Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluating strategies for the early detection of chronic obstructive pulmonary disease Johnson, Kate 2020

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_may_johnson_kate.pdf [ 11.66MB ]
Metadata
JSON: 24-1.0389633.json
JSON-LD: 24-1.0389633-ld.json
RDF/XML (Pretty): 24-1.0389633-rdf.xml
RDF/JSON: 24-1.0389633-rdf.json
Turtle: 24-1.0389633-turtle.txt
N-Triples: 24-1.0389633-rdf-ntriples.txt
Original Record: 24-1.0389633-source.json
Full Text
24-1.0389633-fulltext.txt
Citation
24-1.0389633.ris

Full Text

EVALUATING STRATEGIES FOR THE EARLY DETECTION OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE by  Kate Johnson  B.Sc., The University of Alberta, 2013 M.Sc., The University of British Columbia, 2015  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Pharmaceutical Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  March 2020  © Kate Johnson, 2020  ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:  Evaluating strategies for the early detection of chronic obstructive pulmonary disease  submitted by Kate Johnson in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Pharmaceutical Sciences  Examining Committee: Mohsen Sadatsafavi, Pharmaceutical Sciences Co-supervisor Stirling Bryan, Faculty of Medicine Co-supervisor   Supervisory Committee Member Jacek Kopec, Faculty of Medicine University Examiner Peter Dodek, Faculty of Medicine University Examiner  Additional Supervisory Committee Members:  Supervisory Committee Member  Supervisory Committee Member iii  Abstract  Background: Chronic obstructive pulmonary disease (COPD) is a highly prevalent disease that often goes undiagnosed. Reducing the burden of undiagnosed COPD requires well designed early detection programs that have been formally evaluated.  Objectives: The objective of this thesis was to determine whether there are subgroups of COPD patients in which case detection, followed by evidence-based disease management, would be cost-effective compared with the status quo (no case detection). To answer this question, I 1) identified factors that distinguish patients with undiagnosed from those with diagnosed COPD, 2) assessed heterogeneity in the presence of respiratory symptoms, 3) analysed healthcare encounters prior to COPD diagnosis to identify opportunities for case detection, and 4) evaluated the cost-effectiveness of early detection strategies.   Methods: I performed a systematic review to generate pooled odds ratios for factors associated with a COPD diagnosis. I used data from a population-based Canadian study to assess heterogeneity between individuals in the occurrence of respiratory symptoms. I characterised healthcare encounters before COPD diagnosis using health administrative data from British Columbia. I combined evidence from Objectives 1-3 in a whole disease model of COPD to assess the cost-effectiveness of case detection strategies implemented during routine primary care visits. Results: Patients with diagnosed COPD were less likely to have mild disease (OR 0.30, 95%CI 0.24–0.37) and more likely to report respiratory symptoms (OR 11.45 95%CI 7.20–18.21) than patients with undiagnosed COPD. However, individual-specific probabilities for the occurrence of symptoms indicated substantial heterogeneity between patients. COPD patients frequently iv  visited primary care physicians before diagnosis (mean 10.29, IQR 4–13 visits/year). In the two years prior to diagnosis, 72.1% of COPD patients had a respiratory-related primary care visit that did not result in a diagnosis. In the preferred case detection strategy, all patients ≥40 years received a screening questionnaire during their routine visits to a primary care physician. This strategy had an Incremental Cost-Effectiveness Ratio of $18,791/QALY compared to no case detection. Conclusions: Patients with undiagnosed COPD have identifiable characteristics, they frequently encounter the healthcare system, and strategies for improving their early detection are cost-effective when combined with guideline-recommended treatment.   v  Lay Summary  Chronic obstructive pulmonary disease (COPD) is a progressive lung disease that affects over 2.6 million Canadians. Despite its prevalence, 70% of patients with COPD in Canada have not been diagnosed and are not receiving treatment. The objective of this thesis was to characterise undiagnosed COPD patients, assess their interactions with the healthcare system, and evaluate strategies for improving their diagnosis. I found that patients with undiagnosed COPD tend to have mild disease and few respiratory symptoms. Although current guidelines recommend restricting testing for COPD to patients with symptoms, I found that the presence of symptoms was highly variable between patients. Nearly all COPD patients encountered the healthcare system in the five years before they were diagnosed. Regular use of a questionnaire to assess the risk of COPD, implemented during the routine primary care visits of patients over 40 years, is a cost-effective strategy for improving early detection of COPD. vi  Preface  This thesis is composed of three original studies and one systematic review addressing the burden of undiagnosed COPD and evaluating policies for improving its early detection. I was responsible for developing the research questions, conducting the statistical analyses, developing the decision model, and writing the chapters. My supervisors, Drs. Mohsen Sadatsafavi and Stirling Bryan, and my PhD committee members, Drs. Mark Harrison, Corey Nislow, Larry Lynd, and Don Sin, provided feedback on the research design and interpretation of results.    The following manuscripts resulted from this thesis: 1) Kate M. Johnson, Stirling Bryan, Shahzad Ghanbarian, Don D. Sin, and Mohsen Sadatsafavi. Characterizing Undiagnosed Chronic Obstructive Pulmonary Disease: A Systematic Review and Meta-Analysis. 2018. Respiratory Research 19:26. (published, Chapter 2) MS, SB, and KJ formulated the study idea and designed the study. KJ and SG performed all data analyses and MS, SB and DS contributed to interpretation of findings. KJ wrote the first draft of the manuscript. All authors critically commented on the manuscript and approved the final version. My total contribution to this research was >90%.   2) Kate M. Johnson, Abdollah Safari, Wan C. Tan, Jean Bourbeau, J. Mark FitzGerald, and Mohsen Sadatsafavi. Heterogeneity in the Respiratory Symptoms of Patients with Mild-to-Moderate COPD. 2018. International Journal of Chronic Obstructive Pulmonary Disease 2018:13 pg. 3983-3995. (published, Chapter 3) vii  WT and JB are co-Principal Investigators of the CanCOLD study, which was the source of data for this analysis. MS and KJ formulated the current study idea. KJ performed all data analyses and wrote the first draft of the manuscript. AS provided guidance on the statistical analysis. All authors contributed to the interpretation of findings, critically commented on the manuscript, and approved the final version. My total contribution to this research was >90%.  3) Kate M. Johnson, Amir Khakban, Stirling Bryan, Don D. Sin, and Mohsen Sadatsafavi. Healthcare system encounters before COPD diagnosis: a registry-based longitudinal cohort study. 2020. Thorax 75:108-115. (published, Chapter 4) MS, SB and KJ formulated the study idea. AK prepared the data. KJ performed all data analyses and wrote the first draft of the manuscript. All authors contributed to interpretation of findings, critically commented on the manuscript and approved the final version. My total contribution to this research was >85%.  Ethics approval was required for data related to this thesis: 1) Ethics approval for the CanCOLD study was obtained by the respective university and institutional ethical review boards from each study site: UBC/PHC Research Ethics Board, P05-006 (Vancouver); Biomedical-C Research Ethics Board, BMC-06-002 (Montreal); UHN REB, 06-0421-B (Toronto); Capital Health Research Ethics Board, CDHA-RS/ 2007-255 (Halifax); Conjoint Health Research Ethics Board, ID21258 (Calgary); DMED-1240-09 (Kingston); 2009519-01H (Ottawa); Bio-REB09-162 (Saskatoon); CER20459 (Quebec City). Written informed consent was obtained from all viii  participants prior to study entry. CanCOLD was carried out in accordance with the principles of the Declaration of Helsinki. 2) Ethics approval for obtaining and analysing data from the administrative health databases of British Columbia (BC) Ministry of Health was obtained from Population Data BC (H13-00684). ix  Table of Contents  Abstract ......................................................................................................................................... iii Lay Summary .................................................................................................................................v Preface ........................................................................................................................................... vi Table of Contents ......................................................................................................................... ix List of Tables ................................................................................................................................xv List of Figures ............................................................................................................................. xvi List of Abbreviations ............................................................................................................... xviii Acknowledgements ......................................................................................................................xx Dedication ................................................................................................................................... xxi Chapter 1: Introduction ................................................................................................................1 1.1 COPD epidemiology ....................................................................................................... 2 1.2 Burden of COPD ............................................................................................................. 4 1.3 Definition of COPD ........................................................................................................ 4 1.4 Diagnosis of COPD ........................................................................................................ 5 1.5 Treatment of COPD ........................................................................................................ 6 1.6 Potential benefits of an early diagnosis .......................................................................... 8 1.7 Case detection for COPD ................................................................................................ 9 1.8 Current knowledge gaps ............................................................................................... 12 1.9 Objectives ..................................................................................................................... 13 1.10 Thesis themes ................................................................................................................ 14 1.10.1 Evidence generation .............................................................................................. 14 x  1.10.2 Whole disease modelling ...................................................................................... 16 1.10.2.1 Evaluation Platform in COPD (EPIC) .......................................................... 18 1.11 Thesis summary ............................................................................................................ 20 Chapter 2: Characterising Undiagnosed Chronic Obstructive Pulmonary Disease: a Systematic Review and Meta-Analysis .......................................................................................22 2.1 Introduction ................................................................................................................... 22 2.2 Methods......................................................................................................................... 23 2.2.1 Search strategy and selection criteria .................................................................... 23 2.2.2 Data analysis ......................................................................................................... 25 2.3 Results ........................................................................................................................... 26 2.3.1 Unadjusted analysis .............................................................................................. 32 2.3.2 Adjusted analysis .................................................................................................. 35 2.4 Discussion ..................................................................................................................... 39 2.5 Conclusions ................................................................................................................... 43 Chapter 3: Heterogeneity in the Respiratory Symptoms of Patients with Mild-to-Moderate COPD ............................................................................................................................................45 3.1 Introduction ................................................................................................................... 45 3.2 Methods......................................................................................................................... 46 3.2.1 Statistical analysis ................................................................................................. 48 3.3 Results ........................................................................................................................... 50 3.3.1 Objective 1: Heterogeneity in the occurrence of symptoms ................................. 51 3.3.2 Objective 2: Influence of lung function on symptom heterogeneity .................... 53 3.4 Discussion ..................................................................................................................... 56 xi  3.5 Conclusion .................................................................................................................... 60 Chapter 4: Healthcare System Encounters before COPD Diagnosis: a Registry-Based Longitudinal Cohort Study .........................................................................................................61 4.1 Introduction ................................................................................................................... 61 4.2 Methods......................................................................................................................... 62 4.2.1 Study design .......................................................................................................... 62 4.2.2 Data sources .......................................................................................................... 62 4.2.3 Study population ................................................................................................... 63 4.2.4 Healthcare system encounters ............................................................................... 65 4.2.5 Analysis................................................................................................................. 65 4.3 Results ........................................................................................................................... 67 4.3.1 Rate of outpatient encounters ................................................................................ 69 4.3.2 Probability of outpatient encounters ..................................................................... 75 4.4 Discussion ..................................................................................................................... 76 4.5 Conclusions ................................................................................................................... 79 Chapter 5: Cost-Effectiveness of Case Detection Strategies for the Early Detection of COPD ............................................................................................................................................80 5.1 Introduction ................................................................................................................... 80 5.2 Methods......................................................................................................................... 81 5.2.1 Case detection scenarios ....................................................................................... 82 5.2.2 Evaluating case detection scenarios ...................................................................... 84 5.2.3 Parameter inputs .................................................................................................... 86 5.2.3.1 Treatment effectiveness .................................................................................... 86 xii  5.2.3.2 Costs .................................................................................................................. 87 5.2.3.2.1 Case detection and diagnosis costs ............................................................. 87 5.2.3.2.2 Treatment costs ........................................................................................... 88 5.2.3.3 Health status ...................................................................................................... 88 5.2.4 Analysis................................................................................................................. 90 5.2.4.1 Reference case .................................................................................................. 90 5.2.4.2 Sensitivity analyses ........................................................................................... 91 5.2.5 Model development .............................................................................................. 92 5.2.5.1 Evaluation Platform in COPD (EPIC) .............................................................. 92 5.2.5.2 Symptom module .............................................................................................. 93 5.2.5.3 Primary care visits module ................................................................................ 94 5.2.5.4 Diagnosis module .............................................................................................. 94 5.2.5.5 Treatment module ............................................................................................. 96 5.2.5.6 Case detection module ...................................................................................... 97 5.3 Results ........................................................................................................................... 98 5.3.1 Reference case ...................................................................................................... 98 5.3.2 Sensitivity analyses ............................................................................................. 102 5.4 Discussion ................................................................................................................... 105 5.4.1 Summary of results ............................................................................................. 105 5.4.2 Comparison with other research ......................................................................... 