Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Incidence and determinants of tuberculosis among healthcare workers in Free State, South Africa : a historical… O'Hara, Lyndsay Michelle 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_may_ohara_lyndsay.pdf [ 2.6MB ]
Metadata
JSON: 24-1.0300152.json
JSON-LD: 24-1.0300152-ld.json
RDF/XML (Pretty): 24-1.0300152-rdf.xml
RDF/JSON: 24-1.0300152-rdf.json
Turtle: 24-1.0300152-turtle.txt
N-Triples: 24-1.0300152-rdf-ntriples.txt
Original Record: 24-1.0300152-source.json
Full Text
24-1.0300152-fulltext.txt
Citation
24-1.0300152.ris

Full Text

  INCIDENCE AND DETERMINANTS OF TUBERCULOSIS AMONG HEALTHCARE WORKERS IN FREE STATE, SOUTH AFRICA: A HISTORICAL PROSPECTIVE COHORT STUDY AND EVALUATION OF INFECTION CONTROL   by  LYNDSAY MICHELLE O’HARA  B.Sc., The University of Western Ontario, 2006 M.P.H., The University of British Columbia, 2011     A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF   DOCTOR OF PHILOSOPHY   in   THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES  (Population and Public Health)    THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)    April 2016   © Lyndsay Michelle O’Hara, 2016    ii Abstract Background  Healthcare workers (HCWs) are at high-risk of exposure to tuberculosis (TB) at work, yet the incidence rate of TB disease among HCWs in South Africa, and other high-burden countries, is unknown. The effectiveness of TB infection control (IC) measures in South African hospitals remains unclear and evidence examining the relationship between IC and TB among HCWs is lacking.  Objective 1: Estimate the incidence rate of TB among HCWs in Free State, South Africa from 2002-2012; and Objective 2: Examine the association between TB IC scores in Free State hospitals and the incidence of TB among HCWs in 2012.   Methods A record linkage was conducted to identify HCWs who were registered as TB patients. A historical prospective cohort study was conducted to obtain incidence rate ratios (IRR) of TB among HCWs in Free State from 2002-2012 and to compare patient characteristics. A mixed-effects poisson regression was used to model the association between facility type, occupation, duration of employment, and the rate of TB.  A TB IC workplace assessment tool was used in 28 public hospitals. A generalized linear mixed-effects regression was used to assess the association between TB IC scores and incidence of TB among HCWs in 2012.  Results There were 231,834 people diagnosed with TB in Free State from 2002-2012. Among HCWs, 2,677 cases of TB were diagnosed and 1,280 were expected. IRR ranged from 1.14 in 2012 to 3.12 in 2005. HCWs who were older, male, black, coloured and employed less than 20 years had higher risk of TB.  There is a large variability in TB IC in Free State. As total IC score, environmental and personal protective equipment (PPE) score increased, the probability of TB among HCWs in that hospital decreased.  Conclusions This study objectively estimates the rate of TB among HCWs in South Africa. The findings confirm that HCWs are at high risk of TB - as much as three-times higher than the population they serve. These findings re-affirm that overall IC and PPE are essential to prevent HCWs from acquiring TB. More attention to TB IC is warranted to protect HCWs and to stop the TB epidemic.   iii Preface This dissertation is original, unpublished, independent work by the author, LM O’Hara with input from the author’s supervisor (A Yassi) and committee members (EA Bryce and JM FitzGerald) and international collaborators. All of the work presented henceforth was conducted in the School of Population and Public Health at the University of British Columbia, Point Grey campus.  All components and associated methods were approved by the University of British Columbia’s Behavioural Research Ethics Board [certificate # H12-02489] and the Free State Department of Health.        iv Table of Contents ABSTRACT .................................................................................................................................................. ii PREFACE...................................................................................................................................................iii TABLE OF CONTENTS ............................................................................................................................ iv LIST OF TABLES ..................................................................................................................................... vii LIST OF FIGURES .................................................................................................................................. viii LIST OF ABBREVIATIONS ..................................................................................................................... ix ACKNOWLEDGEMENTS .......................................................................................................................... x DEDICATION ............................................................................................................................................. xi CHAPTER 1- INTRODUCTION .............................................................................................................. 1 BACKGROUND AND RATIONALE ............................................................................................................................. 1 Why focus on tuberculosis (TB)? ..................................................................................................................... 1 Why focus on South Africa? ............................................................................................................................... 2 Why focus on healthcare workers (HCWs)? ............................................................................................... 2 Why focus on infection control (IC)? ............................................................................................................. 4 CONCEPTUAL FRAMEWORK .................................................................................................................................... 5 Host (HCW) ............................................................................................................................................................... 6 Environment (workplace, community, health system) ......................................................................... 6 Agent (Mycobacterium tuberculosis)............................................................................................................ 7 Socio-economic-political factors ..................................................................................................................... 7 DISSERTATION OVERVIEW ...................................................................................................................................... 8 CONTRIBUTIONS OF THIS WORK ............................................................................................................................ 9 CHAPTER 2- TB AMONG HCW AND IC IN HIGH-BURDEN REGIONS: A SCOPING REVIEW TO IDENTIFY EXISTING KNOWLEDGE AND GAPS ...................................................................... 12 BACKGROUND ........................................................................................................................................................ 12 METHODS ............................................................................................................................................................... 14 Search strategy ...................................................................................................................................................... 14 Data extraction ..................................................................................................................................................... 15 RESULTS .................................................................................................................................................................. 15 Study characteristics and summary of findings ..................................................................................... 16 Discussion ................................................................................................................................................................. 17 CHAPTER 3- RECORD LINKAGE METHODOLOGY ...................................................................... 33 STUDY DESIGN........................................................................................................................................................ 33 PARTICIPANTS/ELIGIBILITY CRITERIA .............................................................................................................. 35 VARIABLES (INCLUDING DEFINITION OF OUTCOME) ....................................................................................... 35 Definition of incidence ....................................................................................................................................... 36 DATA SOURCES/MEASUREMENT ........................................................................................................................ 37 DATA COLLECTION PROCEDURES ........................................................................................................................ 38 MATCHING ALGORITHM ....................................................................................................................................... 40 Manual matching procedures......................................................................................................................... 44 Rationale for developing customized linkage software ..................................................................... 44 STATISTICAL METHODS FOR DATA ANALYSIS ................................................................................................... 45 Sensitivity analyses .............................................................................................................................................. 48 Validating linkage algorithm and cut-off points ................................................................................... 49  v ADOPTION OF IMPLEMENTATION RESEARCH METHODS AND DISSEMINATION OF RESULTS ..................... 49 CHAPTER 4- HISTORICAL PROSPECTIVE COHORT STUDY RESULTS ................................. 55 LINKAGE RESULTS ................................................................................................................................................. 55 PARTICIPANTS AND PRIMARY OUTCOME DATA ................................................................................................ 55 DESCRIPTION OF THE HCW COHORT ................................................................................................................ 56 DESCRIPTION OF THE GENERAL POPULATION COHORT .................................................................................. 58 MULTIVARIATE ANALYSES ................................................................................................................................... 58 SENSITIVITY ANALYSES ........................................................................................................................................ 60 CHAPTER 5- TB IC WORKPLACE ASSESSMENT METHODOLOGY ......................................... 78 STUDY SAMPLE ...................................................................................................................................................... 78 DATA COLLECTION PROCEDURES ........................................................................................................................ 79 DATA COLLECTION TOOL ...................................................................................................................................... 80 DATA ANALYSIS ..................................................................................................................................................... 81 VARIABLES ............................................................................................................................................................. 82 Definition of outcome ......................................................................................................................................... 82 Other variables ...................................................................................................................................................... 82 STATISTICAL METHODS ........................................................................................................................................ 83 THEMATIC ANALYSIS OF QUALITATIVE FINDINGS ............................................................................................ 84 DISSEMINATION OF RESULTS............................................................................................................................... 84 CHAPTER 6- TB IC WORKPLACE ASSESSMENT RESULTS ........................................................ 87 DESCRIPTION OF THE 28 HOSPITALS ................................................................................................................. 87 TB IC SCORES......................................................................................................................................................... 88 UNIVARIATE AND BIVARIATE ANALYSES ........................................................................................................... 88 MULTIVARIATE ANALYSES ................................................................................................................................... 90 QUALITATIVE FINDINGS AND QUOTES ............................................................................................................... 91 THEMATIC ANALYSIS ............................................................................................................................................ 95 CHAPTER 7- DISCUSSION ................................................................................................................. 107 DISCUSSION AND INTERPRETATION OF OBJECTIVE 1 RESULTS .................................................................. 107 How do our findings compare to what we already knew? ............................................................. 109 The role of age and HIV status .................................................................................................................... 110 The need for comprehensive OH for HCWs ............................................................................................ 111 The effect of occupation ................................................................................................................................. 112 The linkage process .......................................................................................................................................... 112 Limitations ........................................................................................................................................................... 113 DISCUSSION AND INTERPRETATION OF OBJECTIVE 2 RESULTS .................................................................. 115 Which TB IC measures are most important? ........................................................................................ 116 The value and validity of TB IC assessments/audits ......................................................................... 117 The importance of training and workplace culture .......................................................................... 118 The role of champions ..................................................................................................................................... 120 The need for OH for HCWs ............................................................................................................................. 121 The need for health systems strengthening .......................................................................................... 121 Limitations ........................................................................................................................................................... 122 THE BIGGER PICTURE ......................................................................................................................................... 123 Socioeconomic-political factors .................................................................................................................. 124 Global health research .................................................................................................................................... 130 CHAPTER 8- CONCLUSION ............................................................................................................... 137 CONTRIBUTIONS ................................................................................................................................................. 137 Conceptual framework and scoping review.......................................................................................... 137 Objective 1: Estimating incidence rate and determinants of TB disease among HCWs in Free State, South Africa over a decade .................................................................................................... 138  vi Recommendations related to objective 1 ............................................................................................... 140 Objective 2: Evaluating TB IC in 28 public hospitals in Free State, South Africa and the association with HCW TB incidence ......................................................................................................... 141 Recommendations related to objective 2 ............................................................................................... 142 STRENGTHS AND LIMITATIONS ........................................................................................................................ 143 IMPLICATIONS FOR POLICY AND PRACTICE .................................................................................................... 144 CONCLUSION ....................................................................................................................................................... 145 REFERENCES ......................................................................................................................................... 146 APPENDICES ......................................................................................................................................... 160 APPENDIX A- ICAP TB INFECTION CONTROL PRACTICES FACILITY ASSESSMENT................................ 160 APPENDIX B- KEY STATISTICAL OUTPUTS .................................................................................................... 166         vii List of Tables  Table 2.1 Search terms by research question and concept………………………… 20 Table 2.2 List of 22 high-burden countries for TB………………………………... 21 Table 2.3 Summary of Question 1 findings……………………………………….. 24 Table 2.4 Summary of Question 2 findings……………………………………….. 25 Table 2.5 Included study characteristics and methods for Question 1…………….. 26 Table 2.6 Included study characteristics for Question 2…………………………… 30 Table 3.1 Manual matching decision rules……...…………………………………. 52 Table 3.2 Review of existing linkage programs and their limitations……………… 54 Table 4.1 Incidence rate ratios of HCWs with TB by year (2002-2012)…. 65 Table 4.2 Demographic and clinical characteristics of HCWs and the general population with TB in Free State, South Africa……………………………………..  66 Table 4.3 TB treatment outcomes by facility type, occupation and duration of employment …………................................................................................................  70 Table 4.4 Relative risk estimates from Poisson regression model of HCWs with and without TB …………………………………………………………………………..  71 Table 4.5 Sensitivity analysis by weighting scheme and percentage cut-off (before manual matching)……………………………………………………………………  72 Table 4.6 Number of matches by score and year prior to manual matching……….. 73 Table 4.7 Sensitivity analysis by percentage cut-off……………………………….. 74 Table 6.1 Profile of study hospitals (2012)…………………………………………. 97 Table 6.2 Overall TB infection control workplace assessment scores…………….... 98 Table 6.3 Example of detailed TB infection control assessment results……………. 99 Table 6.4 Logistic regression model with random effect for hospital………………. 101 Table 6.5 Unadjusted and adjusted odds ratios from a generalized linear mixed-effects regression for the association between TB infection control scores and HCW TB rate in 2012………………………………………………..……………………...   105 Table 6.6 Frequency of codes and themes………………………………………….. 106    viii List of Figures  Figure 1.1 Conceptual model of determinants of TB among HCWs based on the epidemiologic triangle……………………………………………….……………….   11 Figure 2.1 Question 1 PRISMA flow diagram of search results………………...….. 22 Figure 2.2 Question 2 PRISMA flow diagram of search results………...………….. Figure 3.1 Overview of record linkage algorithm…………………………………... 23 51 Figure 3.2 Screenshot of customized web-based tool used for manual matching...... 53 Figure 4.1 Free State Province ETR.Net-PERSAL record linkage process…...……. 62 Figure 4.2 TB cases among HCWs by year………………………………...……..... 63 Figure 4.3 TB incidence rate among the general population and HCWs in Free State (2002-2012)…………………………………………………….........................  64 Figure 4.4 Probability of TB among HCWs and the general population………..….. 68 Figure 4.5 Probability of TB among HCWs and the general population by age, sex. 69 Figure 4.6 Incidence rate ratios by percentage cut-off score and year….................... 76 Figure 4.7 Estimation of positive predictive value of matches with subset of scores >90% and <70%...........................................................................................................  77 Figure 5.1 Map of Free State hospitals included in the study by region……………. 86 Figure 6.1 Total infection control score effect plot……………………………….… 102 Figure 6.2 Administrative score effect plot…………………………………………. 102 Figure 6.3 Environmental score effect plot………………………………………..... 103 Figure 6.4 Personal protective equipment (PPE) score effect plot………………...... 103 Figure 6.5 Miscellaneous score effect plot…………………….……………………. 104      ix List of Abbreviations   CI  Confidence interval CXR  Chest x-ray ETR.Net Electronic tuberculosis registry FTE  Full-time equivalent  HCWs  Healthcare workers HIV  Human Immunodeficiency Virus IC  Infection control ILO  International Labour Organization IPC  Infection prevention and control IRR  Incidence rate ratio LTBI  Latent tuberculosis infection MDR-TB Multiple-drug resistant tuberculosis OH  Occupational health OR  Odds ratio PPE  Personal protective equipment RR  Relative risk SARS  Severe acute respiratory syndrome TB  Tuberculosis TST  Tuberculin skin test XDR-TB Extensively-drug resistant tuberculosis WHO  World Health Organization         x Acknowledgements  First and foremost, I would like to thank my supervisor, Dr. Annalee Yassi. I am grateful for your mentorship, guidance and encouragement. I have learned so much from you. I would also like to extend my sincerest gratitude to my wonderfully supportive committee members, Dr. Elizabeth Bryce and Dr. Mark FitzGerald.   Thank you to everyone at the Global Health Research Program, especially Dr. Jerry Spiegel, Stephen Barker and Karen Lockhart. To my many friends and colleagues in South Africa from CHSR&D, NIOH and the Free State Department of Health. It was a great pleasure and honour to collaborate with you. I would also like to acknowledge my friends and classmates from SPPH, the MPH group and all Vancouver friends who kept me smiling throughout this journey.   This work was made possible by financial support from the Canadian Institutes for Health Research, the International Development Research Centre and the UBC Liu Institute for Global Issues.   To my family, Mom and Dad, you have always been my greatest supporters. I am eternally grateful for a lifetime of love and encouragement.  Sarah, Ryan, my grandparents and the O’Haras. Thank you for everything.   Finally- Nate. None of this would have been possible without you by my side.    xi      Dedication   This work is dedicated to Grace Leighton.  You are my greatest accomplishment.   And to the vibrant, kind, resilient people of South Africa.      1 Chapter 1- Introduction  “The struggle against tuberculosis is not dictated from above, and has not always developed in harmony with the rules of science, but it has originated in the people itself, which has finally correctly recognized its mortal enemy. It surges forward with elemental power, sometimes in a rather wild and disorganized fashion, but gradually more and more finding the right paths.” -Dr. Robert Koch, Nobel Laureate, 1905  Background and rationale Why focus on tuberculosis (TB)? Jane Austin, Lois Braille, Emily Bronte, Frederic Chopin, George Orwell, Edgar Allan Poe, and Eleanor Roosevelt are all known historical figures and great minds. They all died from TB. TB is an infectious disease caused by the bacterium Mycobacterium tuberculosis and is spread from person to person via droplet transmission. According to the World Health Organization, approximately 9.6 million people were diagnosed with TB in 2014 and 1.5 million people died from the disease in the same year (1). Treatment for TB has been available since the 1950s, however in some parts of the world, TB morbidity and mortality continues to rise. TB has again become a “burgeoning global health crisis” with the emergence of drug-resistant TB (2) and when coupled with the ongoing challenge to control Human Immunodeficiency Virus (HIV) and resulting co-infections. (3). Prevention, diagnosis and treatment of TB remain significant public health challenges that warrant continued effort and resources from the global community.  2 Why focus on South Africa? The Republic of South Africa is a stunningly beautiful and geographically, racially, socioeconomically, politically and culturally diverse nation of close to 54 million people. Although TB occurs in all parts of the world, it is estimated that 80% of all reported TB cases occurred in only 22 countries in 2014 (4).  South Africa is one of these high-burden countries with a 2014 national incidence rate of 834/100,000 (737–936/100,000) (1). For comparison, the TB incidence rate in Canada is 5.3/100,000 (4.6-5.9/100,000) (1). It is estimated that 96,000 people died from TB or TB-related illness in South Africa in 2014 alone (1). In a country where close to 19% of the adult population is HIV positive (5), TB infection and disease continue to affect all persons but especially upon the immune-compromised.  The overlap of combined infections have created a flourishing syndemic that has pushed many communities and the South African health system itself to the brink of collapse. It is estimated that approximately 80% of patients who are diagnosed with TB disease in KwaZulu-Natal province in South Africa are co-infected with HIV (3). Despite these tragic statistics, or perhaps because of them, South Africans remain hopeful, resilient and committed to overcoming the challenges associated with the ongoing TB epidemic.   Why focus on healthcare workers (HCWs)? The South African healthcare workforce is integral to the fight against TB, yet this population is often over-looked. The 2015 Ebola outbreak in West Africa was an infectious disease tragedy of epic proportions that drew needed attention to the daily occupational risks faced by HCWs.  During the outbreak in Liberia, 0.11% of the general  3 population died from Ebola compared to 8.07% of the country’s doctors, nurses and midwives (6). Prior to this, the SARS outbreak that occurred in late 2002 also garnered global attention and increased focus on the occupational risks faced by HCWs. Among 138 cases of secondary SARS cases in Hong Kong, 85 (62%) occurred among HCWs (7) and among 144 cases in Toronto, 73 (51%) were HCWs (8). One positive aspect that emerged from the SARS epidemic was the increased awareness and commitment to the development and implementation of infection control policies and procedures in healthcare facilities and that more needs to be done to protect the healthcare workforce (9, 10). While Ebola and SARS were both dramatic outbreaks and dominated the headlines, deaths from TB among HCWs continue to occur and unrecognised, despite  occurring  in numbers, far greater than that seen with less prevalent communicable diseases such as Ebola.  In many regions, HCWs are at increased risk of dying from TB due to their occupational exposure. In the Sub-Saharan African region, the HIV and TB epidemics add an additional strain on an already over-burdened healthcare workforce (11). Recent attention to high rates of TB among HCWs (12, 13), as well as hospital-based outbreaks of multidrug and extensively drug-resistant TB among patients and workers (14) have led to increased concern about the risk of Mycobacterium tuberculosis transmission in healthcare settings. A systematic review by Joshi and colleagues confirmed that TB is a significant occupational risk among HCWs in low-and middle-income countries (LMICs) (15). Similarly, another recent systematic review demonstrated that the incidence of TB among HCWs in high burden countries (>100 cases/100,000 population) is 8.4% greater  4 (95% CI 2.7%-14.0%) than the general population (16), yet this high-risk population has not been the focus of systematic research.  While HCWs are recognized as a high-risk population for contracting TB in low-income, high TB-burden settings such as South Africa, there are very few small-scale studies that have been done to quantify the problem (17). Not only are HCWs at increased risk of exposure to infectious diseases, there is a critical shortage of HCWs globally and especially in Africa (18).   The South African healthcare system is composed of parallel public and private care. The vast majority of South Africans depend on the public system, however, public healthcare in the country remains drastically underfunded and understaffed. The South African Department of Health estimates a current shortage of approximately 80,000 healthcare professionals (19). The private health sector, which serves 16% of the country’s population, employs close to 70% of South Africa’s doctors even though close to 85% of South Africa’s population relies on the public health system (19). The South African public healthcare system is already over-burdened. Neglecting the health of HCWs in this fragile setting will undoubtedly have a negative impact on the health system as a whole and its ability to respond to the challenges associated with the TB/HIV syndemic.   Why focus on infection control (IC)? As poignantly illustrated by a well-documented outbreak of MDR-TB and extensively drug-resistant (XDR-TB) in Tugela Ferry, South Africa (20), healthcare facilities are common sites for TB transmission.  A study from a 355-bed hospital in rural KwaZulu-Natal, South Africa utilized genotyping by fingerprinting and spoligotyping on XDR-TB  5 isolates to show that 85% of patients had similar strains. Furthermore, 67% had a recent hospital admission strongly suggesting that TB exposure and transmission occurred in the hospital (20).   The overarching purpose of infection prevention and control is to prevent the occurrence of healthcare acquired infections in patients, HCWs and visitors. More specifically, TB IC consists of a combination of control measures that aim to minimize TB transmission in the healthcare setting. These include administrative and environmental controls and the availability and correct use of personal protective equipment. In addition to these controls, TB IC depends upon rapid diagnosis and proper management of TB patients (21).   It is well accepted that implementation of IC measures can significantly reduce the risk of TB exposure and transmission even in limited-resource, high TB-burden regions (22), however, many published works have stated that further research is necessary to define the most efficient interventions to protect patients and HCWs (22, 23, 24).   Conceptual framework The conceptual framework used for this study (Figure 1.1) is an adapted iteration of the Epidemiologic Triangle (25). The three corners of the triangle are the Host (the HCW), the Agent (Mycobacterium tuberculosis) and the Environment. The environment was further broken down into workplace factors, community factors and health system factors. To eliminate TB among HCWs, it is necessary to break the sides of the triangle,  6 disrupting the connection between the environment, the host, and the agent, and stopping the ongoing transmission of infection and development of disease. Also included in the framework are overarching socio-economic-political factors. This study focuses primarily on the host (the HCW) and the environmental workplace factors.   Host (HCW) There are many host-related factors that may affect a HCW’s likelihood of getting TB infection and TB disease. Some of these include; age, sex, race, HIV status, being diabetic, being a smoker, socioeconomic status, occupation and duration of employment in the healthcare sector. HIV status is perhaps the greatest risk factor for TB disease in HCWs since it is estimated that approximately 61% of TB patients in South Africa are also HIV positive (1). This study compares rates among occupational groups to explore the effect of this host variable. There is also growing evidence that diabetes and smoking are important risk factors for TB that may affect response to treatment (26, 27). Diabetes, smoking and socioeconomic status were not explored here. Although data were not collected to specifically capture socioeconomic status, this study does include information about the patient’s occupation, which is a reasonable proxy variable for income.   Environment (workplace, community, health system) Workplace factors related to risk of TB disease among HCWs are the main focus of this dissertation. A HCW’s risk of TB may be influenced by increased exposure in the workplace due to a large TB patient load, the facility type that they work in and delays in  7 diagnosis and treatment. Inadequate IC (including a lack of administrative, environmental and personal protective controls) is also an important contributing factor. Finally, poor TB and IC knowledge, awareness and support for HCWs may increase rates of TB among staff.  Health system factors such as access to HIV and TB treatment, care support, confidentiality, location, cost and wait-times for services play an important role in keeping HCWs in South Africa healthy. Community factors that may increase the risk of TB transmission include crowded, poorly-ventilated housing and crowded, poorly-ventilated transport (such as minibus taxis), but are beyond the scope of this study.  Agent (Mycobacterium tuberculosis) Some factors related to the agent that have the potential to influence whether a HCW gets TB include the strain’s drug susceptibility profile (sensitive, MDR-TB, XDR-TB), the diagnosis type (pulmonary, extra-pulmonary, both), and the disease classification (new, relapse/re-treatment). These factors, in addition to the bioburden at the time of exposure (smear positive vs. smear negative TB), may also have an effect on the HCW disease outcome, cure and mortality rate.     