ; 21:554 – 9. 154. Spellberg B , Bartlett JG, Gilbert DN. The future of antibiotics and r esistance. N. Engl. J. Med. 2013; 368:299– 302. 155. Arnold SR, Straus SE. Interventions to improve antibiotic prescribing practices in ambulatory care. Cochrane database Syst. Rev. 2005 ; CD003539. 156. Martinez RM, Bauerle ER, Fang FC, Butler Wu SM. Evaluation of three rapid diagnostic methods for direct identification of microorg anisms in positive blood cultures. J. Clin. Microbiol. 2014; 52:2521 – 9. 157. Kser CU, Ellington MJ, Cartwright EJP, et al. Routine use of microbial whole genome sequencing in diagnostic and public health microbiology. PLoS Pathog. 2012; 8:e1002824. 158. Didelot X, Gardy J, Colijn C. Bayesian inference of infectious disease transmission from whole genome sequence data. Mol. Biol. Evol. 2014; 31: 18691879. 159. Greatorex J, Ellington MJ, Koser CU, Rolfe KJ, Curran MD. New methods for identifying infecti ous diseases. Br. Med. Bull. 2014; 112:27– 35. 140

PAGE 141

BIOGRAPHICAL SKETCH Taj Azarian received his Doctor of Philosophy from the Department of Epidemiology in May 2015. He was a member of the Emerging Pathogens Institute at the University of Florida and worked closely with the Department of Pathology, Immunology, and Laboratory medicine. Taj previously received a Master of Public Health from the University of Florida in 2007 and subsequently completed an applied epidemiology fellowship with the Florida Department of Health. During the Florida Epidemic Intelligence Service fellowship, he quickly became acquainted with the local healthcare community through investi gations of notifiable disease and epidemics (including the 2009 H1N1 influenza pandemic). After successful completion of the fellowship, Taj served as surveillance epidemiologist for the Duval County Health Department in Jacksonville, FL prior to returning to the University of Florida to pursue his Doctor of Philosophy. Among broad his public health interest, he is most interested in antibiotic resistance. Currently, his research goals are to understand the epidemiology of bacterial pathogens through the application of whole genome sequencing, phylogenetics, and population genetics. 141