105 5.4.3 Policy implications .............................................................................................. 106 5.4.4 Clinical implications ........................................................................................... 107 5.4.5 Strengths and limitations ..................................................................................... 109 xiii  5.4.6 Conclusions ......................................................................................................... 110 Chapter 6: Conclusions .............................................................................................................111 6.1 Contributions and implications ................................................................................... 111 6.1.1 Improving recognition of risk factors for COPD ................................................ 112 6.1.2 Early detection of COPD is cost-effective .......................................................... 113 6.1.3 Disease heterogeneity should be quantified ........................................................ 115 6.1.4 Whole disease modelling can inform policy development ................................. 115 6.2 Strengths of this research ............................................................................................ 118 6.3 Limitations of this research ......................................................................................... 119 6.4 Future research directions ........................................................................................... 122 6.4.1 Knowledge translation to policy-makers, knowledge users, and patients .......... 122 6.4.2 Extending case detection to pharmacists ............................................................ 124 6.4.3 Empirical evaluation of case detection strategies ............................................... 125 6.5 Conclusions ................................................................................................................. 126 References ...................................................................................................................................128 Appendices ..................................................................................................................................145 Appendix A ............................................................................................................................. 145 A.1 Search strategy ........................................................................................................ 145 A.2 Unadjusted analysis in the general population ........................................................ 148 A.3 Unadjusted analysis using the lower limit of normal criterion ............................... 149 Appendix B ............................................................................................................................. 150 B.1 Symptom assessment questionnaire ........................................................................ 150 xiv  B.2 Characteristics of study participants in the subset of the data used to assess dyspnea and any symptoms ............................................................................................................... 151 Appendix C ............................................................................................................................. 152 C.1 List of medications classified as other respiratory medications ............................. 152 C.2 Parameter estimates from the sensitivity analysis of the subset of participants with spirometry-confirmed moderate to severe COPD near the time of diagnosis .................... 157 C.3 Parameter estimates from the subset of participants with no missing data ............. 158 C.4 Parameters estimates from a new cohort of participants in which comparator subjects never developed COPD during follow-up ............................................................ 159 Appendix D ............................................................................................................................. 160 D.1 Development and validation of the symptom module in EPIC .............................. 160 D.2 Development and validation of the primary care visits module in EPIC ............... 163 D.3 Development and validation of the diagnosis module in EPIC .............................. 165 D.4 Development and validation of the treatment module in EPIC .............................. 171 D.5 Performance characteristics of case detection methods .......................................... 174 D.6 Results of the preferred case detection scenario ..................................................... 176  xv  List of Tables  Table 1-1 Global Initiative for chronic Obstructive Lung Disease (GOLD) classification of COPD severity. ............................................................................................................................... 5 Table 1-2 Global Initiative for chronic Obstructive Lung Disease (GOLD) assessment tool to guide therapeutic decisions in COPD. ............................................................................................ 7 Table 1-3 Definitions of key methods for increasing the probability of a COPD diagnosis. ......... 9 Table 1-4 COPD diagnostic questionnaire (CDQ). ...................................................................... 11 Table 2-1 Characteristics of selected studies. ............................................................................... 29 Table 3-1 Characteristics of study participants at study visits. ..................................................... 50 Table 3-2 Percentage of between-individual variation in symptoms explained by individual’s lung function and all measured characteristics combined. ........................................................... 56 Table 4-1 Characteristics of patients meeting the case definition of COPD and matched comparator subjects. ..................................................................................................................... 68 Table 4-2 Parameter estimates and confidence intervals from the multivariable regression model of the rate of outpatient visits among COPD patients and comparator non-COPD subjects. ....... 72 Table 4-3 Mean number of pharmacist visits per 100 patient years with respiratory and other medications dispensed. ................................................................................................................. 74 Table 5-1 Case detection scenarios evaluated. ............................................................................. 82 Table 5-2 Model parameters specific to the evaluation of case detection. ................................... 89 Table 5-3 Total costs and QALYs for case detection scenarios. .................................................. 99  xvi  List of Figures  Figure 1-1 Diagram of the simulated case detection pathway and sources of evidence. .............. 15 Figure 1-2 Flow diagram of Evaluation Platform in COPD model structure. .............................. 18 Figure 2-1 Preferred reporting items for systematic reviews and meta-analyses flow diagram. .. 27 Figure 2-2 Associations between diagnosed (v. ‘undiagnosed’) COPD and sex, the presence of cough, wheeze, phlegm, dyspnea, any respiratory symptoms, smoking status, smoking history, and COPD severity based on contingency tables (‘unadjusted analysis’). ................................... 34 Figure 2-3 Mean difference (MD) in age, pack-years of smoking, mMRC dyspnea score, and percent of predicted FEV1 between diagnosed and undiagnosed categories. ............................... 35 Figure 2-4 Associations between risk factors and the odds of receiving a COPD diagnosis using the regression coefficients from studies with multivariable regression modelling (‘adjusted analysis’) and persistent airflow limitation defined as post-bronchodilator FEV1/FVC<0.7. ...... 37 Figure 2-5 Associations between risk factors and the odds of receiving a COPD diagnosis using the regression coefficients from studies with multivariable regression modelling (‘adjusted analysis’) and persistent airflow limitation defined as post-bronchodilator FEV1/FVC<LLN. ... 39 Figure 3-1 Sample selection procedure. ........................................................................................ 47 Figure 3-2 The distribution of individual-specific probabilities of the occurrence of symptoms. 52 Figure 3-3 Odds ratios for the associations between independent variables and the presence of a) cough, b) phlegm, c) wheeze, and d) dyspnea. ............................................................................. 55 Figure 3-4 Odds ratios for the associations between independent variables and the presence of any symptoms. .............................................................................................................................. 55 Figure 4-1 Schematic illustration of study design. ....................................................................... 64 xvii  Figure 4-2 Cohort selection process. ............................................................................................ 68 Figure 4-3 Mean number of outpatient visits per patient year in each of the five years before the index date. ..................................................................................................................................... 70 Figure 4-4 Mean number of ICD diagnostic codes per patient year at the primary care visits of COPD patients before the index date. ........................................................................................... 73 Figure 4-5 Proportion of subjects with at least one outpatient visit in any of the five years before index. ............................................................................................................................................. 75 Figure 5-1 Framework for applying case detection strategies. ..................................................... 85 Figure 5-2 Cost-effectiveness plane for case detection scenarios. ............................................. 101 Figure 5-3 Sensitivity analyses of the incremental net monetary benefit of case detection scenarios compared to no case detection. ................................................................................... 104  xviii  List of Abbreviations  AECOPD, Acute Exacerbations of COPD BC, British Columbia BMI, Body Mass Index BOLD, Burden of Obstructive Lung Disease CanCOLD, Canadian Cohort of Obstructive Lung Disease Study CAT, COPD Assessment Test CHEERS, Consolidated Health Economic Evaluation Reporting Standards CI, Confidence Interval  COPD, Chronic Obstructive Pulmonary Disease CDQ, COPD Diagnostic Questionnaire DIN, Drug Identification Number EPIC, Evaluation Platform in COPD EPI-SCAN, Epidemiologic Study of COPD in Spain FEV1, Forced Expiratory Volume in 1 second FVC, Forced Vital Capacity GOLD, Global Initiative for chronic Obstructive Lung Disease ICER, Incremental Cost-Effectiveness Ratio INMB, Incremental Net Monetary Benefit ICS, Inhaled Corticosteroids IQR, Interquartile Range LAMA, Long-Acting Muscarinic Antagonists xix  LABA, Long-Acting Beta-Agonists LLN, Lower Limit of Normal MD, Mean Difference MPR, Medication Possession Ratio MRC, Medical Research Council dyspnea scale mMRC, modified Medical Research Council dyspnea scale NRT, Nicotine Replacement Therapy OR, Odds Ratio PLATINO, Latin American Project for the Investigation of Obstructive Lung Disease PREPOCOL, Prevalence study of COPD in Colombia QALY, Quality Adjusted Life Year RR, Rate Ratio SABA, Short-Acting Beta-Agonists SAMA, Short-Acting Muscarinic Agents SES, Socioeconomic Status SD, Standard Deviation WTP, Willingness to Pay   xx  Acknowledgements  I thank my supervisors, Drs. Mohsen Sadatsafavi and Stirling Bryan, for their time and encouragement in helping to produce this work. I will always be grateful for their mentorship, and their genuine and unwavering support for my success. May every PhD student be as fortunate as I have been.    I received advice and assistance from numerous health economists, clinicians, and statisticians throughout this process. I owe particular thanks to Dr. Donald Sin, St. Paul’s Hospital, and Dr. Mark FitzGerald, Vancouver General Hospital, for their valuable clinical insight, which greatly enhanced this work. I also thank Dr. Abdollah Safari, statistician at the Faculty of Pharmaceutical Sciences, for his knowledge and patience while providing advice on many of the analyses in this thesis. I thank all of the current and former members of the Respiratory Evaluation Sciences Program, including, Amir Khakban, Hamid Tavakoli, Shahzad Ghanbarian, Wenjia Chen, Zafar Zafari, and especially, Amin Adibi. Amin was invariably willing to help; whether by solving a problem or offering a distraction, thank you my friend.   Producing this thesis took hard work and perseverance. I owe my parents, Bibiana Cujec and David Johnson, for teaching me through their example. To my mother, who walks beside me, and to my late father, who will never be left behind: thank you. Finally, I thank David Hanna, Kyle Johnson, Neal Gallagher, my grandmother, Bibiana Cujec, and all my friends who got me off the computer and into the mountains; your advice and encouragement gave me perspective; the most valuable gift of all.  xxi  Dedication For Alex  My partner on the road to Ithaka  1  Chapter 1: Introduction  Early detection is an important component of high-quality care for many diseases. Programs to detect patients with undiagnosed disease have been developed to meet this objective. Two widespread examples of this approach are screening for cervical cancer among adult women1, and screening for hypertension in older adults2, which have both been associated with reduced morbidity and mortality from disease3,4. The goal of early detection programs is to diagnose disease while it is early in its progression. However, in order for early detection to be of value, a diagnosis must be associated with altered disease progression in patients5. It might be this criterion that that has, thus far, prevented widespread implementation of early detection programs for chronic obstructive pulmonary disease (COPD). Many authors argue that early detection of COPD will uncover mostly asymptomatic patients with mild disease, for whom there are limited treatment options and ability to alter the disease progression6,7. On the other hand, objective evaluations of these policies are rare, and respiratory researchers have been accused of nihilism with regards to the value of early detection8. This is in contrast to the key role economic evaluations have played in informing policies related to early intervention for diabetes9, cancer10, and cardiovascular disease11, which have ultimately helped reduce the burden of these diseases3,4,12. A formal evaluation of ‘value for money’ is required before any early detection program is adopted. This evidence is currently lacking for COPD, and although controversial, an earlier diagnosis could have substantial public health benefits13.   The overarching themes of this thesis are evidence generation on patterns of care for COPD and the natural history of disease, followed by decision modelling to evaluate policies for the early detection of COPD. Chronic obstructive pulmonary disease is an extremely 2  heterogeneous disease, and a policy for improving its diagnosis should account for differences in patient characteristics and disease trajectories. I applied the lens of precision medicine to incorporate heterogeneity between individuals when assessing components of the natural history of COPD. Evidence on heterogeneity between individuals can serve as a foundation for individualised prediction tools and enable more precise risk factor and disease management14. This evidence was also synthesised in the disease model that was used to assess the value of early detection for COPD.  In this chapter, I will provide an overview of the epidemiology, burden, and treatment of COPD, and current practices for its diagnosis. I discuss existing evidence on the value of early detection of COPD, and knowledge gaps in these evaluations. I place the specific objectives of this thesis in the context of my research themes: evidence generation for precision medicine, followed by whole disease modelling to evaluate policy. I conclude with background information on the whole disease model that was used to evaluate early detection strategies.    1.1 COPD epidemiology Chronic obstructive pulmonary disease is a disease of the airways that is associated with irreversible airflow obstruction and progressive lung function decline15. Patients with COPD typically experience symptoms such as shortness of breath, cough, sputum production, and periods of intensified disease activity referred to as exacerbations16. Although 80% of the burden of COPD has been attributed to cigarette smoking17, aging, environmental and occupational exposures are also important risk factors18. Because COPD is often caused by chronic and repeated lung injury; patients with COPD tend to be over 40 years old19. Disease progression is characterised by increasing airflow limitation, as measured by forced expiratory volume in 1 3  second (FEV1). Exacerbations are a major cause of morbidity, mortality and economic burden16, and are the leading cause of hospital admissions among chronic diseases in Canada20. A third, often underappreciated component of the burden of COPD is the high incidence of extra-pulmonary comorbidities, including cardiovascular disease, diabetes mellitus21, and depression22. In particular, cardiovascular disease is 2.5 times more prevalent in patients with COPD than in the general population23, and it is the leading cause of mortality in patients with COPD24. This may be due to the underlying role of systemic inflammation, or to shared risk factors. Comorbidities complicate the management of disease25, contribute to disease progression, and increase its burden24.  An important aspect of COPD is that it is very heterogeneous. The extent and rate of lung function decline tends to vary substantially between individuals26–28. In the landmark Lung Health Study, which followed early stage COPD patients for up to 11 years29, patients had an annual rate of change in lung function that ranged from rapidly declining to modestly increasing (interval for the rate of change in FEV1 containing 95% of patients: -83 mL/yr to 15 mL/yr)28. Easily measured clinical characteristics including age, sex and smoking history, explained 88% of this variation in lung function decline28. The rate and severity of exacerbations is also heterogeneous between patients with COPD30,31. In the Azithromycin for Prevention of Exacerbations of COPD (MACRO) clinical trial, 95% of COPD patients had model-based exacerbation rate between 0.47 and 4.22 exacerbations per year32. Heterogeneity in respiratory symptoms, a third major component of the natural history of COPD, has not been well quantified.  4  1.2 Burden of COPD Chronic obstructive pulmonary disease is a highly prevalent disease. In 2014, there were 94,647 cases of diagnosed COPD in British Columbia (BC) and 804,043 cases in Canada33. In BC, the prevalence of COPD is estimated at 19% in adults ≥40 years34. Globally, the number of COPD cases is approximately 400 million35. COPD is also associated with a large economic burden. In BC, the total direct medical costs of COPD were $5,452 (2010 CAD) per patient per year from 2001 to 2010. 