Socio-economic-political factors Some of the socio-economic-political factors related to TB among HCWs in South Africa include structural violence, post-apartheid neoliberalism, syndemics, and stigma (28, 29). These broad, overarching factors were not the focus of this study, however, they are discussed briefly in Chapter 7.    8 Dissertation overview This dissertation is composed of following two separate, yet interconnected objectives and corresponding methods, results and interpretation and discussion of findings:   Objective 1 Estimate the incidence rate and assess the determinants of TB disease among HCWs in the Free State province, South Africa from 2002-2012. Objective 2 Examine the association between TB IC and the incidence of TB among HCWs in 28 hospitals in the Free State province, South Africa in 2012.    Following this introductory chapter is a scoping review of the existing evidence. This chapter maps and summarizes relevant articles and identifies gaps. The third chapter provides a detailed description of the methods used for a record linkage between a health human resource database and a national TB registry and the subsequent 10- year historical prospective cohort (Objective 1). The fourth chapter presents the results of the record linkage and cohort analysis. Chapter 5 describes the methods used to conduct TB IC workplace assessments in 28 public hospitals in Free State, South Africa and to estimate the association between IC scores and HCW TB rates (Objective 2). The sixth chapter presents these findings. Chapter 7 provides a discussion of objective 1 and 2 results and an interpretation of the findings in addition to a broader discussion of: the role of operational research in the fight against TB, the socio-economic and political factors that influence TB transmission, diagnosis and treatment in South Africa, and finally, we  9 discuss the ethics, challenges and opportunities associated with global health research.  The eighth and final chapter provides concluding remarks in addition to a discussion of strengths, limitations and policy implications of this work.  Contributions of this work To our knowledge, this is the first study to link human resource data with a TB database in any high-burden TB setting. Most estimates of rates of TB among HCWs in high incidence regions are based on the results of occupational health (OH) record reviews (13, 17, 30, 31, 32, 33).  Unfortunately this approach misses all HCWs diagnosed and treated elsewhere.  Other studies were limited by self-reporting of TB status in an environment where HIV and TB associated stigma that predisposes to non-disclosure (34). To our knowledge, this is the first study in a low/middle-income, high TB burden country to link confirmed active cases of TB to healthcare human resource records thereby addressing the limitations of self-reporting associated with previous estimates.   Although it is well established that HCWs in high burden countries are at high-risk of exposure to TB at work (15, 16) the incidence rate of TB disease disease among HCWs in South Africa, and other LMICs, remains unclear. Other determinants affecting exposure and outcomes among this population are also poorly understood. This is particularly problematic as the lack of good data precludes the prioritization for resource allocation and evaluation of prevention strategies. This study presents standardized incidence rates of TB among HCWs in the Free State between 2002-2012. Demographic and clinical characteristics were compared between HCWs and TB in the general  10 population.  This study also explores the effect of occupational variables on the likelihood of TB developing among HCWs.  Finally, this study provides a basis for improving TB IC in Free State hospitals. Prior to developing interventions, recommendations, policies, guidelines and even budgeting plans to improve TB IC in South Africa, it is necessary to document and systematically evaluate the current state of administrative, environmental and personal protective equipment controls in the healthcare facilities. This study also provides a baseline on which to evaluate future interventions. Operational research to assess the most appropriate approach to protect HCWs from TB exposure at work should be informed by a comprehensive analysis of current barriers and facilitators to TB IC in this setting.  11 Figure 1.1. Conceptual model of determinants of TB among HCWs based on the epidemiologic triangle     HOST (Healthcare worker) • Age • Sex • Race • HIV status • Diabetes • Smoking • Socioeconomic status • Occupation • # of years employed in healthcare sector AGENT (Mycobacterium tuberculosis) • Drug susceptibility  (sensitive, MDR-TB, XDR-TB) • Diagnosis type (pulmonary, extra-pulmonary, both) • Disease classification  (New, relapse/re-treatment) • Bioburden of the patient  (Smear positive, smear negative) ENVIRONMENT (Community, Workplace Health system) Socio-Economic-Political Factors  Structural violence  Post-apartheid neoliberalism  Syndemics  Stigma  Workplace Factors                 Exposure • # of TB patients • Facility type (hospital, clinic, non-clinical) • Delays in diagnosis and treatment                   Infection Control • Lack of administrative controls • Lack of environmental controls • Lack of personal protective equipment •                  Knowledge, Awareness, Support • Lack of IC and OH training for HCWs • Lack of knowledge re: airborne transmission • Lack of management support • Poor patient education Community  Factors • Crowded, poorly-ventilated housing • Crowded, poorly-ventilated transport (such as minibus taxis) Health System  Factors • Poor access to HIV and TB treatment, care and support • Lack of confidentiality, location, high cost and long wait-times for services • Policies and legal frameworks in place  12 Chapter 2- TB Among HCW and IC in High-Burden Regions: A Scoping Review to Identify Existing Knowledge and Gaps  Background There is extensive published evidence suggesting that HCWs are at higher risk for TB than the general population. For example, a systematic review conducted by Joshi and colleagues in 2006 (15) synthesized and quantified the incidence, prevalence and risk factors for latent tuberculosis infection (LTBI) and TB among HCWs in low and middle-income countries. The authors estimated that the annual incidence of TB in HCWs ranged from 69 to 5,780 per 100,000 and that the attributable risk for TB in HCWs ranged from 25 to 5,361 per 100,000 per year when compared to the general population (15).  Similarly, a more recent meta-analysis conducted by Baussano and colleagues concluded that the median estimated annual incidence of TB among HCWs was 1,180/100,000 persons (IQR 91–3,222) for studies from countries with high TB incidence and suggested that 81% of TB cases among HCWs in high incidence countries were attributable to exposure in health care settings (16).   There are also several published studies that evaluate the level of implementation of infection prevention and control measures in healthcare facilities (35, 36, 37, 38). However, we hypothesized that few studies exist that utilized HCW TB rates as an outcome measure to further quantify effectiveness of TB IC measures. And finally, we suspected that barriers and facilitators to TB IC in high-burden regions are also not well described in the literature.   13  The purpose of this chapter is not to replicate these findings in any way, but instead aims to further explore methodologies used by the studies included in these reviews to estimate incidence of TB disease among HCWs in high-incidence regions and to map research that examines the role that barriers and facilitators to TB IC in the hospital may play in the mitigation of TB disease in this population. This scoping review was undertaken to document what published literature exists related to TB among HCWs and the link with infection control policies and practices in high burden regions. Unlike systematic reviews, scoping reviews do not require an assessment of the quality of the studies included (39).   Based on the conceptual framework outlined in Chapter 1, the following two research questions were developed for the purposes of the scoping review:  Question 1:  What methods have been used to estimate incidence of TB disease among HCWs in high TB-burden regions and particularly in Sub-Saharan Africa? Question 2: What are the barriers and facilitators to TB IC in high TB-burden regions and particularly in Sub-Saharan Africa?     14 Methods Search strategy I conducted an Ovid MEDLINE search on October 1, 2015. Studies published between October 1, 1970 and October 1, 2015 were identified and considered for inclusion. An iterative process was used to generate search terms and the general concepts and specific terms used for each research question are described in detail in Table 2.1. The “AND” operator was used to link the search terms. The search identified terms in the title, abstract and key words fields. Articles with titles and abstracts in English but without full text available in English or French were excluded. As a final step, references of included articles were hand-searched. The following were also excluded:  Question 1  Commentaries (not original research)  Articles from low-burden TB regions (high-burden TB regions listed in Table 2.2)  Articles that describe LTBI and not TB disease  Articles examining TB diagnostics in HCWs  Articles where HCWs (as defined in Chapter 3) were not the main population of interest Question 2  Commentaries (not original research)  Articles from low-burden TB regions (high-burden TB regions listed in Table 2.2)  Articles that do not focus on IC (describe TB only or cost of IC measures)  TB IC guidelines   15 Data extraction Question 1 Studies were grouped into “Sub-Saharan African region” and “Other high TB-burden regions”. Information was recorded regarding year of publication, geographic study location, sample size, study design, and methodological details. Methodological details examined in greater detail included: the time period of the study; information regarding how HCW cases were identified (OH records, surveys, etc.); what denominator was used to calculate incidence rates; effect measures reported.   Question 2 Studies were grouped into “Sub-Saharan African region” and “Other high TB-burden regions”. Information was recorded regarding year of publication, geographic study location, study design, description of any interventions, methods used for data collection and finally barriers and facilitators to TB IC that were identified.   Results Question 1 The search performed in Medline for Question 1 yielded 93 results and an additional 3 articles were found by hand-searching references for a total of 96 articles. As shown in Figure 2.1, 4 duplicates were removed followed by the removal of 3 commentaries and one article that was only available in Russian.  An additional 66 articles that did not meet the inclusion criteria were removed after careful examination by two independent reviewers. In total, 22 articles remained for in-depth review.   16  Question 2 The search performed in Medline for Question 2 yielded 151 results and an additional 2 articles were found by hand-searching references for a total of 153 articles. As shown in Figure 2.2, 7 duplicates were removed followed by the removal of 6 commentaries and 3 articles that were only available in Russian, Japanese or Spanish.  An additional 55 articles that did not meet the inclusion criteria were removed after careful examination. In total, 13 articles remained for in-depth review.   Study characteristics and summary of findings Question 1 Of the 18 original studies identified, 11 were from Sub-Saharan Africa and 7 were from other high-burden regions. Sample sizes ranged from 9-583 HCWs with a median sample size of 108. The time period included in the studies ranged from 1-30 years with a median of 4 years. Most studies (n=13, 72.2%) were retrospective cohorts and most (n=8, 44.4%) identified HCW TB cases by reviewing OH clinic records. The majority (n=16, 88.9%) used the number of HCWs currently employed as the denominator and most (n=10, 55.6%) presented incidence rate ratios that compared HCW rates to the general population rates or another population of interest (such as teachers).   Question 2 Of the 13 studies identified, 8 were from Sub-Saharan Africa and 5 were from other high-burden regions. The number of facilities that were assessed ranged from 1-663 with a  17 median of 26.5. The majority of the studies were cross-sectional (n=9, 69.2%) and only 3 (23.1%) included implementation and evaluation of an intervention. An IC audit or inspection tool was used by 7 (53.8%), questionnaires/surveys were used by 5 (38.5%), interviews were used by 4 (30.8%), records review by 3 (23.1%) and focus groups by 2 (15.4%). Four of the studies (30.8%) estimated the rate of TB among HCWs. Other methods included direct observations (n=1, 7.7%), air sampling (n=1, 7.7%) and a literature review (n=1, 7.7%).  The most frequently identified barriers to TB IC were poor infrastructure and inadequate supplies (n=5, 38.5%) and poor staff motivation (n=5, 38.5%). Other barriers included under staffing (n=3, 23.1%), IC knowledge and awareness deficits (n-3, 23.1%), health system and legal system deficits (n=3, 23.1%) and patient non-adherence (n=1, 7.7%). The most frequently identified facilitators to TB IC were HCW motivation and support (n=5, 38.5%) followed by the development and availability of validated, locally developed assessment/audit/inspection tools (n=4, 30.8%). Other facilitators included training and education (n=3, 23.1%), keeping HCWs healthy by providing accessible and comprehensive TB/HIV services (n=3, 23.1%) ensuring that recommended control measures are simple and low-cost (n=2, 15.4%) and development of mechanisms for monitoring IC in the facility (n=1, 7.7%). Interestingly, one article suggested that fear of punishment and of infecting family members may also act as a facilitator to improved TB IC.   Discussion As shown in Table 2.3, we only found 18 original articles that estimate rates of TB disease among HCWs in high-burden countries and only 13 articles that assess TB IC  18 measures in these settings as shown in Table 2.4. This reaffirms that this is an area that warrants further attention. Although there seems to be consensus that HCWs are a high-risk, and extremely valuable population, they remain neglected in research and in allocation of resources for TB prevention, treatment and support efforts. Furthermore, the World Health Organization identified 22 high-burden countries in the Global TB Report (Table 2.2), however, we were only able to find published articles on TB in HCWs from 8 of these countries (China, Ethiopia, India, Kenya, Malawi, Russia, South Africa, Thailand) suggesting a massive research gap. Although we did identify 4 systematic reviews related to this topic, there were 7 additional articles examined here that were not included in any of these reviews. These articles are identified with an asterisk in Table 2.3.  As mentioned previously, most studies relied on OH records or self-report questionnaires, to identify cases of TB among HCWs. This approach excludes all HCWs with TB who seek care outside their workplace or who choose not to disclose their diagnosis or TB history. The study by Chu and colleagues published in 2014 (40) was the only one that performed a data record linkage to identify 62 cases of TB among HCWs over a 10-year period in Taiwan. Another study conducted in Russia utilized data from an electronic database to compare rates of TB among HCWs working in TB services to those working in general health services and also to the general population (41). It was unclear how HCWs were distinguished from other TB patients in the database.    19 We were only able to identify two studies that evaluated a TB IC intervention. Buregyeya and colleagues assessed the impact of a TB IC training program for HCWs in 51 facilities in Uganda (42).  In 40 hospitals in Malawi, Harries and colleagues evaluated the effect of introducing IC guidelines (43).  Finally, Da Costa conducted a pre-and-post intervention study at one hospital in Brazil to examine the effect of isolating TB suspects, implemented a plan for quick turnaround for acid-fast bacilli sputum tests and development and roll-out of a HCW education program focused on use of PPE (44). We also only found one study that assessed whether TB I C has any impact on the rate of TB in HCWs. Claassens and her colleagues in South Africa estimated the incidence rate ratio of TB among HCWs from questionnaires administered at 133 primary healthcare clinics (45). They also assessed whether these rates were associated with TB IC measures.  Barriers and facilitators to TB IC identified in this scoping review varied considerably. However, common themes centred on lack of resources (including personnel, infrastructure and equipment), the need for motivation, support and empowerment in addition to education, training and awareness.  There are very few large-scale published studies that link occupational data with TB data to estimate rates of TB disease among HCWs in high-burden countries and none conducted in Sub-Saharan Africa. There is a lack of consensus regarding the most important barriers and facilitators to TB IC and measuring effectiveness of TB IC implementation is challenging.   20 Table 2.1. Search terms by research question and concept Question 1: What methods have been used to estimate incidence of TB among HCWs in high TB-burden regions and particularly in Sub-Saharan Africa? Concept Terms Number of studies TB Tuberculosis, Multidrug-Resistant/ OR Extensively Drug-Resistant Tuberculosis/ or Tuberculosis/ OR Tuberculosis, Pulmonary/ OR Mycobacterium tuberculosis/  146,405 HCWs health workers OR healthcare workers OR health personnel or personnel, hospital 39,806 Incidence incidence OR incidence rate 195,900 Question 2: What are the barriers and facilitators to TB IC in high TB-burden regions and particularly in Sub-Saharan Africa? Concept Terms Number of studies TB Tuberculosis, Multidrug-Resistant/ OR Extensively Drug-Resistant Tuberculosis/ or Tuberculosis/ OR Tuberculosis, Pulmonary/ OR Mycobacterium tuberculosis/  146,405 HCWs health workers OR healthcare workers OR health personnel or personnel, hospital 39,806 IC Infection control OR infection prevention and control 21,160        21 Table 2.2. List of 22 high-burden countries for TB*  Sub-Saharan Africa Other high-burden regions Democratic Republic of Congo Ethiopia Kenya Mozambique Nigeria South Africa Uganda United Republic of Tanzania Zimbabwe  Afghanistan Bangladesh Brazil Cambodia China India Indonesia Myanmar Pakistan Philippines Russian Federation Thailand Viet Nam  *As identified in the Global TB Report 2015. Available from: http://www.who.int/tb/publications/global_report/gtbr15_annex02.pdf?ua=1    22 Figure 2.1. Question 1 PRISMA flow diagram* of search results                      *Adapted from: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097  Records identified through database searching (n = 93) Screening Included Eligibility Identification Additional records identified through other sources (n = 3) Records after duplicates removed (n = 92) Records screened (n = 92) Records excluded (Commentaries, n = 3) (Not in English, n=1) Full-text articles assessed for eligibility (n = 88) Full-text articles excluded, (low burden countries, n=43) (not TB disease, n=17) (not HCWs, n=6) (n = 66) Studies included in qualitative synthesis 11 from Sub-Saharan Africa 7 from other high-burden countries 4 reviews (n = 22)  23 Figure 2.2. Question 2 PRISMA flow diagram* of search results                      *Adapted from: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097  Records identified through database searching (n = 151) Screening Included Eligibility Identification Additional records identified through other sources (n = 2) Records after duplicates removed (n = 146) Records screened (n = 77) Records excluded (Commentaries, n = 6) (Not in English, n=3) Full-text articles assessed for eligibility (n = 68) Full-text articles excluded, (low burden countries, n=28) (not IC focused, n=16) (guidelines, n=11) (n = 55)  Studies included in qualitative synthesis 8 from Sub-Saharan Africa 5 from other high-burden countries  (n = 13)  24  Table 2.3.  Summary of Question 1 findings*  N (%) Sample size Range= 9-583 HCW TB cases Median= 108  Time period Range= 1-30 years Median=4 years  Study location Sub-Saharan Africa Other high-burden region  11 (61.1) 7 (38.9) Study design Retrospective cohort Prospective cohort Cross-sectional Case-control Population-based observational cohort  13 (72.2) 2 (11.1) 1 (5.6) 1 (5.6) 1 (5.6)  Data Source Review of OH clinic records Administrative data/registry Self-report (questionnaire) Lab test, Chest x-ray (CXR)  8 (44.4) 6 (33.3) 3 (16.7) 1 (5.6) Denominator All HCWs currently employed Person-years  16 (88.9) 2 (11.1) Effect Measure IRR (compared to gen pop or other pop) Incidence Rate Odds Ratio TB Case notification rate  10 (55.6) 3 (16.7) 3 (16.7) 2 (11.1) *Excluding 4 systematic reviews  25 Table 2.4.  Summary of Question 2 findings  N (%) Number of facilities included Range= 1-663 Median= 26.5  Study location Sub-Saharan Africa Other high-burden region  8 (61.5) 5 (38.5) Study design Cross-sectional Retrospective Record Review Pre-and-post intervention Qualitative  Review  9 (69.2) 1 (7.7) 1 (7.7) 1 (7.7) 1 (7.7) Included an intervention No Yes  10 (76.9) 3 (23.1) Methods used*  IC audit/inspection Questionnaire/survey Determined HCW TB rate Record review Interviews Focus groups Observations Air sampling Literature review  7 (53.8) 5 (38.5) 4 (30.8) 4 (30.8) 3 (23.1) 2 (15.4) 1 (7.7) 1 (7.7) 1 (7.7) Barriers identified* Poor infrastructure/inadequate supplies Poor staff motivation Under staffing IC knowledge/awareness deficits  Health/legal system deficits Patient non-compliance  5 (38.5) 5 (38.5) 3 (23.1) 3 (23.1) 3 (23.1) 1 (7.7) Facilitators identified* HCW motivation and support Validated, locally developed tools Training/education Keeping HCWs healthy Simple, low-cost control measures Mechanisms for monitoring Fear  5 (38.5) 4 (30.8) 3 (23.1) 3 (23.1) 2 (15.4) 1 (7.7) 1 (7.7)  * Studies may have included more than one and percentages therefore do not add up to 100. 26 Table 2.5. Included study characteristics and methods for Question 1   Sub-Saharan African Region (n=11)  Primary Author  (reference #)  Year Published  Study Location  # of TB Cases Among HCWs  Study Design Methodological Details Time Period How were HCW TB cases identified? (Data Source) Denominator used? Effect Measure(s) Tudor C* (46) 2014 KwaZulu-Natal, South Africa 112 Retrospective cohort 5 years Review of employee OH medical records Total # of HCWs employed during each year IRR (comparing HCWs and general population and for risk factors among HCWs) Claassens MM* (45)  2013 133 clinics in 5 provinces of South Africa 47 Cross-sectional 3 years Questionnaire  Total # of HCWs for each of the 3 years in the study SIR and OR (association of TB cases in HCWs with IC audit)   O’Donnell MR* (14) 2010 KwaZulu-Natal, South Africa 231 Retrospective cohort 5 years Self-reported HCW status by questionnaire and chart review  Total # of HCWs employed and Non-HCWs admitted with MDR and XDR TB IRR (Comparing incidence of hospitalization for MDR and XDR-TB for HCWs vs non-HCWs) Galgalo T* (31) 2008 Kenya 227 Case-Control 4 years Review of staff TB treatment register used for this study at the hospital  Total # of staff at the hospital by year aOR (risk factors associated with TB) Naidoo S (13) 2006 Eight public sector hospitals in KwaZulu-Natal, South Africa 583 Retrospective cohort 5 years Review of records in the staff/OH clinic  All HCWs employed in the study hospitals Mean Incidence (TB among HCWs in public sector hospitals vs. community acquired TB) Kanyerere H (47) 2003 Lilongwe Central Hospital, Malawi 33 Retrospective cohort 1 year Review of administration records All HCWs employed between January and December 2001 TB case notification rate  27  Sub-Saharan African Region (n=11) (continued)   Primary Author  (reference #)  Year Published  Study Location  # of TB Cases Among HCWs  Study Design Methodological Details  Time Period How were HCW TB cases identified? (Data Source) Denominator used? Effect Measure(s) Eyob G (33) 2002 Tuberculosis Demonstration and Training Centre, Addis Ababa, Ethiopia 24 Retrospective cohort 9 years Review of OH medical charts Person-time at risk  calculated from study entry for all staff, or employment date for staff joining later to date of tuberculosis diagnosis, date of leaving, or date of study closure IRR (comparing staff to general population) Harries AD (48) 2002 Malawi 196 Retrospective cohort  3 years Review of administrative data and TB registers Total # of HCWs in the study hospitals Relative risk compared to teachers Harries AD (43) 1999 40 hospitals in Malawi 108 Retrospective cohort 1 year Review of hospital TB registers and interviews Total # of HCWs currently employed Annual TB case notification rate of HCWs and RR (HCWs compared to general pop) Wilkinson D (12) 1998 Single District Hospital in South Africa 22 Retrospective cohort 6 years Extracted data from anonymized TB control programme register  Total # of HCWs  IRR (compared IR of TB among HCWs with age specific rate in the community) Harries D (49) 1997 Queen Elizabeth Central Hospital Blantyre, Malawi 12 Retrospective cohort 1 year Nursing Records Total # of nurses at Queen Elizabeth Central Hospital in 1993-1994 IR in percentage (# of nurses with TB at Queen Elizabeth Central Hospital in 1993-1994 *Not included in any review articles on this topic  28   Other High TB-Burden Regions (n=7)   Primary Author (reference #)  Year Published  Study Location  # of TB Cases Among HCWs  Study Design Methodological Details Time Period How were HCW TB cases identified? (Data Source) Denominator used? Effect Measure(s) Chu H* (40) 2014 Taiwan 62 Population based observational cohort 10 years Taiwan National Health Insurance Database (including a registry specifically for medical personnel) All HCWs and randomly selected matched non-HCWs   Incidence Rate  Ratio (HCWs vs matched non-HCW cohort) Kiertiburanakul S* (50) 2012 Ramathibodi Hospital, Mahidol University Bangkok, Thailand 9 Prospective cohort 5 years Tuberculin skin test (TST) using the Mantoux method  All new HCWs at Ramathibodi Hospital from 2005-2008 OR (risk factors associated with prevalence) Dimitrova B (41) 2005 Samara Oblast, Russia 474 Retrospective cohort 9 years Electronic database at Samara Oblast TB Dispensary (OTBD) Average # employed each year at each facility IRR (comparing staff working in TB services to GHS) Gopinath KG (51) 2004 Christian Medical College (CMC) Vellore, India  125 Retrospective cohort 10 years Records maintained in the staff and students health services at CMC All HCWs at CMC and the general population  IR (comparing extra-pulmonary TB among HCWs at CMC with that in the general pop) Kwan SY* (52) 1990 Grantham Hospital, Hong Kong 23 Retrospective cohort 30 years Occupational record review Numbers of total staff-years Cumulative incidence, case detection rate/year Rao KG (53) 2004 Postgraduate Institute of Medical Education and  Research (PIMER), North India  13 Prospective cohort 3 years Structured Questionnaire  All resident doctors working at PIMER Incidence rate *Not included in any review articles on this topic   29   Primary Author (reference #)  Year Published  Number of Studies Included Key Findings Baussano I (16) 2011 43 -Stratified pooled estimates for TB in high-burden countries was 3.7 (95% CI 2.9-4.5) Menzies D (54) 2007 20 (TB disease in low and middle income countries) -The annual risk of TB attributable to nosocomial exposure ranged from 25 to 5361 per 100000 (median 228)  Joshi R (15)  2006  42 -Annual incidence of TB in HCWs ranged from 69 to 5,780 per 100,000 -Attributable risk for TB in HCWs, compared to the general population, ranged from 25 to 5,361 per 100,000 per year  Menzies D (55) 1995 12 (all high-income, low-burden countries) -Incidence of all forms of TB increased by 24 to 34 percent in Denmark, Italy, and Switzerland and by 18.4 percent in the United States between 1985 and 1991     30 Table 2.6. Included study characteristics for Question 2   Sub-Saharan African Region (n=8)   Primary Author (reference #)  Year Published  Study Location  Study Design  Description of Intervention  Methods Used  Barriers  Facilitators Brouwer M (56) 2015 29 health facilities in Mozambique Cross-sectional  None -Inspection with a newly developed study assessment tool  -Interviews with managers  -Lack of guidelines or insufficient implementation of available guidelines -Developing and testing TB-IPC assessment tools locally in different settings -Motivation of HCWs as well as social support from colleagues and superiors -Training Buregyeya E (42) 2013 51 facilities in Mukono and Wakiso districts in Uganda Mixed methods cross-sectional Training on TB IC for HCWs -Facility survey -Participant observations  -Review of facility records -Focus group discussions with HCWs  -Under-staffing  -Lack of space for patient separation,  -Lack of funds to purchase masks  -Health workers not appreciating the importance of TB IC None specified Claassens MM (45) 2013 133 clinics in 5 provinces of South Africa Cross-sectional None -TB IC audit  -Association between TB IC and TB in HCWs -Lack of strategies to document and monitor TB in HCWs -Under-staffing -Validation of audit tools prior to use  Farley JE (35) 2012 All 24 MDR and XDR-TB facilities in South Africa Cross-sectional descriptive None -Structured interviews key informants -Facility inspection  -HCW knowledge, attitudes and practice questionnaire -Poor IC infrastructure -Lack of attention to IC -Nurse empowerment to influence practice change -Training for HCWs with less clinical training -Facility-specific IC training -Inform HCWs of approaches that limit transmission, including alternative models to in-patient care    31  Sub-Saharan African Region (n=8) (continued)  Primary Author (reference #)  Year Published  Study Location  Study Design  Description of Intervention  Methods Used  Barriers  Facilitators Harries A (43) 2002 40 hospitals in Malawi Retrospective Record Review -Introduction of guidelines -TB among HCWs pre and post -Staff and patient interviews -Review of case notes and TB registers to determine time to dx and tx pre and post - Review of lab records -Poor staff motivation -Insufficient staffing -Appointing an IC committee -Employing more laboratory staff  -Faster turn-around on sputum smears -Offering comprehensive OH services for HCWs Kanjee Z (57)  2011 One hospital in KwaZulu-Natal, South Africa Cross-sectional  None -Baseline audit of staff TB IC knowledge, attitude and practice -Stigma and confidentiality concerns -Concerns about the confidentiality of staff health information -Inadequate resources -Patient non-compliance -Conducting simplified TB IC assessments -Implementing multi-faceted TB IC facility and behavioural change interventions Kanjee Z (58) 2012 Two hospitals in KwaZulu-Natal, South Africa Cross-sectional  None -Questionnaire to assess HCW TB IC information, motivation, and behavioral skills (and implementation -Perception by HCWs that TB is not a significant threat  -Poor general TB knowledge -Social support -Emphasizing development of motivation and behavioral skills  Reid (37) 2012 663 HIV care and treatment sites in 9 African countries Cross-sectional  None -Survey to assess the presence of a TB IC plan, triage practices for suspected TB cases, location of sputum collection and availability of particulate and the association with facility characteristics -Unstable/weak health system - Institution of a site-specific IC plan -Developing mechanisms for the supervision and monitoring of IC activities -Providing on-site anti-tuberculosis care     32    Other High TB-Burden Regions (n=5)  Primary Author (reference #)  Year Published  Study Location  Study Design  Description of Intervention  Methods Used  Barriers  Facilitators Da Costa (44) 2009 One hospital, Rio de Janeiro, Brazil Pre-and-post intervention study with HCW TB rate as outcome Implementation of isolation of TB suspects, quick turnaround for  sputum tests and PPE training -Review of medical records and TB skin testing -Evaluation of isolation rooms -Lack of specific laws and conceptual legal framework for the systematic application of effective public health policies in resource-limited nations  -Measures must be easy to implement and low cost - Full support of local and national health authorities He (59) 2010 177 TB centres in Henan Province, China Cross-sectional None -TB IC checklist -TST, sputum smears, CXR -Risk factor questionnaire -Poor/old physical infrastructure -Lack of IC awareness -Promoting awareness of TB disease among HCWs -Simple, cost-effective interventions Luksamijarulkul P (60)  2004 One hospital, Bangkok, Thailand Cross-sectional None -Interviews -Air quality sampling -Lack of IC infrastructure (isolation rooms, poor ventilation) -Over-crowding - Emphasizing health promoting behavior to HCWs to enhance their immunity against infection Pai M (61) 2006 India Literature Review N/A -Review of existing literature -General lack of established IC procedures -Gaps in knowledge  -Patient factors that increase risk for nosocomial exposure (such as prolonged hospitalization, delays in diagnosis and treatment) - Adaptation of international technical guidance to the Indian context Woith (62) 2012 5 TB facilities in Russia Qualitative study None -Focus groups -Knowledge deficits -Negative attitudes related to the discomfort of respirators -Practices with respect to quality and care of respirators -Education and training -Fear of infecting loved ones -Fear of punishment  33 Chapter 3- Record Linkage Methodology  In this chapter, I describe the steps involved in the record linkage and historical prospective cohort study. I discuss the study design and eligibility criteria. I also define the outcome variable in addition to other variables included in the study. The data sources, data collection procedures and development of the linkage algorithm are described in detail. Finally, the statistical methods employed are outlined.  Study design This is a probabilistic record linkage between the South African national human resource database called PERSAL “(Personnel Salaries”) and the national TB registry called ETR.Net. The two data registries do not share a unique identification number. With the database generated from the linked records, we conducted a historical prospective cohort study.  In a historical prospective cohort study, the cohort is constructed in the past based on a common exposure (in this case, employment in the healthcare sector in 2002-2012), and each member of the cohort is followed prospectively from the time that individual began employment until end of the study period to determine if this individual developed the outcome of interest (in this case, TB).  The follow up period to ascertain the outcome of interest can extend beyond the end of the individual’s employment, but could not have occurred prior to entry into the cohort.  This design provided us with information about the HCWs prior to exposure, meaning that we knew if they had TB prior to starting work  34 in the healthcare sector. If an individual had had TB for example in 2003, then began working in healthcare in 2011, this individual would not be in the linked dataset, as the outcome of interest occurred prior to entry into the cohort. This study design is also less frequently referred to as a “non-concurrent study”. In this dissertation, we also use the term “retrospective cohort” interchangeably with “historical prospective cohort” when discussing other cohort studies if this was the term used by the study authors in the published article. Occupational cohort studies are appropriate to examine temporal patterns of disease or injury among workers in a particular industry or in individual facilities. (63). This methodology has played an instrumental role in the identification of numerous occupational hazards and quantification of associated risks. (63) For example, occupational cohort studies have been globally successfully used to investigate numerous morbidity and mortality outcomes such as cancer (64), leukemia (65), respiratory symptoms (66) and adverse reproductive outcomes (67).   Record linkage is the process of combining information from two or more records that are believed to relate to a common entity. (68) Record linkage algorithms may employ deterministic or probabilistic methods (68, 69). This study utilized the probabilistic approach. Unlike deterministic record linkage techniques, probabilistic linkage methods consider that certain values and variables have more discriminatory power than others (69). As illustrated by the model developed by Fellegi and Sunter (70), links can be identified as matched pairs, possible matches (also sometimes referred to as the “grey area”) or non-matches after careful application of sound linkage algorithm and manual matching decision rules. This theory was used to develop and employ the technique for  35 this study described in greater detail below. Methodologies similar to those presented here have been used extensively elsewhere to determine other health outcomes among a wide variety of occupational cohorts (71, 72, 73). More specifically, probabilistic record linkage is a valid approach to combine databases that do not share a common identification number (74).   Participants/eligibility criteria  For this study, HCWs were defined as “all people engaged in the promotion, protection or improvement of the health of the population” (75). This definition is not limited to those who provide direct patient care, but also extends to all who work in a healthcare facility such as cleaners, security personnel, etc.  Although it is recognized that unpaid community health workers and volunteer caregivers are important health providers, for this study, only paid HCWs were included since the human resource database (PERSAL) was used to identify HCWs. This therefore also excludes contracted workers. All employees of the Free State Department of Health from 2002-2012 who had worked in the health department for at least one month were eligible for inclusion. HCWs at all facility levels were included (local clinics, primary, secondary and tertiary hospitals and non-clinical settings such as administrative head offices and central laundry facilities).   Variables (including definition of outcome) The main outcome measure for this study was diagnosis with TB disease. HCWs with laboratory-confirmed M. tuberculosis disease (all forms- including pulmonary, extra-pulmonary, disseminated and miliary TB as well TB meningitis) (76) as identified in  36 ETR.Net were included in the linked dataset. HCWs with confirmed and documented reactivation of TB were included in the linkage. Those with reinfection were included in the incidence calculations if date of diagnosis is after date of employment. Age, sex, race, HIV status, occupation, facility type, duration of employment, diagnosis type, disease classification, outcome and TB drug sensitivity were included as covariates. The total number of TB patients in the Free State (general population and HCWs) was calculated for each year from ETR.Net. The total number of HCWs employed in Free State was recorded from PERSAL and average Full-Time Equivalent (FTE) were calculated for each year (2002-2012). General population estimates for each year (2002-2012) by age and sex were obtained from census data compiled by Statistics South Africa.  Definition of incidence In epidemiology, the incidence rate is a commonly used measure of disease frequency. More specifically, incidence measures the new cases of a disease that develop in a population during a specified period (77). In accordance with this definition, once a HCW is classified as a TB case, she or he is no longer eligible to become a new case, meaning that they should not be included more than once in the linkage. However, it is of course possible for the same pathological event to happen more than once to the same individual. For this reason, the British Medical Journal suggests that in the case of infectious diseases, counting all episodes is more appropriate than the traditional approach where calculation of incidence is restricted to the first event (78). In the case of TB specifically, it is possible to have a re-activation of the disease or to be re-infected.  TB reactivation occurs when dormant bacilli that were present in the body for months or  37 years after primary infection, start to multiply (79). This is often in response to a trigger, such as weakening of the immune system by HIV infection. Re-infection is when an individual incurs a repeat infection after previously having a primary infection. Capturing cases of reactivation and reinfection among HCWs with TB is important, therefore all cases were included in the linked dataset (including individuals who had more than one ETR.Net entry).  For the purposes of this study, TB incidence refers to all episodes and not just first diagnosis.  Data sources/measurement ETR.Net is an electronic TB registry designed for TB and HIV surveillance, program monitoring and evaluation. The system allows for data entry at the district level and is currently used in all nine provinces in South Africa. ETR.Net contains over 1,000,000 records in a secure and encrypted dataset. The database is maintained by the TB department in the Free State Department of Health. The information that is inputted into ETR.Net comes from paper TB registries at the hospitals and clinics. A TB coordinator or a nurse is responsible for keeping these registries up to date at each facility and for submitting case details to the regional or district data entry clerk on a regular basis. The data collection and data entry practices and skills of the individuals in each facility are highly variable; and the quality of ETR.Net data is therefore also assumed to vary by facility, region and province. Data checks and data entry validations including warnings to the user and the use of required fields are built into the ETR.Net system in an effort to improve data quality.    38 PERSAL is an integrated transversal system for the administration of human resource transactions and payment of salary for the South African Government at National and Provincial levels. Approximately 1.1 million government employees are paid through the system each month.  Each employee is assigned a unique ID number within the system (“PERSAL number”). Since this system is used for payroll purposes, the information contained within the database is assumed to be of fair quality and is kept up to date. The system is maintained by the human resource and IT departments at the Free State Provincial Department of Health.   As mentioned previously, data describing the general population (those who are not HCWs and who do not have TB) were obtained from publicly available reports from Statistics South Africa (80) and the Human Sciences Research Council of South Africa (81).   Data collection procedures Prior to traveling to South Africa, the research team worked with local IT experts to evaluate the feasibility of conducting this linkage project.  Although the two databases did not contain a unique identifier, they did share several indirect identifiers such as first name, surname, birthdate, sex and geographic region.  The team concluded that a reliable and accurate linkage was possible, but would require significant time and resources. The record linkage therefore was performed to estimate the probability that a PERSAL and ETR.Net record refers to the same person.    39 Approval from the institutional research ethics board was obtained at the University of British Columbia in September 2012 and at the University of the Free State in November 2012. Written approval was also obtained by the Free State Department of Health in November 2012. The initial record linkage was conducted on site at the Free State Department of Health in Bloemfontein, South Africa in February and March 2013 with assistance from the Information Technology department and the Human Resource Management department.   Raw data were acquired (PERSAL and ETR databases) from 2002-2012. Both databases were in .CSV format. Data were imported into Microsoft SQL Server 2008 and a custom application was used to perform this task. A simple check for data integrity was performed (on ETR) and any rows that did not pass were removed.  Any rows where “last name” was empty were removed and any rows where “facility” was missing or incomplete were removed. This removed 16 rows (16/ 239,204 total rows in ETR= 0.00668%) of data from ETR. A “Region” field was created in both ETR and PERSAL databases using a compiled list of hospital/clinics. In PERSAL, name matching was performed to determine the region of approximately 80% of the entries. Of the remaining entries, the programmer was able to match about half of them manually to a region. In these cases, the name match had failed due to either typos or abbreviated names. ETR had much fewer facilities listed than PERSAL, so 100% of entries were assigned a region. The rationale for including region was that an individual is unlikely to receive TB treatment in a region that is very far from their workplace/home.  However, it is possible that individuals live near regional borders and may live/work in Clocolan for example  40 (Thabo Mofutsanyana region) and drive to Landybrand (Motheo region) for TB treatment. In spite of this limitation, region was deemed to be additional useful information and was included as an unweighted variable in the manual matching.  The linked dataset was then kept in a password-protected file on a UBC server. Data cleaning, development and refinement of the linkage algorithm and protocols and manual review of records were conducted at the University of British Columbia in Vancouver, Canada between April 2013- December 2014.  Reports from Statistics South Africa and the Human Sciences Research Council were downloaded directly from the websites as CSV files. Data analysis was conducted between October 2014- September 2015.   Matching algorithm The matching algorithm was written as a custom application with programming language C#.  The program accessed two different Microsoft SQL Server database tables (ETR, PERSAL) that were imported and cleaned in the previous steps and created a new matching table.  When the ETR and PERSAL data were imported, Microsoft SQL Server randomly assigned a unique index to each row.  These indices were used to reference the ETR and PERSAL rows in the matching table that were generated. This allowed the data to be anonymized after they had been matched; however, if the unique index of the PERSAL ID had been used, it would not have been possible to anonymize the data.   41 Following the theories presented by Clark and Newcombe (82, 68), variables were assigned a linkage weight according to their reliability and discriminatory power. Reliability or m probability is the probability that a matching variable agrees given that the comparison pair is in fact a match.  This is similar in concept to sensitivity. Discriminatory power or u probability is the probability that a matching variable agrees purely by chance (is a non-match). This is similar in concept to specificity and reduces the overall weight. (83) Based on these parameters, the total weight (or “percentage score”) is derived by summing the separate field comparisons across all fields.  For the record linkage described here, surname was evaluated first. If it was an exact match, it was assigned a perfect score of 40 for surname, and the program proceeded to assess given name.  If surname was not an exact match, a coefficient for the inaccuracy of the name was created (“sc”=surname coefficient). Levenstein distance was calculated between the surname as registered in ETR and the surname in PERSAL (“sd”=surname distance).  The number of letters in the longest surname in either ETR or PERSAL was then calculated (“sl”=surname length). By definition, sl >= sd.  We then say sc = sd / sl, and therefore 0 <= sc <= 1.  This means that sc is small when there is little distance between the names matched, and sc is large when they are very different.  Next the program checked if the two names sound the same, based on the double metaphone algorithm (84).  If they do sound the same (eg. Steven and Stephen), 0.05 was subtracted from sc.  Only a small weight was placed on the double metaphone algorithm because it tends to be less accurate than the Levenstein distance. Next, the first initials of  42 the surname were compared (eg. Baker and Barker). If they were the same, 0.1 was subtracted from sc.  If at this point sc < 0, sc was reset to equal 0. The calculation was then finalized for the score of an imperfect surname match.  With 0 <= sc <= 1, with sc closer to 0 when there was a close match, a percentage of a score of a max score of 35 was derived (40 was the score assigned for a perfect match, which is impossible to achieve for an imperfect match.) Therefore, Score = 35 * (1 – sc).  Given name matching was in almost an identical manner to the Surname matching, with the following two exceptions; 1) The score for a perfect match is 30, and the max score for an imperfect match is 25. 2) There could be many given names; for example, PERSAL could have the given names “Stephen John” and ETR might only have “Stephen”.   The algorithm does not penalize the score if a name is missing from one of the databases.  “Stephen” and “Stephen John” would be considered a perfect match. Given Name Score = 25 * (1 – gc).  To generate a score for age, the number of days difference, between the birthdates listed in ETR and PERSAL was calculated. If the difference was 0, an age score of 30 was assigned. If the difference was less than 150, the score was 25. If the difference was less than 366, the score was 20. If the difference was less than 500, the score was 10. If the difference was less than 1000, the score was 5. A difference up to 366 could still indicate a perfect match, as the ETR database does not always have a precise date of birth.  It does  43 contain the age, and date of treatment, which can be used to reverse calculate a possible birthdate.  However, this date could potentially be inaccurate up to 365 days.  These cases were flagged for manual matching so that the reviewer was aware.  For sex, a simple comparison of male vs. female was used.  If it was the same, sex score was 30; otherwise, it was 0.  Based on these parameters, final percentage scores were assigned to each comparison. A total score was calculated as the sum of surname (maximum possible score=40), given name (30), age (30), and sex (30) scores where the maximum possible score was 130.   Any final percentage score less than 70% (91 out of 130) was filtered out and were not included in the final dataset. Scores greater than 90% (117 out of 130) were included without manual review. Scores between 80-90% were reviewed manually. An overview of the matching algorithm and linkage decision-making process in presented in Figure 3.1.   Decision rules for manual matching, as described in detail in Table 3.1, were developed and were employed by two reviewers. After the PERSAL-ETR linkage was complete, the link was reversed to determine how many ETR records were linked with multiple PERSAL entries. Finally, all accepted possible matches were re-assessed by a second reviewer using the same decision rules.    44 Manual matching procedures The computerized steps in the record linkage described above should have generated all possible matches and matched pairs between the ETR and PERSAL databases.  However, this dataset also included many false-positive matches and therefore manual review by a human is necessary to remove as many of these false-positives as possible. A web-based tool, with an example screen shown in Figure 3.2, was used to facilitate manual matching of possible matches in the grey area.  In many cases, a single PERSAL employee was linked to multiple entries in the ETR database, especially if they received multiple TB treatments.  If it was believed that the displayed data shows such a case, all such records were included for reasons discussed previously. It is also possible that the algorithm only returned false positives, for a single PERSAL employee, these cases were not included. After clicking “Save”, a new employee with its matches are loaded onto the screen and this process was repeated for all possible matches.   Rationale for developing customized linkage software The research team decided to develop customized linkage software instead of using existing tools for several reasons. The Free State Department of Health requested that the initial linkage be conducted on site at Bophelo house in Bloemfontein due to data confidentiality issues.  This task was conducted during a tight timeframe since members of the Canadian research team were visiting for a limited amount of time. As manual linkage would not be possible on site within this narrow timeframe, it was necessary to  45 maintain the statistical reasoning behind each individual potential match, without any personally identifying data.  For example, were birthdates an exact match?; were names different (direct comparison)?; did they sound the same (double metaphone)?; how different were the names (Levenshtein distance)?  It was necessary to store these non-identifying values so that the weights of the different variable settings could be adjusted in the future.  To our knowledge, no existing linkage software recorded this information. Furthermore, we did not find any existing software that were designed to consider the nuances of Sesotho, Afrikaans, Xhosa, Zulu and Tswana names as all software were designed for use in North America or Europe. Several additional issues were identified with currently available linkage tools as outlined in Table 3.2.   Methods for privacy protection were considered throughout the process. For example, the team limited exposure of personally identifying information to only two reviewers. Furthermore, that final dataset used for the analyses was stripped of all identifying information and included only encrypted IDs.  Statistical methods for data analysis The total number of TB cases among HCWs in Free State was tabulated for each year (2002-2012). Descriptive statistics were utilized to show demographic and clinical characteristics of HCWs and the general population with TB in the province.  Person-time is the sum of individual units of time that individuals in a study population have been exposed or at risk to the disease of interest. For this study, we estimated the  46 number of person-years at risk for TB for HCWs by assigning a full-time equivalent (FTE) score to each HCW. For example, if they worked part-time, they would receive a 0.5 for this variable. For each individual, their FTE were summed over the 10 years of the study to generate their individual person-time at risk. Average FTE was then calculated for each year to generate denominators for subsequent calculations.   The number of observed cases of TB among HCWs and person years at risk were identified (from HR database PERSAL) for each year (2009-2012). Expected number of cases for each year were calculated by multiplying the number of person years at risk each year by the corresponding national TB incidence rate in the general adult population. To calculate the incidence rate ratios, observed numbers of cases of TB among HCWs were divided by the expected numbers in the general adult population for each year. The exact 95% confidence interval was defined, assuming that the number of observed cases follows a Poisson distribution.   Poisson regression was used to model the association between employment as a HCW (compared to the general population) and the probability of TB. Age and sex were also included in the model.  Using the cohort of HCWs only (those with TB and those without TB), bivariate analyses were conducted using a chi-square test to determine whether there is a significant difference between the expected frequencies and the observed frequencies in treatment  47 outcomes and workplace variables (facility type, occupation, and duration of employment).   With the same dataset (HCWs only with TB and without TB), Poisson regression was used to model the association between facility type, occupation and duration of employment and the rate of TB, with the relative risk being a measure of this association. Birth year, race and sex were entered as independent variables in the multivariate regression to obtain adjusted effects.  A random effect for hospital was included in the model to account for the fact that HCWs are naturally clustered by facility. This means that there is a possibility that HCWs in the same facility are correlated with each other. This recognizes that it is possible that one HCW having TB increases the chance of others getting it, or they are correlated because they are exposed to the same group of patients. To account for this potential clustering effect, we added random effect to the regression model producing a generalized mixed-effects regression. The extra random effect allows us to distinguish two levels of variation, within facility and between facilities, and thus assess the significance of effects of interest at the appropriate level of variation. For both multivariate regression models, log-likelihood ratio test was used to assess the significance of difference between observed and fitted event counts or to assess the significance of the association fitted by the model and to select most important predictors. Significance of individual regression estimates was tested by Wald statistics (t-test).    48 Sensitivity analyses The impact of alternative weighting schemes was explored by re-running the computer linkage with eight different combinations of linkage variables and seven different cut-off scores as outlined below. The number of partial matches within each category was calculated. Accepting all scores >90% (without manual review) 1- Accepting all scores >85% (without manual review) 2- Accepting only scores of 100% (without manual review) 3- Rejecting scores <75% 4- Rejecting scores <70% 5- Change weights to:  Surname / 50, Given name / 30, birthdate / 40, and sex / 10   6- Change weights to:  Surname / 40, Given name / 40, birthdate / 40, and sex / 10   7- Change weights to:  Surname / 50, Given name / 40, birthdate / 40, and sex / 0  (eliminated) 8- Change weights to:  Surname / 40, Given name / 30, birthdate / 40, and sex / 20   9- Change weights to:  Surname / 50, Given name / 20, birthdate / 40, and sex / 20   10- Change weights to:  Surname / 50, Given name / 30, birthdate / 30, and sex / 20     49 Validating linkage algorithm and cut-off points Incidence rate ratios were calculated in the same manner described above using the number of observed cases at ≤70%, ≤75%, ≤80%, ≤85%, ≤95% and ≤100% as a comparison to the selected cut-off of 90%. Trends over time (from 2002-2012) were examined for matches scoring 80%, 85%, 90% and 95%.  Finally, a sub-set of 390 possible matches that scored within the 90-100% range and a sub-set of 411 possible matches that scored less than 70% were examined manually to validate the computer algorithm and the selection of 70% and 90% as the lower and upper cut-off points. Sensitivity, specificity, positive and negative predictive values were calculated from these values by using a 2x2 table.  Adoption of implementation research methods and dissemination of results Implementation science is defined by the National Institute of Health as “the study of methods to promote the integration of research findings and evidence into healthcare policy and practice. It seeks to understand the behaviour of healthcare professionals and other stakeholders as a key variable in the sustainable uptake, adoption, and implementation of evidence-based interventions.” (85). In line with this approach, we sought input from frontline HCWs, hospital managers, local TB, IC and OH experts and decision-makers at both the provincial and national levels within the South African Department of Health at multiple stages throughout this study.  Representatives from these groups assisted with development of the research protocol prior to submission for ethics review, revision of the data collection procedures, interpretation of results and development of recommendations from findings. Regular communication with key  50 stakeholders throughout the duration of the project allowed for frequent updates and opportunities for input and feedback while ensuring that knowledge translation was ongoing. These extensive consultations were possible and well supported due to 8 years of prior work in the Free State province. The PhD student’s supervisor had an existing large collaborative research program based in the Free State where networks and relationships with collaborators and stakeholders were already well established.   When creating the final report for this study, we considered principles from the Diffusion of Innovations Theory as presented by Rogers (86). This framework offers a theory of how, why and at what rate practices such as TB IC spread through defined populations. This theory outlines characteristics of interventions, ideas or practices that determine their rate of adoption. According to this theory, uptake depends on knowledge, persuasion, decision, implementation and confirmation (86). Furthermore, the final report of findings included significant input from the South African-Canadian steering committee of the larger research program in which this PhD study was embedded. Findings were presented to key decision makers in South Africa at the provincial and national levels. Eventual publication of these findings in a peer-reviewed journal and presentation at scientific conferences will allow for international dissemination and discussion.    51 Figure 3.1. Overview of record linkage algorithm Surname	(40)	Given	Name	(30)	Age		(30)	Gender	(30)	Total	Score	%	(130)	• Excluded	without	review	<80%	• Manually	reviewed	80-90%	• Included	without	review	>90%	           52 Table 3.1. Manual matching decision rules Included Not Included If difference between birthdates greater than 1 with all other variables an exact match If only first initial (full first/given name not available) (even if other variables match) If month and day of birthdate were reversed If difference between birthdates/year was greater than 2 If one digit of day or month of birth were reversed If difference between birthdates/year was greater than 1 with at least 1 other variable that was not an exact match If there was an obvious typo or spelling mistake If surname was very common (e.g. Mofokeng) and other variables were not perfect matches If the surnames differed but one could be assumed to be a nickname of the other If date of birth was different by one or two days If the given names differed but one could be assumed to be a nickname of the other If the Surname, Given Name/s and year of birth match perfectly but month and day of birth are different If all other variables matched perfectly but sex differed  If the Surname, Given Name/s and year of birth match but month/day in birthdate are missing  If surname sounds the same but is spelled differently and at least one of the given names and year of birth match    53 Figure 3.2. Screenshot of customized web-based tool used for manual matching   54 Table 3.2. Review of existing linkage programs and their limitations Linkage Tool Limitations FRIL http://fril.sourceforge.net/ Does not have double meataphone. Registry Plus™ Link Plus http://www.cdc.gov/cancer/npcr/tools/registryplus/lp.htm Does not support SQL Server. May not be efficient or capable of processing large datasets.  LinkageWiz http://www.linkagewiz.net/ Expensive. Single license capable of processing our datasets is $3000 USD. Data-Matching.com http://www.data-matching.com/ Data security issue. Requires uploading data to 3rd party website. The Link King http://www.the-link-king.com/ Requires first and last name, as well as Social Security Number or date of birth.     55 Chapter 4- Historical Prospective Cohort Study Results This chapter presents key findings from the ETR.Net-PERSAL record linkage and the historical prospective cohort study to estimate incidence of TB disease among HCWs in Free State, South Africa. We also describe results from various sensitivity analyses to evaluate record linkage cut-off scores.   Linkage results As shown in Figure 4.1, there were 23,924 partial ETR.Net-PERSAL matches. Prior to further review, 8,615 records where date of TB treatment initiation preceded date of employment were removed.  All matches with scores <70% were not included. There were 8,224 partial matches that scored between 70-80%. After manually examining a sub-set of these partial matches, these were determined to be of poor quality and were therefore also all excluded. Matches with scores between 81-90% (n=5,133) were reviewed manually. Matches with scores >90% (n=1,952) were automatically included without manual review. After manual matching by two independent reviewers, there were 857 accepted matches in the dataset. Finally, 132 records were removed where single ETR.Net record matched to multiple PERSAL records.  The final dataset consisted of 2,677 cases of TB among HCWs from 2002-2012.   Participants and primary outcome data Overall, there were 231,834 people diagnosed with TB in Free State from 2002-2012. There were 32,039 HCWs employed by 258 facilities during this timeframe. Average FTE ranged from 13,473 in 2002 to 19,491 in 2012. During these 10 years, 2,677 cases of  56 TB were diagnosed among HCWs, compared with 1,280 expected cases. The mean TB incidence rate among HCWs during the study period was 1,494.48 (SD=569.50).  The number of observed cases of TB among HCWs was greater than the expected number of cases for every year during the study period (Table 4.1). The number of observed cases among HCWs ranged from 80 in 2002 to 371 in 2007 (Figure 4.2).  The number of TB cases diagnosed among HCWs between 2002-2012 followed a similar trend over time when compared to the general population (Figure 4.3) however, the number of TB cases among HCWs was much higher than the general population between 2005-2008 and then decreased significantly in 2009. Incidence rate ratios ranged from 1.14 (95% CI: 0.98 to 1.32) in 2012 to 3.12 in 2005 (95% CI: 2.81 to 3.45). Since 1.14 is 14% greater than 1.0, the incidence rate ratio indicates an excess of 14% or at least a 14% increase in 2012.  Similarly, the incidence rate ratio in 2005 indicates a 312% excess of TB among HCWs.    Description of the HCW cohort As shown in Table 4.2, most (n= 1,989, 74.3%) HCWs that were diagnosed with TB from 2002-2012 were aged between 30-49 years old at the time of diagnosis. There were more females (n=1,574, 58.8%) and the overwhelming majority were African/Black (n=2,546, 95.1%). About half (n=1,551, 57.9%) were employed in a hospital, while 882 (32.9%) worked in a clinic and 244 (9.2%) were employed in “other” settings such as the provincial department of health or central laundry facilities.  Most HCWs with TB were nurses (n=1,113, 41.6%), 767 (28.7%) were support staff such as maintenance workers,  57 laundry workers, food service workers, security personnel, cleaners and porters, 282 (10.5%) were physicians and surgeons, and 407 (15.2%) were administrative staff. There were 108 (4.2%) allied health professionals (physical therapists, audiologists, technologist/technicians, pharmacists, social workers and dieticians). Most HCWs with TB did not know their HIV status or did not disclose it (n=2,035, 75.6%) but 498 were known to be HIV positive (18.6%).  Pulmonary TB was diagnosed in 2,039 (76.2%) cases. There were 560 cases (20.9%) of extra-pulmonary TB and 78 (2.9%) were classified as having both pulmonary and extra-pulmonary disease.  Most (n=2,149, 80.3%) were newly diagnosed cases with the rest being relapses or re-treatment cases. There were only 18 documented cases in ten years that were classified as being multiple-drug resistant TB (MDR-TB) among HCWs. The majority (n=1,742, 65.1%) of HCWs completed their course of treatment and were classified as “cured” while 306 (11.3%) died. One hundred and thirty-six (5.1%) defaulted or failed treatment, 405 (15.1%) transferred or moved out of province and the outcome was unknown for 90 HCWs (3.4%).   Table 4.3 shows the results from a Chi-Square test of independence. These findings suggest that TB treatment outcomes among HCWs are related to duration of employment and the type of facility that they work in (hospital, clinic or other non-clinical setting). Occupation was not related to TB treatment outcome.     58 Description of the general population cohort Similar to the HCW cohort, the majority of TB patients from the general population (not employed as a HCW during the study period), were in the 30-39 age group at the time of diagnosis (n=76,463, 33.3%) followed closely by the 40-49 year old age group (n=58,223, 25.4%).  In the general population, 130,285 (56.0%) TB patients were female. HIV status was unknown for 98,959 patients (43.2%), 71,778 (31.3%) were HIV positive and 22,420 (9.8%) knew that they were HIV negative. Among the general population, 179,579 (78.4%) TB cases were classified as pulmonary, 44,952 (19.6%) were classified as extra-pulmonary and 4,626 (2.0%) were both pulmonary and extra-pulmonary. Most (n=186,679, 81.5%) were new TB cases, however, 42,478 (18.5%) were relapses or re-treatment cases. There were 1,430 cases of MDR-TB among the general population (0.6%) during the ten-year study period. The majority of the general population with TB (n=134,857, 58.8%) completed their course of treatment and were classified as “cured” while 29,156 people died (12.7%). In the general population, 14,783 (6.5%) defaulted or failed treatment, 38,585 (16.8%) transferred or moved out of province and the outcome was unknown for 11,776 (5.1%) TB patients who were not HCWs.  Multivariate analyses As shown in Table 4.4, the risk of TB disease was greater among older HCWs, than those who were born in the 1980’s (For those born in 1970-1979 RR=3.84, 95% CI: 2.89 to 5.09; 1960-1969 RR=7.29, 95% CI: 5.48 to 9.72; 1950-1959 RR=7.11, 95% CI: 5.22 to 9.69; 1940-1949 RR=4.03, 95% CI: 2.78 to 5.83).  Results for HCWs born in the 1930’s were not significant. Black or African HCWs had a greater than 5-fold increased risk of TB when compared to their white colleagues (RR=5.30, 95% CI: 3.90 to 7.20). Similarly,  59 coloured HCWs had an almost 3-fold increased risk of TB (RR= 2.90, 95% CI: 1.87 to 4.50). The risk of TB was greater among male compared to female HCWs (RR= 1.51, 95% CI: 1.38 to 1.64). In the unadjusted analysis, TB risk was greater among support staff (RR= 2.36, 95% CI: 1.79 to 3.11), nursing staff (RR=2.07, 95% CI: 1.58 to 2.71) and administrative staff (RR=1.74, 95% CI: 1.29 to 2.32) when compared to allied health professionals.  