57% of these costs were due to hospital admissions36. In Canada, expenditures related to COPD were estimated to cost the healthcare system $4.25 billion in 2011 alone37.  The aging population is expected to result in a 150% increase in the prevalence of COPD by 2030, and a 220% increase in prevalence in the population 75 years of age or older38. This is despite a 43% decrease in the prevalence of smoking in BC between 1999 and 201339. Similarly, hospitalisations due to COPD are projected to increase by 182% between 2010 and 203038. The economic burden of COPD is predicted to increase in conjunction with its prevalence. It is critical to develop disease management strategies to alleviate the increasing economic and public health burden of COPD.   1.3 Definition of COPD The Global Initiative for chronic Obstructive Lung Disease (GOLD) recommends considering a diagnosis of COPD in patients with recurring respiratory symptoms such as dyspnea, cough, or sputum, and a history of exposure to risk factors15. Spirometry, a lung function test that measures airflow limitation, is used to diagnose COPD. Spirometry is a simple and safe diagnostic procedure that is highly reliable if performed correctly8. GOLD defines COPD as the ratio of FEV1 to forced vital capacity (FVC) of less than 0.7 after the administration of a bronchodilator. 5  An alternative definition of COPD is FEV1/FVC below the lower limit of normal (LLN), which is the lower fifth percentile of an age and gender-stratified healthy reference population40. The severity of COPD is divided into four stages based on the ratio of FEV1 to its predicted reference value for a healthy individual41 (Table 1-1).   Table 1-1 Global Initiative for chronic Obstructive Lung Disease (GOLD) classification of COPD severity. COPD Severity GOLD 1 (mild COPD) FEV1 ≥ 80% predicted FEV1/FVC < 0.70 GOLD 2 (moderate COPD) 50% ≤ FEV1 < 80% predicted GOLD 3 (severe COPD) 30% ≤ FEV1 < 50% predicted GOLD 4 (very severe COPD) FEV1 < 30% predicted   1.4 Diagnosis of COPD Patients with COPD generally seek medical attention that results in a diagnosis because they are experiencing respiratory symptoms, most frequently chronic and progressive dyspnea15. However, the presence and severity of symptoms varies between patients, and also within patients over time42. In some cases, respiratory symptoms may precede airflow limitation, and conversely, symptoms may not be present in patients with early stage COPD15. As a result, and because adults rarely use the full capacity of their lungs43, progressive airflow limitation is often undetected in its early stages. In many cases, patients do not receive a diagnosis of COPD until after being hospitalised due to a severe exacerbation44. The Canadian Cohort of Obstructive Lung Disease (CanCOLD) is a prospective, population-based, longitudinal cohort study. Individuals with COPD were selected from the general population for long-term follow-up in CanCOLD, along with a matched cohort of at-risk subjects and healthy controls45. 70% of subjects with COPD in CanCOLD had not previously 6  received a physician diagnosis of COPD46. Global estimates of the burden of undiagnosed COPD are even higher. Lamprecht et al.47 reported an average underdiagnosis rate of 81% in a prevalence study that included 30,874 participants across 44 countries. The undiagnosed patients in this study tended to be younger, male, never smokers, with fewer respiratory symptoms, and less severe COPD.   1.5 Treatment of COPD The mainstays of treatment for COPD are interventions for smoking cessation and inhaled therapies. Smoking is the only major risk factor for COPD that is modifiable, and aside from oxygen therapy in patients with severe disease15, smoking cessation is the only intervention that can slow the progression of lung function decline29. Interventions for smoking cessation increase the likelihood of sustained quitting and reduce the rate of FEV1 decline29,48,49. Pulmonary rehabilitation, which aims to improve functional ability through exercise and behavioural modification, is a second effective therapy, although it is most commonly used in patients with severe COPD50. A recent review of randomised controlled trials found that pulmonary rehabilitation led to clinically important improvements in lung function, health-related quality of life and functional ability50.  In addition to lifestyle interventions, inhaled pharmacotherapies can reduce symptoms, the risk and severity of exacerbations, and improve quality of life in patients with COPD15. Maintenance pharmacotherapy is generally used in patients with moderate to very severe disease, with the most important medications being short and long-acting bronchodilators (beta-agonists and anticholinergics), and inhaled corticosteroids (ICS)15. ICS has also been found to reduce the risk of lung cancer in patients with COPD51. 7  In 2011, GOLD refined its assessment to include the burden of symptoms and risk of exacerbations as part of its criteria for guiding management decisions. The resulting ‘ABCD’ groups determine the recommended treatment for COPD patients after assessing airflow limitation (Table 1-2)15. Patients with few symptoms and a low risk of exacerbations (‘group A’) are treated with a short-acting bronchodilator (SABA), and stepped up to a long-acting muscarinic antagonist (LAMA) if symptoms increase (‘group B’). Patients with few symptoms but a high risk of exacerbations (‘group C’) are treated with combination long-acting beta-agonist and muscarinic antagonist (LAMA/LABA). This is increased to triple combination therapy with LAMA, LABA, and ICS if symptoms are also high (‘group D’)15.  Table 1-2 Global Initiative for chronic Obstructive Lung Disease (GOLD) assessment tool to guide therapeutic decisions in COPD. Exacerbation history ≥2 moderate exacerbations* or 1 hospitalisation due to exacerbation in the previous year C Few symptoms, High exacerbation risk D Many symptoms, High exacerbation risk ≤1 moderate exacerbation in the previous year A Few symptoms, Low exacerbation risk B Many symptoms, Low exacerbation risk  mMRC† 0-1 or CAT‡ score <10 mMRC ≥2 or  CAT score ≥10 Symptoms  * Moderate exacerbations are defined as those requiring treatment with antibiotics or oral corticosteroids15  † modified Medical Research Council dyspnea scale52 ‡ COPD Assessment Test53  8  1.6 Potential benefits of an early diagnosis A late diagnosis of COPD can be a missed opportunity to modify the course of disease through evidence-informed risk factor management and treatment54. In most cases, undiagnosed patients have mild or early stage COPD. The goals of treatment are similar in these patients: modify risk factors and reduce both symptoms and the risk of exacerbations. To accomplish this, guidelines recommend administering short or long-acting bronchodilators and smoking cessation (if applicable)15. Receiving a COPD diagnosis might itself lead to smoking cessation if the extent of lung obstruction is conveyed to the patient in easily understood terms (e.g., the ‘lung age’)55. Although smoking cessation is the only intervention with a direct disease-modifying effect in patients with mild disease, inhaled therapies may indirectly slow the rate of disease progression by reducing the rate of exacerbations, which are in turn associated with accelerated lung function decline56. In a prospective cohort study, moderate exacerbations were associated with an additional 23 mL/year decline in FEV1 among patients with mild COPD, which was greater than was observed in moderate or severe COPD57. Emerging evidence suggests that early administration of long-acting bronchodilators can also have a disease-modifying effect in patients with mild COPD58, and ICS might reduce the risk of lung cancer51. However, the conventional wisdom is that pharmacotherapy does not directly reduce the rate of lung function decline at any disease stage15. Therefore, modifying risk factors such as smoking and hazardous exposures while the disease is early in its progression is the most important component of treatment59.   9  1.7 Case detection for COPD Given the high rate of underdiagnosis in COPD, many authors have called for screening or case detection programs to diagnose patients with COPD earlier8,60–62. Screening and case detection are interventions applied in addition to routine diagnosis that increase the likelihood of detecting patients with undiagnosed COPD. Screening involves actively seeking out asymptomatic individuals in the general population to test for COPD. In contrast, case detection is opportunistic8. For example, a case detection strategy could involve testing individuals as they encounter the healthcare system for reasons related to cardiovascular disease. The definitions of case detection, screening, and routine diagnosis used in this thesis are summarised in Table 1-3.  Table 1-3 Definitions of key methods for increasing the probability of a COPD diagnosis.  Definition Case detection An intervention that increases the likelihood of detecting individuals with undiagnosed COPD through opportunistic testing of patients who encounter the healthcare system for any reason.  Screening An intervention that increases the likelihood of detecting individuals with undiagnosed COPD by systematically seeking out and testing individuals from the general population. Routine diagnosis A physician determining whether a patient has COPD based on the degree of airflow limitation, respiratory symptoms, and risk factors for the disease.    Despite the success of some programs, most guidelines do not recommend screening for airflow limitation in healthy adults with no respiratory symptoms6,63. These guidelines suggest that the high rate of false positives and questionable benefits of treating asymptomatic patients are not worth the economic costs to society. These costs can include subsequent diagnostic spirometry, and repeat follow-up visits with comprehensive pulmonary function tests and lung imaging64. 10  The U.S. Preventive Services Task Force determined that 883 asymptomatic smokers would need to be screened to prevent one exacerbation in the next 6-36 months. They concluded that this would not be an efficient use of resources65. However, this was a simplified analysis that did not assess the impact of population characteristics, healthcare setting, or diagnostic modality on the number needed to screen. Targeted testing of patients with a high risk of undiagnosed COPD is possible through case detection, and many authors have suggested that this has the potential to improve health outcomes through earlier detection and intervention8,60–62.   Existing case detection modalities are composed of screening spirometry or questionnaire-based methods8. Handheld flow meters measure airflow limitation using the ratio of FEV1 to forced expiratory volume in 6 seconds (FEV6). They are often used in case detection because they are cheaper and quicker to administer than diagnostic spirometry66. Patients with evidence of airflow limitation using a handheld flow meter are referred for high-quality diagnostic spirometry to confirm a diagnosis. In five studies in which patients were opportunistically recruited from general practices (irrespective of respiratory symptoms) and screened with handheld flow meters, the yield of new COPD cases per subject screened was between 6 and 20%59, suggesting that the efficiency of these programs is variable.  Questionnaire-based case detection methods typically use one of two validated questionnaires: the COPD diagnostic questionnaire (CDQ)67,68, or the lung function questionnaire69. The CDQ is most extensively used and has been externally validated four times59,70.  It assesses risk factors and respiratory symptoms to determine whether a patient should undergo diagnostic spirometry. Different thresholds for referral have been used66; typically, a total score  ≥19.5 indicates a high risk of COPD, and a score between 16.5 and 19.5 indicates medium risk68 (Table 1-4). 11  Table 1-4 COPD diagnostic questionnaire (CDQ). Questions CDQ Score Age (years)     40-49 0    50-59 4    60-69 8    ≥70 10 BMI     <25.4 5    25.4-29.7 1    >29.7 0 Pack-years of smoking     0-14 0    15-24 2    25-49 3    ≥50 7 Does the weather affect your cough?     Yes 3    No/No cough 0 Do you ever cough up phlegm (sputum) from your chest when you don’t have a cold?     Yes 3    No 0 Do you usually cough up phlegm (sputum) from your chest first thing in the morning?     Yes 0    No 3 How frequently do you wheeze?     Occasionally or more often 4    Never 0 Do you have or have you ever had allergies?     Yes 0    No 3  This table is reproduced from Stanley et al.70 and licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.  12  1.8 Current knowledge gaps A major challenge for improving the early detection of COPD is that it is not clear which individuals are most likely to have undiagnosed COPD. Identifying subgroups of patients with a high prevalence of underdiagnosis can facilitate targeted case detection, which would increase the value of any early detection strategy. To date, there have been several studies comparing the clinical characteristics and risk factors of patients with a previous diagnosis to those who are undiagnosed46,47,71. However, the data collection and analysis standards of these studies are variable, and a formal meta-analysis to generate pooled estimates has not been performed. I hypothesise that there are easily identifiable patient characteristics that can predict the likelihood of a patient having undiagnosed COPD.   Respiratory symptoms are increasingly recognised as an important component of the burden of COPD on patients and healthcare systems72. However, in contrast to other components of the natural history of COPD such as lung function28 and exacerbations30,32, heterogeneity in symptoms has not been well studied. Patients with COPD generally seek medical attention when they experience respiratory symptoms, and evaluating the natural history of symptoms in undiagnosed patients can inform opportunities for case detection. The traditional paradigm is that respiratory symptoms increase linearly in conjunction with lung function decline73. However, more recent evidence suggests that symptoms are only weakly correlated with objective measures of airflow obstruction74. There is a need to characterise heterogeneity in the presence of respiratory symptoms among patients with COPD, and to quantify the extent to which lung function and other easily measured clinical characteristics can explain this heterogeneity. Case detection strategies rely on encounters between undiagnosed COPD patients and the healthcare system. Understanding patterns of healthcare services use among patients with 13  undiagnosed COPD can provide critical information for developing case detection strategies. Previous studies have reported an increase in healthcare utilisation two75 and five76 years prior to an initial diagnosis of COPD. Jones et al.76 found a missed opportunity for an earlier diagnosis in 85% of COPD patients. However, these studies did not assess the type of healthcare encounters that occur most frequently, and in comparison to a control cohort. In order to devise more efficient case detection strategies, the types of visits that offer high value opportunities for case detection should be determined.  Finally, despite information on the performance of existing case detection modalities59, there is very little evidence on which of many proposed case detection techniques will provide the best value for money in different patient subgroups and healthcare settings. The lack of formal evaluations of the cost-effectiveness of case detection strategies is in stark contrast to other disease areas, where economic evaluations have played a major role in the adoption and implementation of screening and case finding programs9–11. It is increasingly accepted that evidence on cost-effectiveness is required before implementation of any early detection strategy77.   1.9 Objectives Given these knowledge gaps, my objectives are as follows: 1) evaluate existing evidence on the relationship between patient characteristics and the presence of undiagnosed COPD; 2) assess heterogeneity in the respiratory symptoms of patients with COPD and the extent to which easily measured patient and disease factors can explain this heterogeneity; 3) characterise the routine healthcare encounters of COPD patients before diagnosis; 4) evaluate the cost-effectiveness of 14  case detection strategies for COPD, including initial criteria for selecting patients, case detection method, and disease management strategies for newly diagnosed patients.   1.10 Thesis themes The overarching themes of my work are evidence generation to understand patterns of care for COPD and the natural history of disease, followed by whole disease modelling to evaluate policy.   1.10.1 Evidence generation To evaluate case detection strategies, I generated evidence on the characteristics of undiagnosed COPD patients, heterogeneity in respiratory symptoms, and patterns of care that give rise to opportunities for an earlier diagnosis. In addition to addressing stand-alone research questions related to undiagnosed COPD, each chapter was also used as an input in the disease model that was used to evaluate the cost-effectiveness of case detection (Figure 1-1).   15  Figure 1-1 Diagram of the simulated case detection pathway and sources of evidence.  My first objective (Chapter 2) was to summarise existing evidence on the characteristics of undiagnosed COPD patients. Identifying these patients can increase our understanding of current practices around diagnosis of COPD in the community, and help identify subgroups of the population that should be targeted for case detection. This evidence was also used to identify components of disease that should be included in the whole disease model in order to simulate routine diagnosis of COPD. This was necessary to evaluate early detection strategies in Chapter 5. My second objective (Chapter 3) was to quantify heterogeneity between individuals in the occurrence of respiratory symptoms. The results of Chapter 2 indicated that symptoms were highly influential in determining whether a patient with COPD had received a diagnosis. Therefore, the respiratory symptoms of COPD patients needed to be simulated in the whole disease model. Symptoms are also a core component of the natural history of COPD, and the highly anticipated paradigm shift from population-level to individual-level decision making in healthcare requires robust evidence on heterogeneity in disease outcomes78. In many cases, sufficient evidence to enable this paradigm shift does not exist. Information on heterogeneity between individuals in the natural history of disease is fundamental to developments in precision medicine, such as identifying COPD phenotypes to characterise disease trajectory and guide treatment decisions. By explicitly incorporating heterogeneity between simulated individuals in disease characteristics, the disease model can also be used as a platform to evaluate policies in precision medicine. 