These estimates attenuated in the adjusted model and were no longer statistically significant. The risk of TB was lower among doctors and surgeons in the adjusted model (RR= 0.85, 95% CI: 0.62 to 1.15). HCWs who had worked in the healthcare sector for less than 20 years had a greater risk of TB compared to those who had been employed for more than 20 years. In particular, HCWs who were employed for 11-15 years had a more than 3-fold increased risk of TB (RR=3.60, 95% CI: 2.97 to 4.37). Facility type was not associated with increased risk of TB.    Figure 4.4 shows the mean probability of TB for each group over the 10-year study period. Overall, the probability of getting TB is greater in HCW when compared to the general population. The confidence intervals vary considerably around the exact point estimate which corresponds to the range in IRR that we saw previously. As illustrated in Figure 4.5, the final model comparing the probability of TB in HCWs and the general population shows that the difference in TB rates between HCWs and the general population depends on age and sex. For young females (aged 20-29 years), the TB rate in HCWs is lower than the general population.  The peak TB rate occurs between 30-39 years for females in both the general population and the HCW population. The overall TB rate for females declines as they get older but not as swiftly for HCWs as the general  60 population. For females over 50 years old, the TB rate in HCWs becomes significantly larger than the general population. The TB rates in males are significantly larger than females, except for the 20-29 year old age group. The TB rate peaks between 40-49 years for males. The TB rate for males is similar for HCWs and the general population except between 40-59 where their TB rates are lower than the general population.   Sensitivity analyses The research team considered putting more weight on surname and less on sex (since as a binary variable it was “all or nothing”).  However, after examining the output from various weighting schemes as shown in Table 4.5, it was determined that changing the surname to 50 from 40 and sex to 10 from 30, did not drastically alter the number of matches identified (N=2,143 vs. N=1,915 before manual matching at 90% cut-off). Table 4.6 shows the number of matches by score and year prior to manual matching and Table 4.7 provides details regarding how incidence rate ratios were calculated for TB, among HCWs, compared to the general public by year and for varying cut-off scores. Figure 4.6 then provides a visual illustration of these incidence rate ratios over time for cut-off scores of 80%, 85%, 90% and 95%.  With an 80% cut-off point, incidence rate ratios ranged from 12.62 in 2005 to 2.00 in 2012. With an 85% cut-off point, the range was from 7.23 in 2006 to 1.52 in 2011. With a 95% cut-off point, incidence rate ratios ranged from 1.99 in 2006 to 0.64 in 2012. Values for 90% do not exactly match results presented above because matches where date of TB treatment initiation preceded employment were removed as one of the first steps in an effort to minimize the number of records to me manually reviewed, but were not removed in the sensitivity analysis.   61  Figure 4.7 shows results from manual matching of partial matches that scored under 70% and over 90%. This step was undertaken to validate cut-off points and to estimate the quality of the links that were excluded without full review (<70%) and those that were included without full review (90%). Of the 411 records that scored <70%, we found only 2 that were deemed to be false negatives that should have been included in the linked dataset and 409 that were appropriately discarded. Similarly, of the 390 records that scored >90%, we determined that 383 were true positive matches and that only 7 were false positives that should have been excluded. Using these values, we estimated that the sensitivity of the record linkage was estimated to be 99.5% and specificity was 98.3%. The positive predictive value was 98.2% and the negative predictive value was 99.5%.  62 Fig 4.1. Free State Province ETR.Net- PERSAL record linkage process n=	238,721	Pa ents	registered	in	ETR.net	(2002-2012)	n=	16,264	Average	FTE	in	PERSAL	(2002-2012)	n=	23,924	Dis nct	ETR-PERSAL	matches	(>70%)	n=	8,224	Matches	with	scores	70-80%	(All	excluded)		n=	1,952	Matches	with	scores	>90%	(All	included)		n=	5,133	Matches	with	scores	80-90%	(Manually	reviewed)	n=	857	Matches	a er	manual	review	(All	included)	n=	2,677	Accepted	matches	in	final	dataset	n=	8,615	Matches	removed	where	date	of	TB	treatment	ini a on	preceded	date	of	employment	(All	excluded)		n=	132	Matches	where	single	ETR	record	matched	to	mul ple	PERSAL	records	(All	excluded)		n=	476	Matches	with	scores	=100%	(All	included)	   63 Fig 4.2. TB cases among HCWs by year (N=2,677)     64 Figure 4.3.  TB incidence rate among the general population and HCWs in Free State (2002-2012)    65 Table 4.1. Incidence rate ratios of HCWs with TB by year (2002-2012)    Year  HCW Person Years (FTE)  Observed Cases  Expected Cases Incidence in General Population  (per 100,000)* Incidence in Healthcare Workers  (per 100,000) Incidence Rate Ratio  (95% CI)  P-Value** 2002 13,473 80 34 248.5    593.8 2.35 (1.88-2.91) <0.001 2003 13,964 198 77 552.2    1,417.9 2.57 (2.23-2.95) <0.001 2004 15,101 255 98 648.2    1,688.6 2.60 (2.30-2.94) <0.001 2005 16,448 365 117 708.1    2,219.1 3.12 (2.81-3.45) <0.001 2006 16,391 362 129 787.6   2,208.5 2.81 (2.53-3.11) <0.001 2007 16,641 371 136 815.1    2,229.4 2.73 (2.46-3.02) <0.001 2008 16,722 326 149 891.6    1,949.5 2.19 (1.96-2.44) <0.001 2009 16,404 157 138 842.1 957.1 1.39 (0.97-1.33) 0.106 2010 16,298 203 137 840.0 1,245.6 1.48 (1.29-1.70) <0.001 2011 17,973 191 148 822.7 1,062.7 1.29 (1.12-1.49) <0.001 2012 19,491 169 148 757.0 867.1 1.14 (0.98-1.32) 0.084 Mean TB incidence (SD)   1,494.48 (569.50)  *As reported by the South African National Department of Health (from ETR.Net) and Statistics South Africa **Chi-Square Test 66 Table 4.2. Demographic and clinical characteristics of HCWs and the general population with TB in Free State, South Africa     Variable Healthcare  Workers (N=2,677) General Population (N=229,157) Frequency (%) Frequency (%) Age (at diagnosis)   <19 20-29 0 (0.0) 205 (7.7) 4714 (2.1) 51409 (22.4) 30-39 40-49 992 (37.1) 997 (37.2) 76463 (33.3) 58223 (25.4) 50-59 60+ Sex 429 (16.0) 54 (2.0) 27598 (12.0) 10750 (4.7) Male Female Race African White Coloured Indian HIV Status HIV+ HIV- Unknown Facility Type Hospital Clinic Other 1103 (41.2) 1574 (58.8)  2546 (95.1) 70 (2.6) 59 (2.2) 2 (0.08)   498 (18.6) 154 (5.8) 2025 (75.6)   1551 (57.9) 88 (32.9) 244 (9.2) 98872 (43.1) 130285 (56.0)  - - - -   71778 (31.3) 22420 (9.8) 98959 (43.2)   - - -  Occupation   Doctor/Surgeon 282 (10.5) - Nurse 1113 (41.6) - Allied Health Professional Administrative/Clerical Support Services   108 (4.2) 407 (15.2) 767 (28.7) - - -    67   Table 4.2 (continued)     Variable Healthcare  Workers (N=2,677) General Population (N=229,157)  Frequency (%) Frequency (%) Duration of Employment  1-5 years 6-10 years 11-15 years 16-20 years 20+ years Disease Classification  529 (20.0) 359 (13.4) 536 (20.0) 498 (18.6) 461 (17.2)   - - - - - Pulmonary 2039 (76.2) 179579 (78.4) Extra-pulmonary 560 (20.9) 44952 (19.6) Both 78 (2.9) 4626 (2.0) Diagnosis Type New  2149 (80.3)  186679 (81.5) Relapse/Re-treatment 528 (19.7) 42478 (18.5) MDR-TB (pre-treatment) Yes No  18 (0.7) 2659 (99.3)  1430 (0.6) 227727 (99.4) Outcome    Cured/Completed Defaulted/Failed Transferred/Moved Died Unknown 1742 (65.1) 136 (5.1) 405 (15.1) 306 (11.3) 90 (3.4) 134857 (58.8) 14783 (6.5) 38585 (16.8) 29156 (12.7) 11776 (5.1) *Physicians and Surgeon= Surgeon, Radiologist, Anaesthesiologist, Other Physicians, Medical Registrar; Nurse= Professional nurse, Assistant or auxiliary Nurse, Staff Nurse; Allied Health Professional= Therapist (i.e. audiologist), Technologist/technician, Pharmacist, Social Worker; Administrative/Clerical= Manager/administrator, Clerk, General Assistant; Support Staff= Maintenance Worker, Laundry Worker, Food Service Worker, Security, Cleaner, Porter    68 Figure 4.4. Probability of TB among HCWs and the general population     Legend   Mean point estimate   Upper confidence limit (95%)   Lower confidence limit (95%)  69 Figure 4.5. Probability of TB among HCWs and the general population by age and sex  Legend   Mean point estimate   Upper confidence limit (95%)   Lower confidence limit (95%)  70 Table 4.3. TB treatment outcome by facility type, occupation and duration of employment  Cured/Completed Frequency (%) Defaulted/Failed Frequency (%) Transferred/Moved Frequency (%) Died Frequency (%) Unknown Frequency (%) Chi-Square* (P-value) Facility Type Hospital  986 (56.6)  83 (61.9)  265 (65.4)  172 (56.2)  43 (47.8) 49.0 (<0.001) Clinic 591 (33.9) 38 (28.4) 85 (21.0) 94 (30.7) 23 (25.6)  Other 165 (9.5) 13 (9.7) 85 (13.6) 40 (13.1) 24 (26.7)  Occupation Doctor/Surgeon  190 (10.9)  17 (12.7)  34 (8.4)  33 (10.8)  8 (8.9) 14.8 (0.54) Nurse 727 (41.7) 50 (37.3) 168 (41.5) 134 (43.8) 34 (37.8)  Allied Health Professional 74 (4.3) 6 (4.5) 16 (4.0) 8 (2.6) 4 (4.4)  Administrative/Clerical 262 (15.0) 27 (20.2) 53 (13.1) 46 (15.0) 19 (21.1)  Support Staff 489 (28.1) 34 (25.4) 134 (33.1) 85 (27.8) 25 (27.8)  Duration of Employment (yrs)      70.6 (<0.001) <1 176 (10.1) 14 (10.5) 48 (11.9) 39 (12.8) 17 (18.9)  1-5 335 (19.2) 28 (20.9) 95 (23.5) 50 (16.3) 21 (23.3)  6-10 237 (13.6) 14 (10.5) 49 (12.1) 36 (11.8) 23 (25.6)  11-15 374 (14.0) 32 (23.9) 53 (17.0) 73 (23.9) 4 (4.4)  16-20 327 (18.8) 32 (23.9) 76 (18.8) 56 (18.3) 7 (7.8)  20+ 293 (16.8) 14 (10.5) 84 (20.7) 52 (17.0) 18 (20.0)  *Likelihood Ratio   71 Table 4.4.  Relative risk estimates from mixed-effects poisson regression model of HCWs with TB and those without TB (N= 32,039)    Unadjusted  P-Value  Adjusted  P-Value  RR (95% CI)  RR (95% CI)   Birth Year     1980-1989 1970-1979 1960-1969 1950-1959 1940-1949 1930-1939 1.00 4.28 (3.24 to 5.65) 7.95 (6.06 to 10.42) 6.10 (4.59 to 8.10) 3.43 (2.43 to 4.85) 1.79 (0.16 to 20.84)  <0.0001 <0.0001 <0.0001 <0.0001 0.99 1.00 3.84 (2.89 to 5.09) 7.29 (5.48 to 9.72) 7.11 (5.22 to 9.69) 4.03 (2.78 to 5.83) 1.39 (0.13 to 15.13)  <0.0001 <0.0001 <0.0001 <0.0001 0.99  Race     White African Coloured Asian 1.00 6.74 (4.98 to 9.11) 3.59 (2.32 to 5.56) 1.15 (0.18 to 7.02)  <0.0001 <0.0001 0.99 1.00 5.30 (3.90 to 7.20) 2.90 (1.87 to 4.50) 1.05 (0.18 to 6.07)  <0.0001 <0.0001 0.99  Sex Female Male  Facility Type  1.00 1.41 (1.31 to 1.53)   <0.0001  1.00 1.51 (1.38 to 1.64)    <0.0001  Non-Clinical  Hospital Clinic 1.00 2.04 (0.79 to 5.26) 2.47 (0.67 to 9.12)  0.19 0.24 1.00 1.27 (0.59 to 2.74) 1.48 (0.54 to 2.52)   0.76 0.67  Occupation     Allied Health Professional Doctor/Surgeon Nurse Administrative/Clerical Support Staff 1.00 1.13 (0.83 to 1.53) 2.07 (1.58 to 2.71) 1.74 (1.29 to 2.32) 2.36 (1.79 to 3.11)  0.84 <0.0001 <0.0001 <0.0001 1.00 0.85 (0.62 to 1.15) 1.24 (0.93 to 1.63) 1.13 (0.84 to 1.51) 1.28 (0.96 to 1.70)  0.60 0.25 0.82 0.14  Duration of Employment (yrs) 20+ 16-20 11-15 6-10 1-5 <1   1.00 2.05 (1.71 to 2.47) 3.36 (2.80 to 4.04) 1.21 (0.99 to 1.48) 1.06 (0.89 to 1.28) 2.74 (2.24 to 3.53)    <0.0001 <0.0001 0.09 0.97 0.03   1.00 1.97 (1.63 to 2.39) 3.60 (2.97 to 4.37) 1.72 (1.38 to 2.14) 1.92 (1.56 to 2.37) 1.60 (1.44 to 1.82)    <0.0001 <0.0001 <0.0001 <0.0001 <0.0001       Beta-coefficient estimate from univariate and multivariate Poisson regression analysis (p<0.01) 72 Table 4.5.  Sensitivity analysis by weighting scheme and percentage cut-off (before manual matching)    Variable and Weight   Number of matches by score    Weighting Scheme Surname Given Name Birthdate Sex >= 100% >= 95% >= 90% >= 85% >= 80% >= 75% >= 70% A 40 30 30 30 460 1737 2143 5623 9432 12326 12458 B 35 35 30 0 476 1796 1956 3199 7624 9637 14847 C 50 30 40 10 460 1735 1915 5252 6344 11127 12515 D 40 40 40 10 460 1720 1901 3563 7069 9697 12547 E 50 40 40 0 476 1796 1956 3539 7233 9910 13582 F 40 30 40 20 460 1737 1911 5383 7329 10426 12138 G 50 20 40 20 460 1783 2786 5425 6802 10373 12158 H 50 30 30 20 460 1735 2092 5384 8735 11795 13587   73 Table 4.6. Number of matches by score and year prior to manual matching  Year ≥ 100% ≥ 95% ≥ 90% ≥ 85% ≥ 80% ≥ 75% ≥ 70% 2002 0 56 70 224 374 428 432 2003 0 125 149 516 839 980 990 2004 0 155 192 639 1100 1334 1357 2005 0 226 276 839 1477 1751 1782 2006 0 257 295 933 1541 1776 1796 2007 1 250 301 967 1632 1862 1882 2008 59 225 272 683 1113 1461 1478 2009 80 101 135 214 363 714 717 2010 120 128 172 225 361 710 711 2011 113 120 153 208 335 660 661 2012 87 94 128 175 296 648 650 Total 460 1737 2143 5623 9431 12324 12456     74 Table 4.7. Sensitivity analysis by percentage cut-off    Year   Cut-off score (%) HCW Person Years (FTE)  Observed Cases  Expected Cases Incidence in General Population  (per 100,000)* Incidence rate ratio   2002 80 85 90 95   13,473 374 224 80 56   34   248.5    11.00 6.59 2.35 1.65  2003 80 85 90 95   13,964 839 516 198 125   77   552.2    10.90 6.70 2.57 1.62  2004 80 85 90 95   15,101 1100 639 255 155   98   648.2    11.22 6.52 2.60 1.58  2005 80 85 90 95   16,448 1477 839 365 226   117   708.1    12.62 7.17 3.12 1.93  2006 80 85 90 95   16,391 1541 933 362 257   129   787.6   11.95 7.23 2.81 1.99  2007 80 85 90 95   16,641 1632 967 371 250   136   815.1    12.00 7.11 2.73 1.84   75 Table 4.7 (continued)     Year   Cut-off score (%) HCW Person Years (FTE)  Observed Cases  Expected Cases Incidence in General Population  (per 100,000)* Incidence rate ratio   2008 80 85 90 95   16,722 1113 683 326 225   149   891.6    7.47 4.58 2.19 1.51  2009 80 85 90 95   16,404 363 214 157 101   138   842.1 2.65 1.55 1.39 0.73  2010 80 85 90 95   16,298 361 225 203 128   137   840.0 2.64 1.64 1.48 0.93  2011 80 85 90 95   17,973 335 208 191 120   148   822.7 2.26 1.52 1.29 0.81  2012 80 85 90 95   19,491 296 175 169 94   148   757.0 2.00 1.18 1.14 0.64     76 Figure 4.6. Incidence rate ratios by percentage cut-off score and year    Incidence rate ratio  77 Fig 4.7. Estimation of positive predictive value of matches with subset of scores >90% (n=390) and <70% (n=411)    Matches Non-Matches  Linked  383 True +   7 False +    390     (>90%)             included Unlinked  2 False -  409 True –   411     (<70%)             excluded  78 Chapter 5- TB IC Workplace Assessment Methodology  A facility assessment tool or audit tool is an appropriate tool to document workplace hazards and risk factors for TB exposure in a systematic and unbiased manner. Methods using general IC audit tools have been successfully employed in various settings to objectively record IC hazards and to make recommendations for improvement (87, 88). For example, Bryce and colleagues found that implementation of a standardized audit tool allows for benchmarking of practices across the institution and enhances standards of care (87).  Comprehensive guidelines specific to TB IC are presented in the “WHO policy on TB infection control in health-care facilities, congregate settings and households” (21) and provide a solid basis for developing a TB-specific IC audit tool such as the one used for this study. As mentioned previously, prolonged exposure at work makes HCWs, especially those in high-incidence regions, at higher risk than the general population of developing LTBI and TB (15, 16, 54). The rate of TB among hospital staff can therefore be used as a proxy for measuring rates of nosocomial spread and overall effectiveness and implementation of IC measures.    Study sample   The TB IC facility assessment conducted for this study included 28 hospitals in the Free State Province, South Africa (21 district, 5 regional, 1 tertiary, and 1 specialized hospital). In total, there are 32 public hospitals in Free State, however, four were excluded since they were not part of the larger Randomized Control Trial (RCT) within which this PhD project was nested. These four facilities were not included in the RCT  79 because at the time of the study, they did not have a functioning OH clinic, therefore making it impossible to implement the intervention (strengthening OH services to improve access to HIV and TB care among HCWs).   Data collection procedures On-site visits and data collection occurred between June 17, 2013 and July 15, 2013.  The PhD candidate travelled over 4500 km across the province by car with a local research colleague from the University of the Free State. Visits were pre-arranged months in advance and were confirmed by phone prior to arrival. Appointments were organized in a manner that was geographically and logistically convenient when possible (see Figure 5.1).  As part of the TB IC workplace assessment tool, all visits included a face-to-face interview with the facility IC officer or the nurse who is in charge of TB at each hospital. At many hospitals, the IC nurse, OH nurse, TB coordinator and quality assurance coordinator all chose to participate in the data collection process. These focus groups proved to be especially fruitful and highlighted the multi-disciplinary nature of TB IC.  In addition to focus groups with these key informants and interactions with front-line staff and managers, the hospital visits also included an observational walk-through the outpatient department and a review of the TB register. Some facilities did not have an outpatient department, and in these cases, observations were made in the emergency/casualty area or wherever TB suspects enter the hospital. The only facility that did not have an outpatient department or emergency/casualty was the specialized hospital. Patients are admitted to his psychiatric facility by referral only, therefore walk-through observations were made on one ward and in common areas where patients eat  80 and participate in recreational activities. In general, hospital management/executives who authorized the visits and healthcare professionals who participated in the TB IC assessments were welcoming and forthcoming. Many IC and OH nurses and hospital TB coordinators provided valuable insight and recommendations beyond what was captured in the data collection tool itself. This information was captured in detailed field notes and was transcribed as quotations.   Data collection tool The ICAP TB Infection Control Practices: Facility Assessment is a comprehensive tool that was developed by a team from the International Centre for AIDS Care and Treatment Programs (ICAP) at Columbia University in the United States and has been used extensively in similar settings. Implementation of this tool was evaluated in 663 HIV care sites in nine sub-Saharan African countries including South Africa (37). Written permission was obtained to use the tool for this study. Small yet pertinent adaptations and revisions to the tool were made with input from colleagues responsible for IC, OH and TB at the provincial level at the Free State Department of Health and from IC, OH and TB experts in Vancouver. Refinement of the tool was also informed by an 8-year collaborative research partnership between the Global Health Research Program in the School of Population and Public Health at the University of British Columbia and the Free State Department of Health, the National Institute for OH (South Africa) and the Centre for Health Systems Research & Development at the University of the Free State (89, 90).    81 This tool allows for systematic documentation of the presence or absence of TB IC measures. More specifically, it assesses environmental controls (such as natural and mechanical ventilation), administrative controls (such as patient triage and management), personal protective equipment (such as availability and appropriate use of N95 respirators), estimated TB case-load and average time to diagnosis and treatment.  The final version of the tool included 83 items. As part of the workplace assessment, key informants (IC nurses, TB coordinators, OH nurses and quality assurance coordinators) were asked open-ended questions regarding their perceived barriers and facilitators to TB infection control in their hospital. Responses to these questions were audio-recorded and transcribed verbatim. Names or other personal identifiers were excluded to preserve confidentiality. The tool took between 2-5 hours to complete (see Appendix A for the modified ICAP TB Infection Control Practices: Facility Assessment Tool) and completion time varied by size of hospital, openness and willingness of healthcare professionals to talk and quality and completeness of record. The same tool was used in all 28 hospitals.   Data analysis   Data entry was performed by the PhD candidate and findings were recorded in an Excel spreadsheet. All TB IC assessment forms were checked three times for accuracy and consistency. Based on the findings from the ICAP TB IC practice tool, an overall facility level TB IC score was calculated for each hospital out of a total possible score of 75.     82 Variables Definition of outcome The incidence rate of TB among HCWs was calculated for each facility for 2012 from the record linkage that is described in detail in Chapters 3 and 4. TB cases were identified by matching records that appeared in both the Free State health human resource database (PERSAL) and the provincial TB registry (ETR.Net). Rates were obtained by dividing the number of TB cases in each hospital by the total number of HCWs employed in each hospital in 2012.    Other variables Findings from the ICAP TB IC practice tool were tallied and an overall facility level TB IC score was calculated for each hospital out of a total possible score of 75. Eight items of the 83 total were not included in the score. These included items such as “Respondents name”, “Date of assessment”, etc. Total scores were also broken down by four categories (administrative control score - out of a possible score of 32, environmental control score- out of a possible score of 23, personal protective equipment control score, out of a possible score of 11 and miscellaneous control score, out of a possible score of 9). The miscellaneous control score included turn-around time for AFB smear and culture results, time to transfer for MDR patients, availability of TB screening for staff, availability of IPT for staff, and re-assignment of HIV+ staff to low-risk area. Hospitals were ranked according to total TB IC Scores.    83 The number of registered TB patients, the number of smear positive TB patients and the number of registered MDR-TB patients registered at each hospital in 2012 were obtained directly from the ETR.Net.  These covariates all contained outliers and were therefore categorized into groups. The number of TB patients variable was divided into 3 groups (“Low”= 2-19 patients; “Medium”= 51-121 patients; “High”=142-759 patients).  Similarly, the number of smear positive TB patients variable was divided into 3 groups (“Low”= 0-7 patients; “Medium”= 16-31 patients; “High”=35-225 patients).  The number of MDR-TB patients variable was divided into 2 groups (“None”=0 patients; “Some”= ≥1 patient). Geographic location of each hospital was classified as either urban or rural.   Statistical methods Basic descriptive statistics such as medians and interquartile ranges were utilized to describe the study hospitals. Logistic regression was used to model the association between IC scores (total score and all sub-categories- administrative controls, environmental controls, personal protective equipment and miscellaneous) and the likelihood of a HCW having TB in that hospital, with the odds ratio being a measure of this association. As in the analysis for the record linkage, to account for a potential clustering effect, a random effect on facility was added to the logistic regression resulting in a generalized linear mixed-effects regression.   The number of TB patients, the number of smear positive TB patients, the number of MDR-TB patients and geographic location (urban or rural) were considered to obtain adjusted effects. Log-likelihood ratio test was used to assess the significance of the  84 association fitted by the model and to select the most important predictors. Significance of individual regression estimates was tested by Wald statistics (t-test).  Thematic analysis of qualitative findings Thematic analysis was utilized to analyze the qualitative findings that emerged from the open-ended questions included in the workplace assessment tool. This is an appropriate method to extract unique information from unstructured interviews such as those described in this study (91, 92). Thematic analysis approach offers fewer preconceptions than other qualitative methods and is therefore subject to less bias (93). Two reviewers independently coded data collected from the transcripts. Emergent themes and patterns in the data were identified by initial inductive analysis. Through reviewer consensus, themes were condensed into overarching control group categories for barriers and facilitators separately. The frequency of each code by category was recorded. Nvivo 9 Software was used to help facilitate management and analysis of the data.  Dissemination of results Similar to the implementation research methods employed for objective 1, we also sought input from frontline HCWs, hospital managers, local TB, IC and OH experts and decision-makers at both the provincial and national levels within the South African Department of Health for this work.  Representatives from these groups assisted with development of the research protocol prior to submission for ethics review, revision of the workplace assessment tool, interpretation of results and development of recommendations from findings to improve TB infection control. Each hospital received  85 an individualized “feedback package” (see Table 6.3 for an example) in August 2014 where they were able to compare their total IC score and scores for each category (administrative, environmental, personal protective equipment and miscellaneous) to the other 27 hospitals. Hospital names were removed to protect confidentiality. It was stressed in writing that findings should be used to make recommendations for improvement, but are not meant to assign blame or fault. A detailed breakdown of the scores was provided for each variable that was included in the tool. These results were presented using a “dashboard” display to simplify presentation and interpretation of results.  The following colour coding was used for the dashboards: Green= Done, observed, desired outcome; Yellow= Not Applicable; and Red= Not done, Not done right or Not observed.  Specific recommendations and suggestions to improve TB IC were formulated for each hospital based on the findings.    86 Fig 5.1. Map of Free State hospitals included in the study by region  1	2	3	4	5	6	7	8	9	10	11	12	21	13	14	15	16	17	18	19	27	20	22	23	24	25	26	28	  List of Hospitals  1 Boitumelo Hospital                            15 Mofumahadi Manapomopeli Hospital 2 Bongani Hospital  16 Mohau Hospital 3 Botshabelo Hospital 17 National District Hospital 4 Diamond/Diamant Hospital  18 Parys Hospital 5 Dihlabeng Hospital  19 Pelonomi Hospital 6 Dr JS Moroka Hospital 20 Phekolong District Hospital 7 Elizabeth Ross Hospital 21 Phumelela Hospital  8 Free State Psychiatric Complex 22 Phutholoha Hospital  9 Itemoheng Hospital  23 Stoffel Coetzee Hospital  10 John Daniel Newberry Hospital  24 Thebe Hospital  11 Katleho Hospital  25 Thusanong Hospital 12 Mafube Hospital  26 Tokollo Hospital  13 Mantsopa Hospital  27 Universitas Hospital 14 Metsimaholo Hospital  28 Winburg Hospital      87 Chapter 6- TB IC Workplace Assessment Results This chapter presents a summary of findings from the TB IC workplace assessments that were conducted across the Free State province. All twenty-eight hospitals in the sample agreed to participate in the study. The name, designation/position and contact information for each respondent was recorded.  Description of the 28 hospitals Sixteen hospitals (57%) were located in rural areas and 12 (43%) were located in urban centres. As shown in Table 6.1, hospitals ranged in size from 22 beds to 870 beds (Median= 80, IQR=146.3).  The patient population and burden of TB disease at the 28 hospitals also varied considerably.  The number of registered TB patients in 2012 ranged from 2 to 759 (Median= 100, IQR= 233.0). Similarly, the number of registered smear positive TB patients was from 0 to 225 (Median= 19.5, IQR= 34.8) and the number of MDR-TB patients ranged from 0 to 20 (Median= 0, IQR= 4.0).    In total, there were 86 cases (as recorded in the ETR.Net national TB surveillance system) of TB diagnosed in 2012 among HCWs who worked in the 28 public hospitals included in the study. The number of HCWs employed at each hospital in 2012 ranged from 54 to 2,577 (Median= 198, IQR= 440.0) and the incidence rate of TB among HCWs in 2012 was between 0 to 3,225.8 per 100,000 (Median= 744 per 100,000, IQR= 736.4).       88 TB IC scores Table 6.2 presents the individual hospital scores from the TB IC workplace assessment tool. The lowest was 23/75 while the highest total TB IC score was 52/75 (Median= 40, IQR= 10.8).  Workplace assessments revealed that administrative control scores ranged from 15/32 to 28/32 (Median= 21, IQR= 6.0), environmental control scores ranged from 3/23 to 14/23 (Median= 6, IQR= 2.8), PPE scores ranged from 1/11 to 10/11 (Median= 5, IQR= 4.8) and miscellaneous scores ranged from 2/9 to 9/9 (Median= 6, IQR= 2.8).  Detailed reports showing all 75 items were created for each hospital using a colour-coded scoring system as shown in Table 6.3.  If the item was “done, observed or desired outcome” it was coloured green. Items that were “not done, not done right or not observed” were coloured red. Items that were “not applicable” were coloured yellow. Hospital names were omitted to protect confidentiality.  Univariate and bivariate analyses There was a significant association between the likelihood of a HCW having TB in a hospital and the total TB IC score.  Results from the logistic regression model with random effect for hospital are presented in Table 6.4 below and in Figures 6.1-6.5. Slopes are the log odds ratio for a one-unit change in the predictor.   As shown in Table 6.5, as the total IC score increases, the probability of there being a HCW with TB at that hospital decreases (OR= 0.94, 95% CI: 0.91 to 0.97). Similarly, as the administrative score increases, the probability of there being a HCW with TB at that hospital decreases (OR= 0.94, 95% CI: 0.87 to 1.02). The probability of having a HCW  89 with TB at that hospital also decreases when the environmental score (OR= 0.88, 95% CI: 0.80 to 0.96), PPE score (OR= 0.86, 95% CI: 0.78 to 0.95) and miscellaneous score (OR= 0.86, 95% CI: 0.73 to 0.99) increase. All were statistically significant except administrative score (p=0.12).  The number of TB patients registered at the hospital and the number of smear positive TB patients were also both associated with the outcome. When compared to the 2-19 TB patient group, the odds ratio was 0.39 (95% CI: 0.19 to 0.78) for the group with 51-121 TB patients and the odds ratio was 0.46 (95% CI: 0.24 to 0.89) for the 142-759 TB patient group. Similarly, when compared to the 0-7 smear positive TB patient group, the odds ratio was 0.72 (95% CI: 0.34 to1.50) for the 16-31 smear positive TB patient group and the odds ratio was 0.48 (95% CI: 0.27 to 0.85) for the 35-225 smear positive TB patient group. This means that hospitals with more TB patients had a lower probability of TB among their staff. The same was seen for the number of smear positive patients. The number of TB patients registered and the number of smear positive patients were highly correlated to each other, and therefore only the “number of TB patients” variable was included in the multivariate model. The number of MDR-TB patients and geographic location of the hospital were not statistically significant and were therefore not included in the final model.  When compared with rural geographic location, the effect of urban location of the hospital was not statistically significant (OR=1.36, 95% CI: 0.81 to 2.26).      90 Multivariate analyses The relationships observed in the bivariate analysis persisted in the multivariable model when adjusted for the number of TB patients at each hospital, however, the odds of a HCW having TB increased only slightly for all IC score categories including the total score.   As shown in Table 6.5, as the total IC score increases, the probability of there being a HCW with TB at that hospital decreases (OR= 0.95, 95% CI: 0.91 to 0.99). This can also be interpreted as the percent change in the odds. For example, in the univariate analysis, this means that if the total score of a hospital increases by one unit then the odds of a HCW having TB decreases by 6.3% (95% CI: 3.1 to 9.4). This percentage is derived from the presented odds ratios as follow: 6.3 = 100*(1-0.937).   When adjusted for other covariates in the multivariate analysis, if the total score of a hospital increases by one unit then the odds of a HCW having TB decreases by 4.9% (95% CI: 0.9 to 8.8).  Significant associations were also seen for PPE score where odds decreased by 11.5% (95% CI: 1.8 to 20.1) for each unit increase in score.   Effects for administrative score, environmental score and miscellaneous score were not statistically significant in the multivariate model. If the administrative score of a hospital increases by one unit then the odds of a HCW having TB decreases by 3.3% (95% CI: -3.9 to 10.2) when adjusted for other covariates. Similarly, if the environmental score of a hospital increases by one unit then the odds of a HCW having TB decreases by 8.3% (95% CI: -0.9 to 17.4) when adjusted for other covariates. Finally, if the miscellaneous  91 score of a hospital increases by one unit then the odds of a HCW having TB decreases by 7.3% (95% CI: -3.0 to 20.3) when adjusted for other covariates.  Qualitative findings and quotes The quotations presented below were compiled from responses to the open-ended questions in the workplace assessment tool from IC professionals, OH professionals and TB coordinators who assisted with data collection.  Quotations are grouped by theme.    “We have a GeneXpert on-site. But MDR results usually come after the patient is already gone from here and at the clinic. Often MDR patients come to hospital from clinics where resistance is known, and are admitted but do not tell staff they have MDR-TB, so they are not isolated. Isolation happens very late, if ever.”  “We have only 4 single rooms in the whole hospital to isolate patients. The sisters try to put patients with respiratory tract infections in the same room but true isolation is not possible due to air flow. Coughing patients sit together in a different cubicle while waiting.  We had a patient admitted last year who escaped from MDR facility. As soon as he was identified, he was sent back. But it took some time.”  “There are plexi-glass screens at triage and admissions to protect the workers from coughing patients. The door in the room for MDR patients is broken and no longer closes so we try to put TB patients near the door when they must share a room with other patients. MDR patients are referred from clinics regularly. We get so many. The MDR  92 facility is sometimes full so MDR patients are put on a waiting list and must wait with others on the ward.”  “Stock-outs of N95s occur quite often. IC and OH do not work together. Workers do not offer masks to coughing patients because patients feel stigmatized. Our OPD has been moved to a small clinic due to ongoing construction- this is not good. This space is too small and does not allow us to separate coughing patients.”  “There are no rooms at all suitable to isolate patients. Stock outs of N95s happen. It’s true. Sometimes we do not have them.”  “This is an old building and our OPD waiting area is very small and crowded with no ventilation (just a narrow corridor with chairs). Casualty has zero ventilation (4 beds in very small congested room with no windows). But we are lucky to have our IC nurse. She is so motivated and very proactive, she does her best. We now have two ladies from the community who work full-time to do screening and cough patrol (sit at desk in entry to hospital to ensure all patients are screened for TB symptoms). We do our best, but last year we had a child with MDR and experienced delays in transfer. The child died in hospital after being there for 7 days. We did contact tracing for family and other children. It took so much time and is not part of our duties, but it must be done.”  “We are so lucky to have a new OPD that is well-organized for IC (large waiting area, plexi-glass screens at triage and admissions). Suspected TB patients wait in separate  93 area. All MDR follow-ups from the district are done in this hospital. MDR patients are often referred to hospital from clinics.”  “There is a very good culture here with support from management. There is a cleaner who received TB treatment several years ago who now motivates others to go to the OH department when they are coughing. When she hears someone coughing on the wards, she tells them her story and walks them to OH. All managers on the ward collect sputum from staff if coughing and send it directly to OH. Results only go to occupation health. We have isolation rooms- but they are only single rooms and patients must go into the hallway to use the bathroom.”  “There was a staff member diagnosed with MDR last year. Patients wait outside in separate area from the casualty and we always try to keep the windows open even if the people say it is too cold. We have two ladies who work 8-12 every day to educate patients about TB, cough etiquette and hand hygiene.”  “On this very day, we have no masks of any kind available. Not even in the TB ward. We admit MDR patients who are already on treatment if they have complications. There are no toilets in isolation rooms, therefore patients must go out in to the hall to use the toilet. This can be a problem.”  “More than 95 patients come through casualty each day (waiting area is a small, enclosed room with windows on one wall. All windows were closed at time of visit and a  94 ceiling fan was not turned on). Many times patients give sputum right on the bed even if there are other patients (there are 4 beds in each room).”  “There are so many miners with TB who come here. So we now have four DOTS supporters for TB prevention and work from 7-12 every day. MDR patients are kept on-site if they have another condition. Each ward has 2 isolation rooms with extractor fans. There are no toilets inside the room but we always give masks to wear when they must use the toilet.”  “The queue marshals pull coughing patients out of the waiting room and move them to an isolation room.”  “Right now we have 20 MDR patients admitted. Most come referred from district hospital. Rooms in MDR ward share bathrooms and must go out into the hallway to use the toilet. We track what proportion of sputum smears take more than 48 hours to get results from lab. This is almost all of them. It is too slow.”  “This hospital has high rates of defaulters because some patients are allowed hospital leave. They go back into the community for short times and do not take meds due to mental illness. It causes great difficulties.”  “The nurses are very good to wear N95 and patients understand to wear a surgical mask in OPD when nurse is taking info at triage. MDR patients are transferred in to this  95 hospital from other areas. They are cared for in the MDR area. Here there are UV lights and extractor fans. Bed occupancy is 70. Today we have 56 patients admitted with MDR. This is so many. Our IC nurse is very knowledgeable and motivated, makes IC a priority in the hospital so this is good.”  “The IC committee is not active because the IC coordinator is working on the wards due to staff shortages.”  Thematic analysis Themes and sub-themes for perceived barriers and facilitators to TB IC in the hospital derived from qualitative analysis were developed.  In total, there were 10 sub-themes describing barriers to IC and 8 sub-themes describing facilitators to IC. Frequencies of each sub-theme grouped by theme (type of control measure) are presented in Table 6.6.   In total, there were 26 sub-themes classified as barriers to IC in the hospital. Barriers in the administrative control category included “lack of patient education” and “poor/unsupportive workplace culture”. Both were mentioned once. The most commonly identified environmental control was “inadequate isolation facilities” (n=7), followed by “Physical infrastructure limitations” (n=4), “Poor ventilation/air flow” (n=2) and “no safe sputum collection area” (n=1).  For PPE controls, “inadequate supply of particulate respirators” was referred to by 3 participants. In total, there were 7 miscellaneous controls identified by participants as barriers to IC in their hospital. The most commonly  96 mentioned item was “health system constraints” (n=4), followed by “diagnostic delays” (n=2) and “human resource shortages” (n=1).  In total, there were 19 sub-themes classified as facilitators to IC in the hospital. Twelve of these were facilitators in the administrative control category. These controls included “triaging TB suspects” (n=3), “patient education” (n=2), “positive/supportive workplace culture” (n=2), “motivated IC professional” (n=2), “queue marshals” (n=2) and “regular IC training” (n=1). “Updated infrastructure” was identified by 4 participants as an environmental control that is a facilitator of IC. “Appropriate use of PPE” was mentioned by 3 participants; there were no miscellaneous controls identified as facilitators to TB IC in the hospital.   Interestingly, the word “MDR” or “MDR-TB” was mentioned 19 times by participants. Although this does not fit into a control group category, the high frequency of use warrants mention here.    97 Table 6.1. Profile of study hospitals (2012)  Hospital # Beds Registered TB Patients  Smear + TB Patients  MDR Patients    HCWs with TB  Total Staff  Rate of TB among HCWs     (per 100,000) 1 460 568 123 4 5 860 581 2 45 103 37 20 0 138 0 3 33 121 50 4 0 149 0 4 171 759 225 18 2 478 420 5 478 60 35 0 9 2,577 349 6 29 13 1 0 0 109 0 7 180 275 41 1 5 416 1,201 8 71 112 16 0 0 171 0 9 86 83 3 0 3 210 1,428 10 78 142 28 4 0 186 0 11 270 53 7 0 4 516 780 12 82 285 66 0 4 318 1,260 13 50 116 31 5 1 110 910 14 22 19 1 0 0 68 0 15 720 552 37 0 23 1,927 1,190 16 265 555 36 10 4 725 551 17 33 101 16 0 0 54 0 18 55 19 1 0 2 86 2,330 19 110 54 4 0 4 228 1,750 20 135 292 27 0 4 640 625 21 31 99 7 19 2 85 2,350 22 140 51 23 0 8 545 1,470 23 28 17 1 0 0 84 0 24 24 2 0 0 3 62 4,838 25 870 16 2 2 11 751 1,460 26 27 12 0 0 2 131 1,530 27 150 183 31 5 2 225 890 28 24 9 2 0 1 94 1,060   98 Table 6.2. Overall TB IC workplace assessment scores  Hospital Admin Control Score     (/32) Enviro Control Score  (/23) PPE   Score  (/11) Misc Score*    (/9) TOTAL TB IC Score (/75) 1 25 14 5 8 52 2 25 11 6 8 50 3 26 6 9 8 49 4 27 12 5 5 49 5 20 10 8 7 45 6 22 6 10 7 45 7 24 4 8 8 44 8 23 7 8 6 44 9 28 5 5 5 43 10 26 7 3 7 43 11 23 7 7 5 42 12 23 7 4 7 41 13 23 6 7 5 41 14 17 9 8 7 41 15 19 6 5 9 39 16 24 7 2 5 38 17 15 6 8 9 38 18 20 6 5 6 37 19 16 10 5 5 36 20 16 7 6 6 35 21 18 4 4 8 34 22 17 8 1 7 33 23 18 6 4 4 32 24 23 4 2 3 32 25 18 5 3 5 31 26 18 5 2 5 30 27 18 5 2 4 29 28 16 3 2 2 23 *Misc Score includes: turn-around time for AFB smear and culture results, time to transfer for MDR patients, availability of TB screening for staff, availability of IPT for staff, re-assignment of HIV+ staff to low-risk area   99 Table 6.3. Example of detailed TB IC assessment results  TB IC Indicator Variable Hospital 5 Administrative Controls There is a written infection control plan on-site  Tuberculosis control is included in this plan  Patients are screened for cough when they arrive  Cough surveillance/screening takes place in an appropriate area  There is a trained and dedicated person responsible for TB screening  Tissues are available for coughing patients  Face masks are available for coughing patients  Appropriate colour-coded waste containers are available   Patients are asked whether they have a history of TB  Patients are asked about the duration of their cough  Coughing patients are placed at the front of the queue  Coughing patients are separated from other patients while waiting  Patients are requested to produce a sputum sample if TB is suspected  Coughing patients are sent home  A symptom checklist is used  The symptom checklist is comprehensive   The checklist is used at each visit at admission or triage  There is an up-to-date TB register  Patients receive education about TB signs and symptoms  Patients receive education about respiratory hygiene and cough etiquette  Patient education occurs regularly  Educational pamphlets about TB are available to patients  Posters displaying cough etiquette are displayed  Staff receives infection control training  Infection control training is offered to all staff (not just clinical)  Infection control training occurs regularly   Infection control training is mandatory  Infection control training attendance is recorded  The hospital has an infection control committee  The committee contains representation from all depts/occupations  The committee meets regularly (at least semi-annually)  The infection control committee has a budget to support activities  Environmental Controls There is a separate area for TB patients and TB suspects  The separate waiting area is well-ventilated and away from other patients  There is a designated sputum collection area  The sputum collection area is well-ventilated and away from other patients  There are windows on opposite walls with unrestricted air flow  There are windows are on one wall (restricted air flow)  There are vents to improve air flow  The ceiling height is high (>3m)  Windows are kept open during the day  Windows are kept open at night  Windows are kept open during the summer  Windows are kept open during the winter   100 TB IC Indicator Variable  Ceiling fans are present to increase air mixing  Ceiling fans are in good working order  Extractor fans are present  Extractor fans are inspected and maintained regularly (at least once/year)  Negative pressure rooms are available  Air exchange rate in negative pressure rooms is appropriate  Ultraviolet germicidal irradiation (UVGI) is available  UVGI lights are inspected and maintained regularly (at least once/year)  HEPA filtration is available  HEPA filtration is inspected and maintained regularly (at least once/year)  Electricity is reliable in the hospital  Personal Protective Equipment Controls There are adequate supplies of face masks available for coughing patients  N95 respirators are not given to coughing patients  Staff wear PPE (any type) when assisting patients with sputum collection  Staff wear N95 respirators when assisting patients with sputum collection  N95 respirators are used for high-risk procedures (i.e. intubation, bronchoscopy)  N95 respirators are currently available   N95 respirators are always available in adequate amounts  Staff perform “fit-checks” when using N95 respirators  “Fit tests” are available for N95 respirators  N95 respirators are not re-used if soiled or damaged  Staff know when, how and why to use masks and respirators correctly  Miscellaneous Controls Lab turn-around time for AFB smear is <72 hours  Lab turn-around time for culture is <4 weeks  Time to transfer MDR-TB patients is <24 hours  TB screening is available on-site for staff  TB screening for staff is conducted regularly  HIV counselling and testing is available on-site for staff  Comprehensive HIV treatment and care is available on-site for staff  HIV+ staff members are offered re-assignment to lower risk areas  Isoniazid prophylaxis is available to staff if appropriate   Legend:       Done, observed, desired outcome Not done or not right, not observed Not applicable  101 Table 6.4. Logistic regression model with random effect for hospital   Dependent variable Estimate Standard Error P-Value Total IC score -0.065 0.017 0.00016 Administrative score -0.059 0.037 0.117 Environmental score  PPE score -0.130 -0.151 0.046 0.048 0.005 0.002 Miscellaneous score -0.153 0.076 0.044           102 Fig. 6.1. Total IC score effect plot   Fig. 6.2. Administrative score effect plot   103  Fig. 6.3. Environmental score effect plot  Fig. 6.4. PPE score effect plot   104   Fig. 6.5. Miscellaneous score effect plot         105 Table 6.5. Unadjusted and adjusted odds ratios from a generalized linear mixed-effects regression for the association between TB IC scores and HCW TB rate in 2012   Variable Unadjusted  Adjusted*  OR (95% CI) P-Value  OR (95% CI) P-Value Total TB IC Score Admin Score Enviro Score PPE Score Misc Score **Number of TB Patients         2-19         51-121         142-759 Number of Smear+ Patients         0-7         16-31         35-225 Number of MDR- TB Patients         0         At least 1   Geographic Location          Rural          Urban 0.94 (0.91 to 0.97) 0.94 (0.87 to 1.02) 0.88 (0.80 to 0.96) 0.86 (0.78 to 0.95) 0.86 (0.73 to 0.99)  1.00 (Ref) 0.39 (0.19-0.78) 0.46 (0.24-0.89)  1.0 (Ref) 0.72 (0.34-1.50) 0.48 (0.27-0.85)  1.0 (Ref) 1.05 (0.61-1.82)  1.0 (Ref) 1.36 (0.81-2.26) 0.0002 0.12 0.005 0.002 0.04 0.017 - - - 0.026 - - - 0.86 - - 0.27 - - 0.95 (0.91 to 0.99) 0.97 (0.90 to 1.04) 0.92 (0.83 to 1.01) 0.86 (0.80 to 0.98) 0.93 (0.80 to 1.08)  1.00 (Ref) 0.65 (0.28-1.52) 0.72 (0.33-1.54) - - - - - - - - - - 0.017 0.36 0.09 0.02 0.32 0.49 - - - - - - - - - - - - - *Only covariates with p>0.05 in the univariate analysis were included in the multivariate model **Number of TB patients and Number of smear+ patients were highly correlated and therefore only Number of TB patients was included in the final model   106 Table 6.6. Frequency of codes and themes  Line-by-line code Frequency Barriers  Administrative controls 2 Lack of patient education 1 Poor/unsupportive workplace culture 1 Environmental controls 14 Inadequate isolation facilities 7 Physical infrastructure limitations 4 Poor ventilation/air flow 2 No safe sputum collection area 1 PPE controls 3 Inadequate supply of particulate respirators 3 Miscellaneous controls 7 Health system constraints 4 Diagnostic delays 2 Human resource shortages 1 Total 26 Facilitators  Administrative controls 12 Triaging TB suspects 3 Patient education 2 Positive/supportive workplace culture 2 Motivated IC professional 2 Queue marshals 2 Regular IC training  1 Environmental controls 4 Updated infrastructure 4 PPE Controls 3 Appropriate use of PPE 3 Miscellaneous controls 0 None mentioned 0 Total 19    107 Chapter 7- Discussion  Discussion and interpretation of objective 1 results Objective 1: Estimate the incidence rate and assess the determinants of TB disease among HCWs in the Free State province, South Africa from 2002-2012  The results of the record linkage confirm that HCWs in Free State, South Africa have higher rates of TB than the general population.  Although the rates were higher than the general population in all study years, the excess of cases was particularly high between 2002-2008 and highest in 2005. For this year, there was an alarming 312% more cases of TB among HCWs than expected.  This means that the incidence of TB was more than 3-fold greater among HCWs than the general population in this year.    We observed a dramatic drop among HCW TB rates around 2009. We consulted with colleagues from the National Department of Health and the National Institute for Occupational Health in South Africa to ascertain whether they had any insight into this observation.  Although we were not able to identify any new policies or practice changes that were put into place to explain these findings, it is possible that there may have been changes in the data capture system or reporting practices during the peak years. It is also possible that the drop in TB rates could be explained by the aggressive role out of a free antiretroviral treatment program in the country in 2004.  It is estimated that there were 919,923 HIV patients enrolled in the public program by November 2009- a drastic increase from only 32,895 in January 2005 (94).  We are also aware of the fact that there  108 was some recent re-structuring in the Free State regarding cleaning and environmental services in the public hospitals. In the past, these services were provided by staff that was employed by the department of health. The department of health now contracts out these services meaning that these HCWs would no longer appear in the PERSAL database. Although this is a plausible explanation, we did not observe a drop in our denominators. Finally, it is also possible that the case definitions used in the ETR.Net system were changed. For example, if they changed the way they entered re-infections for the same person, we might see a drop in incidence.  Further investigations beyond the scope of this dissertation are necessary to fully explore the cause of the drop in incidence rates in 2009.  The Bacille Calmette-GuDerin, or BCG vaccine is the only currently available vaccine for TB. It was developed in 1920’s and is still used to immunize infants in many regions, including South Africa (1). There are several reasons as to why offering this vaccine is not appropriate for healthcare workers in South Africa despite their high risk of exposure to TB. First of all, the effectiveness of the BCG vaccine is poor overall. Secondly, BCG vaccination is not usually offered to people over the age of 16, and never to individuals over the age of 35 because the vaccine does not perform well in adults.  Furthermore, BCG vaccine is not recommended for: people who have already had a BCG vaccination, people with a past history of TB, people with a positive tuberculin skin test, people with HIV or other immune system disorders, and pregnant women (1, 4).     109 How do our findings compare to what we already knew? Our estimates of TB among HCWs are consistent with other reports from the region. A recent study by O’Donnell and colleagues from South Africa estimated rates of MDR and XDR-TB related hospital admissions. This retrospective chart review was conducted at one public referral hospital in KwaZulu-Natal and examined admissions that occurred between 2003-2008. The estimated incidence of MDR-TB was 64.8/100,000 among HCWs compared to 11.9/100,000 among the general population. Similarly, the estimated incidence of XDR-TB was 7.2/100,000 among HCWs compared to 1.1/100,000 among the general population (14). Although it presented extremely important results, this study had several limitations: HCW status was self-reported; no information was collected regarding occupation or workplace setting (i.e. laboratory, community HIV clinic, outpatient clinic in a hospital, etc.); and no information was collected regarding duration of employment in the healthcare sector.  A retrospective cohort study reviewed routinely maintained records in one hospital in Kenya to document TB case notification rates among hospital staff (95).  The authors estimated that TB case notification rates among HCWs ranged from 645 to 1,115/100,000 compared to 301/100,000 among the general population.  The authors admitted that HCWs who choose to seek TB treatment in the private health sector would not be included in this study.   A 25-year study conducted at a teaching hospital in Romania found an 11-fold increased risk of TB disease for HCWs when compared to the general population (mean TB incidence of 942.8/100,000 among HCWs compared to 98.6/100,000 among the general population) (38).  This study provided dramatic evidence that even in Romania, a middle- 110 income country in Europe, TB is a major occupational risk for HCWs and that increased attention to hospital IC is essential.  Another study from Eastern Europe focused on Belarus aimed to estimate the prevalence of TB among HCWs from 2008-2012 in 25 TB hospitals, by conducting a retrospective national record review. Similar to our findings and those from Romania, this study found that TB case notification rates were higher among HCWs than the general population (349/100,000 compared to 40/100,000). They also determined that most HCWs who were diagnosed with TB were female nurses between the ages of 25-44 (96).   The role of age and HIV status According to our findings, almost 30,000 people died from TB in Free State during the study period. More than 300 of those who died were HCWs. This loss of skilled personnel is a huge detriment to a health system that is over-burdened by the TB/HIV syndemic and where health human resource shortages are common. We found that there were more TB patients in the 60+ age category and in the age categories <29 in the general population group than in the HCW group. This is likely due to the fact that many HCWs retire in their sixties and may still be completing their education and training in their twenties and therefore are not yet employed. WHO estimates that 61% of TB patients in South Africa are co-infected with HIV (1). In our study, only 31.3% of the general population (non-HCWs) were known to be HIV positive. It was also interesting to note that the rate of TB patients who were known to be HIV positive was still much higher in the general population group when compared to the HCW group. This is because the HIV status was unknown for the majority of HCWs with TB (75.6%). This  111 suggests that HCWs in Free State are either not receiving adequate access to HIV counselling and testing or that they are afraid to disclose their status.   The need for comprehensive OH for HCWs Similar to the study from Kenya (95), HCWs in Free State had sub-optimal cure and completion rates even though their incidence of disease was higher than the general population. HCWs must therefore receive early diagnosis and treatment for TB in addition to improved infection prevention and control efforts (97) in accordance with international guidelines (98). Naidoo and colleagues administered semi-structured questionnaires to sixty-two medical doctors in South Africa who had been diagnosed with TB between 2007-2009. They found that a prompt diagnosis within 7 days was only made in 20% of participants. Ninety-five percent of those surveyed expressed concerns regarding a lack of IC in the workplace and negative attitudes of senior administrators and some colleagues (99).   HCWs should also be screened regularly for TB by programs that are free, confidential and available in the workplace (100). As mentioned previously, HIV status was unknown for 75.6% of health workers with TB compared to 43.2% of other TB patients. This suggests that stigma remains a major deterrent to disclose and perhaps evaluate HIV status among HCWs.  112  The effect of occupation The relative risk estimates for TB disease were highest for support staff and nurses. The health system in South Africa is largely nurse-driven and nurses have the most direct patient contact, perhaps putting them at increased risk of exposure. Support staff (such as cleaners, porters and security) also interact with patients regularly even though they are non-clinical staff, and they therefore require sufficient IC training and resources necessary to adequately protect themselves from TB in the workplace. These findings propose that education and training specifically tailored to support staff may be beneficial.  Interestingly, doctors and surgeons had a lower risk of TB when compared to allied health professionals. This could be due to the fact that doctors have the opportunity to self-treat or access treatment privately are thus not entered as TB cases in the ETR.Net system. These results also suggest that occupation and facility type are not as strongly associated with increased incidence of TB among HCWs as expected. This suggests that all HCWs who work in hospitals, clinics and even administrative settings are at risk of exposure to TB in the workplace.   The linkage process A probabilistic record linkage between ETR.Net and PERSAL was an appropriate method to create a database to identify HCWs with TB in South Africa however, similar to other record linkages described in the literature, this methodology was extremely time consuming and complex.  Re-evaluating and re-running the linkage algorithm several times and manually reviewing thousands of partial matches took additional time, but  113 improved the quality of the linked dataset.  As shown in Figure 2, 90% was a reasonable cut-point to accept all matches. The quality of the matches seems to decrease dramatically at 85% as evidenced by the large jump in incidence rate ratios. With all cut-off scores (80%, 85%, 90% and 95%) there is a noticeable drop in incidence rate ratios in 2008-2009 as mentioned previously.   Limitations Although this probabilistic record linkage study is the first in the region to objectively estimate TB incidence among HCWs, it does have several limitations. Primarily, the ETR.Net register does not necessarily contain all records of patients diagnosed with TB as HCWs in particular may be less likely to report their disease. It is possible that the estimates of TB among HCWs are under-reported here. Although TB treatment is available for free in the public health system, many HCWs who have health insurance (and particularly doctors and surgeons) may seek diagnosis, treatment and care for TB from a private provider. TB is a reportable disease in South Africa, however TB case reporting and subsequent entry into the ETR.Net system is difficult to enforce and monitor in the private system. Similarly, HCWs who are diagnosed with TB are eligible for compensation in South Africa, which is an additional incentive to report. However, the stigma associated with TB and HIV may override the desire to receive compensation and may prevent HCWs from reporting their disease and being registered in ETR.Net.   The quality of the data in ETR.Net is also variable as the system relies on input from paper forms collected by nurses at each facility. One study suggested that over one third  114 of South African TB patients who were identified in ETR.Net as smear positive did not actually have a laboratory record since ETR.Net is not linked to the laboratory system (101). This creates uncertainty in regards to the validity of smear results and treatment outcomes recorded in this database.  Another study that aimed to determine the completeness and reliability of ETR.Net data found that agreement between the electronic registry and the paper record was excellent for sex, moderate for patient type (whether it was a new, relapse or retreatment case), moderate for treatment outcome and poor for HIV status.  Furthermore, the information contained in ETR.Net did not allow us to distinguish between relapse and retreatment cases. We recognize that the major risk factors for relapse include inadequate therapy due to irregularity, high disease burden in the population, inadequate duration of therapy and underlying drug resistance. Recurrence of disease due to true relapse would ideally be distinguished from reinfection.  Occupational cohort studies are vulnerable to several biases such as misclassification bias. Misclassification bias on exposure is not likely here, however, misclassification of the outcome (disease status) is possible. Since HCWs are employed, they may be more likely than the general population to seek treatment in private hospitals and clinics and my therefore not be registered in ETR.Net.  Registration of TB cases in ETR is not adequately enforced in the private system. This could result in HCWs who would have been incorrectly classified as TB-free in our cohort. This study may also be influenced by the healthy worker effect. This bias suggests that the least healthy workers may leave the workforce and will not be included in the study. Conversely, at the time of hire, most HCWs are relatively healthy. To minimize this bias, selecting an internal comparison  115 group instead of the general population is recommended (63), however, identifying HW groups where exposures levels are known and non-uniform is difficult. Confounding by non-occupational risk factors (such as diabetes or smoking) was not examined here.   Discussion and interpretation of objective 2 results Objective 2: Examine the association between TB IC and the incidence of TB among HCWs in 28 hospitals in the Free State province, South Africa in 2012   These findings show that there is large variability in TB infection prevention and control measures in public hospitals in Free State, South Africa. In spite of differences in staffing levels, TB patient load and facility size, implementation of TB IC measures is associated with the incidence of TB disease among HCWs.  The total IC score, the environmental score and the personal protective equipment score were shown to have the greatest association with HCW TB rates.  As scores in these three categories increased, the probability of having a HCW with TB at that hospital decreased when controlling for the number of TB patients in each hospital.   Although we know that TB spread in hospitals is in large part due to a lack of infection prevention and control measures and that the HCW who work in these facilities are at increased risk of exposure, little research has focused on this issue and many gaps remain (102). More specifically, few studies have used HCW TB disease incidence as a marker of TB transmission in hospitals in low and middle-income countries (15, 54).   116  Which TB IC measures are most important? As suggested by the results presented here, administrative controls, environmental controls and correct use of personal protective equipment are all proven and effective measures, that when used in combination, can drastically decrease TB transmission in healthcare facilities (22) and protect HCWs from getting sick.  Other studies have also examined the effectiveness of hospital IC measures and have shown that even the most basic IC interventions in resource-limited settings can decrease the spread of TB in hospitals (22, 44, 103). Based on the findings presented here, our primary recommendations include: 1- Develop and implement of targeted education for TB IC for all staff groups 2- Emphasize workplace culture and the role of champions within the institution 3- Implement of regular TB IC workplace assessments using a validated tool 4- Purchase adequate supplies of PPE (both N95 respirators and surgical masks) and provide appropriate education and training on proper use These recommendations were generating from the qualitative and quantitative findings and the published literature. We also focused on low-cost measures that have the potential to have high impact.   Findings from a study conducted in a large, inner-city Chicago hospital showed that annual HCW skin test conversion rates fell significantly when simple administrative controls such as educating HCWs about how to recognize TB suspects, were implemented (36).  Although Chicago is located in a high-resource, low TB-burden  117 region, general lessons learned are relevant to low-resource, high TB-burden regions as well. Simple, low-cost measures to reduce the spread of tuberculosis are often the most effective, but least used (102). Increasing the index of suspicion and promptly separating suspected TB patients from others, encouraging cough etiquette among patients, offering masks to coughing patients, opening windows and constructing outdoor waiting areas are some examples of low-cost control measures that have the potential for high impact.   Albuqerque de Costa and colleagues studied the impact of implementing administrative measures on rates of LTBI among HCWs in Brazil, which is also a low-resource, high TB-burden setting. More specifically, they examined the effect of isolation of TB suspects, HCW education and turnaround times for sputum tests. They found that TST conversion rates were reduced from 5.8/1,000 to 3.7/1,000 person-months (p=0.006). (44). Previous research such as this has suggested that administrative control measures should be the first priority when it is not feasible to implement the full hierarchy of recommended IC measures. Although we agree that ICs measures in settings such as South Africa should be simple and low-cost, the findings presented here indicate that implementation of all categories of controls is most important and that implementation of administrative controls should not be prioritized over other measures when aiming to decrease transmission of TB to HCWs, patients and visitors.   The value and validity of TB IC assessments/audits One study by Kanjee and colleagues recommended that facilities should conduct regular TB IC assessments and implement multi-faceted TB IC and behavioural change  118 interventions to decrease nosocomial TB transmission (57, 58).  An article by Claassens and colleagues from South Africa found that scores from an IC audit were significantly associated with reported cases of TB among HCWs in primary health clinics in a univariate logistic regression. However, in contrast to our results, this did not hold true in the multivariable analysis leading them to conclude that scores from the IC audit tool were not adequate measures of TB risk for HCWs (45). The authors go on to speculate that IC assessments or audit tools may only be “a process management tool” and may not be worthwhile doing in an already overwhelmed health system such as South Africa’s. The findings presented here provide additional evidence to support policies that encourage regular IC assessments. Our study results also indicate that TB IC scores from these assessments or audits can be used as a proxy for nosocomial transmission of TB.    The importance of training and workplace culture The barriers and facilitators identified in the qualitative analysis were similar to those compiled from articles included in the scoping review in Chapter 2. Knowledge and attitudes of the healthcare workforce are critical to effective implementation of IC measures in any facility. Farley and colleagues conducted a cross-sectional study in all multidrug-resistant tuberculosis (MDR-TB) and extensively drug-resistant tuberculosis (XDR-TB) facilities in South Africa.  Like our study, they found that facility infrastructure and adherence to IC recommendations varied considerably between hospitals (35).  They also concluded that observed IC practices were poor across all disciplines and that greater infection knowledge was associated with higher levels of clinical training (35).  Similarly, a study from Russia concluded that TB disease was  119 generally well understood by HCWs, but that knowledge of IC measures related to TB was poor among all employees (62). A study conducted in 29 healthcare facilities in Mozambique found that three quarters of the HCWs had N95 respirators however, only 36% reported knowing how to use them correctly, again highlighting deficiencies in existing training and education programs (56).  