16  My third objective (Chapter 4) was to assess the type and frequency of encounters between undiagnosed COPD patients and the healthcare system. Case detection strategies rely on opportunistic encounters, and identifying patterns of care among undiagnosed COPD patients is therefore critical to their effectiveness. In addition to informing the feasibility of case detection, I used the results of this chapter to identify the type of healthcare visits to simulate in the disease model. These visits were opportunities for simulated individuals to receive case detection. My final objective (Chapter 5) was to evaluate the cost-effectiveness of case detection for COPD. Case detection strategies were composed of eligibility characteristics to receive case detection, the case detection modality (questionnaire- or spirometry-based case detection methods), and the frequency at which case detection was administered (3- or 5-year intervals). I assessed the total costs and quality adjusted life years (QALYs) of patients receiving case detection, compared to the status quo scenario with no case detection. I determined the cost-effectiveness of each case detection strategy, identified a preferred strategy, and assessed the influence of various aspects of treatment on the value of early detection. The whole disease model that was used to evaluate case detection strategies is discussed in the next section.  1.10.2 Whole disease modelling An objective assessment of case detection strategies requires a decision-analytic model that is capable of translating the performance of combinations of selection characteristics and case detection method into the long-term economic and health consequences of early detection. In this context, the overarching goals of decision-analytic modelling are threefold: 1) to incorporate evidence from diverse sources on the performance of case detection methods among various subgroups of patients; 2) to generate predictions over a long-time frame that may exceed the 17  assessment periods of empirical studies; and, 3) to translate the available evidence from intermediate outcomes (e.g., FEV1) into final outcome measures of interest (e.g., costs and quality of life). Decision modelling can provide objective and evidence-informed comparisons of the value of different case detection strategies in terms of their overall costs and benefits.  Any strategy for improving a diagnosis of COPD needs to be combined with disease management to be impactful. Therefore, the efficiency of case detection is closely linked to the effectiveness of treatment for diagnosed patients. An improvement in treatment would greatly increase the value of early detection. In order to account for this potential feedback, both diagnosis and treatment need to be modelled simultaneously. A class of models, referred to as ‘whole disease models’, are particularly well suited to these types of interactions between model components79. Whole disease models simulate the entire trajectory of disease from biological inception until death. This is critical for policy design, as it allows a fuller exploration of the decision space created by a multitude of patient- and disease-related factors. Whole disease models can be used to evaluate interventions at any point along the care pathway. They can account for the downstream impacts of decisions, and the potential for changes that occur later in the care pathway to feed back into the decision point of interest79. I expanded a whole disease model of COPD, referred to as Evaluation Platform in COPD (EPIC)80, by using the evidence generated in Chapters 2-4 to add diagnosis, symptoms, and primary care visits to the model. This allowed me to simulate case detection strategies in EPIC and evaluate their cost-effectiveness.  18  1.10.2.1 Evaluation Platform in COPD (EPIC) EPIC simulates the demographics, risk factors, development, and progression of COPD in the entire Canadian population over 40 years of age from 2015 onwards. The core function of EPIC is to simulate individuals who experience events, such as the occurrence of COPD, at probabilities that are quantified through risk equations. These risk equations are developed using empirical data and serve as inputs to the model. Each component of EPIC has been checked for internal validity, meaning the model produces results that match the patterns observed in the input dataset. A second and more rigorous test of model validity is external validity, which tests whether the model correctly predicts data that was not used in its development. The core functions and structure of EPIC are depicted in Figure 1-2.   Figure 1-2 Flow diagram of Evaluation Platform in COPD model structure. 19  This figure is adapted from Sadatsafavi et al.80.   EPIC models a dynamic (open) population using discrete event simulation79. Individuals are simulated and followed through time as they die, emigrate, or the time horizon is reached. There are five core components to the model. Their development and validation is detailed in a previous study80 and summarised here. (1) In the demographic module, the entire Canadian population ≥40 years of age in 2015 is created. Each simulated individual is assigned an age, sex, and baseline smoking history. The resulting demographic trends were validated against Statistics Canada projections81. (2) The smoking module assesses the smoking behaviour of individuals over time. Individuals can initiate or quit smoking at rates predicted by the Canadian Community Health Survey and CanCOLD45. This module was validated against overall trends in smoking prevalence in Canada, including a 4% annual decline in current smokers39. (3) The COPD module simulates prevalent and incident COPD. The proportion of individuals with COPD at the start of the model time horizon was validated with the Canadian COPD prevalence study (COLD)45. (4) When COPD occurs, the lung function module assigns a baseline FEV1 using regression equations developed from COLD. The progression of FEV1 over time is determined from previously published equations that were developed using data from the Lung Health Study and have been extensively externally validated28. (5) Simulated individuals could experience exacerbations at rates and severities that were validated with a meta-analysis82 and the MACRO clinical trial83. There was also a probability of mortality associated with severe exacerbations. (5) Finally, the payoff module assigns background costs84 and utilities84–86 stratified by GOLD grade. A cost87,88 and disutility84 is also associated with exacerbations. Overall mortality, and the 20  rate and severity of exacerbations generated by EPIC were validated against two external cohort studies (TORCH and UPLIFT) with robust results80.  1.11 Thesis summary In this thesis, I identified patient and disease characteristics associated with a diagnosis of COPD, quantified heterogeneity in the respiratory symptoms of COPD patients, which are a key component of the natural history of disease, and assessed healthcare encounters among undiagnosed COPD patients. The results of each chapter informed the simulation of case detection in EPIC, which was used to evaluate the cost-effectiveness of early detection strategies for COPD. In Chapter 2, I performed a systematic review and meta-analysis to compare patient and disease characteristics between individuals with undiagnosed persistent airflow limitation and those with diagnosed COPD. I extracted and pooled summary data on the proportion or mean of each risk factor among diagnosed and undiagnosed patients, and coefficients for the adjusted association between risk factors and diagnosis status. This analysis was used to determine which aspects of the natural history of COPD needed to be added to EPIC in order to simulate current practices around COPD diagnosis. In Chapter 3, I used a longitudinal cohort study of the general Canadian population with persistent airflow limitation to assess overall heterogeneity in the occurrence of symptoms. I determined the proportion of heterogeneity in the burden of symptoms that could be explained by lung function versus all other measured clinical characteristics of participants. The risk equations developed in this chapter were used to simulate respiratory symptoms in EPIC. 21  In Chapter 4, I used administrative health data for the province of British Columbia to create a retrospective cohort of COPD patients before diagnosis. I assessed the rate of visits to pharmacists, primary care physicians, and specialist physicians in the 5 years prior to an initial diagnosis of COPD. I compared this to non-COPD subjects matched on age, sex, and socioeconomic status. The results of this chapter were used to identify routine healthcare encounters that could be used as opportunities for case detection of undiagnosed patients.  In Chapter 5, I used EPIC to assess the cost-effectiveness of 16 case detection scenarios. Individuals meeting selection criteria based on age, smoking history, or symptoms, received the CDQ or screening spirometry (using a handheld flow meter) during their routine visits to a primary care physician. I determined the case detection strategy with the highest value, and provided evidence on whether early detection of COPD is a worthwhile investment of resources. In Chapter 6, I conclude by summarizing my findings on the value of early detection for COPD, and suggest specific policies that should be targeted for implementation. I identify key messages for improving existing practices around the diagnosis of COPD in the community, and suggest changes to current guidelines that reflect this. Finally, I critique my research in terms of its strengths and limitations, and discuss future work to facilitate policy development and adoption. The results of this thesis can help identify undiagnosed COPD patients, improve our understanding of disease heterogeneity, inform opportunities for diagnosis, and provide an objective evaluation of the value for money of early detection strategies for COPD. Evidence on the impact of an early diagnosis on the long-term health outcomes of patients with COPD, and the cost-effectiveness of strategies for improving detection, can help reduce the large burden of COPD on patients and the healthcare system. 22  Chapter 2: Characterising Undiagnosed Chronic Obstructive Pulmonary Disease: a Systematic Review and Meta-Analysis  2.1  Introduction Chronic obstructive pulmonary disease (COPD) is a lung disorder that is characterised by persistent airflow limitation15 and associated with symptoms of shortness of breath, cough, and sputum production16. Patients with COPD generally seek medical attention when they experience respiratory symptoms, most notably dyspnea that is persistent and progressive15. However, owing to under-utilisation of lung function measurements and non-specific nature of the symptoms, COPD is often not recognised until late in the disease process. Indeed, many patients do not receive a diagnosis of COPD until after being hospitalised due to a severe exacerbation44.  Lamprecht et al.47 reported an average underdiagnosis rate of 81% in a prevalence study that included 30,874 participants across 44 countries. Reducing risk factors such as smoking and occupational risk factors while the disease is early in its progression is an important component of treatment for COPD59. As such, late diagnosis of COPD represents a missed opportunity to modify the course of the disease through evidence-informed risk factor management and treatment54,61. The extent of this missed opportunity is a function of the number of COPD patients who are undiagnosed, and the burden of disease (e.g., symptom burden, lung function status) in this population.  Quantifying the true burden of undiagnosed COPD and the benefit of screening and case detection can be informed by a comparative assessment of patient and disease factors between diagnosed and undiagnosed patients. Numerous studies have compared the characteristics of 23  patients with undiagnosed and diagnosed COPD, but to the best of my knowledge, these studies have never been systematically compiled and pooled. The objective of this review was to synthesise existing evidence on patient and disease factors in patients with undiagnosed persistent airflow limitation compared to those with diagnosed COPD. I hypothesised that the characteristics of patients, their risk factors, respiratory symptoms, and disease stage influence the likelihood of receiving a diagnosis of COPD.  2.2 Methods 2.2.1  Search strategy and selection criteria The protocol for this study is registered on the PROSPERO register of systematic reviews (CRD42017058235)89. I conducted a systematic review and meta-analysis to compare patient characteristics, risk factors, and symptoms in diagnosed and undiagnosed patients. I searched MEDLINE and EMBASE using the Ovid interface for eligible articles. The search strategy (Appendix A.1) was developed in MEDLINE and adapted to EMBASE using appropriate vocabulary terms. I included longitudinal or cross-sectional studies published in English between 1980 and April 11, 2017 that were based on original analysis of individual data. I did not assess grey literature but conference abstracts were eligible if they provided all the required information. I extracted summary data from the eligible articles and contacted the authors to obtain additional information when required (one author group provided us with additional information). Title and abstract screening were initially performed, followed by full-text analysis to determine article eligibility. I extracted data using a customised Excel spreadsheet after the eligible articles had been compiled. I initially performed the selection procedure, and a second reviewer independently repeated each step on a subset (10%) of articles. Discrepancies were 24  resolved through a discussion with the second reviewer. Duplicate articles found in both MEDLINE and EMBASE were identified using a reference manager and manually removed. I used the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies developed by the National Institutes of Health90 to assign an overall quality rating (good, fair, or poor) to each study. I extracted relevant data and assessed the quality of the included studies, and the second reviewer replicated the assessment on 10% of articles. Both reviewers determined the overall quality of each article by assigning 'yes', 'no', or 'other' (cannot determine, not applicable, or not reported) to 14 questions relating to external validity, bias in the measurements of the risk factors or outcomes, and confounders present in the study. The results of this assessment were interpreted qualitatively.  The population of interest in this review was adult patients (≥18 years old) with persistent airflow limitation at the time of assessment. Persistent airflow limitation was defined when the study subjects demonstrated a ratio of forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) < 0.7 (fixed ratio definition)15 or FEV1 to FVC lower than the lower limit of normal (LLN definition)40 after the administration of a bronchodilator during spirometry. Study subjects who had airflow limitation and also a prior diagnosis of COPD or an obstructive lung disease (emphysema, chronic bronchitis, asthma) from a healthcare professional were considered to have ‘diagnosed’ COPD, whereas those with persistent airflow limitation but without a prior health professional diagnosis of COPD were considered to be ‘undiagnosed’. Studies in which COPD was not the primary disease of interest were excluded. I included studies that used either population-based (random sampling of the general population) or convenience (e.g., recruiting patients from a healthcare setting) sampling.  25  Given the exploratory nature of the observational studies included in this review, I used a broad definition of risk factors that included any observable factor that could be associated with the probability of having received a diagnosis of COPD. Risk factors included patient-reported respiratory symptoms (cough, wheeze, phlegm, dyspnea), sex, age, current smoking status, smoking history (pack-years), and disease severity classified using the Global Initiative for chronic Obstructive Lung Disease (GOLD) grades. The relationship of interest was the association between these risk factors and the probability of having ‘diagnosed’ COPD among patients with persistent airflow limitation.  I extracted summary data from each eligible article, which included study characteristics, the definition of persistent airflow limitation that was employed in each of the studies, the method of COPD diagnosis, sampling methodology, and sample size. I also extracted the proportion or mean of risk factors between the diagnosed and undiagnosed groups, as well as the odds ratios (ORs) and their confidence intervals in studies that used regression modelling to assess the independent impact of the risk factors on diagnosis status. Studies reporting the characteristics of diagnosed and undiagnosed patients using means or proportions in contingency tables were pooled in an ‘unadjusted analysis.’ Studies that reported associations using multivariable regression modelling were pooled in a separate ‘adjusted analysis.’  2.2.2 Data analysis I used data extracted from articles measuring categorical data to generate ORs and standard errors for the association between risk factors and the probability of having received a diagnosis of COPD. In articles assessing continuous data, I calculated the mean difference (MD) in risk factors and their standard errors among diagnosed and undiagnosed patients. I pooled the ORs or 26  MDs from individual studies using the inverse variance method implemented with the ‘meta’ package91 in R Statistical Software92 (version 3.3.3). I used fixed-effects models when estimates from only two studies were being pooled, or if the null hypothesis that all studies evaluated the same effect was not rejected (at 0.05 significance level) using Cochran’s Q statistic93. Otherwise, I used random-effects models. I quantified heterogeneity between studies using the I2 statistic94. I did not pool together studies that used alternate definitions of persistent airflow limitation (fixed ratio and LLN) given the demonstrated differences in patients meeting these criteria95. When separate studies used subsets of the same dataset (i.e., the Latin American Project for the Investigation of Obstructive Lung Disease [PLATINO] dataset47,96–98), I used the estimate from the study with the largest sample size. I conducted a sensitivity analysis to determine the association between risk factors and COPD diagnosis only among population-based studies (as opposed to convenience sampling).   2.3 Results The search resulted in 1,857 references after excluding duplicates. 1,788 references were excluded by screening their titles and abstracts, and 69 remained for full-text review to determine eligibility. A total of 18 articles met the inclusion criteria following the screening process, but only 16 articles were included in the quantitative synthesis (Figure 2-1). Two eligible articles were excluded from the meta-analysis because they were missing the necessary information99, or did not measure any risk factors in common with other studies100. The overall agreement between reviewers was high (90%).  27   Figure 2-1 Preferred reporting items for systematic reviews and meta-analyses flow diagram.  A summary of the 16 eligible articles is presented in Table 2-1. The majority of the 16 eligible articles were cross-sectional (n=14), and population-based (n=9). Other studies sampled patients from primary care clinics (n=4), or among hospitalised patients (n=3). Studies originated from Europe (n=6), Latin America (n=5), Canada (n=2), and Asia (n=1). Data from the Epidemiologic Study of COPD in Spain (EPI-SCAN)47,101,102, PLATINO, and the Burden of Obstructive Lung Disease (BOLD)47,103, were used in three, four, and two different studies, 28  respectively, but only one study from each dataset was included in pooled analyses. The definition of persistent airflow limitation varied between articles; 13 studies defined it as the fixed ratio, two studies used the LLN definition, and one study reported results using both definitions. Only one of the eligible articles71 included asthma as a risk factor for assessing a previous diagnosis. The percentage of patients with undiagnosed persistent airflow limitation was greater than 50% in all but two studies (which sampled from healthcare settings).  