A paper published in the European Respiratory Journal describes a survey tool similar to the ICAP tool that was developed and used in MDR-TB and XDR-TB facilities in five European countries. Although respirators were always available to HCWs in this setting, the results of this study demonstrated that even in high-resource settings, IC measures do not fully comply with international recommendations (38) further stressing the need for behaviour change interventions and more focus on HCW motivation and support as suggested by the qualitative findings presented here. An assessment of TB IC knowledge, attitudes and practices conducted in a South African hospital with documented MDR-TB and XDR-TB transmission revealed that 49% of employees felt that the hospital did not care about them and other HCWs. Respondents also reported several other barriers to TB IC such as lack of confidentiality of OH information, TB and HIV stigma in the facility and inadequate resources available for them to adequately protect themselves (57, 58).    A study from Uganda found that universal precautions were more acceptable among HCWs and patients than targeted ones, with the exception of separating TB patients. They also found a lack of community awareness about airborne transmission as a primary barrier to effective implementation of TB IC measures (42).  The authors concluded that  120 halting the spread of TB in healthcare facilities requires a patient-centered, rights-based and evidence-based approach. A systematic review from Europe identified elements that are essential to effective infection prevention and control programs in the healthcare setting. After reviewing 92 studies, some of the essential components of successful IC implementation identified included: appropriate staffing and workload; education and training; access to materials and equipment; appropriate use of guidelines; surveillance; multimodal and multidisciplinary prevention programs with a behaviour change component; engagement of champions; and a positive organization culture (104). Like our study, these published articles stress the need for a positive and supportive workplace culture to enable proper implementation of TB IC policy and practice.   The role of champions Both the results from the workplace assessments and the discussions with the HCWs suggest that having a TB IC “champion” in a hospital improves IC and reduces TB exposures for both HCWs and patients. A champion could be a queue marshal who makes sure that all patients entering the hospital are screened for TB, a dedicated and determined IC nurse, a community health worker who educates patients about TB signs and symptoms and cough etiquette, or simply a HCW on a ward who has been affected by TB and has made it her/his priority to share their story and to encourage their colleagues to protect themselves.  A similar approach has been successful in improving compliance with hand hygiene (105). Encouraging TB champions is a low-cost recommendation that has the potential for great impact.    121 The need for OH for HCWs It is clear that not enough is being done in low-resource, high TB-burden countries to care for the healthcare workforce who may also be living with TB and HIV. The study by Reid and colleagues (37) found that PPE for HCWs were only available at 16% of sites surveyed. Furthermore, TB screening for employees was only available at 13% of sites and isoniazid preventive therapy for those who were HIV positive, and co-infected with TB, was only available in 8% of participating facilities. The low miscellaneous scores and qualitative findings presented here highlight similar deficiencies across the province related to availability and frequency of on-site TB screening for staff and availability and accessibility of on-site HIV and TB testing, counselling and treatment (ARVs as well as isoniazid prophylaxis as appropriate).  More needs to be done to ensure that HCWs in South Africa are being taken care of as recommended by the joint WHO ILO UNAIDS policy guidelines for improving health workers' access to HIV and TB prevention, treatment, care and support services (98).   The need for health systems strengthening Implementation of IC measures to prevent the spread of TB in healthcare facilities requires that the fundamental components of a fully functioning health system be in place (106). Without appropriate governance, financing, infrastructure, supply chain management and service delivery, implementation of IC is not possible (37).  Although these measures of health system functioning were not explicitly evaluated and were beyond the scope of this study, health system constraints and human resource shortages were themes that emerged from the qualitative analysis. A study by Cohen and colleagues  122 from Canada examined the relationship between workload and injury, pain, burnout and self-reported health among Canadian nursing staff. They found that workload and staffing levels are important determinants of injury at work and burnout among staff (107). Although this study focused on musculoskeletal injury, the same reasoning can be applied to infectious disease transmission. For example, a study from the Free State found that 21.2% of HCWs surveyed reported needle-stick injuries and other body fluid exposures at work (108).  Staff shortages and subsequent workloads in an already over-burdened health system such as in South Africa make it challenging for HCWs to take the time necessary to adequately protect themselves from exposure to infectious diseases such as TB.    Limitations This study included almost all public hospitals in the Free State province of South Africa. Although it is home to only approximately 6% of the country’s population (109), the Free State province is similar to other provinces in terms of demographic composition (109).  Free State has the third highest overall adult HIV prevalence rate, behind only KwaZulu-Natal and Mpumalanga provinces (110). Interestingly, the Free State has the lowest life expectancy at birth (110). These results are of relevance to other provinces in the country and also to other high-incidence HIV and TB countries in Sub-Saharan Africa.  Although this study included almost all public hospitals in the Free State, it is limited by the fact that data were collected at only one time-point. The results presented here therefore provide only a snapshot, and the situation may provide differing results if data  123 collection was repeated during a different timeframe. Another major limitation of this study is that only one year of disease incidence was used as the outcome measure. Since occupational infection is cumulative and there is usually a delay to development of TB disease, future similar studies may consider using more than one year of HCW TB rate data if it is available. Like most of the studies included in Question 2 of the scoping review, this study did not include an intervention. Developing, implementing and evaluating an intervention by re-administering the data collection tool was beyond the scope and budget of this project, however, the possibility of doing so was discussed with the larger research team and remains a possibility as part of future research.  Non-differential misclassification is another possible limitation of this work. It is possible that the observed, not observed or presence, absence of control measures may have been missed or incorrectly identified due to human error. Finally, the ICAP tool was only able to assess the availability of IC measures and may not have adequately captured information regarding quality and consistency of implementation.   The bigger picture This dissertation has explored host (HCW), agent (TB) and environment (workplace and health systems) interactions as they relate to TB disease among HCWs in South Africa. As illustrated by the conceptual framework presented in Chapter 1 (Figure 1.1), we recognize that it is also critical to acknowledge that there are broader socio-economic and political factors that impact both the rate of TB among HCWs and the implementation of TB IC in this setting. Considering context in global health research is critical. Some of  124 these factors will be briefly discussed in this chapter, along with some reflections on the opportunities, challenges and ethics associated with global health research.   Socioeconomic-political factors The Republic of South Africa is a multi-ethnic country with 11 languages officially recognized in the constitution. South Africa, sometimes referred to as the “Rainbow Nation”, has endured a history that is far from sunny and remains tainted by the legacy of apartheid. The word “apartheid” comes from the Afrikaans language and means “separateness”, “the state of being apart” or literally, “apart-hood” (111). Although approximately 80% of the South African population is black, since the 1940s the population was controlled and repressed by the white minority. After a long struggle with institutionalized racism, Nelson Mandela, leader of the African National Congress, was elected to office in South Africa’s first democratic election in 1994. Although the Rainbow Nation has made great strides forward, the legacy of apartheid has left deep wounds that remain visible more than 20 years later.  “Giving people medicine for TB and not giving them food is like washing your hands and drying them in the dirt.” -Haitian proverb Structural violence The term structural violence refers to the scenario where a social structure or social institution poses the potential to harm people by preventing them from meeting their basic needs (112). South Africa is classified by the World Bank as an “upper-middle  125 income economy”, yet the disparity in wealth, resources and access to healthcare services is staggering. In 2002, life expectancy was 60 years for black South Africans and 74 years for white South Africans (113). It is not uncommon to see expensive cars parked outside the large, luxurious homes in the elite suburbs of Cape Town and Johannesburg. Yet thousands of South Africans still live in informal townships or slums. In 2000, it was estimated that 30% of South Africans do not have electricity in their homes and this number is closer to 50% among those living outside of the major cities (113).   Poverty and TB are inextricably linked. The following statement from the World Health Report released in 1995 by the WHO acknowledges that poverty is perhaps the world's greatest killer: "Poverty wields its destructive influence at every stage of human life, from the moment of conception to the grave. It conspires with the most deadly and painful diseases to bring a wretched existence to all those who suffer from it." (114) Likewise, in his aptly titled book Infections and Inequalities, Dr. Paul Farmer writes:  "Physicians again need to think hard about poverty and inequality, which influence any population’s morbidity and mortality patterns and determine, especially in a fee-for-service system, who will have access to care… All of the processes leading to sickness and then to diagnosis and treatment, are related to a series of large-scale social factors.”  (115). Reactivation of TB occurs frequently when people are malnourished and live in crowded, poorly ventilated housing. Not having enough nutritional food to eat impairs cell-mediated immunity, making malnutrition an important risk factor for the development of TB. One study from India examined national estimates of the population-attributable fraction for malnutrition and concluded that addressing chronic malnutrition  126 in this country would complement the current TB control strategy and would ultimately help reduce the incidence of TB (116). Similarly, a study from an Inuit community in Northern Canada found that a poor diet was associated with TB infection (117). Some may suggest that HCWs in South Africa are of a higher socioeconomic status than much of the general population simply because they are employed, however, many nurses struggle to pay bills, to buy healthy food for their families and almost all live in the townships. Not even HCWs are immune to inequality and are equally affected by the social determinants of health. No person, regardless of where they live or how much money they have, should die of an infectious disease that is treatable, yet far too many people continue to die of TB.  Post-apartheid neoliberalism Neoliberalism refers to state policies that are market-oriented, favour commercialization and privatization and reduce expenditure of social services. This approach, especially in the post-apartheid era, has had a direct impact on health and health services in South Africa. It has been suggested that structural adjustment programs and neoliberalism in Sub-Saharan Africa have led to declines in health-seeking behaviour (118).  Neoliberal policies also had an impact on HCWs, including cuts in the size of public service leading to understaffing, salary decreases making it hard to recruit and retain HCWs in the public healthcare sector, declining morale and brain-drain of doctors and researchers in particular (118). At the health system level, South Africans have witnessed declining fiscal support, difficulties in procuring medications and medical equipment and a diminished ability of the system to deal with the ongoing HIV epidemic.   127 Syndemics A syndemic is defined as the scenario when two or more diseases or conditions interact synergistically and contribute to the excess burden of disease in a population (119). This theory also delves beyond the biological interaction and stresses that individual epidemics are sustained in a community or population due to “harmful social conditions and injurious social connections” (120).  The intertwining of TB and HIV is a perfect example of a syndemic. Biologically, being HIV positive greatly increases chances of developing TB. If a person is HIV positive and not on effective treatment, they have an estimated 10% annual risk of developing TB disease. For individuals who are HIV negative, they have 5% chance within first 1-2 years following infection then another 5% risk for the remainder of their life. This means that a person who is HIV positive has a 10% annual risk as opposed to 10% lifetime risk for someone who is HIV negative. Most recent WHO statistics estimate that 61% of TB patients in South Africa are HIV positive and that 1,148,477 HIV patients were screened for TB (1).   Although HIV can affect anyone, those who are poor may be less likely to have access to healthcare services, get tested and know their status. Being unaware that their immune system is compromised as a result of HIV infection may prevent individuals from taking the necessary additional precautions to avoid getting TB. Those who live in poverty are also more likely to have frequent exposures to Mycobacterium tuberculosis (in crowded housing or while using shared transportation) and may be more susceptible to developing TB disease due to preexistent immune system damage from other infections and malnutrition (115). One study from South Africa reported a strong association between  128 TB incidence and the education level of children and income of adults (121). Lack of consistent income and education may also be associated with disadvantaged access to diagnosis and treatment for TB (115). Like housing, nutrition and education, a healthy workplace is an important social determinant of health.  As the Commission on the Social Determinants of Health has highlighted, working conditions contribute considerably to health inequities (122). Despite this, inadequate attention has been devoted to improving working conditions – including for HCWs. Many nurses in particular are forced to cope with high levels of stress. A study conducted by Ratner and Sawatzky found that Canadian nurses are more likely than other employed female postsecondary graduates to report work stress and back pain (123). With the continuing struggle against the HIV and TB syndemic, HCWs globally have witnessed massive shifts in their day-to-day practices and are facing increasingly difficult working conditions (124). In settings such as South Africa, increasing workload and inadequate supplies contribute to burnout, stress, and low morale, driving HCWs out of the public health sector, increasing pressure on the remaining workers, resulting in more stress. The results of the TB IC assessments presented in the previous chapter showed that some hospitals in Free State are actively making an effort to create healthy workplaces by prioritizing IC efforts and the health of the HCWs by offering HIV and TB counseling, testing, treatment and support to their employees. Some facilities are also trying to keep their staff healthy by accommodating HCWs who are HIV positive by discretely allowing them to be transferred to lower-risk areas for TB in the hospital. In order to successfully combat infectious diseases such as TB, it is important to consider not only biological interactions and synergies such as that  129 with HIV, but also the precipitating and resultant social conditions at the population level.  Stigma  The term stigma refers to a negative stereotype or a mark of disgrace. Stigma and fear of discrimination deters people from getting tested for HIV, seeking treatment if positive and disclosing their status to friends and relatives. Since TB and HIV co-infections are common, the stigma, shame and fear associated with being HIV positive manifest amongst TB patients as well, regardless of their HIV status. A study that interviewed community health workers in KwaZulu-Natal, South Africa concluded that stigma, denial, and lack of education are the primary reasons for failure to complete TB treatment and for avoiding HIV counselling and testing (125).  Stigma reduction efforts must therefore be integrated into TB testing and treatment plans.  Stigma in the workplace presents a unique set of challenges. HCWs who are HIV positive may fear losing their jobs or losing the respect and trust of colleagues and patients if they disclose their status. The workplace is a vital front in the struggle against HIV and TB since the majority of HIV and TB patients are in their most productive years. In 2001, the International Labour Organization adopted a code of practice on HIV/AIDS and the world of work. This document focuses on recognizing of HIV/AIDS as a workplace issue, and also discusses non-discrimination in employment, sex equality, confidentiality, social dialogue and prevention and care and support (126).  A study from the Free State, South Africa examined barriers to uptake of HIV counselling and testing offered to  130 HCWs at on-site OH clinics. Approximately 40% of respondents identified stigma as a barrier (127).  This suggests that although HCWs may have more knowledge and access to information than most of the general population, they remain fearful of what an HIV or TB diagnosis may mean for them, their families and their careers. Recent WHO/ILO/UNAIDS guidelines recommended improving access to HIV and TB diagnosis, treatment, care and support for HCWs by offering these services in their hospital OH clinics (98). Offering these services in the OH clinic ensures that the care is free and convenient, but efforts to reduce stigma must also be considered to improve utilization. An article by Siegel and colleagues described the development of a stigma reduction campaign specific to the healthcare sector. Among other things, the authors determined that stigma reduction among HCWs must be embedded in institutional management and must also consider the localised needs of the health workforce themselves (34).   In order to make significant strides towards reducing and ultimately eliminating TB among HCWs in South Africa, it is critical to address host, environment and agent characteristics while also considering socioeconomic factors such as structural violence, post-apartheid neoliberalism, syndemics and stigma.   Global health research With ongoing globalization and greater global inter-connectedness, international collaborations are becoming more common in the field of public health and interest in global health research continues to grow.  131 Opportunities and benefits  Conducting PhD fieldwork in an international setting such as South Africa provided me with the opportunity to be immersed in the local culture, thereby facilitating a deeper understanding of the broader socioeconomic determinants of health. This international travel not only allowed me to collect the data necessary for my dissertation, but also provided an opportunity for me to develop my contextual knowledge of the research setting in which I am fortunate enough to work. The diversity of people, cultures and languages in South Africa is both fascinating and somewhat overwhelming, however, I feel that understanding the complex history and the roots of these traditions are vital to understanding the barriers and facilitators to health in this country. Furthermore, I believe that the many months I spent in South Africa over an 8-year period helped generate trust and respect from the HCWs who graciously agreed to participate in this study. Without their courage to come forward to speak about a disease that carries significant stigma, this work would be impossible.   For the most part, I felt welcome in the hospitals that I visited while conducting fieldwork and the nursing staff, in particular, was warm and inviting. However, there were some instances where my work and motivations were scrutinized simply because I was a foreigner. I made a concerted effort to meet with hospital managers, CEOs and frontline to explain my study and seek input while also building trust. The ultimate goal of a global health partnership is to contribute to improvements in health outcomes and health equity (128). Developing and nurturing those relationships and partnerships (with participants, institutions, local researchers) is integral to conducting successful population health  132 research, and particularly when the study is based outside of your home city/region/country. I hope that my global health experience also contributed towards building the capacity of my local research colleagues with whom I collaborated. With the support of faculty from UBC, I did my best to foster equitable partnership dynamics.   In addition to learning about epidemiologic study design and statistical methods, I had the opportunity to learn about cultural sensitivity, diversity of opinion, diplomacy and international-interdisciplinary collaboration and partnership. These skills are difficult to acquire in a classroom setting, and I am grateful for the complexities that shaped my experiences while working in South Africa. I believe that I am a more competent public health researcher because of my global health experience.  Challenges  Maintaining good quality surveillance data systems is challenging in all regions, but particularly so in resource-strained settings that are also over-whelmed by a surging syndemic. The lack of electronic records and poor quality paper-based surveillance systems in many hospitals in Free State made it difficult, and sometimes impossible, to accurately estimate TB patient case-load and other related parameters such as time to diagnosis, etc.  On several occasions, my expectations had to be scaled back to be consistent with the local realities.   Conducting research in a foreign setting is often logistically complicated and expensive. Coordinating meetings and finding the hospitals (some located in densely populated  133 townships in remote areas) was frustrating and time-consuming, however, I am eternally grateful to my colleagues from Bloemfontein who played a critical role in facilitating and supporting this research. Partnering with a local researcher who knows the geography of the area as well as the local language was extremely invaluable.   Although I have already mentioned the benefits of a productive international collaboration, partnership dynamics can also present a myriad of challenges when priorities, personalities and cultural norms clash. The Canadian Coalition for Global Health Research developed a practical tool (called the Partnership Assessment Toolkit) that enables partners to openly discuss the ethics of their partnership and to put in place structures that create ethical accountability (128). Similarly, our South African-Canadian research team reviewed our experiences of working together and identified the factors that population health researchers must explicitly consider when planning a randomized control trial in particular within a North-South partnership (90). We concluded that global health partnerships require significant investment of time and resources and that it is important to recognize the iterative nature of the process. Researchers should also be prepared to revise protocols as challenges emerge (90).  Although I had the distinct privilege of working in South Africa for several years prior to starting my PhD work and therefore had a basic understanding of the historical, political, and sociological complexities of this setting, I always did my best to show respect for local traditions, cultures and ways of doing things. I quickly learned that transdisciplinarity, equity and participation are critical to success.    134 Ethics  As more researchers become involved in global health initiatives, ethical considerations pertaining to such collaborations are receiving warranted additional attention. International documents have been recently developed to provide guidance and standardized recommendations for conducting research in international settings. For example, the Nuffield Council on Bioethics Report, published in 2002, examines ethical issues related to healthcare research conducted in developing countries and funded by sponsors from developed nations (129). In 2000, the WHO developed the Operational Guidelines for Ethics Committees that Review Biomedical Research. This document provides guidelines based on the requirements for ethical review and presents an evaluation of existing practices of ethical review in countries around the world (130). In the United States, the National Institute of Allergy and Infectious Diseases within the National Institutes of Health created a website called ClinReg.  This tool enables researchers to plan international clinical trials more efficiently by consolidating country-specific clinical research regulatory information and allowing comparisons across countries (131).   Health research should be relevant to the country and community in which it is conducted and should promote equity while also developing local capacity (132). In order to ensure that participants will benefit from the research, global health projects must actively involve the community in all stages of the study. It has been suggested that community engagement must go beyond community participation and should involve “building authentic partnerships, including mutual respect and active, inclusive participation; power  135 sharing and equity; and mutual benefit” (133). Yassi and colleagues take this concept one step further and suggest that ethics reviews of global health studies must also consider beneficence and non-maleficence at the community level as well as social justice issues and local capacity-building (134). A study conducted with First Nations participants in Canada successfully utilized community engagement to inform and implement a community-randomized controlled trial for cervical cancer screening. The findings from this study were used to inform culturally appropriate screening strategies in this population (135). Likewise, research that aims to decrease TB among HCWs, must actively include opinions and experiences of frontline HCWs, particularly those who have been diagnosed with HIV and/or TB and are willing to share their stories.   There has also been debate surrounding the extent of obligations of the researchers who are engaged in global health work. After reviewing international research ethics guidelines and related documents, Lairumbi and colleagues found that there are notable differences between the "global opinion" and the views of local individuals within resource poor settings regarding research benefits and beneficiaries (136). The authors suggest that global research ethics require more reflection, and a careful assessment of the practical realities of implementing the ethical principles in real world context (136). For example, many feel that institutional ethics committees should require researchers who initiate TB treatment for study purposes to report final treatment outcomes, even if this is beyond the scope of their study (137). Researchers should not be exempt from adhering to international standards of care. Although this study focused on prevention rather than treatment of TB among HCWs, our research team felt similarly obligated to  136 communicate findings from the TB IC workplace assessments to hospital management and TB, IC and OH professionals. We made a concerted effort to work with those who were interested to develop recommendations and plans for improvement based on our findings. This was not part of the official study since it did not include a formal intervention, however, it was imperative from an ethical perspective that we do so.  The ethics surrounding student involvement in international initiatives have been well explored, particularly in the field of medical education (138, 139). As a student working in global health, I was conscious of the fact that I was in South Africa to learn, and that the local HCWs and patients had a far deeper understanding than me of TB in this setting (and related host, agent and environment factors). Overall, my research experience in South Africa was productive, positive, challenging and illuminating.      137 Chapter 8- Conclusion The overall goal of this dissertation is to describe the burden of TB disease among HCWs in the Free State, South Africa and to evaluate the state of infection prevention and control in public hospitals. To our knowledge, this is the most rigorous attempt to date to quantify the burden of occupational TB among HCWs in a high-burden setting. This is also one of few studies to use rates of TB among HCWs as an outcome measure to study the relationship between infection control and TB transmission in hospitals. Ultimately, this work aims to protect HCWs from TB in the workplace.   Contributions Conceptual framework and scoping review The conceptual framework presented in Chapter 1 demonstrates that TB among HCWs is a complex, multifaceted issue influenced by many direct and overarching factors. The framework provides a structure around which researchers, policy makers, and frontline HCWs responsible for TB/IC/OH management could develop TB prevention and control efforts in the healthcare workplace. Interventions to decrease the spread of TB in hospitals must consider host (HCW), agent (Mycobacterium tuberculosis), environment (workplace, community and health system), in addition to socioeconomic factors (structural violence, post-apartheid neoliberalism, syndemics and stigma).   The scoping review in Chapter 2 maps and summarizes existing literature related to TB among HCWs and IC in hospitals. More specifically, the scoping review shows that there are only 18 original research articles and 4 systematic reviews that attempt to quantify the  138 rate of occupational TB among HCWs. The review also assesses the methodologies used by these studies focusing on the study design, data source, denominator used and effect measure presented. We found that 11/18 (61.1%) of studies were from Sub-Saharan Africa with a median study period of 4 years. Most studies (72.2%) were retrospective or historical prospective cohort studies and most used IRR as their measure of effect (55.6%).   Furthermore, our results include several studies that were not part of the 4 systematic reviews previously published on this topic and thus provide an updated overview of the literature. The scoping review also identifies 13 studies that present barriers and facilitators to TB IC in high TB-burden regions and particularly in Sub-Saharan Africa. Study design, methods used, description of the intervention (if applicable) and a detailed list of barriers and facilitators are discussed. We found that 8/12 studies (61.5%) were conducted in Sub-Saharan Africa and that the median number of facilities included was 26.5. The majority of studies were cross-sectional (n=9, 69.2%) and only 3 included an intervention (23.1%). The most commonly used methods were an IC audit or inspection tool (n=7, 53.8%) followed by questionnaires or surveys (n=5, 38.5%). Overall, the scoping review shows that more high-quality research on this topic is warranted.  Objective 1: Estimating incidence rate and determinants of TB disease among HCWs in Free State, South Africa over a decade This study utilizes record linkage techniques to objectively quantify the magnitude of TB among HCWs in the Free State, South Africa over a decade. We identified 2,677 cases of  139 tuberculosis among HCWs over the ten-year study period compared with 1,280 expected cases.  The mean incidence for TB among HCWs over the study period was 1,494.48 (SD= 569.50). Overall, the incidence of TB among HCWs was greater than non-HCWs in Free State re-affirming that occupational exposures are contributing to the excess of TB disease in this population. Incidence rate ratios ranged from 1.14 (95% CI: 0.98 to 1.32) in 2012 to 3.12 in 2005 (95% CI: 2.81 to 3.45). The findings show a significant drop in TB incidence among HCWs in 2009. Although there are several possible explanations for this observation as discussed in the previous chapter, it warrants further research as we were not able to come to any clear conclusions or associations based on our findings and observations. Age, sex, race and duration of employment were associated with rates of TB among HCWs. HCWs who were older (especially those born in the 50s and 60s), male, black, coloured and were employed for less than 20 years (especially those employed for 11-15 years) had higher rates of TB. Occupation and facility type were significant in the univariate analysis but were not significant in the multivariate model.   Establishing a baseline incidence rate of TB disease among HCWs in South Africa is necessary to allow for monitoring effectiveness of interventions and to measure progress. By providing an objective estimate of the burden of TB disease, this study describes the magnitude of the problem among this high-risk, yet essential group.  Documenting and understanding the magnitude of the problem is the first step towards eliminating nosocomial transmission of TB. Exploring other risk factors for TB including workplace factors such as occupation, facility type and duration of employment in addition to non-workplace factors such as age, sex, race, and HIV status provides an opportunity to  140 understand the workplace hazards and socioeconomic disparities that are so prevalent in South Africa.  Including these potential risk factors that emerge from unfair social processes in the study provides a more holistic examination of issues that affect the community of HCWs. These findings can be used to inform health and wellness programs targeted specifically to HCWs.  Recommendations related to objective 1 In addition to international guidelines from WHO, ILO and UNAIDS (21, 91), South Africa has adequate policies in place for the protection of HCWs from TB and other workplace conditions (95, 96, 97, 98). This study illustrates the urgent need for the implementation of these policies, in particular TB infection prevention and control measures and OH and safety practices (32).  There is a need for better workplace as well as workforce surveillance, with prompt follow-up of cases of HCWs with TB to ensure that all IC measures are being followed in areas in which staff who contracted TB worked and to ensure that there is appropriate follow-up amongst other HCWs who may have been exposed.  HCWs work in stressful environments where they are at high risk of exposure to infectious diseases such as hepatitis, HIV, TB and even Ebola. Many HCWs who are diagnosed with TB, report feeling stigmatized (99) and unsupported in their journey back to health (144). The findings presented here re-affirm the urgent call for action to protect the healthcare workforce. HCWs are frustrated with the ineffectual global response to occupational TB and the lack of infrastructure and resources to prevent transmission in healthcare settings (99).  A published South African report entitled “OH Challenge Facing the Department of Health” suggests that protection of HCWs must go  141 beyond infection prevention and control. The authors conclude that the following are all required to adequately protect South African HCWs from TB: an integrated management system that includes commitment from top management; practical and locally relevant polices; training; surveillance; and provision of comprehensive and convenient OH services (145).   HCWs in Free State, South Africa have higher rates of TB than the general population. HCWs are the backbone of health systems worldwide and this study shows that we are not doing enough to protect them. Additional efforts must be made to protect this high-risk, high-value population by implementing effective IC measures and providing timely TB screening, diagnosis, treatment and support. Probabilistic record linkage is a valid tool to combine occupational data with infectious disease registry data in absence of a unique identifier.  Objective 2: Evaluating TB IC in 28 public hospitals in Free State, South Africa and the association with HCW TB incidence Nosocomial and occupational risks were documented in 28 public hospitals using a comprehensive TB IC facility assessment tool. The TB IC facility assessment scores show that there is large variability in TB infection prevention and control measures in public hospitals in Free State, South Africa. We also determined that implementation of TB IC measures is associated with the incidence of TB disease among HCWs.  The total IC score, the environmental score and the personal protective equipment score were shown to have the greatest effect on HCW TB rates.  As scores in these three categories  142 increased, the probability of having a HCW with TB at that hospital decreased when controlling for the number of TB patients in each hospital. Qualitative analysis identified numerous barriers and facilitators to implementation and uptake of TB prevention and control in the hospital. Some of the barriers identified included inadequate isolation facilities, physical infrastructure limitations, inadequate supply of PPE and health systems constraints. Some of the facilitators identified included triaging TB suspects, updated infrastructure and appropriate use of PPE. Overall, this study reiterates that good infection control in hospitals is essential to prevent HCWs from acquiring TB.  More specifically, we found that environmental controls and adequate supply of and proper use of personal protective equipment are especially important.  These findings can be used to inform improved infection prevention and control recommendations and will allow for evidence-based prioritization of scarce resources to strengthen OH and IC in hospitals. Consistent application of IC principles across the healthcare spectrum is important not only in resource-strained settings such as South Africa, but in Canada as well with the emergence of pandemic influenza, Ebola virus disease, SARS, and other infections.  Recommendations related to objective 2 We recommend that TB IC assessments using a locally relevant standardized tool such as the one used for this study, be conducted regularly in high-burden regions such as South Africa. Data related to rates of TB among HCWs should be collected in a systematic manner and can be used as an indicator of TB IC in the facility. Corrective measures and  143 action plans generated from this process should include environmental, administrative, miscellaneous and PPE controls. There are many lost-cost, high-impact simple solutions that can be implemented to improve TB infection control in South African hospitals. For example, we recommend utilizing “champions” such as queue marshals or highly motivated HCWs to encourage improved infection prevention and control. Healthcare facilities should also develop interventions and training programs that focus on behaviour change and motivation of HCWs. Support from management and a positive workplace culture is also critical to improving IC practices. Special attention should be paid to improving OH services for HCWs including access to confidential, on-site TB and HIV services. Finally, health systems strengthening is required to ultimately eliminate TB transmission from healthcare facilities in South Africa.  Strengths and limitations Strengths and limitations of both parts of this study are discussed in detail in the previous chapter. Notably, our study does not rely on OH records or self-reporting of TB among HCWs and the findings presented here are therefore a robust estimate of TB incidence. This is also one of the few published studies to link HCW TB incidence to TB IC. This study had a large sample size of 32,039 HCWs and 229,157 TB patients from the general population and we identified 2,677 cases of TB among HCWs during a long study period of ten years.  Despite these strengths, we recognize that our estimates are only as good as the quality of the data entered in both PERSAL and ETR.Net and that this quality likely varies considerably between hospitals and regions. Another major limitation is that the TB IC workplace assessments were only conducted at one time point and therefore may  144 not adequately reflect the current situation in all hospitals. The assessments were however conducted in almost all public hospitals in the province ensuring that the findings are reflective of all facilities of varying sizes, geographic locations, staffing levels and TB patient loads.  Implications for policy and practice Epidemiology has played a major role in shaping public health policy and prevention, with examples spanning from Snow’s classic 19th century cholera studies leading to the removal of the closing of the Broad Street water pump in the United Kingdom (146) to the current American Heart Association recommendations for prevention and control of cardiovascular diseases in the United States (147). As mentioned previously, describing the problem is necessary before meaningful interventions can be formulated. Secondly, understanding risk factors is necessary to develop and implement appropriate prevention measures. Finally, the incidence rates of TB calculated for the purposes of this study can be used as a baseline to evaluate effectiveness of interventions and changes implemented.   If this work is successful, the true beneficiaries will be the healthcare workforce in South Africa and the populations they serve. As mentioned previously, the results and recommendations from this work have the potential to improve working conditions and to ultimately eliminate exposure to TB disease. Decision makers at the facility level and provincial level will also benefit from this research if they are able to use the findings to inform improved policies and to improve resource allocation. Finally, the international health community may also benefit from these results. These findings and conclusions  145 from South Africa are directly transferable to others regions that are also struggling to prevent morbidity and mortality associated with the global TB/HIV syndemic.   Conclusion Rates of TB are much higher among HCWs than the general population in Free State, South Africa. TB infection prevention and control policies and practices vary in Free State hospitals and poor implementation of controls is associated with higher rates of TB among staff. Additional attention and resources should be allocated to protect the healthcare workforce from TB in low-resource, high-burden settings.    146 References   1. World Health Organization. Global TB Report. 2015. Available from: http://www.who.int/tb/publications/global_report/gtbr15_annex02.pdf?ua=1 Accessed September 10, 2015.  2. Dheda K, Gumbo T, Gandhi NR, Murray M, Theron G, Udwadia Z, Migliori GB, Warren R. Global control of tuberculosis: from extensively drug-resistant to untreatable tuberculosis. The Lancet Respiratory Medicine. 2014 Apr 30;2(4):321-38.  3. Gandhi NR, Moll A, Sturm AW, Pawinski R, Govender T, Lalloo U, Zeller K, Andrews J, Friedland G. Extensively drug-resistant tuberculosis as a cause of death in patients co-infected with tuberculosis and HIV in a rural area of South Africa. The Lancet. 2006 Nov 10;368(9547):1575-80.  4. World Health Organization. Ten facts about TB. October 2015. Available from: http://www.who.int/features/factfiles/tuberculosis/en/ Accessed October 20, 2015.  5. UNAIDS. South Africa country profile 2014. Available from: http://www.unaids.org/en/regionscountries/countries/southafrica Accessed October 20, 2015.  6. Evans DK, Goldstein M, Popova A. Health-care worker mortality and the legacy of the Ebola epidemic. The Lancet Global Health. 2015 Jan 8;3(8):e439-40.  7. Lee N, Sung JJ. Nosocomial transmission of SARS. Current infectious disease reports. 2003 Dec 1;5(6):473-6.  8. Booth CM, Matukas LM, Tomlinson GA, Rachlis AR, Rose DB, Dwosh HA, Walmsley SL, Mazzulli T, Avendano M, Derkach P, Ephtimios IE. Clinical features and short-term outcomes of 144 patients with SARS in the greater Toronto area. Jama. 2003 Jun 4;289(21):2801-9.  9. Moore D, Gamage B, Bryce E, Copes R, Yassi A, BC Interdisciplinary Respiratory Protection Study Group. Protecting health care workers from SARS and other respiratory pathogens: organizational and individual factors that affect adherence to infection control guidelines. American Journal of Infection Control. 2005 Mar 31;33(2):88-96.  10. Gamage B, Moore D, Copes R, Yassi A, Bryce E, BC Interdisciplinary Respiratory Protection Study Group. Protecting health care workers from SARS  147 and other respiratory pathogens: a review of the infection control literature. American Journal of Infection Control. 2005 Mar 31;33(2):114-21.  11. Tawfik L, Kinoti, S.  The impact of HIV/AIDS on health systems and the health workforce in Sub-Saharan Africa.  Report for; Support for Analysis and Research in Africa Project. 2003 June.  12. Wilkinson D, Gilks CF. Increasing frequency of tuberculosis among staff in a South African district hospital: impact of the HIV epidemic on the supply side of health care. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1998 Sep 1;92(5):500-2.  13. Naidoo S, Jinabhai CC. TB in health care workers in KwaZulu-Natal, South Africa. The International Journal of Tuberculosis and Lung Disease. 2006 Jun 1;10(6):676-82.  14. O'Donnell MR, Jarand J, Loveday M, Padayatchi N, Zelnick J, Werner L, Naidoo K, Master I, Osburn G, Kvasnovsky C, Shean K. High incidence of hospital admissions with multidrug-resistant and extensively drug-resistant tuberculosis among South African health care workers. Annals of Internal Medicine. 2010 Oct 19;153(8):516-22.  15. Joshi R, Reingold AL, Menzies D, Pai M. Tuberculosis among health-care workers in low-and middle-income countries: a systematic review. PLoS Med. 2006 Dec 1;3(12):e494.  16. Baussano I, Nunn P, Williams B, Pivetta E, Bugiani M, Scano F. Tuberculosis among health care workers. Emerging Infectious Diseases. 2011 Mar;17(3):488.  17. Adams S, Ehrlich R, Baatjies R, van Zyl-Smit RN, Said-Hartley Q, Dawson R, Dheda K. Incidence of occupational latent tuberculosis infection in South African healthcare workers. European Respiratory Journal. 2015 Feb 19:ERJ-01384.  18. World Health Organization. World Health Report: Working together for health. 2006. Available from: http://www.who.int/whr/2006/en/ Accessed July 20, 2015.  19. National Department of Health, South Africa. Available from: http://www.health.gov.za/ Accessed July 20, 2015.  20. Gandhi NR, Weissman D, Moodley P, Ramathal M, Elson I, Kreiswirth BN, Mathema B, Shashkina E, Rothenberg R, Moll AP, Friedland G. Nosocomial transmission of extensively drug-resistant tuberculosis in a rural hospital in South Africa. Journal of Infectious Diseases. 2013 Jan 1;207(1):9-17.   148 21. World Health Organization. WHO policy on TB infection control in health-care facilities, congregate settings and households. 2009. Available at: http://apps.who.int/iris/bitstream/10665/44148/1/9789241598323_eng.pdf Accessed August 1, 2015.  22. Bock NN, Jensen PA, Miller B, Nardell E. Tuberculosis Infection Control in Resource-Limited Settings in the Era of Expanding HIV Care and Treatment. Journal of Infectious Diseases. 2007; 196, 108-13.  23. Kompala T, Shenoi SV, Friedland G. Transmission of tuberculosis in resource-limited settings. Current HIV/AIDS Reports. 2013; 10, 264-272.  24. Frieden TR, Brudney KF, Harries AD. Global tuberculosis: perspectives, prospects, and priorities. JAMA. 2014; 12(14), 1393-4.   25. Nelson KE, Williams CM. Infectious disease epidemiology: theory and practice. Jones & Bartlett Publishers; 2014.  26. Dooley KE, Chaisson RE. Tuberculosis and diabetes mellitus: convergence of two epidemics. The Lancet Infectious Diseases. 2009 Dec 31;9(12):737-46.  27. Maurya V, Vijayan V, Shah A. Smoking and tuberculosis: an association overlooked. The International Journal of Tuberculosis and Lung Disease. 2002 Nov 1;6(11):942-51.  28. Mason PH, Degeling C, Denholm J. Sociocultural dimensions of tuberculosis: an overview of key concepts. The International Journal of Tuberculosis and Lung Disease. 2015 Oct 1;19(10):1135-43.  29. Lönnroth K, Jaramillo E, Williams BG, Dye C, Raviglione M. Drivers of tuberculosis epidemics: the role of risk factors and social determinants. Social Science & Medicine. 2009 Jun 30;68(12):2240-6.  30. Tudor C, Van der Walt M, Margot B, Dorman SE, Pan WK, Yenokyan G, Farley JE. Tuberculosis among health care workers in KwaZulu-Natal, South Africa: a retrospective cohort analysis. BMC Public Health. 2004;14:891.  31. Galgalo T, Dalal S, Cain KP, Oeltmann J, Tetteh C, Kamau JG, Ijaz K. Tuberculosis risk among staff of a large public hospital in Kenya. The International Journal of Tuberculosis and Lung Disease. 2008; 12(8), 949-954.  32. McCarthy KM, Scott LE, Gous N, Tellie M, Venter WD, Stevens WS, Van Rie A. High incidence of latent tuberculous infection among South African health workers: an urgent call for action. The International Journal of Tuberculosis and Lung Disease. 2015 Jun 1;19(6):647-53. 2.    149 33. Eyob G, Gebeyhu M, Goshu S, Girma M, Lemma E, Fontanet A. Increase in tuberculosis incidence among the staff working at the Tuberculosis Demonstration and Training Centre in Addis Ababa, Ethiopia: a retrospective cohort study (1989-1998). International Journal of Tuberculosis and Lung Disease. 2002; 6(1):85-8.  34. Siegel J, Yassi A, Rau A, Buxton JA, Wouters E, Engelbrecht MC, Uebel KE, Nophale LE. Workplace interventions to reduce HIV and TB stigma among health care workers–Where do we go from here?. Global Public Health. 2015 Mar 13(ahead-of-print):1-3.  35. Farley JE, Tudor C, Mphahlele M, Franz K, Perrin NA, Dorman S, Van der Walt M. A national infection control evaluation of drug-resistant tuberculosis hospitals in South Africa. International Journal of Tuberculosis and Lung disease. 2012; 16(1), 82-89.  36. Welbel SF, French AL, Bush P, DeGuzman D, Weinstein RA. Protecting health care workers from tuberculosis: A 10-year experience. American Journal of Infection Control. 2009; 37(8), 668-673.  37. Reid MJA, Saito D, Nash D, Scardigli A, Casalini C, Howard AA. Implementation of tuberculosis infection control measures at HIV care and treatment sites in sub-Saharan Africa. International Journal of Tuberculosis and Lung disease. 2012; 16(12), 1605-1612.   38. Stogiu G, Arbore AS, Cojocariu V, Piana A, Ferrara G, Cirillo DM, Matteelli A, Castiglia P, Ditiu L, Spanevello A, Zellweger JP, Mihaescu T, Migliori GB. High risk of tuberculosis in health care workers in Romania. International Journal of Tuberculosis and Lung disease. 2008; 12(6), 606-611.  39. Arksey H, O'Malley L. Scoping studies: towards a methodological framework. International Journal of Social Research Methodology. 2005 Feb 1;8(1):19-32.  40. Chu H, Shih CJ, Lee YJ, Kuo SC, Hsu YT, Ou SM, Shih YN, Tarng DC, Li SY, Chen YT, Chen RC. Risk of tuberculosis among healthcare workers in an intermediate-burden country: A nationwide population study. Journal of Infection. 2014 Dec 31;69(6):525-32.  41. Dimitrova B, Hutchings A, Atun R, Drobniewski F, Marchenko G, Zakharova S, Fedorin I, Coker RJ. Increased risk of tuberculosis among health care workers in Samara Oblast, Russia: analysis of notification data. The International Journal of Tuberculosis and Lung Disease. 2005 Jan 1;9(1):43-8.  42. Buregyeya E, Mitchell EMH, Rutebemberwa E, Colebunders R, Criel B, Kiguli J, Nuwaha F. Acceptability of masking and patient separation to control nosocomial tuberculosis in Uganda: a qualitative study. Journal of Public Health. 2012; 20, 599-606.  150  43. Harries AD, Nyirenda TE, Banerjee A, Boeree MJ, Salaniponi FM. Tuberculosis in health care workers in Malawi. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1999 Jan 1;93(1):32-5.  44. Albuquerque da Costa P, Trajman A, Carvalho de Queiroz Mello F, Goudinho S, Monteiro Vieira Silva MA, Garret D, Ruffino-Netto A, Linue Kritski A. Administrative measures for preventing Mycobacterium tuberculosis infection among healthcare workers in a teaching hospital in Rio de Janeiro, Brazil. Journal of Hospital Infection. 2009; 72, 57-64.  45. Claassens MM, Van Schalkwyk C, du Toit E, Roest E, Lombard CJ, Enarson DA, Beyers N, Borgdorff MW. Tuberculosis in healthcare workers and infection control measures at primary healthcare facilities in South Africa. PloS One. 2013 Oct 2;8(10):e76272.  46. Tudor C, Van der Walt M, Margot B, Dorman SE, Pan WK, Yenokyan G, Farley JE. Tuberculosis among health care workers in KwaZulu-Natal, South Africa: a retrospective cohort analysis. BMC Public Health. 2014 Aug 30;14(1):891.  47. Kanyerere HS, Salaniponi FM. Tuberculosis in health care workers in a central hospital in Malawi. The International Journal of Tuberculosis and Lung Disease. 2003 May 1;7(5):489-92.  48. Harries AD, Hargreaves NJ, Gausi F, Kwanjana JH, Salaniponi FM. Preventing tuberculosis among health workers in Malawi. Bulletin of the World Health Organization. 2002 Jul;80(7):526-31.  49. Harries AD, Kamenya A, Namarika D, Msolomba IW, Salaniponi FM, Nyangulu DS, Nunn P. Delays in diagnosis and treatment of smear-positive tuberculosis and the incidence of tuberculosis in hospital nurses in Blantyre, Malawi. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1997 Jan 1;91(1):15-7.  50. Kiertiburanakul S, Suebsing S, Kehachindawat P, Apivanich S, Somsakul S, Sathapatayavongs B, Malathum K. Five-year prospective study of tuberculin skin testing among new healthcare personnel at a university hospital in Thailand. Journal of Hospital Infection. 2012 Feb 29;80(2):173-5.  51. Gopinath KG, Siddique S, Kirubakaran H, Shanmugam A, Mathai E, Chandy GM. Tuberculosis among healthcare workers in a tertiary-care hospital in South India. Journal of Hospital Infection. 2004 Aug 31;57(4):339-42.  52. Kwan SY, Yew WW, Chan SL. Nosocomial tuberculosis in hospital staff. The size of the problem in a Hong Kong chest hospital. Chinese Medical Journal. 1990 Nov;103(11):909-14.   151 53. Rao KG, Aggarwal AN, Behera D. Tuberculosis among physicians in training. The International Journal of Tuberculosis and Lung Disease. 2004 Nov 1;8(11):1392-4.  54. Menzies D, Joshi R, Pai M. Risk of tuberculosis infection and disease associated with work in health care settings [State of the Art Series. Occupational lung disease in high-and low-income countries, Edited by M. Chan-Yeung. Number 5 in the series]. The International Journal of Tuberculosis and Lung Disease. 2007 Jun 1;11(6):593-605.  55. Menzies D, Fanning A, Yuan L, Fitzgerald M. Tuberculosis among health care workers. New England Journal of Medicine. 1995 Jan 12;332(2):92-8.  56. Brouwer M, Coelho E, das Dores Mosse C, van Leth F. Implementation of tuberculosis infection prevention and control in Mozambican health care facilities. The International Journal of Tuberculosis and Lung Disease. 2015 Jan 1;19(1):44-9.  57. Kanjee Z, Catterick K, Moll AP, Amico KR, Friedland GH. Tuberculosis infection control in rural South Africa: survey of knowledge, attitude and practice in hospital staff. Journal of Hospital Infection. 2011 Dec 31;79(4):333-8.  58. Kanjee Z, Amico KR, Li F, Mbolekwa K, Moll AP, Friedland GH. Tuberculosis infection control in a high drug-resistance setting in rural South Africa: information, motivation, and behavioral skills. Journal of infection and public health. 2012 Feb 29;5(1):67-81.  59. He GX, van den Hof S, van der Werf MJ, Wang GJ, Ma SW, Zhao DY, Hu YL, Yu SC, Borgdorff MW. Infection control and the burden of tuberculosis infection and disease in health care workers in china: a cross-sectional study. BMC infectious diseases. 2010 Oct 28;10(1):313.  60. Luksamijarulkul P, Supapvanit C, Loosereewanich P, Aiumlaor P. Risk assessment towards tuberculosis among hospital personnel: administrative control, risk exposure, use of protective barriers and microbial air quality.  61. Pai M, Kalantri S, Aggarwal AN, Menzies D, Blumberg HM. Nosocomial tuberculosis in India. Emerging infectious diseases. 2006 Sep;12(9):1311.  62. Woith WM, Volchenkov G, Larson JL. Russian health care workers’ knowledge of tuberculosis and infection control. International Journal of Tuberculosis and Lung disease. 2010; 14(11), 1489-1492.  63. Checkoway H, Eisen EA. Developments in occupational cohort studies. Epidemiologic Reviews. 1998 Jan 1;20(1):100-11.   152 64. Wu W. Occupational cancer epidemiology in the People's Republic of China. Journal of Occupational and Environmental Medicine. 1988 Dec 1;30(12):968-74.  65. Rinsky RA, Smith AB, Hornung R, Filloon TG, Young RJ, Okun AH, Landrigan PJ. Benzene and leukemia. New England Journal of Medicine. 1987 Apr 23;316(17):1044-50.  66. Eisen EA, Wegman DH, Kriebel D, Woskie SR, Hu X. An epidemiologic approach to the study of acute reversible health effects in the workplace. Epidemiology. 1991 Jul 1:263-70.  67. Dimich-Ward H, Hertzman C, Teschke K, Hershler R, Marion SA, Ostry A, Kelly S. Reproductive effects of paternal exposure to chlorophenate wood preservatives in the sawmill industry. Scandinavian Journal of Work, Environment & Health. 1996 Aug 1:267-73.  68. Newcombe HB. Handbook of record linkage: methods for health statistical studies, administration and business. Oxford, UK: Oxford University Press, 1988.  69. Dusetzina SB, Tyree S, Meyer AM, Meyer A, Green L, Carpenter WR. Agency for Healthcare Research and Quality. Linking data for health services research: A framework and instructional guide.  2014. Available from: http://www.ncbi.nlm.nih.gov/books/NBK253313/pdf/Bookshelf_NBK253313.pdf Accessed October 1, 2014.  70. Fellegi IP, Sunter AB. A theory for record linkage. Journal of the American Statistical Association. 1969 Dec 1;64(328):1183-210.  71. Parker L, Smith J, Dickinson H, Binks K, Scott L, McElvenny D, Jones S, Wakeford R. The creation of a database of children of workers at a nuclear facility: an exercise in record linkage. Applied Occupational and Environmental Hygiene. 1997 Jan 1;12(1):40-5.  72. Pukkala E, Aspholm R, Auvinen A, Eliasch H, Gundestrup M, Haldorsen T, Hammar N, Hrafnkelsson J, Kyyrönen P, Linnersjö A, Rafnsson V. Incidence of cancer among Nordic airline pilots over five decades: occupational cohort study. BMJ. 2002 Sep 14;325(7364):567.  73. Boffetta P, Gridley G, Gustavsson P, Brennan P, Blair A, Ekström AM, Fraumeni JF. Employment as butcher and cancer risk in a record-linkage study from Sweden. Cancer Causes & Control. 2000 Jul 1;11(7):627-33.  74. Méray N, Reitsma JB, Ravelli AC, Bonsel GJ. Probabilistic record linkage is a valid and transparent tool to combine databases without a patient identification number. Journal of Clinical Epidemiology. 2007 Sep 30;60(9):883-e1.   153 75. World Health Organization. Counting health workers: Definitions, data, methods and global results. 2007. Available from: http://www.who.int/hrh/documents/counting_health_workers.pdf Accessed September 6, 2014.  76. World Health Organization. International Union Against Tuberculosis and Lung Disease, Royal Netherlands Tuberculosis Association. Revised international definitions in tuberculosis control. The International Journal of Tuberculosis and Lung Disease. 2001 Mar;5(3):213-5.  77. Rothman KJ, Greenland S, Lash TL, editors. Modern epidemiology. Lippincott Williams & Wilkins; 2008.  78. The British Medical Journal. Quantifying disease in populations. 2015. Available from: http://www.bmj.com/about-bmj/resources-readers/publications/epidemiology-uninitiated/2-quantifying-disease-populations Accessed September 5, 2015.  79. South African National Department of Health. Tuberculosis Control Programme Practical Guidelines. 2004. Available from: http://www.kznhealth.gov.za/chrp/documents/Guidelines/Guidelines%20National/Tuberculosis/SA%20TB%20Guidelines%202004.pdf Accessed September 5, 2015.  80. Statistics South Africa. Mid-year population estimates. Available from: http://www.statssa.gov.za/?page_id=1866&PPN=P0302&SCH=6334  Accessed September 5, 2015.  81. Human Sciences Research Council of South Africa. South Africa National HIV Prevalence, Incidence and Behaviour Survey. Available from: 2002: http://curation.hsrc.ac.za/Dataset-169-datafiles.phtml, 2005: http://curation.hsrc.ac.za/Dataset-214-datafiles.phtml, 2008: http://curation.hsrc.ac.za/Dataset-314-datafiles.phtml Accessed September 5, 2015.  82. Clark DE. Practical introduction to record linkage for injury research. Injury Prevention. 2004 Jun 1;10(3):186-91.  83. Centers for Disease Prevention and Control. Record linkage concepts. Slides adapted from training materials developed by CDC–NPCR Faculty.  Available from: http://www.hcup-us.ahrq.gov/datainnovations/raceethnicitytoolkit/or19.jsp Accessed September 15, 2014.  84. Black PE. ‘NYSIIS’. In: US National Institute of Standards and Technology. Dictionary of algorithms and data structures. Gaithersburg, MD, USA: NIST,  154 2009. Available from: www.nist.gov/dads/HTML/nysiis.html Accessed September 15, 2014.  85. NIH Fogarty International Center. Implementation Science Information and Resources. 2015. Available from: http://www.fic.nih.gov/researchtopics/pages/implementationscience.aspx. Accessed March 2016.  86. Rogers EM. Diffusion of innovations. 5th edition. New York, London: Free Press; 2003.  87. Bryce EA, Scharf S, Walker M, Walsh A. The infection control audit: the standardized audit as a tool for change. American Journal of Infection Control. 2007 May 31;35(4):271-83.  88. Malik RE, Cooper RA, Griffith CJ. Use of audit tools to evaluate the efficacy of cleaning systems in hospitals. American Journal of Infection Control. 2003 May 31;31(3):181-7.  89. Yassi A, Nophale LE, Dybka L, Bryce E, Kruger W, Spiegel J. Building capacity to secure healthier and safer working conditions for healthcare workers: A South African-Canadian collaboration. International journal of occupational and environmental health. 2009 Oct 1;15(4):360-9.  90. Yassi A, Zungu M, Spiegel J, Kistnasamy B, Lockhart K, Jones D, O'Hara LM, Nophale L, Bryce EA, Darwin L. Protecting health workers from infectious disease transmission: An exploration of a Canadian-South African partnership of partnerships. Globalization and Health. 2015. In Press.  91. Fereday J, Muir-Cochrane E. Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. International Journal of Qualitative Methods. 2008 Nov 26;5(1):80-92.  92. Braun V, Clarke V. Using thematic analysis in psychology. Qualitative Research in Psychology. 2006 Jan 1;3(2):77-101.  93. Sandelowski M. Telling stories: Narrative approaches in qualitative research. Image: Journal of Nursing Scholarship. 1991 Aug 31;23(3):161-6.  94. Wouters E, et al. The long road to universal antiretroviral treatment coverage in South Africa. Future Virology 6.7. 2011: 801-812.  95. Wahome E, Makori L, Gikera M, Wafula J, Chakaya J, Edginton ME, Kumar AMV. Tuberculosis treatment outcomes among hospital workers at a public teaching and national referral hospital in Kenya. Public Health Action. 2013; 3(4), 323-7.  155  96. Klimuk D, Hurevich H, Harries AD, Babrukevich A, Kremer K, Van den Bergh R, Acosta CD, Astrauko A, Skrahina A. Tuberculosis in health care workers in Belarus. Public Health Action. 2014; 4(2), 29-36.  97. Zungu M, Malotle M. Do we know enough to prevent occupationally acquired tuberculosis in healthcare workers? OH Southern Africa. 2011;17-12.  98. World Health Organization, International Labour Office, UNAIDS the Joint United Nations Programme on HIV/AIDS. The joint WHO-UNAIDS policy guidelines on improving health workers’ access to HIV and TB prevention, treatment, care and support services. Geneva WHO, 2011. Available from: Available from: http://www.who.int/occupational_health/publications/hiv_tb_guidelines/en/ Accessed July 15, 2014.  99. Naidoo A, Naidoo SS, Gathiram P, Lalloo UG. Tuberculosis in medical doctors - a study of personal experiences and attitudes. South African Medical Journal. 2013;301(3):176-189.   100. O’Hara NN, Roy L, O’Hara LM, Spiegel J, Lynd L, Fitzgerald JM, Yassi A, Nophale LE, Marra C. Active tuberculosis screening preferences of healthcare workers in South Africa: A best-worst scaling experiment. PloS one. 2015;10(7):e0133304.  101. Dilraj A, Bristow CC, Connolly C, Margot B, Dlamini S, Podewils LJ. Validation of sputum smear results in the Electronic TB Register for the management of tuberculosis, South Africa. The International Journal of Tuberculosis and Lung Disease.  2013;17(10):1317-1321.  102. Friedland G. Protecting health care workers: the critical role of airborne infection control [Editorial]. The International Journal of Tuberculosis and Lung Disease. 2008 Jun 1;12(6):585-.  103. Dharmadhikari AS, Mphahlele M, Stoltz A, Venter K, Mathebula R, Masotla T, Lubbe W, Pagano M, First M, Jensen PA, van der Walt M. Surgical face masks worn by patients with multidrug-resistant tuberculosis: impact on infectivity of air on a hospital ward. American journal of respiratory and critical care medicine. 2012 May 15;185(10):1104-9.  104. Zingg W, Holmes A, Dettenkofer M, Goetting T, Secci F, Clack L, Allegranzi B, Magiorakos AP, Pittet D. Hospital organisation, management, and structure for prevention of health-care-associated infection: a systematic review and expert consensus. The Lancet. Infectious Diseases. 2015; 15(2), 212-224.    156 105. Pantle AC, Fitzpatrick KR, McLaws ML, Hughes CF. A statewide approach to systematising hand hygiene behaviour in hospitals: Clean hands save lives, Part I. Medical Journal of Australia. 2009 Oct 19;191(8):S8.  106. Fennelly KP, Iseman MD. Health care workers and tuberculosis: the battle of a century [Editorial]. The International Journal of Tuberculosis and Lung Disease. 1999 May 1;3(5):363-4.  107. Cohen M, Village J, Ostry AS, Ratner PA, Cvitkoyich Y, Yassi A. Workload as a determinant of staff injury in intermediate care. International Journal of Occupational and Environmental Health. 2004 Oct 1;10(4):375-83.  108. Engelbrecht MC, Yassi A, Spiegel JM, Van Rensburg AJ, O'Hara LM, Bryce EA, Nophale LE, Rau A. Tuberculosis and blood-borne infectious diseases: workplace conditions and practices of healthcare workers at three public hospitals in the Free State: original research. Southern African Journal of Infectious Diseases. 2015;30(1):23-8.  109. Statistics South Africa. 2011. Mid-year population estimates. Available from: http://www.statssa.gov.za/publications/P0302/P03022011.pdf Accessed January 20, 2015.  110. Simbayi LC, Shisana O, Rehle T, Onoya D, Jooste D, Zungu N, Labadarios D, Zuma K. South Africa National HIV Prevalence, Incidence and Behaviour Survey, 2012. Health Sciences Research Council. Available at: http://www.hsrc.ac.za/en/research-data/ktree-doc/15031 Accessed January 20, 2015.  111. Dictionary.com. Apartheid. Accessed September 2015. Accessed July 10, 2015.  112. Galtung J. Violence, peace, and peace research. Journal of peace research. 1969 Jan 1;6(3):167-91.  113. van Rensburg HC, editor. Health and health care in South Africa. Van Schaik Publishers; 2004.  114. World Health Organization. The World Health Report 1995. Available from: http://www.who.int/whr/1995/media_centre/en/whr95_press_release_en.pdf Accessed July 10, 2015.  115. Farmer P, Walton D, Tarter L. Infections and inequalities. Global Change and Human Health. 2000 Dec 1;1(2):94-109.  116. Bhargava AB, Oxlade O, Pai MD. Undernutrition and the incidence of tuberculosis in India: National and subnational estimates of the population  157 attributable fraction related to undernutrition. National Medical Journal of India. 2014 May 1;27:128-33.  117. Fox GJ, Lee RS, Lucas M, Khan FA, Proulx JF, Hornby K, Jung S, Benedetti A, Behr MA, Menzies D. Inadequate Diet Is Associated with Acquiring Mycobacterium tuberculosis Infection in an Inuit Community. A Case–Control Study. Annals of the American Thoracic Society. 2015 Aug;12(8):1153-62.  118. Bond P, Dor G. Uneven outcomes and neoliberalism in Africa. Discussion paper for the Regional Network for Equity In Health, Equity in Southern Africa (Equinet) 2003; 2003.  119. Syndemics Prevention Network, CDC. Syndemics Overview.http://www.cdc.gov/syndemics/overview.htm. Updated January 30, 2008. Accessed February 25, 2015.  120. Singer M, Clair S. Syndemics and public health: reconceptualizing disease in bio-social context. Medical anthropology quarterly. 2003 Dec 1:423-41.  121. van Rie A et al. Childhood tuberculosis in an urban population in South Africa: burden and risk factor. Archives of Disease in Childhood, 1999, 80:433–437.  122. WHO. Closing the gap in a generation: health equity through action on the social determinants of health.  Final report of the commission on social determinants of health Geneva: World Health Organization, 2008.  123. Ratner PA, Sawatzky R. Health status, preventive behaviour and risk factors among female nurses. Statistics Canada; 2009 Aug.  124. Moodley PP, Bachmann MO. Inequity in OH services for government hospital workers in South Africa. Occupational Medicine. 2002 Oct 1;52(7):393-9.  125. Suri A, Gan K, Carpenter S. Voices from the field: perspectives from community health workers on health care delivery in rural KwaZulu-Natal, South Africa. Journal of Infectious Diseases. 2007 Dec 1;196(Supplement 3):S505-11.  126. International Labour Organization. 2001. The ILO Code of Practice on HIV/AIDS and the World of Work. Available from: http://www.ilo.org/aids/Publications/WCMS_113783/lang--en/index.htm Accessed September 25, 2014.  127. Khan R, Yassi A, Engelbrecht MC, Nophale L, van Rensburg AJ, Spiegel J. Barriers to HIV counselling and testing uptake by health workers in three public hospitals in Free State Province, South Africa. AIDS care. 2015 Feb 1;27(2):198-205.   158 128. Murphy J, Hatfield J, Afsana K, Neufeld V. Making a Commitment to Ethics in Global Health Research Partnerships: A Practical Tool to Support Ethical Practice. Journal of bioethical inquiry. 2015 Mar 1;12(1):137-46.  