The quality of the 16 eligible articles was variable. Nine studies were assigned a quality rating of ‘good’, six studies were assigned a rating of ‘fair’, and one was deemed poor in quality. Studies that were not assigned a ‘good’ quality rating generally had a primary study focus that was not my question of interest. The use of regression modelling to examine the independent impact of risk factors on the likelihood of receiving a COPD diagnosis was performed in seven studies, but the risk factors that were adjusted for varied substantially across studies. 29  Table 2-1 Characteristics of selected studies.  Country Study type Population Definition of COPD Definition of undiagnosed COPD Participants with COPD Percentage undiagnosed Quality rating Ancochea et al. (2013)101 Spain Cross-sectional (EPI-SCAN*) General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of COPD (self-reported) 386 73% Good Balcells et al. (2015)44 Spain Prospective cohort study  Hospitalised patients, all eligible patients were invited Post-bronchodilator FEV1/FVC<0.7, 3 months after discharge Spirometric obstruction and no diagnosis of respiratory disease or regular use of pharmacological respiratory treatment (self-reported) 342 34% Good Herrera et al. (2016)104 Argentina, Colombia, Venezuela, Uruguay Cross-sectional  Primary care clinics, convenience sample Post-bronchodilator FEV1/FVC<0.7 and LLN Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 309 77% Fair Hill et al. (2010)105 Canada Cross-sectional Primary care clinics, convenience sample Post-bronchodilator FEV1/FVC<0.7 and FEV1<80% predicted Spirometric obstruction and no previous diagnosis of COPD based on medical chart review over the previous 12-months 107 46% Good Hvidsten et al. (2010)71 Norway Cross-sectional  General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spriometric obstruction and being treated by a physician or admitted to hospital for a diagnosis of obstructive lung disease (asthma, chronic bronchitis, emphysema, or COPD) in the previous 12-months (self-reported) 303 66% Good Labonté et al. (2016)46 Canada Prospective cohort study  General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 505 70% Fair Lamprecht et al. (2015)47 Global Cross-sectional (BOLD†, PLATINO‡, General Population, random sample Post-bronchodilator FEV1/FVC<LLN Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 2992 81% Good 30   Country Study type Population Definition of COPD Definition of undiagnosed COPD Participants with COPD Percentage undiagnosed Quality rating EPI-SCAN, PREPOCOL§) Llordes et al. (2015)106 Spain Cross-sectional Primary care clinic, all eligible patients were invited Post-bronchodilator FEV1/FVC<0.7 in 2 tests 4 weeks apart (the 2nd after 4 weeks of pharmacological treatment) Spirometric obstruction and no previous diagnosis of COPD in medical reports 422 57% Fair Mahishale et al. (2015)107 Not reported Cross-sectional Hospitalised patients, convenience sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of COPD (self-reported) 404 56% Poor Miravitlles et al. (2009)102 Spain Cross-sectional (EPI-SCAN) General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 408 73% Good Moreira et al. (2013)96 Brazil Cross-sectional (PLATINO) General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 53 62% Fair Nascimento et al. (2007)97 Brazil Cross-sectional (PLATINO) General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 144 88% Fair Queiroz et al. (2012)108 Brazil Cross-sectional Primary care clinics, convenience sample  Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 63 71% Good Schirnhofer et al. (2011)103 Austria Cross-sectional (BOLD) General Population, random sample Post-bronchodilator FEV1/FVC<LLN Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 199 86% Good 31   Country Study type Population Definition of COPD Definition of undiagnosed COPD Participants with COPD Percentage undiagnosed Quality rating Talamo et al. (2007)98 Brazil, Chile, Mexico, Uruguay, Venezuela Cross-sectional (PLATINO General Population, random sample Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and no previous diagnosis of chronic bronchitis, emphysema, or COPD (self-reported) 758 89% Good Zhang et al. (2013)109 China Cross-sectional Hospitalised patients, all eligible patients were invited Post-bronchodilator FEV1/FVC<0.7 Spirometric obstruction and COPD not recorded as a discharge diagnosis in medical records 705 93% Fair  * Epidemiologic Study of COPD in Spain (EPI-SCAN)  † Burden of Obstructive Lung Disease (BOLD) ‡ Latin American Project for the Investigation of Obstructive Lung Disease (PLATINO) § Prevalence study of COPD in Colombia (PREPOCOL)  32  2.3.1 Unadjusted analysis The characteristics of diagnosed and undiagnosed patients meeting the fixed ratio definition of airflow limitation were compared using contingency tables (‘unadjusted analysis’) in 12 studies. The pooled results are shown in Figure 2-2. Patients with ‘diagnosed’ COPD were more likely to be experiencing respiratory symptoms such as wheezing (OR 3.51, 95% CI 2.19-5.63, 3 studies), phlegm (OR 2.16, 95% CI 1.38-3.38, 3 studies), dyspnea (OR 4.67, 95% CI 2.62-8.35, 3 studies), or any respiratory symptoms (OR 11.45 95% CI 7.20-18.21, 3 studies). They were much less likely to have mild (GOLD grade I) COPD than moderate to very severe COPD (grade II-IV) as measured by GOLD grades (OR 0.30 95% CI 0.24-0.37, 7 studies). The heterogeneity between studies was relatively low (I2<35.0% for all symptoms and COPD severity). Patient sex, current smoking status, and smoking history were not associated with a COPD diagnosis. Having a cough was also not significantly associated with diagnosis status; however, variability between the three studies measuring this risk factor was particularly high (I2 77.9%).   33   34  Figure 2-2 Associations between diagnosed (v. ‘undiagnosed’) COPD and sex, the presence of cough, wheeze, phlegm, dyspnea, any respiratory symptoms, smoking status, smoking history, and COPD severity based on contingency tables (‘unadjusted analysis’).  Persistent airflow limitation was defined as post-bronchodilator FEV1/FVC<0.7. Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).   Sensitivity analysis of only the population-based studies (excluding those that used convenience sampling) revealed very similar results (n=5 studies, Appendix A.2). Pooled analysis of two studies103,104 using the LLN definition of airflow limitation was consistent with the findings based on fixed ratio results (Appendix A.3); however, cough was marginally associated with diagnosis status in this analysis (OR 1.65, 95% CI 1.02-2.66). Similarly, patients with ‘diagnosed’ COPD (fixed ratio definition) were more impaired by dyspnea (modified Medical Research Council [mMRC] dyspnea scale52 MD 0.52, 95% CI 0.40-0.64, 3 studies) and had greater airflow obstruction (percent predicted FEV1 MD -12.85%, 95% CI -15.26% to -10.44%, 4 studies) than undiagnosed patients (Figure 2-3). Patients with ‘diagnosed’ COPD also had a slightly greater smoking history (pack-years MD 8.39, 95% CI 0.68-16.44, 4 studies); however, there was high variability between the study means (I2 84.2%). There was no difference in mean age between diagnosed and undiagnosed patients.     35   Figure 2-3 Mean difference (MD) in age, pack-years of smoking, mMRC dyspnea score, and percent of predicted FEV1 between diagnosed and undiagnosed categories.   Persistent airflow limitation was defined as post-bronchodilator FEV1/FVC<0.7. Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).  * modified Medical Research Council (mMRC) Dyspnea scale52 means and standard errors (SE) for the diagnosed and undiagnosed categories are multiplied by a factor of 10.   2.3.2 Adjusted analysis Articles using regression modelling to assess the independent impact of risk factors on COPD diagnosis (‘adjusted analysis’) were pooled by risk factor type, and the results are presented in Figure 2-4 for the fixed ratio definition of persistent airflow limitation (5 articles), and Figure 2-5 36  for the LLN definition (2 articles with 5 datasets). Compared with the unadjusted analysis, the effect sizes of the risk factors were attenuated in these analyses. The presence of phlegm had a weak independent association with the diagnosis of COPD (OR 1.16, 95% CI 1.00-1.35, 2 studies) using the fixed ratio definition. The presence of wheezing (OR 1.20, 95% CI 0.99-1.44, 2 studies) and dyspnea (OR 1.13 95% CI 0.99-1.29, 2 studies) showed a trend towards association. Mild COPD (GOLD grade I OR 0.72, 95% CI 0.58-0.80) or moderate COPD (GOLD grade II, OR 0.71, 95% CI 0.58-0.86) were independently associated with a lower likelihood of diagnosis, compared with severe or very severe (reference GOLD grades III-IV). Sex and the presence of cough did not influence the likelihood of being diagnosed in the adjusted analyses. Overall, heterogeneity in the effect estimates between studies was very high (I2>70.0% for all risk factors except sex).  37   Figure 2-4 Associations between risk factors and the odds of receiving a COPD diagnosis using the regression coefficients from studies with multivariable regression modelling (‘adjusted analysis’) and persistent airflow limitation defined as post-bronchodilator FEV1/FVC<0.7.   The reference categories were female, the absence of cough, wheeze, dyspnea, phlegm, and GOLD grades 3 and 4, respectively. Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).  Regression models were adjusted for age (Herrera, Hill, Hvidsten, Miravitlles, Talamo), sex (Herrera, Hill, Hvidsten, Miravitlles, Talamo), ethnicity (Herrera, Talamo), body mass index (Herrera, Hvidsten), education (Herrera, Hvidsten, Miravitlles, Talamo), income (Hvidsten), employment (Talamo), risk factor to dust (Herrera), 38  smoking (Herrera, Hill, Hvidsten, Miravitlles, Talamo), respiratory symptoms, (Herrera, Hill, Hvidsten, Talamo), self-rated health (Hvidsten, Miravitlles), COPD severity (Herrera, Miravitlles, Talamo), comorbidities (Herrera, Hvidsten), prior healthcare use (Herrera, Hill), and exacerbations (Herrera).  *The reference category was changed from GOLD grade 1 to GOLD grades 3 and 4 by assuming a covariance of 0 between the dummy variables representing GOLD grades 1 and 2. †Herrera et al.104 reported prevalence ratios from Poisson regression models.   Three risk factors were pooled in my assessment of studies using adjusted analysis based on the LLN definition of persistent airflow limitation. This analysis indicated a more strongly positive association between the presence of phlegm and being diagnosed with COPD (OR 1.71, 95% CI 1.15-2.56), although there was heterogeneity between datasets (I2 75.2%). Patient sex and the presence of cough had no independent effects.    39  Figure 2-5 Associations between risk factors and the odds of receiving a COPD diagnosis using the regression coefficients from studies with multivariable regression modelling (‘adjusted analysis’) and persistent airflow limitation defined as post-bronchodilator FEV1/FVC<LLN.   The reference categories were female, and the absence of cough and phlegm, respectively. The results for each dataset (BOLD, PLATINO, EPI-SCAN, PREPOCOL) analysed in Lamprecht et al.47 were pooled separately. Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).  Regression models were adjusted for age (Herrera, Lamprecht), sex (Herrera, Lamprecht), ethnicity (Herrera), body mass index (Herrera), education (Herrera, Lamprecht), risk factors to dust (Herrera), smoking (Herrera, Lamprecht), respiratory symptoms (Herrera, Lamprecht), COPD severity (Herrera, Lamprecht), comorbidities (Herrera), and prior healthcare use (Herrera, Lamprecht).  2.4 Discussion I conducted a systematic review and meta-analysis to compare patient and disease factors between patients with undiagnosed persistent airflow limitation and those with diagnosed COPD. In the 16 observational studies included in this review, the presence of respiratory symptoms and GOLD grade III or IV disease severity were strongly associated with a prior diagnosis of COPD among individuals with persistent airflow limitation on spirometry. These findings were relatively consistent across analysis methods and alternate definitions of persistent airflow limitation. Greater disease severity was the most important characteristic of diagnosed patients in two out of three pooled analyses. In particular, patients with mild or moderate COPD (as measured by GOLD grades) were 78% less likely to have received a diagnosis than patients with severe or very severe COPD in the unadjusted analysis (based on contingency tables), and mean percent predicted FEV1 was 13% lower in diagnosed than undiagnosed patients. Disease severity 40  was also the only risk factor that was associated with a diagnosis in both the unadjusted and adjusted (based on regression modelling) analyses. In the adjusted analysis, patients with moderate COPD were 29% less likely to have received a diagnosis than patients with severe or very severe COPD. Among respiratory symptoms, the presence of dyspnea was the most strongly associated with a previous diagnosis in the unadjusted analysis. Patients with ‘diagnosed’ COPD scored 0.52 points higher on the mMRC dyspnea scale. One study108 provided evidence that the mean score on the mMRC scale could have been used to distinguish undiagnosed from diagnosed patients using commonly accepted criteria (‘more dyspnea’ if mMRC score ≥2 v. ‘less dyspnea’ if mMRC score <2)15. Following dyspnea, the presence of wheeze and phlegm were also strongly associated with ‘diagnosed’ COPD in the unadjusted analysis. These associations were weaker in the adjusted analysis but were still present. Interestingly, the presence of coughing was not well associated with a previous diagnosis in any of the pooled analyses. Overall, aside from the attenuated results in the adjusted analysis (discussed in detail below), my findings suggest a strong association between the presence of dyspnea, phlegm, or wheeze and a COPD diagnosis. This association is likely because patients with respiratory symptoms are more likely to seek care, and doctors are more likely to test for COPD in patients with symptoms15.  Overall, these findings suggest that a diagnosis of COPD is likely to be delayed in patients with a lower burden of respiratory symptoms and lung function decline that progresses more slowly, meaning that opportunities for early intervention may be lower in these patients. Patients with fewer respiratory symptoms are also less likely to be screened for COPD following the U.S. Preventive Services Task Force recommendation65. However, my results suggest that screening or case detection methods that rely exclusively on the presence of respiratory 41  symptoms may be missing undiagnosed patients. Symptoms should be assessed in combination with a history of exposure to risk factors for COPD when considering a diagnosis15.   Importantly, sex, age, and smoking status were not independently associated with receiving a diagnosis of COPD in any of the pooled analyses. Because these are major risk factors for the presence of COPD15, physicians should be more likely to diagnose COPD in older patients who are current or past smokers, which should have resulted in an independent association between these factors and a diagnosis of COPD. The lack of such association in any of my analyses may indicate that the relation between these risk factors and the presence of COPD is not sufficiently recognised by physicians. This is in contrast to the strong association of GOLD grade and symptoms with diagnosis status, which suggests that a diagnosis of COPD is more commonly made based on the degree of impairment in lung function and/or patient respiratory symptoms. This hypothesis might also explain my finding that smoking history in terms of pack-years was associated with a previous diagnosis of COPD while current smoking status was not, as only smoking history is related to cumulative lung function impairment.  The effects of risk factors on the likelihood of being diagnosed were weaker in the adjusted analyses than in the unadjusted analyses. The adjusted analyses were based on pooled coefficients from regression modelling. Although the inclusion of covariates is expected to reduce the effects sizes compared to odds ratios derived from contingency tables (as in the unadjusted analysis), one study in the adjusted analysis104 had unusual results that received disproportionate weighting. In contrast to all other studies in this review, Herrera et al.104 found that respiratory symptoms were not associated with the likelihood of having received a diagnosis of COPD. In the adjusted analysis, these results were pooled with one other study98, which found that the presence of respiratory symptoms were strongly associated with the likelihood of 42  receiving a diagnosis. This discrepancy between studies may be due to differences in the population that was sampled (primary care clinic104 versus general population98). In general, studies in clinic settings might have observed smaller differences between undiagnosed and diagnosed patients because they sampled from a subset of patients that were prompted to seek care because of symptoms. This systematic review has several strengths. First, I used data from a total of 16 articles in the meta-analysis, and these articles were mostly population-based studies that scored high in quality. Second, there were a robust number of studies for many risk factors; patient sex was assessed in 10 studies in total, followed by disease severity in 9 studies, and respiratory symptoms and smoking history in 8 studies. The methods used to measure disease severity, respiratory symptoms, and smoking history were relatively consistent across studies, which facilitated pooling of their findings. Lastly, I conducted several pooled analyses to assess the sensitivity of my findings to alternate definitions of COPD (fixed ratio and LLN) and analysis methods (unadjusted and adjusted analyses). Except for one study104, my findings were consistent. This systematic review also has several limitations. First, half of the pooled samples were based on data from three large prevalence studies (EPI-SCAN, PLATINO, and BOLD). This resulted in overrepresentation of patients in Spain and Latin America, and differences in patient and physician behaviour and healthcare services use can result in findings that vary across settings. Second, although the total number of studies for each risk factor was robust, the number of studies assessing each risk factor within pooled analyses tended to be small. This was partly because separate articles using the same dataset could not be combined in our pooled analyses. Third, with the exception of dyspnea, all other respiratory symptoms in the pooled analyses were 43  measured as binary variables (either present or absent), and all symptoms were measured cross-sectionally. Given my finding that symptoms are characteristic of a COPD diagnosis, a more nuanced assessment of their severity and temporal variability might result in an even greater ability to distinguish between undiagnosed and diagnosed patients. In addition, because respiratory symptoms were self-reported in all studies, reporting bias might have exaggerated the difference in symptoms between the undiagnosed and diagnosed groups. Finally, several studies reported additional comparisons of risk factors between diagnosed and undiagnosed patients. Examples include education, income, comorbidities, quantity and type of care, and a previous diagnosis of asthma. However, due to inconsistent definitions and different categorizations, I could not pool these estimates across studies. An important knowledge gap is the impact of environmental risk factors on the likelihood of receiving a diagnosis, which was rarely measured in the included studies.   2.5 Conclusions The findings from this systematic review have important implications for research and policy around COPD diagnosis, for example, in estimating the return on investment in screening and case detection strategies for COPD. The true burden of COPD is the sum of the disease burden in diagnosed and undiagnosed patients. My results indicate that undiagnosed patients generally have milder disease and therefore a lower disease burden, meaning that strategies aiming to reduce the underdiagnosis problem are unlikely to result in immediate and dramatic improvements in patient-related outcomes such as symptoms. However, the gap in disease severity and symptom burden between diagnosed and undiagnosed patients also indicates a delay in COPD diagnosis. Given the potential for disease modification at early stages of COPD, 44  reducing this delay could be associated with substantial improvement in long-term patient outcomes and a reduction in mortality and costs. 45  Chapter 3: Heterogeneity in the Respiratory Symptoms of Patients with Mild-to-Moderate COPD  3.1 Introduction Chronic obstructive pulmonary disease (COPD) is a common inflammatory lung condition that affects close to 400 million people worldwide35. COPD is characterised by persistent airflow limitation and symptoms such as breathlessness, chronic cough, sputum production, wheezing, and chest tightness15. Respiratory symptoms are a major burden in many patients, and are associated with an increased frequency of exacerbations110, worse disease prognosis72,111,112, lower health status113,114, reduced quality of life115, and higher healthcare resource utilisation116.   The three major components of the natural history of COPD are lung function status, patterns of exacerbations, and symptom burden15. Modern guidelines such as the Global Initiative for Chronic Obstructive Lung Disease (GOLD) appreciate the importance of all three components in disease management decisions. The GOLD guidelines recommend evaluating symptoms separately from airflow limitation and history of exacerbations in providing therapeutic recommendations15.  It is increasingly recognised that COPD is a heterogeneous disease. Individuals can vary markedly in their rate of lung function decline28 and frequency of exacerbations117,118 over the course of their disease. For example, COPD patients in the Lung Health Study had an annual rate of change in FEV1 that ranged from rapidly declining to modestly increasing (95% CI -83 mL/yr to +15 mL/yr)28. Similarly, the annual rate of exacerbations observed in the MACRO clinical 46  trial varied from 0.47 to 4.22118. Quantifying this variation at an individual level is critical to enabling precise risk factor and disease management14.  In contrast, heterogeneity in the burden of symptoms has not received the same level of attention as these other disease components. This is despite the fact that the degree of symptom impairment is increasingly recognised as an important determinant of patient management strategies, and one that is only partially dependent on the severity of airflow limitation72,74,113,119. Previous studies have reported that patient symptoms tend to vary over the day, week, or season74,120–122, but variation between individuals in the occurrence of symptoms has been less well characterised. Understanding the extent and drivers of this heterogeneity can help improve our understanding of the natural history of COPD and ultimately help formulate disease management strategies that provide optimal therapeutic strategies for each patient. Using data from a population-based prospective cohort, I assessed the burden of self-reported respiratory symptoms in patients with persistent airflow limitation in order to (1) characterise variation in the occurrence of symptoms between individuals, and (2) determine the proportion of between-individual variability in symptoms that can be explained by lung function versus all other observable characteristics. I hypothesised that there is high variability in the occurrence of symptoms between individuals, and that an individual’s clinical and demographic characteristics explain a larger fraction of this heterogeneity than lung function alone.    3.2 Methods I used data from the Canadian Cohort of Obstructive Lung Disease (CanCOLD), which is a multicentre prospective longitudinal cohort study conducted across Canada45. Individuals ≥40 years old were recruited using random digit dialing and multi-level sampling to ensure 47  representativeness of the general Canadian population. Participants were followed for a maximum of 3 years with in-person visits at baseline and 18-month intervals. From the entire cohort of CanCOLD participants (N=1,561), I selected any visits in which the participant had persistent airflow limitation, defined as post-bronchodilator FEV1/FVC< lower limit of normal (LLN)41. As a result, I could have included any combination of the three study visits per participant. Participants who had airflow limitation at one visit but not at any of their subsequent visits were excluded, as their airflow limitation was not considered persistent. I also excluded participants who did not meet the clinical definition of COPD because they were asymptomatic throughout follow-up and had no smoking history15. The sample selection procedure is shown in Figure 3-1.  Figure 3-1 Sample selection procedure. 48  Information was collected during each visit on the presence of cough, phlegm, wheeze, and dyspnea using separate questions for each symptom. Participants reported whether they (1) usually coughed in the absence of a cold, (2) brought up phlegm from the chest in the absence of a cold, and (3) experienced any wheezing or whistling in the chest. Dyspnea (4) was measured using the Medical Research Council (MRC) dyspnea scale52, which was converted to a binary variable by assuming that a score of 2-5 indicated the presence of dyspnea. The questionnaire used to assess symptoms is reproduced in Appendix B.1. Dyspnea and whether the participant experienced any symptoms were assessed in a subset of the data that included 856 visits from 415 participants because 34 participants were unable to walk and therefore did not complete the MRC dyspnea test. Other variables that were assessed at each visit and included in this analysis were: demographic information, smoking status and history, number of comorbidities, previous diagnosis of anxiety, major or minor depression, history of physician-diagnosed COPD (including emphysema and chronic bronchitis), and history of physician-diagnosed asthma, all self-reported using validated questionnaires with a recall period spanning the length of time between visits45. Participants also reported the frequency and type of all respiratory-related medication use, and previous exacerbations of any severity15, both with a 12-month recall period.  3.2.1 Statistical analysis I used separate mixed effect logistic regression models for cough, phlegm, wheeze, dyspnea, and any symptoms to model heterogeneity. A random effect term captured the variability among individuals (heterogeneity) that was not attributable to the independent variables in the model. I initially determined the total heterogeneity in the occurrence of symptoms using an intercept-only random-effects logistic regression model for each symptom (the null model, i.e., no 49  independent variables). I used this model to determine the individual-specific probability of experiencing each symptom, and estimated the interquartile (25%-75%) range of probabilities to measure heterogeneity in the occurrence of symptoms. I subsequently assessed the proportion of the total heterogeneity in symptoms that could be explained by all measured characteristics of individuals. For this, I included patient age, sex, body mass index (BMI), ethnicity, number of comorbidities, diagnosis of anxiety or minor/major depression, smoking status, pack-years of smoking, any exacerbations in the past 12 months, medication possession ratio (MPR)123 for all respiratory related medications, post-bronchodilator forced expiratory volume in 1 second (FEV1), previous diagnosis of asthma, and COPD diagnosis status as independent variables in each model (the full models). In order to determine the variance explained by the independent variables (i.e., participants’ measured characteristics), I calculated the proportion of the estimated variance in the random effect of the full model for each symptom (with all the independent variables), compared to the estimated variance in the random effect of the null model with no independent variables124. I repeated this process using a reduced model with FEV1 as the only independent variable (as opposed to the full model) to determine the proportion of total heterogeneity explained by lung function alone. I conducted sensitivity analyses in which percent predicted FEV1 and GOLD grade were used in place of FEV1 as indicators of lung function (collinearity prevented these variables from being included in the model at the same time). Seasonality was not included in the main analysis because the recall period was >1 year and therefore spanned all seasons; however, season was assessed in a sensitivity analysis to account for the possibility that patients were more likely to recall their recent symptom burden (which could be affected by the current season). All analyses were performed in SAS (version 9.4, 2016).   50  3.3 Results The characteristics of participants are shown in Table 3-1. There were 968 visits from 449 participants in the final sample (53% male, mean age 67 years). 91% of participants had mild to moderate disease (grade I-II), 8% had severe disease (grade III), and 1% had very severe disease (grade IV) as measured by GOLD grades15. 71% of participants with persistent airflow limitation on spirometry had not been previously diagnosed. The average follow-up time was 36 months; 28% of participants underwent only one study visit, and 44% of participants were assessed at all three study visits. There were 390, 280, and 298 participants at visit 1 (baseline), 2, and 3 respectively. The characteristics of the subset of the data used to analyse dyspnea and any symptoms were very similar (Appendix B.2).  Table 3-1 Characteristics of study participants at study visits.   Visit 1 (n=390) Visit 2 (n=280) Visit 3 (n=298) Age 65.3 (10.3) 67.1 (10.1) 68.2 (9.5) Male (vs. female) 54.1% 53.2% 52.3% BMI 27.3 (5.2) 27.3 (5.1) 27.3 (5.1) Caucasian (vs. non-Caucasian) 97.4% 97.1% 98.0% Comorbidities*       0 comorbidities 59.2% 53.2% 54.4%    1 comorbidity 32.3% 33.6% 26.8%    ≥2 comorbidities 8.5% 13.2% 18.8% Anxiety/Depression (vs. no) 19.5% 21.8% 20.8% Smoking between visits (vs. no) 76.2% 29.3% 26.5% Lifetime pack-years smoked 28.0 (26.5) 26.6 (25.5) 27.3 (25.4) Any exacerbations† (vs. no) 7.9%  11.1% 15.4% Medication Possession Ratio‡ 79.9% (108.4) 81.5% (112.4) 79.3% (110.3) FEV1 (L) 2.1 (0.8) 2.1 (0.8) 2.1 (0.7) % Predicted FEV1 74.4 (18.1) 74.9 (19.1) 75.7 (18.8) Diagnosed COPD (vs. undiagnosed) 24.6% 28.9% 32.2% Asthma (vs. no) 20.8% 27.1% 29.9% Symptoms (present vs. absent)       Cough  47.7% 47.1% 40.6% 51   Visit 1 (n=390) Visit 2 (n=280) Visit 3 (n=298)    Phlegm 33.8% 31.8% 28.2%    Wheeze 51.3% 43.6% 42.6%    Dyspnea§ 53.3% 46.5% 44.8%    Any symptoms§ 83.9% 77.2% 74.6%  Means (and standard deviations) are reported unless otherwise indicated. * Participants reported whether they had ever been diagnosed with coronary artery disease, hypertension, diabetes, lung cancer, stroke, and tuberculosis at each study visit † COPD exacerbations of any severity (mild, moderate, severe15) over the past 12 months  ‡ 12-month medication possession ratio123 for all respiratory-related medications § Determined for the subset of participants in which dyspnea was measured (N=415)  3.3.1 Objective 1: Heterogeneity in the occurrence of symptoms  Most participants did not report having cough, phlegm, wheeze, or dyspnea at each study visit, but only 11% of participants were completely asymptomatic throughout the study period. The asymptomatic participants tended to have mild airflow obstruction (mean of 87% predicted FEV1, 18% SD). The proportion of patients that reported a given symptom at least once during follow-up ranged from 43% for phlegm (95% CI 38-47%, the least common symptom) to 61% for dyspnea (95% CI 57-66%, the most common symptom). Symptoms were generally stable within participants: 64% of participants reported the same level of cough throughout their follow-up (95% CI 59-70%, the least stable symptom), and 74% for phlegm (95% CI 69-78%, the most stable symptom).  There was substantial variation in the individual-specific probabilities for the occurrence of symptoms that were estimated from the models (Figure 3-2). The median probabilities of an individual experiencing cough, wheeze, and dyspnea were 0.43, 0.42, and 0.50, respectively. In 52  contrast, the median probability of experiencing phlegm was 0.16 and >0.99 for any symptoms. The interquartile range of probabilities was 0.17-0.73 for cough, 0.03-0.53 for phlegm, 0.13-0.78 for wheeze, 0.19-0.81 for dyspnea, and 0.78->0.99 for any symptoms. Median probabilities are depicted with blue lines and interquartile ranges are depicted with grey boxes in Figure 2.    Figure 3-2 The distribution of individual-specific probabilities of the occurrence of symptoms.   The box spans the lower and upper quartiles (25%-75%) of individuals around the median (blue line). Individual random effects are drawn from a normal distribution with a mean of 0 and standard deviation of the fitted random effects. The statistics shown by the boxes were determined from 1000 repetitions for each individual, and the points show the results of one repetition.  53  3.3.2 Objective 2: Influence of lung function on symptom heterogeneity The logistic regression models revealed relatively consistent associations between patient and disease characteristics and the presence of cough, phlegm, wheeze, dyspnea and any symptoms. Comparisons of the strength of associations across individual symptoms are shown in Figure 3-3, and with any symptoms in Figure 3-4. Lung function, sex, pack-years of smoking, BMI, and MPR were associated with most patient-reported symptoms. Lung function was most strongly associated with the presence of any symptoms (OR per 100 mL increase in FEV1 0.85, 95% CI 0.79-0.91), and least strongly associated with the presence of cough (OR 0.98, 95% CI 0.93-1.03). These results were similar when lung function was assessed as percent predicted FEV1 or GOLD grade in sensitivity analyses (results not shown). Higher pack-years of smoking, BMI, and MPR were all associated with an increased odds of reporting most symptoms. Males were more likely than females to report the presence of phlegm (OR 4.14, 95% CI 1.89-9.03), wheeze (OR 2.07, 95% CI 1.04-4.11), and any symptoms (OR 3.78, 95% CI 1.54-9.31). Summer (vs. winter) was associated with increased reporting of cough (OR 2.04 95% CI 1.13-3.68), phlegm (OR 2.04, 95% CI 1.01-4.13), wheeze (OR 2.95 95% CI 1.54-5.66) and any symptoms (OR 3.13, 95% CI 1.40-6.97) when it was included in sensitivity analyses.54   55  Figure 3-3 Odds ratios for the associations between independent variables and the presence of a) cough, b) phlegm, c) wheeze, and d) dyspnea.   Figure 3-4 Odds ratios for the associations between independent variables and the presence of any symptoms.  The proportion of between-individual variation in the occurrence of symptoms that could be attributed to participants’ measured characteristics (all independent variables in the full models) was 26%, 39%, 39%, 49%, and 91%, for cough, phlegm, wheeze, dyspnea, and any symptoms, respectively (Table 3-2). The proportion of variation explained by FEV1 alone ranged from 2% (for wheeze) to 78% (for any symptoms, Table 3-2).    56  Table 3-2 Percentage of between-individual variation in symptoms explained by individual’s lung function and all measured characteristics combined.    Cough Phlegm Wheeze Dyspnea Any Symptoms FEV1 3% 8% 2% 28% 78% All measured characteristics* 26% 39% 39% 49% 91%  * Visit, Age, Sex, BMI, Caucasian, Comorbidities, Anxiety/Depression, Smoking status, Pack-years of smoking, any exacerbations in the past 12 months, MPR, Diagnosed COPD, Asthma Diagnosis, FEV1  3.4 Discussion In this chapter, I characterised heterogeneity in the occurrence of respiratory symptoms between patients with COPD and assessed the extent to which commonly measured patient and disease characteristics explained the observed heterogeneity in symptoms. Respiratory symptoms were very common in this sample despite over 90% of patients having mild to moderate COPD, and only 29% of them having been diagnosed with COPD. Dyspnea was the most common symptom, followed by cough and wheeze. Individual-specific probabilities for the occurrence of symptoms were highly variable between individuals and for different symptoms. The interquartile range of probabilities was the largest for wheeze and dyspnea, indicating greater variability between individuals in the presence of these symptoms than for cough and phlegm. For phlegm, the majority of individuals had a probability of experiencing phlegm near 0 (visible in Figure 3-2 as a higher density of points at the left edge of the plot). In contrast, the individual-specific probabilities for cough, wheeze, and dyspnea were more evenly spread across the range of possible values. This indicates that phlegm is more stable in nature, and that individuals who do not currently have phlegm are unlikely to report it in the future. Indeed, a pan-European study 57  reported that daily and weekly variability in dyspnea, wheeze, and cough were higher than that for phlegm125. My findings extend these observations on symptom variability within individuals to variability between individuals in the occurrence of symptoms. As a result, tools for assessing COPD severity that involve the measurement of symptoms (such as the GOLD ABCD assessment tool),15 are likely to be more or less variable over time, depending on the symptom measured. The proportion of heterogeneity explained by the measured characteristics of participants differed substantially between symptoms. Most heterogeneity in the occurrence of any symptoms, and half the heterogeneity in dyspnea was explained by the demographic and clinical characteristics of participants included in the models. In contrast, these characteristics explained less than half the heterogeneity in phlegm and wheeze, and only a quarter of the heterogeneity in cough, indicating that other characteristics not included in these models are more important drivers of these symptoms. Cough in particular may be less predictable than dyspnea or the presence of any symptoms using many easily measured patient characteristics. Indeed, age, sex, BMI, smoking history, and lung function were weakly correlated with cough frequency126 in previous studies. Instead, cough frequency was driven by current smoking intensity and percentage of sputum neutrophils126. A unique aspect of this study is that the assessment of heterogeneity was not limited to the patient characteristics included in these models. I quantified total between-individual variation in the burden of symptoms independent of the measured characteristics of participants. The proportion of heterogeneity that was not explained by measured characteristics highlights the potential for other factors, such as biologic disease markers, to improve predictions of symptom burden. 