129. Nuffield Council on Bioethics Report, 2002. Available from: http://www.nuffieldbioethics.org/research-developing-countries  Accessed September 30, 2015.  130. World Health Organization, 2000. Operational Guidelines for Ethics Committees that Review Biomedical Research. Available from: http://www.who.int/tdr/publications/training-guideline-publications/operational-guidelines-ethics-biomedical-research/en http://www.who.int/tdr/publications/publications/pdf/ethics.pdf Accessed September 30, 2015.  131. National Institute of Allergy and Infectious Diseases. ClinReg website. Available from: https://clinregs.niaid.nih.gov/index.php Accessed September 30, 2015.  132. Bhutta ZA. Ethics in international health research: a perspective from the developing world. Bulletin of the World Health Organization. 2002 Jan;80(2):114-20.  133. Tindana PO, Singh JA, Tracy CS, Upshur RE, Daar AS, Singer PA, Frohlich J, Lavery JV. Grand challenges in global health: community engagement in research in developing countries. PLoS Med. 2007 Sep 1;4(9):e273.  134. Yassi A, Breilh J, Dharamsi S, Lockhart K, Spiegel JM. The ethics of ethics reviews in global health research: Case studies applying a new paradigm. Journal of Academic Ethics. 2013;11(2):83-101.  135. Wood B, Burchell AN, Escott N, Little J, Maar M, Ogilvie G, Severini A, Bishop L, Morrisseau K, Zehbe I. Using community engagement to inform and implement a community-randomized controlled trial in the Anishinaabek cervical cancer screening study. Frontiers in oncology. 2014;4.  136. Lairumbi GM, Michael P, Fitzpatrick R, English MC. Ethics in practice: the state of the debate on promoting the social value of global health research in resource poor settings particularly Africa. BMC Medical Ethics. 2011 Nov 15;12(1):22.  137. Menzies D. Finishing What They Started: When Do Researchers’ Responsibilities End?. American journal of respiratory and critical care medicine. 2015 May 1;191(9):977-8.   159 138. Dharamsi S, Osei-Twum J, Whiteman M. Socially responsible approaches to international electives and global health outreach. Medical Education. 2011;45(5):530-531.  139. Crump JA, Sugarman J. Ethics and best practice guidelines for training experiences in global health. The American journal of tropical medicine and hygiene. 2010 Dec 6;83(6):1178-82.  140. Constitution of the Republic of South Africa, 1996 (Act No. 108 of 1996).  141. National Department of Health South Africa. Draft national infection prevention and control policy for TB, MDR-TB and XDR-TB. Pretoria, South Africa: National Department of Health, South Africa, 2007.  142. National Department of Health. National Tuberculosis Management Guidelines 2014. Pretoria, South Africa: National Department of Health, South Africa.  143. Department of Labour.  OH and Safety Act, 1993 (Act No. 85 of 1993) and its Hazardous Biological Agents Regulations.  144. Von Delft A, Dramowski A, Khosa C, Kotze K, Lederer P, Mosidi T, Peters JA, Smith J, van der Westhuizen HM, von Delft D, Willems B, Bates M, Craig G, Maeurer M, Marais BJ, Mwaba P, Nunes EA, Nyirenda T, Oliver M, Zumla A. Why healthcare workers are sick of TB. International Journal of Infectious Diseases. 2015;32:147-151.  145. Adams S, Ehrlich R, Ismail N, Quail Z, Jeebhay MF. OH challenges facing the Department of Health: protecting employees against tuberculosis and caring for former mineworkers with OH disease: service delivery. South African Health Review. 2012:67-82.  146. Paneth N. Assessing the contributions of John Snow to epidemiology: 150 years after the broad street pump handle. Epidemiology. 2004;15:514-516.  147. Grundy SM, Balady GJ, Criqui MH et al. Primary evaluation of coronary heart disease: Guidance from Framingham: A statement for healthcare professionals from the AHA task force on risk reduction: American Heart Association. Circulation. 1998;97:1876-1887.    160 Appendices   Appendix A- ICAP TB Infection Control Practices Facility Assessment      161       162       163      164       165        166 Appendix B- Key Statistical Outputs   Do healthcare workers with TB differ from those without TB?  Univariate Analyses- Cluster Adjusted  Excluding 1920s & 1990s birth Decades because no TB cases ============================================================= BirthDecade TBCase BirthDecade No Yes 1930s 107 1 1940s 2179 145 1950s 4830 606 1960s 6850 1167 1970s 7982 642 1980s 7004 116  Probability of TB Prob LB UB 1930s 0.0102 0.0018 0.0585 1980s 0.0183 0.0145 0.0231 1940s 0.0626 0.0506 0.0775 1970s 0.0781 0.0665 0.0918 1950s 0.1115 0.0949 0.1310 1960s 0.1451 0.1247 0.1688  Log relative risk between groups contrast estimate SE z.ratio p.value 1980s - 1930s 0.585 0.8941 0.654 0.9867 1940s - 1930s 1.817 0.8924 2.036 0.3218 1940s - 1980s 1.232 0.1262 9.758 <.0001 1970s - 1930s 2.038 0.8900 2.290 0.1981 1970s - 1980s 1.453 0.1016 14.300 <.0001 1970s - 1940s 0.221 0.0924 2.393 0.1585 1950s - 1930s 2.394 0.8903 2.689 0.0775 1950s - 1980s 1.809 0.1032 17.521 <.0001 1950s - 1940s 0.577 0.0924 6.241 <.0001 1950s - 1970s 0.356 0.0574 6.190 <.0001 1960s - 1930s 2.657 0.8899 2.986 0.0337 1960s - 1980s 2.072 0.0990 20.940 <.0001 1960s - 1940s 0.840 0.0881 9.540 <.0001 1960s - 1970s 0.619 0.0496 12.486 <.0001 1960s - 1950s 0.264 0.0502 5.253 <.0001     Relative Risk between groups  167 Estimate lwr upr 1980s - 1930s 1.79 0.155 20.84 1940s - 1930s 6.15 0.532 71.10 1940s - 1980s 3.43 2.425 4.85 1970s - 1930s 7.68 0.669 88.13 1970s - 1980s 4.28 3.236 5.65 1970s - 1940s 1.25 0.968 1.61 1950s - 1930s 10.95 0.953 125.83 1950s - 1980s 6.10 4.598 8.10 1950s - 1940s 1.78 1.382 2.29 1950s - 1970s 1.43 1.219 1.67 1960s - 1930s 14.26 1.242 163.61 1960s - 1980s 7.94 6.055 10.42 1960s - 1940s 2.32 1.820 2.95 1960s - 1970s 1.86 1.621 2.13 1960s - 1950s 1.30 1.134 1.49 ============================================================= Race TBCase Race No Yes ASIAN 138 2 BLACK 22289 2546 COLOURED 1211 59 WHITE 5314 70  Probability of TB Prob LB UB WHITE 0.0162 0.0123 0.0213 ASIAN 0.0186 0.0045 0.0767 COLOURED 0.0581 0.0433 0.0780 BLACK 0.1091 0.0941 0.1264  Log relative risk between groups contrast estimate SE z.ratio p.value ASIAN - WHITE 0.136 0.732 0.186 0.9977 COLOURED - WHITE 1.278 0.177 7.237 <.0001 COLOURED - ASIAN 1.141 0.734 1.556 0.4041 BLACK - WHITE 1.908 0.122 15.666 <.0001 BLACK - ASIAN 1.772 0.722 2.455 0.0672 BLACK - COLOURED 0.630 0.132 4.756 <.0001  Relative Risk between groups Estimate lwr upr ASIAN - WHITE 1.15 0.187 7.02 COLOURED - WHITE 3.59 2.318 5.56 COLOURED - ASIAN 3.13 0.509 19.27 BLACK - WHITE 6.74 4.984 9.11 BLACK - ASIAN 5.88 0.984 35.12 BLACK - COLOURED 1.88 1.352 2.61 ============================================================= Sex  168 TBCase Sex No Yes FEMALE 19872 1574 MALE 9080 1103  Probability of TB Prob LB UB FEMALE 0.0809 0.0687 0.0953 MALE 0.1143 0.0967 0.1350  Log relative risk between groups contrast estimate SE z.ratio p.value MALE - FEMALE 0.345 0.0393 8.79 <.0001  Relative Risk between groups Estimate lwr upr MALE - FEMALE 1.41 1.31 1.53 ============================================================= FacilityType TBCase FacilityType No Yes Clinic 6463 831 Hospital 16364 1549 Other 6125 297  Probability of TB Prob LB UB Other 0.0462 0.0211 0.1011 Hospital 0.0942 0.0804 0.1103 Clinic 0.1139 0.0525 0.2471  Log relative risk between groups contrast estimate SE z.ratio p.value Hospital - Other 0.713 0.408 1.747 0.1879 Clinic - Other 0.903 0.563 1.602 0.2448 Clinic - Hospital 0.190 0.403 0.471 0.8851  Relative Risk between groups Estimate lwr upr Hospital - Other 2.04 0.791 5.26 Clinic - Other 2.47 0.667 9.12 Clinic - Hospital 1.21 0.474 3.08 ============================================================= Occupational.Group TBCase Occupational.Group No Yes Administration 5472 407 Allied Health 2432 108 Doctor / Medical 4593 282 Nursing 10702 1113 Support Services 5753 767  169  Probability of TB Prob LB UB Allied Health 0.0478 0.0375 0.0611 Doctor / Medical 0.0538 0.0442 0.0656 Administration 0.0831 0.0692 0.0997 Nursing 0.0990 0.0840 0.1169 Support Services 0.1130 0.0955 0.1336  Log relative risk between groups contrast estimate SE z.ratio p.value Doctor / Medical - Allied Health 0.118 0.1133 1.04 0.8367 Administration - Allied Health 0.551 0.1079 5.11 <.0001 Administration - Doctor / Medical 0.434 0.0793 5.47 <.0001 Nursing - Allied Health 0.728 0.1005 7.24 <.0001 Nursing - Doctor / Medical 0.610 0.0676 9.03 <.0001 Nursing - Administration 0.176 0.0586 3.00 0.0224 Support Services - Allied Health 0.859 0.1029 8.34 <.0001 Support Services - Doctor / Medical 0.741 0.0707 10.48 <.0001 Support Services - Administration 0.308 0.0628 4.89 <.0001 Support Services - Nursing 0.131 0.0473 2.78 0.0436  Relative Risk between groups Estimate lwr upr Doctor / Medical - Allied Health 1.13 0.829 1.53 Administration - Allied Health 1.74 1.298 2.32 Administration - Doctor / Medical 1.54 1.246 1.91 Nursing - Allied Health 2.07 1.579 2.71 Nursing - Doctor / Medical 1.84 1.534 2.21 Nursing - Administration 1.19 1.018 1.40 Support Services - Allied Health 2.36 1.789 3.11 Support Services - Doctor / Medical 2.10 1.734 2.54 Support Services - Administration 1.36 1.148 1.61 Support Services - Nursing 1.14 1.004 1.30 ============================================================= ExposureYearsGroup TBCase ExposureYearsGroup No Yes [0 - 1) 5418 294 [1 - 5) 7596 529 [5 - 10) 4331 359 [10 - 15) 1921 536 [15 - 20) 3084 498 [20+) 6602 461  Probability of TB Prob LB UB [0 - 1) 0.0662 0.0545 0.0804 [20+) 0.0668 0.0559 0.0798 [1 - 5) 0.0711 0.0598 0.0846 [5 - 10) 0.0806 0.0671 0.0967  170 [15 - 20) 0.1372 0.1152 0.1635 [10 - 15) 0.2245 0.1887 0.2672  Log relative risk between groups contrast estimate SE z.ratio p.value [20+) - [0 - 1) 0.0088 0.0784 0.112 1.0000 [1 - 5) - [0 - 1) 0.0715 0.0746 0.958 0.9311 [1 - 5) - [20+) 0.0627 0.0644 0.972 0.9267 [5 - 10) - [0 - 1) 0.1962 0.0815 2.407 0.1538 [5 - 10) - [20+) 0.1874 0.0711 2.635 0.0890 [5 - 10) - [1 - 5) 0.1247 0.0686 1.819 0.4536 [15 - 20) - [0 - 1) 0.7287 0.0771 9.456 <.0001 [15 - 20) - [20+) 0.7199 0.0651 11.066 <.0001 [15 - 20) - [1 - 5) 0.6572 0.0630 10.438 <.0001 [15 - 20) - [5 - 10) 0.5325 0.0695 7.667 <.0001 [10 - 15) - [0 - 1) 1.2213 0.0761 16.052 <.0001 [10 - 15) - [20+) 1.2125 0.0646 18.763 <.0001 [10 - 15) - [1 - 5) 1.1498 0.0618 18.591 <.0001 [10 - 15) - [5 - 10) 1.0251 0.0682 15.028 <.0001 [10 - 15) - [15 - 20) 0.4926 0.0625 7.886 <.0001  Relative Risk between groups Estimate lwr upr [20+) - [0 - 1) 1.01 0.807 1.26 [1 - 5) - [0 - 1) 1.07 0.869 1.33 [1 - 5) - [20+) 1.06 0.886 1.28 [5 - 10) - [0 - 1) 1.22 0.965 1.53 [5 - 10) - [20+) 1.21 0.985 1.48 [5 - 10) - [1 - 5) 1.13 0.932 1.38 [15 - 20) - [0 - 1) 2.07 1.664 2.58 [15 - 20) - [20+) 2.05 1.707 2.47 [15 - 20) - [1 - 5) 1.93 1.613 2.31 [15 - 20) - [5 - 10) 1.70 1.398 2.08 [10 - 15) - [0 - 1) 3.39 2.731 4.21 [10 - 15) - [20+) 3.36 2.797 4.04 [10 - 15) - [1 - 5) 3.16 2.648 3.77 [10 - 15) - [5 - 10) 2.79 2.296 3.38 [10 - 15) - [15 - 20) 1.64 1.370 1.95 ============================================================= Multivariate Analysis ------------------ BirthDecade Probability of TB Prob LB UB 1930s 0.0058 0.0010 0.0338 1980s 0.0081 0.0049 0.0133 1940s 0.0324 0.0198 0.0531 1970s 0.0309 0.0193 0.0494 1950s 0.0573 0.0357 0.0919 1960s 0.0587 0.0368 0.0938 Log relative risk between groups  171 contrast estimate SE z.ratio p.value 1980s - 1930s 0.3260 0.8718 0.374 0.9991 1940s - 1930s 1.7186 0.8701 1.975 0.3566 1940s - 1980s 1.3927 0.1352 10.300 <.0001 1970s - 1930s 1.6705 0.8675 1.926 0.3863 1970s - 1980s 1.3445 0.1032 13.033 <.0001 1970s - 1940s -0.0482 0.1003 -0.480 0.9969 1950s - 1930s 2.2878 0.8676 2.637 0.0885 1950s - 1980s 1.9618 0.1127 17.402 <.0001 1950s - 1940s 0.5691 0.0927 6.138 <.0001 1950s - 1970s 0.6173 0.0676 9.137 <.0001 1960s - 1930s 2.3129 0.8670 2.668 0.0818 1960s - 1980s 1.9869 0.1046 18.998 <.0001 1960s - 1940s 0.5943 0.0910 6.530 <.0001 1960s - 1970s 0.6424 0.0543 11.834 <.0001 1960s - 1950s 0.0251 0.0533 0.471 0.9971  Relative Risk between groups Estimate lwr upr 1980s - 1930s 1.385 0.127 15.13 1940s - 1930s 5.577 0.513 60.62 1940s - 1980s 4.026 2.778 5.83 1970s - 1930s 5.315 0.492 57.37 1970s - 1980s 3.836 2.891 5.09 1970s - 1940s 0.953 0.724 1.25 1950s - 1930s 9.853 0.913 106.37 1950s - 1980s 7.112 5.221 9.69 1950s - 1940s 1.767 1.370 2.28 1950s - 1970s 1.854 1.540 2.23 1960s - 1930s 10.104 0.938 108.89 1960s - 1980s 7.293 5.475 9.72 1960s - 1940s 1.812 1.412 2.33 1960s - 1970s 1.901 1.638 2.21 1960s - 1950s 1.025 0.886 1.19 ------------------ Race Probability of TB Prob LB UB WHITE 0.0116 0.0073 0.0185 ASIAN 0.0121 0.0029 0.0511 COLOURED 0.0337 0.0207 0.0546 BLACK 0.0615 0.0407 0.0928  Log relative risk between groups contrast estimate SE z.ratio p.value ASIAN - WHITE 0.0454 0.708 0.064 0.9999 COLOURED - WHITE 1.0646 0.177 6.004 <.0001 COLOURED - ASIAN 1.0193 0.712 1.432 0.4790 BLACK - WHITE 1.6674 0.123 13.516 <.0001 BLACK - ASIAN 1.6220 0.700 2.318 0.0940 BLACK - COLOURED 0.6027 0.133 4.532 <.0001  172  Relative Risk between groups Estimate lwr upr ASIAN - WHITE 1.05 0.180 6.07 COLOURED - WHITE 2.90 1.867 4.50 COLOURED - ASIAN 2.77 0.474 16.21 BLACK - WHITE 5.30 3.901 7.20 BLACK - ASIAN 5.06 0.891 28.77 BLACK - COLOURED 1.83 1.313 2.54 ------------------ Sex Probability of TB Prob LB UB FEMALE 0.0189 0.0110 0.0327 MALE 0.0285 0.0165 0.0493  Log relative risk between groups contrast estimate SE z.ratio p.value MALE - FEMALE 0.41 0.0431 9.51 <.0001  Relative Risk between groups Estimate lwr upr MALE - FEMALE 1.51 1.38 1.64 ------------------ FacilityType Probability of TB Prob LB UB Other 0.0188 0.0087 0.0409 Hospital 0.0239 0.0148 0.0384 Clinic 0.0279 0.0128 0.0611  Log relative risk between groups contrast estimate SE z.ratio p.value Hospital - Other 0.236 0.333 0.710 0.7575 Clinic - Other 0.395 0.459 0.859 0.6662 Clinic - Hospital 0.158 0.331 0.478 0.8815  Relative Risk between groups Estimate lwr upr Hospital - Other 1.27 0.585 2.74 Clinic - Other 1.48 0.511 4.31 Clinic - Hospital 1.17 0.544 2.52 ------------------ Occupational.Group Probability of TB Prob LB UB Allied Health 0.0214 0.0121 0.0379 Doctor / Medical 0.0182 0.0104 0.0316 Administration 0.0241 0.0139 0.0418 Nursing 0.0265 0.0153 0.0458 Support Services 0.0273 0.0158 0.0474  173  Log relative risk between groups contrast estimate SE z.ratio p.value Doctor / Medical - Allied Health -0.1644 0.1140 -1.442 0.6002 Administration - Allied Health 0.1180 0.1099 1.074 0.8201 Administration - Doctor / Medical 0.2824 0.0826 3.419 0.0057 Nursing - Allied Health 0.2119 0.1037 2.044 0.2452 Nursing - Doctor / Medical 0.3763 0.0752 5.004 <.0001 Nursing - Administration 0.0939 0.0613 1.531 0.5421 Support Services - Allied Health 0.2445 0.1056 2.315 0.1399 Support Services - Doctor / Medical 0.4089 0.0758 5.398 <.0001 Support Services - Administration 0.1265 0.0644 1.965 0.2833 Support Services - Nursing 0.0326 0.0505 0.646 0.9674  Relative Risk between groups Estimate lwr upr Doctor / Medical - Allied Health 0.848 0.624 1.15 Administration - Allied Health 1.125 0.837 1.51 Administration - Doctor / Medical 1.326 1.061 1.66 Nursing - Allied Health 1.236 0.934 1.63 Nursing - Doctor / Medical 1.457 1.189 1.78 Nursing - Administration 1.098 0.931 1.30 Support Services - Allied Health 1.277 0.960 1.70 Support Services - Doctor / Medical 1.505 1.227 1.85 Support Services - Administration 1.135 0.954 1.35 Support Services - Nursing 1.033 0.902 1.18 ------------------ ExposureYearsGroup Probability of TB Prob LB UB [0 - 1) 0.0258 0.0148 0.0450 [20+) 0.0121 0.0070 0.0210 [1 - 5) 0.0233 0.0134 0.0404 [5 - 10) 0.0208 0.0119 0.0362 [15 - 20) 0.0239 0.0138 0.0415 [10 - 15) 0.0436 0.0252 0.0756  Log relative risk between groups contrast estimate SE z.ratio p.value [20+) - [0 - 1) -0.7567 0.0859 -8.813 <.0001 [1 - 5) - [0 - 1) -0.1027 0.0740 -1.388 0.7341 [1 - 5) - [20+) 0.6539 0.0739 8.853 <.0001 [5 - 10) - [0 - 1) -0.2162 0.0817 -2.646 0.0865 [5 - 10) - [20+) 0.5404 0.0781 6.921 <.0001 [5 - 10) - [1 - 5) -0.1135 0.0694 -1.636 0.5744 [15 - 20) - [0 - 1) -0.0765 0.0816 -0.937 0.9368 [15 - 20) - [20+) 0.6801 0.0666 10.209 <.0001 [15 - 20) - [1 - 5) 0.0262 0.0689 0.381 0.9990 [15 - 20) - [5 - 10) 0.1397 0.0735 1.902 0.4009 [10 - 15) - [0 - 1) 0.5254 0.0786 6.687 <.0001 [10 - 15) - [20+) 1.2820 0.0679 18.877 <.0001  174 [10 - 15) - [1 - 5) 0.6281 0.0652 9.636 <.0001 [10 - 15) - [5 - 10) 0.7416 0.0699 10.609 <.0001 [10 - 15) - [15 - 20) 0.6019 0.0633 9.505 <.0001  Relative Risk between groups Estimate lwr upr [20+) - [0 - 1) 0.469 0.368 0.599 [1 - 5) - [0 - 1) 0.902 0.731 1.114 [1 - 5) - [20+) 1.923 1.559 2.372 [5 - 10) - [0 - 1) 0.806 0.639 1.016 [5 - 10) - [20+) 1.717 1.375 2.143 [5 - 10) - [1 - 5) 0.893 0.733 1.087 [15 - 20) - [0 - 1) 0.926 0.735 1.168 [15 - 20) - [20+) 1.974 1.634 2.386 [15 - 20) - [1 - 5) 1.027 0.844 1.249 [15 - 20) - [5 - 10) 1.150 0.933 1.417 [10 - 15) - [0 - 1) 1.691 1.353 2.114 [10 - 15) - [20+) 3.604 2.971 4.371 [10 - 15) - [1 - 5) 1.874 1.557 2.256 [10 - 15) - [5 - 10) 2.099 1.721 2.561 [10 - 15) - [15 - 20) 1.826 1.525 2.186   Comparing HCW to the general population (with and without TB)  Sex Age Year HCW N TB nTB %TB 1 Female 20-29 2002 No 265649 876 264773 0.330 2 Female 20-29 2002 Yes 781 1 780 0.128 3 Female 20-29 2003 No 263209 1754 261455 0.666 4 Female 20-29 2003 Yes 827 6 821 0.726 5 Female 20-29 2004 No 261111 2125 258986 0.814 6 Female 20-29 2004 Yes 1140 14 1126 1.228 7 Female 20-29 2005 No 259746 2413 257333 0.929 8 Female 20-29 2005 Yes 1439 21 1418 1.459 9 Female 20-29 2006 No 258294 2722 255572 1.054 10 Female 20-29 2006 Yes 1539 11 1528 0.715 11 Female 20-29 2007 No 257974 2934 255040 1.137 12 Female 20-29 2007 Yes 1704 17 1687 0.998 13 Female 20-29 2008 No 258132 3185 254947 1.234 14 Female 20-29 2008 Yes 1673 27 1646 1.614 15 Female 20-29 2009 No 258210 3115 255095 1.206 16 Female 20-29 2009 Yes 1474 8 1466 0.543 17 Female 20-29 2010 No 257894 3279 254615 1.271 18 Female 20-29 2010 Yes 1737 2 1735 0.115 19 Female 20-29 2011 No 257496 2976 254520 1.156 20 Female 20-29 2011 Yes 2064 9 2055 0.436 21 Female 20-29 2012 No 257253 2506 254747 0.974 22 Female 20-29 2012 Yes 2228 13 2215 0.583 23 Male 20-29 2002 No 245861 652 245209 0.265 24 Male 20-29 2002 Yes 543 2 541 0.368 25 Male 20-29 2003 No 246282 1473 244809 0.598  175 26 Male 20-29 2003 Yes 585 6 579 1.026 27 Male 20-29 2004 No 247108 1643 245465 0.665 28 Male 20-29 2004 Yes 824 3 821 0.364 29 Male 20-29 2005 No 248236 1759 246477 0.709 30 Male 20-29 2005 Yes 1062 13 1049 1.224 31 Male 20-29 2006 No 250932 1852 249080 0.738 32 Male 20-29 2006 Yes 987 9 978 0.912 33 Male 20-29 2007 No 253495 1945 251550 0.767 34 Male 20-29 2007 Yes 1041 8 1033 0.768 35 Male 20-29 2008 No 255716 2056 253660 0.804 36 Male 20-29 2008 Yes 1039 9 1030 0.866 37 Male 20-29 2009 No 257290 2206 255084 0.857 38 Male 20-29 2009 Yes 893 1 892 0.112 39 Male 20-29 2010 No 258217 2412 255805 0.934 40 Male 20-29 2010 Yes 1028 9 1019 0.875 41 Male 20-29 2011 No 259080 2187 256893 0.844 42 Male 20-29 2011 Yes 1237 5 1232 0.404 43 Male 20-29 2012 No 260227 2145 258082 0.824 44 Male 20-29 2012 Yes 1319 11 1308 0.834 45 Female 30-39 2002 No 204818 940 203878 0.459 46 Female 30-39 2002 Yes 3347 26 3321 0.777 47 Female 30-39 2003 No 207051 2157 204894 1.042 48 Female 30-39 2003 Yes 3287 52 3235 1.582 49 Female 30-39 2004 No 208696 2797 205899 1.340 50 Female 30-39 2004 Yes 3434 69 3365 2.009 51 Female 30-39 2005 No 209625 3129 206496 1.493 52 Female 30-39 2005 Yes 3515 82 3433 2.333 53 Female 30-39 2006 No 210262 3661 206601 1.741 54 Female 30-39 2006 Yes 3259 74 3185 2.271 55 Female 30-39 2007 No 211480 3837 207643 1.814 56 Female 30-39 2007 Yes 3221 80 3141 2.484 57 Female 30-39 2008 No 212386 4199 208187 1.977 58 Female 30-39 2008 Yes 3118 50 3068 1.604 59 Female 30-39 2009 No 212960 4123 208837 1.936 60 Female 30-39 2009 Yes 2884 26 2858 0.902 61 Female 30-39 2010 No 213286 4029 209257 1.889 62 Female 30-39 2010 Yes 3277 31 3246 0.946 63 Female 30-39 2011 No 213409 3547 209862 1.662 64 Female 30-39 2011 Yes 3745 39 3706 1.041 65 Female 30-39 2012 No 213648 2935 210713 1.374 66 Female 30-39 2012 Yes 3880 31 3849 0.799 67 Male 30-39 2002 No 191727 1449 190278 0.756 68 Male 30-39 2002 Yes 1603 17 1586 1.061 69 Male 30-39 2003 No 191867 3105 188762 1.618 70 Male 30-39 2003 Yes 1617 40 1577 2.474 71 Male 30-39 2004 No 191980 3668 188312 1.911 72 Male 30-39 2004 Yes 1795 46 1749 2.563 73 Male 30-39 2005 No 192230 3946 188284 2.053 74 Male 30-39 2005 Yes 2117 53 2064 2.504 75 Male 30-39 2006 No 193160 4169 188991 2.158 76 Male 30-39 2006 Yes 2036 70 1966 3.438  176 77 Male 30-39 2007 No 195308 4344 190964 2.224 78 Male 30-39 2007 Yes 2038 54 1984 2.650 79 Male 30-39 2008 No 197774 4732 193042 2.393 80 Male 30-39 2008 Yes 1965 53 1912 2.697 81 Male 30-39 2009 No 200437 4542 195895 2.266 82 Male 30-39 2009 Yes 1859 29 1830 1.560 83 Male 30-39 2010 No 203219 4801 198418 2.362 84 Male 30-39 2010 Yes 1925 19 1906 0.987 85 Male 30-39 2011 No 205926 4359 201567 2.117 86 Male 30-39 2011 Yes 2157 22 2135 1.020 87 Male 30-39 2012 No 208719 4015 204704 1.924 88 Male 30-39 2012 Yes 2243 29 2214 1.293 89 Female 40-49 2002 No 149592 466 149126 0.312 90 Female 40-49 2002 Yes 3846 12 3834 0.312 91 Female 40-49 2003 No 151315 1043 150272 0.689 92 Female 40-49 2003 Yes 3976 36 3940 0.905 93 Female 40-49 2004 No 152702 1444 151258 0.946 94 Female 40-49 2004 Yes 4151 52 4099 1.253 95 Female 40-49 2005 No 153692 1706 151986 1.110 96 Female 40-49 2005 Yes 4350 76 4274 1.747 97 Female 40-49 2006 No 154224 2047 152177 1.327 98 Female 40-49 2006 Yes 4364 76 4288 1.742 99 Female 40-49 2007 No 154982 2179 152803 1.406 100 Female 40-49 2007 Yes 4375 96 4279 2.194 101 Female 40-49 2008 No 155741 2496 153245 1.603 102 Female 40-49 2008 Yes 4368 73 4295 1.671 103 Female 40-49 2009 No 156839 2459 154380 1.568 104 Female 40-49 2009 Yes 4285 37 4248 0.863 105 Female 40-49 2010 No 158362 2634 155728 1.663 106 Female 40-49 2010 Yes 4360 48 4312 1.101 107 Female 40-49 2011 No 160351 2395 157956 1.494 108 Female 40-49 2011 Yes 4385 42 4343 0.958 109 Female 40-49 2012 No 163053 2044 161009 1.254 110 Female 40-49 2012 Yes 4424 27 4397 0.610 111 Male 40-49 2002 No 143269 1239 142030 0.865 112 Male 40-49 2002 Yes 1270 12 1258 0.945 113 Male 40-49 2003 No 143346 2688 140658 1.875 114 Male 40-49 2003 Yes 1410 37 1373 2.624 115 Male 40-49 2004 No 142996 3412 139584 2.386 116 Male 40-49 2004 Yes 1618 39 1579 2.410 117 Male 40-49 2005 No 142234 3974 138260 2.794 118 Male 40-49 2005 Yes 1836 62 1774 3.377 119 Male 40-49 2006 No 141142 4089 137053 2.897 120 Male 40-49 2006 Yes 1816 67 1749 3.689 121 Male 40-49 2007 No 140639 4083 136556 2.903 122 Male 40-49 2007 Yes 1822 56 1766 3.074 123 Male 40-49 2008 No 140312 4301 136011 3.065 124 Male 40-49 2008 Yes 1781 46 1735 2.583 125 Male 40-49 2009 No 140396 3990 136406 2.842 126 Male 40-49 2009 Yes 1725 19 1706 1.101 127 Male 40-49 2010 No 140917 4142 136775 2.939  177 128 Male 40-49 2010 Yes 1767 29 1738 1.641 129 Male 40-49 2011 No 141902 3932 137970 2.771 130 Male 40-49 2011 Yes 1838 38 1800 2.067 131 Male 40-49 2012 No 143595 3583 140012 2.495 132 Male 40-49 2012 Yes 1861 17 1844 0.913 133 Female 50-59 2002 No 100626 157 100469 0.156 134 Female 50-59 2002 Yes 2119 6 2113 0.283 135 Female 50-59 2003 No 102485 327 102158 0.319 136 Female 50-59 2003 Yes 2378 15 2363 0.631 137 Female 50-59 2004 No 104342 527 103815 0.505 138 Female 50-59 2004 Yes 2686 22 2664 0.819 139 Female 50-59 2005 No 106213 668 105545 0.629 140 Female 50-59 2005 Yes 2948 36 2912 1.221 141 Female 50-59 2006 No 108083 751 107332 0.695 142 Female 50-59 2006 Yes 3030 31 2999 1.023 143 Female 50-59 2007 No 110094 855 109239 0.777 144 Female 50-59 2007 Yes 3086 35 3051 1.134 145 Female 50-59 2008 No 112061 1033 111028 0.922 146 Female 50-59 2008 Yes 3184 42 3142 1.319 147 Female 50-59 2009 No 113987 1054 112933 0.925 148 Female 50-59 2009 Yes 3276 15 3261 0.458 149 Female 50-59 2010 No 115895 1187 114708 1.024 150 Female 50-59 2010 Yes 3459 36 3423 1.041 151 Female 50-59 2011 No 117892 1093 116799 0.927 152 Female 50-59 2011 Yes 3575 24 3551 0.671 153 Female 50-59 2012 No 120143 1012 119131 0.842 154 Female 50-59 2012 Yes 3714 16 3698 0.431 155 Male 50-59 2002 No 89832 521 89311 0.580 156 Male 50-59 2002 Yes 582 4 578 0.687 157 Male 50-59 2003 No 91297 1099 90198 1.204 158 Male 50-59 2003 Yes 642 5 637 0.779 159 Male 50-59 2004 No 92840 1421 91419 1.531 160 Male 50-59 2004 Yes 757 9 748 1.189 161 Male 50-59 2005 No 94416 1651 92765 1.749 162 Male 50-59 2005 Yes 892 19 873 2.130 163 Male 50-59 2006 No 95956 1731 94225 1.804 164 Male 50-59 2006 Yes 913 18 895 1.972 165 Male 50-59 2007 No 97855 1978 95877 2.021 166 Male 50-59 2007 Yes 938 16 922 1.706 167 Male 50-59 2008 No 99494 2121 97373 2.132 168 Male 50-59 2008 Yes 1012 19 993 1.877 169 Male 50-59 2009 No 100833 2119 98714 2.101 170 Male 50-59 2009 Yes 1030 20 1010 1.942 171 Male 50-59 2010 No 101879 2386 99493 2.342 172 Male 50-59 2010 Yes 1048 20 1028 1.908 173 Male 50-59 2011 No 102751 2249 100502 2.189 174 Male 50-59 2011 Yes 1122 7 1115 0.624 175 Male 50-59 2012 No 103905 2179 101726 2.097 176 Male 50-59 2012 Yes 1181 14 1167 1.185 177 Female 60+ 2002 No 64298 63 64235 0.098 178 Female 60+ 2002 Yes 122 0 122 0.000  178 179 Female 60+ 2003 No 65425 119 65306 0.182 180 Female 60+ 2003 Yes 192 0 192 0.000 181 Female 60+ 2004 No 66759 144 66615 0.216 182 Female 60+ 2004 Yes 298 1 297 0.336 183 Female 60+ 2005 No 68428 155 68273 0.227 184 Female 60+ 2005 Yes 478 3 475 0.628 185 Female 60+ 2006 No 70354 208 70146 0.296 186 Female 60+ 2006 Yes 542 2 540 0.369 187 Female 60+ 2007 No 72526 228 72298 0.314 188 Female 60+ 2007 Yes 617 5 612 0.810 189 Female 60+ 2008 No 74812 257 74555 0.344 190 Female 60+ 2008 Yes 714 6 708 0.840 191 Female 60+ 2009 No 77035 291 76744 0.378 192 Female 60+ 2009 Yes 771 2 769 0.259 193 Female 60+ 2010 No 79091 375 78716 0.474 194 Female 60+ 2010 Yes 836 6 830 0.718 195 Female 60+ 2011 No 80992 357 80635 0.441 196 Female 60+ 2011 Yes 913 2 911 0.219 197 Female 60+ 2012 No 82939 371 82568 0.447 198 Female 60+ 2012 Yes 1026 5 1021 0.487 199 Male 60+ 2002 No 46541 99 46442 0.213 200 Male 60+ 2002 Yes 44 0 44 0.000 201 Male 60+ 2003 No 47154 238 46916 0.505 202 Male 60+ 2003 Yes 71 1 70 1.408 203 Male 60+ 2004 No 47824 298 47526 0.623 204 Male 60+ 2004 Yes 133 0 133 0.000 205 Male 60+ 2005 No 48566 342 48224 0.704 206 Male 60+ 2005 Yes 234 0 234 0.000 207 Male 60+ 2006 No 49363 390 48973 0.790 208 Male 60+ 2006 Yes 241 4 237 1.660 209 Male 60+ 2007 No 50524 476 50048 0.942 210 Male 60+ 2007 Yes 258 4 254 1.550 211 Male 60+ 2008 No 51762 531 51231 1.026 212 Male 60+ 2008 Yes 297 1 296 0.337 213 Male 60+ 2009 No 53020 512 52508 0.966 214 Male 60+ 2009 Yes 300 0 300 0.000 215 Male 60+ 2010 No 54280 638 53642 1.175 216 Male 60+ 2010 Yes 313 3 310 0.958 217 Male 60+ 2011 No 55567 644 54923 1.159 218 Male 60+ 2011 Yes 331 3 328 0.906 219 Male 60+ 2012 No 57201 684 56517 1.196 220 Male 60+ 2012 Yes 322 6 316 1.863 -------------------------------- HCW Only LR Chisq Df Pr(>Chisq) HCW 0.10422 1 0.7468  Probability of TB HCW lsmean asymp.LCL asymp.UCL No 0.01305760 0.01300382 0.01311160 Yes 0.01313936 0.01265409 0.01364324  179 -------------------------------- HCW and Sex Lilelihood Ratio Test (Type 2) LR Chisq Df Pr(>Chisq) HCW 13.2 1 0.0002745 Sex 8017.8 1 < 2.2e-16 HCW:Sex 2.7 1 0.0977887  Probability of TB HCW Sex lsmean asymp.LCL asymp.UCL No Female 0.01071981 0.01065236 0.01078768 Yes Female 0.01120915 0.01067182 0.01177353 No Male 0.01561043 0.01552555 0.01569577 Yes Male 0.01742001 0.01643022 0.01846942  Log Relative Risk HCW Sex estimate z.ratio p.value Yes - No Female 0.04463729 1.766448 0.0773 Yes - No Male 0.10968019 3.659020 0.0003 -------------------------------- HCW and Age Likelihood Ratio Test (Type 2) LR Chisq Df Pr(>Chisq) HCW 47.0 1 7.169e-12 Age 30999.7 4 < 2.2e-16 HCW:Age 20.7 4 0.0003619  Probability of TB HCW Age lsmean asymp.LCL asymp.UCL No 20-29 0.008552683 0.008477006 0.008629036 Yes 20-29 0.007546753 0.006584700 0.008649366 No 30-39 0.017479857 0.017359059 0.017601496 Yes 30-39 0.017009019 0.015991329 0.018091474 No 40-49 0.018445403 0.018300176 0.018591784 Yes 40-49 0.015191686 0.014284148 0.016156885 No 50-59 0.012317341 0.012175096 0.012461249 Yes 50-59 0.009845773 0.008961003 0.010817900 No 60+ 0.005438045 0.005316046 0.005562844 Yes 60+ 0.005964874 0.004572206 0.007781741  Log Relative Risk HCW Age estimate z.ratio p.value Yes - No 20-29 -0.12512767 -1.7945962 0.0727 Yes - No 30-39 -0.02730548 -0.8619999 0.3887 Yes - No 40-49 -0.19406686 -6.1247276 <.0001 Yes - No 50-59 -0.22396596 -4.6268368 <.0001 Yes - No 60+ 0.09246829 0.6791408 0.4970 -------------------------------- HCW and Age and Sex Likelihood Ratio Test (Type 2) LR Chisq Df Pr(>Chisq)  180 HCW 1.5 1 0.215752 Age 30746.1 4 < 2.2e-16 Sex 7756.9 1 < 2.2e-16 HCW:Age 9.5 4 0.049997 HCW:Sex 4.4 1 0.034990 Age:Sex 10039.7 4 < 2.2e-16 HCW:Age:Sex 15.8 4 0.003259  Probability of TB HCW Age Sex lsmean asymp.LCL asymp.UCL No 20-29 Female 0.009767185 0.009653771 0.009881931 Yes 20-29 Female 0.007768277 0.006541486 0.009225139 No 30-39 Female 0.015254435 0.015097456 0.015413047 Yes 30-39 Female 0.015148646 0.013953322 0.016446369 No 40-49 Female 0.012223727 0.012060177 0.012389495 Yes 40-49 Female 0.012264312 0.011307438 0.013302160 No 50-59 Female 0.007149571 0.007001127 0.007301162 Yes 50-59 Female 0.008309670 0.007391707 0.009341632 No 60+ Female 0.003199366 0.003078178 0.003325325 Yes 60+ Female 0.004916270 0.003479639 0.006946039 No 20-29 Male 0.007306526 0.007207140 0.007407283 Yes 20-29 Male 0.007198333 0.005753671 0.009005729 No 30-39 Male 0.019854102 0.019669462 0.020040476 Yes 30-39 Male 0.020229454 0.018426685 0.022208598 No 40-49 Male 0.025265450 0.025020445 0.025512853 Yes 40-49 Male 0.022513871 0.020487247 0.024740971 No 50-59 Male 0.018164282 0.017913123 0.018418964 Yes 50-59 Male 0.014925373 0.012740145 0.017485418 No 60+ Male 0.008636495 0.008397894 0.008881875 Yes 60+ Male 0.008647800 0.005705008 0.013108561  Log Relative Risk HCW Age Sex estimate z.ratio p.value Yes - No 20-29 Female -0.228979940 -2.605012960 0.0092 Yes - No 30-39 Female -0.006959151 -0.164647179 0.8692 Yes - No 40-49 Female 0.003314676 0.078898231 0.9371 Yes - No 50-59 Female 0.150367539 2.478127009 0.0132 Yes - No 60+ Female 0.429597370 2.421113896 0.0155 Yes - No 20-29 Male -0.014918476 -0.130284321 0.8963 Yes - No 30-39 Male 0.018729023 0.391319289 0.6956 Yes - No 40-49 Male -0.115306223 -2.383145188 0.0172 Yes - No 50-59 Male -0.196394502 -2.422202078 0.0154 Yes - No 60+ Male 0.001308114 0.006149765 0.9951          181  TB infection control score and rate of TB among hospital employees  Logistic Regression model with random effect for Hospital  Univariate analysis for each score   Total_Score Log odds ratio for 1 unit increase              Estimate Std. Error  z value Pr(>|z|) Total_Score -0.064887   0.017209 -3.77049 0.000163  Odds ratio with 95%CI for 1 unit increase                OR    LB    UB Total_Score 0.937 0.906 0.969   Admin_Score Log odds ratio for 1 unit increase              Estimate Std. Error   z value Pr(>|z|) Admin_Score -0.058645   0.037422 -1.567133 0.117083  Odds ratio with 95%CI for 1 unit increase                OR    LB    UB Admin_Score 0.943 0.874 1.015   Enviro_Score Log odds ratio for 1 unit increase               Estimate Std. Error  z value Pr(>|z|) Enviro_Score -0.130143   0.046034 -2.82711 0.004697  Odds ratio with 95%CI for 1 unit increase                 OR  LB    UB Enviro_Score 0.878 0.8 0.963   PPE_Score Log odds ratio for 1 unit increase            Estimate Std. Error   z value Pr(>|z|) PPE_Score -0.151392   0.047744 -3.170881  0.00152  Odds ratio with 95%CI for 1 unit increase             OR    LB    UB PPE_Score 0.86 0.782 0.945   Misc_Score Log odds ratio for 1 unit increase             Estimate Std. Error   z value Pr(>|z|)  182 Misc_Score -0.153004   0.076091 -2.010793 0.044347  Odds ratio with 95%CI for 1 unit increase               OR    LB    UB Misc_Score 0.858 0.734 0.999  ======================================================== Adding the covariates  I categorized the covariates as follows  Location Rural Urban     16    12   TB_Patients    2-19  51-121 142-759        8      11       9   SmearPlus    0-7  16-31 35-225      12      7      9   MDR  0 1+  17 11    Checking one covariate at a time  Location                 OR    LB   UB Urban - Rural 1.36 0.812 2.26  TB_Patients                     OR    LB    UB 51-121 - 2-19    0.387 0.194 0.776 142-759 - 2-19   0.464 0.243 0.887 142-759 - 51-121 1.198 0.665 2.156  SmearPlus                   OR    LB    UB 16-31 - 0-7    0.715 0.342 1.493 35-225 - 0-7   0.476 0.269 0.845 35-225 - 16-31 0.666 0.334 1.327  MDR          OR   LB   UB 1+ - 0 1.05 0.61 1.82    183 Overall test for each covariate             Df    LRT Pr(Chi) Location     1 1.2368 0.26609 TB_Patients  2 8.1309 0.01716 SmearPlus    2 7.2880 0.02615 MDR          1 0.0330 0.85592  TB_Parients and Smear+ and both important in a univariate analysis. However they are highly corelated to each other so no need to include both.             SmearPlus TB_Patients 0-7 16-31 35-225     2-19      8     0      0     51-121    4     4      3     142-759   0     3      6  Multivariate model             Df    AIC    LRT Pr(Chi) TB_Patients  2 98.123 3.7468  0.1536 SmearPlus    2 97.280 2.9039  0.2341  We will procced with TB_Patients included in the model   Final Multivariate Model  Total_Score Log odds ratio for 1 unit increase              Estimate Std. Error   z value Pr(>|z|) Total_Score -0.050065   0.021059 -2.377407 0.017435  Odds ratio with 95%CI for 1 unit increase                OR    LB    UB Total_Score 0.951 0.912 0.991  Effect of Covariate             Df  LRT Pr(Chi) TB_Patients  2 1.43    0.49                      OR    LB   UB 51-121 - 2-19    0.647 0.276 1.52 142-759 - 2-19   0.717 0.333 1.54 142-759 - 51-121 1.108 0.613 2.00   Admin_Score Log odds ratio for 1 unit increase              Estimate Std. Error   z value Pr(>|z|) Admin_Score -0.033693   0.037171 -0.906432 0.364707  Odds ratio with 95%CI for 1 unit increase  184                OR    LB    UB Admin_Score 0.967 0.898 1.039  Effect of Covariate             Df  LRT Pr(Chi) TB_Patients  2 6.48   0.039                      OR    LB    UB 51-121 - 2-19    0.414 0.203 0.846 142-759 - 2-19   0.513 0.257 1.025 142-759 - 51-121 1.238 0.684 2.240   Enviro_Score Log odds ratio for 1 unit increase               Estimate Std. Error   z value Pr(>|z|) Enviro_Score -0.086556   0.050832 -1.702802 0.088605  Odds ratio with 95%CI for 1 unit increase                 OR    LB    UB Enviro_Score 0.917 0.826 1.009  Effect of Covariate             Df  LRT Pr(Chi) TB_Patients  2 4.09    0.13                      OR    LB   UB 51-121 - 2-19    0.524 0.235 1.17 142-759 - 2-19   0.565 0.284 1.13 142-759 - 51-121 1.080 0.587 1.99   PPE_Score Log odds ratio for 1 unit increase            Estimate Std. Error   z value Pr(>|z|) PPE_Score -0.122392    0.05257 -2.328152 0.019904  Odds ratio with 95%CI for 1 unit increase              OR    LB    UB PPE_Score 0.885 0.799 0.982  Effect of Covariate             Df  LRT Pr(Chi) TB_Patients  2 4.93   0.085                      OR    LB   UB 51-121 - 2-19    0.546 0.255 1.17 142-759 - 2-19   0.528 0.272 1.02 142-759 - 51-121 0.966 0.522 1.79    185 Misc_Score Log odds ratio for 1 unit increase             Estimate Std. Error   z value Pr(>|z|) Misc_Score -0.076256   0.077225 -0.987451 0.323422  Odds ratio with 95%CI for 1 unit increase               OR    LB   UB Misc_Score 0.927 0.797 1.08  Effect of Covariate             Df LRT Pr(Chi) TB_Patients  2 5.2   0.074                     OR    LB    UB 51-121 - 2-19    0.44 0.206 0.939 142-759 - 2-19   0.55 0.258 1.172 142-759 - 51-121 1.25 0.690 2.261    

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0300152/manifest

Comment

Related Items