58  Although lung function has traditionally been regarded as the primary driver of respiratory symptoms73, I found that FEV1 explained the majority of between-individual variation in only the occurrence of any symptoms, and a substantial minority of variation in dyspnea. This finding is in line with the observation of high symptom variability within levels of disease severity127, and high short-term variability in symptoms that is not due to changes in lung function72,119. My results extend these previous studies by examining the role of FEV1 in each symptom individually. These findings suggest that lung function is an important but not dominant driver of the occurrence of dyspnea, but it explains very little between-individual variation in the occurrence of phlegm, cough, and wheeze. These symptoms are expected to be more variable within levels of disease severity defined by FEV1. My results highlight the importance of moving beyond FEV1 to incorporate other disease attributes, such as the presence of specific symptoms and exacerbation risk, when classifying disease severity. Given the large differences in the drivers of each symptom that I observed, it is likely that variation between patients in the burden of specific symptoms arises from different disease mechanisms. This can provide insights for refined COPD phenotyping. In addition to analysing heterogeneity in symptoms, I documented associations between symptoms and many demographic and clinical characteristics of participants. In particular, I observed substantial sex-based differences in the reporting of all symptoms apart from cough and dyspnea. Controlled for disease severity, smoking history, and other variables, male patients were over three times more likely to be symptomatic, and four times more likely to report experiencing phlegm. Whether this is a biological phenomenon, or due to gender-related differences in the experience of symptoms128, remains to be further assessed. I also observed positive associations between the occurrence of all symptoms and MPR. The direction of this 59  association is likely due to the long recall period over which symptoms were assessed, and reflects the underlying disease activity rather than short-term variation in symptoms due to treatment. As a result, high treatment intensity was indicative of greater disease activity and therefore more symptomatic disease.  Unique features of this study are its reliance on a large, nationally representative sample of the general Canadian population, the use of standardised spirometry in lung function assessment, validated questionnaires, and a long follow-up time. My sample consisted primarily of patients with mild to moderate COPD, a population that is often underrepresented in large cohort studies. Further, the majority of participants in this study had undiagnosed airflow obstruction. Because patients with a higher symptom burden are more likely to seek care that leads to a diagnosis129, samples of diagnosed patients do not accurately represent symptoms in the entire population with permanent airflow obstruction. Finally, the associations determined from conventional regressions describe the relation between patient characteristics and the presence of symptoms for an average participant. My use of a random effect term in these models enabled me to extend these results by describing the extent to which these population-level associations apply to a given individual. I found that variation between individuals in the presence of any symptoms and dyspnea were reasonably well described by these population-level associations, but this was not the case for cough, phlegm, and wheeze. The assessment of variation at an individual-level is critical to fully characterising heterogeneity in the natural history of COPD, and ultimately to enabling effective use of symptoms in risk prediction tools and case finding algorithms for COPD.  This study also has several limitations. Patients reported their respiratory symptoms with a recall period that spanned the length of time between study visits, which could reach a 60  maximum of three years. The long duration of the recall period is likely to have resulted in inaccuracies in symptom reporting. However, my inclusion of comorbid anxiety and depression reduces the likelihood that psychological factors resulted in reporting bias. In addition, I only assessed the presence of symptoms, not their intensity. A more granular measurement of patient symptoms could provide a more nuanced assessment of symptom variability. Future studies should conduct similar analyses of symptom heterogeneity in patients with severe to very severe COPD, and in subgroups of patients defined by phenotypes or the GOLD ABCD grading system. The impact of individual symptoms on disease outcomes, such as the rate of exacerbations and the slope of lung function decline, should also be assessed. Given the tremendous heterogeneity in the burden of symptoms and their drivers, a detailed classification of patients according to their symptoms might enable better risk stratification to inform treatment decisions.   3.5 Conclusion I assessed a sample of the general population with mostly mild-moderate COPD and found substantial variation in the occurrence of respiratory symptoms between individuals. Lung function explained a small proportion of between-individual variation in the occurrence of cough, phlegm, and wheeze. Commonly measured patient and disease characteristics explained very little heterogeneity in the occurrence of cough in particular. Overall, the observed differences in symptom variation may reflect the divergent etiology of symptoms associated with COPD. Defining phenotypes based on symptoms and evaluating their relation to disease outcomes is a key area of future research.61  Chapter 4: Healthcare System Encounters before COPD Diagnosis: a Registry-Based Longitudinal Cohort Study  4.1 Introduction Chronic obstructive pulmonary disease (COPD) is a progressive disease of the airways that affects approximately 384 million people globally35, and resulted in the deaths of 3.2 million people in 2015 alone130. Despite the high prevalence of COPD, an estimated 81% of COPD patients worldwide have not received a diagnosis at any given time47. These patients are not receiving treatment and risk factor mitigation for the disease, and are likely to experience worse outcomes over the long-term54. In high income countries, undiagnosed COPD patients tend to have less severe disease than diagnosed patients129. However, previous studies indicate that healthcare services use among these patients is substantial46,131.  Numerous case detection and screening strategies have been proposed to diagnose patients with COPD earlier59. These strategies typically rely on opportunistic encounters between undiagnosed COPD patients and the healthcare system59. The patterns of healthcare services use among undiagnosed COPD patients are therefore critical factors in determining their success. In a population-based study of nearly 40,000 COPD patients, a sharp increase in respiratory-related physician consultations in the five years preceding a diagnosis of COPD was documented76. The authors concluded that there was a missed opportunity for an earlier diagnosis of COPD in 85% of patients, and suggested implementing a case finding approach based on patient risk factors76. However, this study only examined respiratory-related resource use, and most interactions between undiagnosed COPD patients and the healthcare system likely occur for non-respiratory 62  reasons131. This study also did not assess healthcare utilisation in a comparator group. In order to enable efficient case detection strategies, it is necessary to identify the encounters that are best at distinguishing COPD from non-COPD patients. To address this knowledge gap, I documented the type and frequency of outpatient visits in a population-based sample of COPD patients before diagnosis. I assessed the rate and probability of primary care, specialist, and pharmacist visits in the five years preceding a diagnosis of COPD. I compared these results to a matched population of non-COPD subjects to identify the types of visits that were most predictive of a patient having COPD.   4.2 Methods 4.2.1 Study design I conducted a population-based, retrospective cohort study using health administrative data. Ethics approval was obtained from Population Data BC (H13-00684). All inferences, opinions and conclusions drawn in this research are those of the authors and do not reflect the opinions or policies of the Data Steward(s).  4.2.2 Data sources The province of British Columbia (BC), Canada, had a population of 4.8 million in 2017132. To administer the public healthcare system in BC, healthcare utilisation records of all legal residents are collected in centralised databases. These administrative databases comprehensively capture information on (1) hospitalisations133, which include admission date, discharge date, and diagnoses coded using the International Classification of Diseases 9th or 10th revisions (ICD-9, ICD-10) revisions, (2) physician billing claims134, which include service date, diagnoses, and 63  practitioner type, and (3) medication dispensation records135, which include service date and type for all drugs dispensed outside a hospital setting. In BC, general practitioners and specialists can practice within the community or be employed by hospitals. There are no private healthcare institutions that provide similar services. Demographic and census information are also recorded136,137, including date of birth, sex, and socioeconomic status (SES) based on income quintiles in the geographic neighbourhood of residence. These databases were linked at the individual level using a unique, anonymous identifier. Administrative databases in Canada have been shown to have low rates of missing, unreported, or misclassified information138,139.  4.2.3 Study population Figure 4-1 depicts the study design. I created a retrospective longitudinal cohort of patients who fulfilled a validated case definition for COPD between January 1, 1996 and December 31, 2015. Patients with three or more physician billing claims and/or one hospitalisation for COPD in any 2-year window were defined as having physician diagnosed COPD140. This case definition has a specificity of 95%140. I defined the index date as the date of first healthcare resource use of any type (outpatient or inpatient) with COPD as the main diagnostic code. Patients who fulfilled the case definition but were <40 years on the index date were excluded, as there is a higher likelihood of misdiagnosis in these patients.  Following the principle of incident density sampling141, the comparator non-COPD cohort was composed of subjects who, as of the same day of follow-up for each COPD patient, were not diagnosed with COPD. Controls either never developed COPD during follow-up, or were diagnosed with COPD later in the study period. A comparator subject was matched to each COPD patient using the following criteria: 1) year of birth within 5 years, 2) sex, 3) 64  sociodemographic status within one income quintile, and 4) no COPD diagnosis date, or a diagnosis date that occurred later than their matched COPD patient. Comparator subjects were assigned the same index date as their matched COPD patient. COPD patients and comparator subjects were required to be in the dataset for at least five years prior, and 1 year following the index date. To ensure that comparator subjects were in the province around the index date, COPD/comparator pairs in which the comparator had no healthcare resource use of any type in the year following the index date were excluded. COPD patients had an instance of resource use at the index date by definition, thus satisfying this criterion. I assessed healthcare encounters in the five years prior to the index date for COPD and comparator subjects. I excluded all pairs in which the COPD patient had a diagnostic code for spirometry within 1 to 5 years of index, but the test did not result in a diagnosis of COPD.    Figure 4-1 Schematic illustration of study design.   The 1-year assessment window for comparators was to ensure comparator subjects had at least one instance of resource use in the year after the index date. 65  4.2.4 Healthcare system encounters For each subject, I determined the number of outpatient visits in each of the five one-year periods prior to the index date. I assessed visits to (1) pharmacists, (2) primary care physicians, (3) specialist physicians, and (4) primary care physicians for respiratory-related reasons. Physician visits types were distinguishable using specialty codes, and all specialties, regardless of their field, were included in the specialist physician assessment. Respiratory-related visits were those with a primary diagnostic code for respiratory diseases other than COPD (ICD-9 460-466, 470-478, 480-487, 490-496, 500-508, 510-514, 516-519, ICD-10 J00-J99) or for respiratory symptoms (ICD-9 786, ICD-10 R05-R07). I did not assess hospitalisations as the focus of this study was on community-based case detection.  I categorised the types of COPD-related inhaler medications dispensed at pharmacist visits using Health Canada’s unique Drug Identification Numbers. I defined the following medication groups: inhaled corticosteroids (ICS), long-acting beta-agonists (LABA), long-acting muscarinic agents (LAMA), short-acting beta-agonists (SABA), and short-acting muscarinic agents (SAMA). Combination therapies were single inhaler ICS/LABA, SABA/SAMA, and LAMA/LABA, or LAMA and LABA dispensed as separate inhalers on the same date. I used a master drug list (Appendix C.1) to identify all other respiratory medications. Over-the-counter medications were not recorded in this dataset.  4.2.5 Analysis All analyses were conducted in SAS version 9.4 (SAS Institute Inc., Cary, NC). I initially created a descriptive summary of the rate of outpatient visits per patient year, stratified by COPD status. Second, I assessed the medical indications that prompted COPD patients to visit a primary care 66  physician before diagnosis. Medical indications were grouped by disease type based on the main diagnostic code for each visit. Within disease types, main diagnostic codes could be for either symptoms or disease diagnoses. In a small minority of cases, a visit was attributed to more than one medical indication (i.e. >1 main diagnostic code), and these were counted as separate visits. Third, I calculated the mean number of pharmacist visits per patient year in which a respiratory medication was dispensed. When more than one medication of different classes were dispensed at the same visit, these were counted separately. Last, I assessed the probability that COPD patients had any opportunity for case detection in the five years before diagnosis. This was done by determining the proportion of COPD and comparator subjects with at least one visit of each type in the cumulative years before diagnosis (i.e., year 1, then years 1 and 2, etc.). I tested for differences in the rate of outpatient visits between COPD patients and comparator subjects before the index date. I constructed separate generalised linear regression models (negative binomial distribution, log link) for each of the four visit types. The rate ratio (RR) gives the ratio of visit rates between COPD patients and comparators. I used generalised estimating equations to account for clustering of data between matched pairs of observations. Period number (1-5 range), calendar year at the beginning of the period, age, sex, and SES on the index date were included in all models to adjust the exposure estimates for the overall effects of these variables over the study period. To assess trends in the rate of healthcare visits as COPD patients approached diagnosis, I constructed a second model for each of the dependent variables. The independent variables were identical to the original model, with the addition of an interaction term between period and the group-defining variable (COPD/comparator). The sum of the coefficients for period and the interaction term gave the rate of change in healthcare visits in the COPD group relative to the comparator group. 67  I conducted a sensitivity analysis on the subset of COPD patients who were prescribed LAMA in the year before or after their index date. In BC, physicians prescribing this drug must provide evidence that the ratio of forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) is < 0.7 and FEV1 £ 65% of predicted142, meaning that nearly all patients in this subgroup had moderate to very severe airflow obstruction according to the Global Initiative for chronic Obstructive Lung Disease (GOLD) guidelines15. Therefore, the specificity of the case definition for COPD is close to 100% in this subgroup. I performed a second sensitivity analysis in which all participants with missing data were excluded. In a third sensitivity analysis, I constructed a new cohort with the additional criterion that comparator subjects never developed COPD at any point during the study period.   4.3 Results The sample selection flow chart is shown in Figure 4-2. 149,213 patients fulfilled the case definition of COPD and had data available for at least five years prior to the index date. 112,635 COPD/comparator pairs were included following the selection process. 28.3% of COPD patients received an initial diagnosis of COPD during hospital admission; the remaining were diagnosed in outpatient settings. 12.3% of COPD patients were matched to a comparator subject who developed COPD later in follow-up. 51.0% of pairs were male, and the mean age at the index date was 68.6 (SD 12.2) years. A descriptive summary of the sample characteristics is shown in Table 4-1.  68   Figure 4-2 Cohort selection process.  Table 4-1 Characteristics of patients meeting the case definition of COPD and matched comparator subjects.  COPD patients  (N= 112,635) Comparator subjects  (N= 112,635) Age 68.6 (12.2) 68.7 (12.2) Male (vs. female)* 51.0% 51.0% SES†‡     Quintile 1 25.9% 25.9%    Quintile 2 21.3% 21.3%    Quintile 3 18.7% 18.7%    Quintile 4 17.3% 17.3%    Quintile 5 15.0% 15.0% Year of index date      2001-2005 28.4% 28.4%    2006-2010 40.1% 40.1%    2011-2015 31.5% 31.5%  * Sex was unknown in 0.05% (N=57) of COPD patients and 0.05% (N=57) of comparators.  69  † SES was unknown in 1.80% (N=2,023) of COPD patients and 1.80% (N=2,023) of comparators. Unknown observations for sex and SES were coded as missing and included as a separate category in regression analysis. ‡ Quintile 5 is the highest socioeconomic status.  4.3.1 Rate of outpatient encounters Pharmacist, primary care, and specialist physician visits were more common among COPD patients than comparator subjects throughout the five-year assessment period (Figure 4-3). In the year prior to being diagnosed, COPD patients had an average of 17.0 (Interquartile Range [IQR] 5.0-20.0) visits to a pharmacist, 12.6 (IQR 5.0-16.0) primary care visits, 10.5 (IQR 3.0-14.0) specialist visits, and 1.8 (IQR 0.0-2.0) visits to a primary care physician for respiratory-related reasons. In contrast, comparator subjects had an average of 10.2 (IQR 3.0-13.0), 8.2 (IQR 3.0-11.0), 6.9 (IQR 2.0-9.0), and 0.5 (IQR 0.0-1.0) pharmacist, primary care, specialist, and respiratory-related primary care visits in the year before index, respectively.  70   Figure 4-3 Mean number of outpatient visits per patient year in each of the five years before the index date.  COPD patients and their comparator non-COPD subjects are shown.  The regression analysis indicated that COPD patients prior to diagnosis incurred all types of visits at higher rates than comparator subjects (Table 4-2). Relative to comparator subjects, COPD patients were most likely to visit a primary care physician for respiratory related reasons (rate ratio [RR] 2.17 95%CI 2.14–2.19). COPD and comparator subjects with lower SES had a higher rate of primary care and pharmacist visits. For all outcomes, healthcare visits increased as subjects approached the index date. Models with an additional interaction term between period and the group-defining variable indicated that the rate of primary care visits increased by 8.5% more visits per year (95%CI 8.1–8.9%) in COPD than comparator subjects. This figure was 7.3% (95%CI 6.8–7.8%) for pharmacist visits, 10.8% (95%CI 10.2–11.4%) for specialist visits, and 71  21.9% (95%CI 20.8–23.0%) for respiratory-related primary care visits. Coefficients for female, age, SES, and calendar year were similar to the main models and are therefore not shown. I observed similar results in the sensitivity analysis of 9,079 patients with spirometry-confirmed moderate to very severe COPD near the index date15 (Appendix C.2), for the 110,556 COPD/comparator pairs with no missing data (Appendix C.3), and for the cohort of 106,439 COPD/comparator pairs in which the comparator never developed COPD during follow-up (Appendix C.4). 72  Table 4-2 Parameter estimates and confidence intervals from the multivariable regression model of the rate of outpatient visits among COPD patients and comparator non-COPD subjects.  All-cause primary care visits Respiratory-related primary care visits Specialist physician visits Pharmacist visits  Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI COPD (vs. comparator) 1.40 1.39-1.41 2.17 2.14-2.19 1.35 1.34-1.37 1.62 1.60-1.63 Period 1.06 1.06-1.06 1.14 1.14-1.14 1.07 1.07-1.08 1.06 1.06-1.06 Female (vs. male) 1.16 1.15-1.16 1.27 1.26-1.29 1.08 1.07-1.09 1.18 1.17-1.20 Age 1.01 1.01-1.01 1.00 1.00-1.00 1.02 1.02-1.02 1.01 1.01-1.01 SES     1 (vs. 5) 1.17 1.16-1.19 1.11 1.09-1.13 1.02 1.00-1.03 1.32 1.30-1.35    2 (vs. 5) 1.10 1.09-1.11 1.08 1.06-1.10 0.99 0.98-1.01 1.12 1.10-1.14    3 (vs. 5) 1.06 1.04-1.07 1.05 1.03-1.07 0.99 0.97-1.00 1.06 1.04-1.08    4 (vs. 5) 1.03 1.02-1.05 1.02 1.00-1.04 0.99 0.97-1.00 1.04 1.03-1.06 Calendar year 1.00 1.00-1.00 0.97 0.96-0.97 1.01 1.01-1.01 1.02 1.02-1.02  Parameter estimates for the category of observations that were missing for sex and socioeconomic status are not shown due to their small numbers (see Table 4-1 for details).73  The most common reason for visits to a primary care physician among COPD patients was circulatory disease (mean of 1.96 diagnostic codes/year, SD 3.77), followed by respiratory disease (1.16 diagnostic codes/year, SD 2.14, Figure 4-4). The frequency of circulatory disease diagnoses increased by 20.4% from two years before COPD diagnosis to year one before, and by 62.9% for respiratory disease diagnoses over the same time period.    Figure 4-4 Mean number of ICD diagnostic codes per patient year at the primary care visits of COPD patients before the index date.   In 97.7% out of 5,665,378 total visits analysed, there was only one diagnostic code per visit. Disease categories with less than <0.1 diagnostic codes per year are not shown.  74  COPD patients had a mean of 1.61 (SD 4.48) pharmacist visits per year in which respiratory medications were dispensed (Table 4-3). This increased to 2.46 (SD 5.57) visits in the final year before diagnosis. In contrast, these figures were 0.31 visits per year (SD 1.66) and 0.36 visits (SD 1.86) for comparator subjects, respectively. SABA was the respiratory medication most commonly dispensed to COPD patients before diagnosis; with a mean of 0.58 (SD 1.96) pharmacist visits with a SABA dispensed per year. Non-respiratory medications were dispensed 53 percent more frequently to COPD patients than to comparator subjects.  Table 4-3 Mean number of pharmacist visits per 100 patient years with respiratory and other medications dispensed.  COPD patients Comparator subjects Years before index 5 4 3 2 1 5 4 3 2 1 ICS 26 28 29 33 43 6 6 6 6 6 LABA 3 3 4 4 5 1 1 1 1 1 ICS/LABA 10 12 15 20 31 2 2 2 3 4 LAMA 1 1 2 3 5 0 0 0 0 0 LABA 3 3 4 4 5 1 1 1 1 1 LAMA/LABA 0 0 0 0 0 0 0 0 0 0 SABA 44 48 52 61 85 8 8 8 9 9 SAMA 8 10 12 16 27 1 1 1 2 2 SAMA/SABA 2 3 4 4 6 0 0 0 0 0 Other Respiratory Medications 21 24 26 30 42 10 11 11 13 14 Non-Respiratory Medications 1141 1226 1320 1441 1612 766 817 871 937 1007   75  4.3.2 Probability of outpatient encounters The proportion of COPD patients that incurred at least one visit in any of the five years before the index date was ≥99.0% for pharmacist, primary care, and specialist visits, and 85.3% for respiratory-related primary care visits (Figure 4-5). These proportions were similar for comparator subjects, with the exception of fewer respiratory-related primary care visits (66.2%). The highest proportion of COPD patients visited a primary care physician in the year prior to diagnosis (97.2%), followed by pharmacist (95.4%), specialist (93.6%), and respiratory-related primary care visits (60.3%).    Figure 4-5 Proportion of subjects with at least one outpatient visit in any of the five years before index.   COPD patients and their comparator non-COPD subjects are shown.  76  4.4 Discussion In this chapter, I conducted a longitudinal study of healthcare encounters before COPD diagnosis in a sample of the general population. Nearly all patients had at least one encounter in an outpatient setting in the five years before they were diagnosed with COPD. They interacted most frequently with pharmacists, followed by primary care physicians, and then with specialist physicians. The rate of visits of every type increased as COPD patients approached diagnosis, and healthcare encounters were particularly high in the year before diagnosis. Outpatient visits occurred at a substantially higher rate in COPD patients than in matched non-COPD comparator subjects. My findings indicate that case detection technologies implemented in primary care settings have frequent opportunities to diagnose COPD earlier.   COPD is a progressive disease that is characterised by airway, and in certain patients, systemic inflammation143. Due to this inflammation and also to shared risk factors, patients with COPD have higher rates of comorbidities than subjects without the disease144. Cardiovascular comorbidities are especially common among COPD patients23. The most common reason for visiting a primary care physician before being diagnosed with COPD was circulatory disease, followed by respiratory diseases other than COPD or respiratory symptoms. Gershon et al.131 also found that circulatory diseases were the most common reason for outpatient visits in undiagnosed COPD patients. In my analysis, the frequency of circulatory and respiratory disease diagnoses spiked in the year before COPD diagnosis. The increase in healthcare encounters leading up to diagnosis might reflect these systemic manifestations of the disease. It is also possible that COPD was the underlying cause of some of these circulatory-related visits, as symptoms such as dyspnea and fatigue are shared between diseases145. It is likely that these increased encounters were what eventually led to the COPD diagnosis. Increased awareness of 77  the risk of COPD among patients with cardiovascular disease in particular might lead to earlier detection. However, the presence of frequent healthcare encounters is not in itself sufficient to warrant suspicion of COPD, as the pattern of increased healthcare services use leading up to diagnosis is common to other disease areas146,147. Case detection methods such as risk assessment questionnaires70,148 or screening spirometry66 should be implemented at these types of routine visits in order to identify patients at high risk of having COPD.   These results indicate that there are substantial opportunities to diagnose COPD earlier. 85% of COPD patients visited a primary care physician for a respiratory-related reason in the five years before they were diagnosed with COPD. This proportion is almost identical to that found by Jones et al.76 in the UK. Moreover, due to my exclusion criteria, none of the COPD patients in the main analysis received spirometry despite visiting a primary care physician for either respiratory symptoms or other respiratory diseases. Many of these patients were also receiving treatment for respiratory symptoms despite not having been diagnosed with COPD. Influential guidelines, such as GOLD, recommend considering a diagnosis of COPD in patients with respiratory symptoms or risk factors for COPD15. Risk factors can include other respiratory diseases such as asthma, environmental, or early childhood exposures, for example to childhood respiratory infections, that can result in lung damage long before the clinical manifestation of COPD149. This suggests that programs systematically attempting to diagnose COPD earlier through the use of case detection methods have the potential to substantially increase testing for COPD. Given that almost all COPD patients had at least one respiratory consultation in the five years before diagnosis, and many were already receiving respiratory medications, respiratory-related primary care visits might provide particularly high value opportunities for case detection. 78  This type of visit was much more common among COPD patients than in comparator subjects, which would increase the prior predictive value of any diagnostic test.  This study has several strengths. I characterised healthcare services use in a large sample of the general population. I used routinely collected data rather than self-reporting by physicians or patients, which is particularly important given that information or recall bias might affect the accuracy of self-reports. A limitation of this study is the lack of data on individual patient characteristics such as spirometry and disease severity, which limited us to studying healthcare services use before COPD diagnosis, rather than in undiagnosed COPD patients. However, I observed similar results in the subset of patients who had spirometry-confirmed moderate to severe COPD near the time of diagnosis. My use of a highly specific case definition may have resulted in a sample of patients with more severe COPD and higher healthcare service use than in the general population of undiagnosed COPD patients. The definition of comparator subjects allowed subjects to develop COPD after the index date. However, due to the high prevalence of undiagnosed COPD46,150, some comparator subjects may have already had COPD at the index date. If the comparator group included patients with COPD, my results would be biased towards the null. Another limitation is that I only followed patients for 5 years before diagnosis. I imposed this criterion to ensure identical follow-up between matched COPD and comparator subjects while still maintaining a large sample size. The probability of a patient having undiagnosed COPD also decreases as time to diagnosis increases. Finally, my results are most relevant to high income countries, as more limited access to care could result in lower or no healthcare services use among undiagnosed COPD patients in low- and middle- income countries.   79  4.5 Conclusions I documented frequent interactions between COPD patients and the healthcare system prior to COPD diagnosis. This suggests that there are many opportunities to implement case detection strategies in outpatient settings, and that it is possible to improve early detection while relying on only opportunistic healthcare encounters. COPD patients interacted most frequently with pharmacists before diagnosis, but the greatest difference between COPD and non-COPD subjects was in the rate of primary care visits for respiratory-related reasons. Either of these types of visits might present high value opportunities for case detection. Increased testing for COPD during the routine visits of at-risk patients has the potential to substantially improve early detection. The cost-effectiveness of case detection strategies implemented during pharmacist or primary care visits should be investigated in a simulation experiment.80  Chapter 5: Cost-Effectiveness of Case Detection Strategies for the Early Detection of COPD  5.1 Introduction Chronic obstructive pulmonary disease (COPD) is the third most common cause of mortality among chronic diseases globally151, and is the leading medical cause of hospital admission in Canada152. Patients with COPD typically experience periods of increased impairment known as exacerbations, as well as a progressive loss of lung function that is typically irreversible15. As a result, risk factor management to preserve lung function early in the disease progression is regarded as the most important component of treatment54. However, early interventions to reduce the burden of COPD require an early diagnosis, and estimates from Canada indicate that 70% of patients with COPD are currently undiagnosed46,150, with up to one third of COPD patients initially diagnosed in hospital following an exacerbation44,153. A late diagnosis of COPD can represent a missed opportunity, as treatment for smoking cessation and pharmacotherapy to reduce symptoms and the risk of exacerbations can improve quality of life and long-term outcomes54,59. Screening and case detection strategies offer potential solutions to the problem of underdiagnosis in COPD. Screening involves testing healthy individuals in the general population for COPD8. In contrast, case detection is opportunistic, and involves targeted testing of high-risk individuals during their routine healthcare encounters8,60–62. Influential clinical guidelines recommend against screening for COPD among asymptomatic adults due to the lack of evidence that it improves disease outcomes6,63. However, two recent randomised controlled 81  trials found that opportunistic case finding strategies resulted in a higher yield of new COPD cases than usual care154,155. A subsequent economic analysis showed that an active case finding approach, in which symptomatic patients at risk of COPD were recruited for diagnostic spirometry, can be cost-effective155,156. However, existing case detection strategies are extremely heterogeneous59, and the decision space around selection characteristics, case detection technology, and frequency of testing should be more fully explored.  The objective of this chapter was to address these questions using a ‘whole disease’ model of COPD. The Evaluation Platform in COPD (EPIC) was recently developed and has been extensively validated to project the outcomes of COPD-related policies for the general Canadian population80. In contrast to existing trial-based economic evaluations154,155, model-based decision analysis offers a more flexible approach to policy design through a fuller evaluation of scenarios and a broader scope of evidence. The objective of this study was to determine the cost-effectiveness of primary care-based case detection scenarios, which vary in terms of their initial eligibility criteria, case detection technology, follow-up intervals, and subsequent disease management.   5.2 Methods This chapter was conducted in two phases, and the methods are reported accordingly. In the first section, I describe my methods for evaluating the cost-effectiveness of case detection; in the second section, I describe my expansion of the EPIC model to simulate the case detection pathway. The case detection scenarios evaluated, their implementation in EPIC, and key parameters for the costs of case detection and effectiveness of treatment are summarised in Sections 5.2.1 to 5.2.3. I describe the cost-effectiveness analysis in Section 5.2.4. In the second 82  phase (Section 5.2.5), I describe the development of modules for simulating respiratory symptoms, primary care visits, COPD diagnosis, treatment, and case detection in EPIC. Diagnosis, symptoms, and primary care visits were the subjects of Chapters 2–4, respectively, and the results of these chapters informed their development in EPIC. I have adhered to the consolidated health economic evaluation reporting standards (CHEERS) statement for reporting economic evaluations of health interventions157.  5.2.1 Case detection scenarios I evaluated 16 strategies for improving early detection of COPD during routine primary visits. Scenarios were defined by three sets of selection characteristics for inclusion, two types of case detection methods, and two different time intervals between case detection.  Scenarios were grouped based on their eligibility criteria for selecting patients to receive case detection: all patients (Scenario 1: ‘S1’), patients reporting any respiratory symptoms at baseline (Scenario 2: ‘S2’), and current or former smokers (‘ever smokers’) ≥50 years of age at baseline (Scenario 3: ‘S3’). Within each set of eligibility criteria, scenarios were further characterised by the case detection method employed: the COPD diagnostic questionnaire (CDQ)158 or the handheld flow meter, and the time interval between implementation (Table 5-1).  Table 5-1 Case detection scenarios evaluated. Case detection method Testing interval Sensitivity, Specificity Reference S0: No case detection (S1) All patients S1a: CDQ ≥17 points 3 years, 5 years 91%, 49% 159 S1b: Screening spirometry (with bronchodilator) 3 years, 5 years 80%, 94% 83  S1c: CDQ  + Screening spirometry (with bronchodilator) 3 years, 5 years 72%, 97% (S2) Symptomatic patients (any respiratory symptoms) S2a: Screening spirometry (without bronchodilator) 3 years, 5 years 81.5%, 88.9% 6 (S3) Smoking history (ever smokers ≥ 50 years) S3a: CDQ ≥19.5 points 3 years, 5 years 64.5%, 65.2% 66 S3b: CDQ ≥16.5 points 3 years, 5 years 87.5%, 38.8% S3c: Screening spirometry (without bronchodilator) 3 years, 5 years 79.9%, 84.4% S3d: CDQ  + Screening spirometry (with bronchodilator) 3 years, 5 years 74.4%, 97% 159   I evaluated two types of case detection methods: the handheld flow meter and the COPD diagnostic questionnaire (CDQ). Handheld flow meters are used for performing screening spirometry based on the ratio of FEV1 to forced expiratory volume in 6 seconds (FEV6) < 0.766. I assessed handheld flow meters with prior administration of a bronchodilator (scenarios: S1b, S1c, S3d)159, and without a bronchodilator (scenarios: S2a, S3c)45,59, depending on their implementation in the study from which I derived their performance characteristics.  The CDQ is a 5-item symptom and risk factor assessment questionnaire with a total score of 38 points68,158. Different score cut points have been used to identify patients for referral to diagnostic spirometry. In separate scenarios, I evaluated scoring cut points of  ≥19.5 (S3a) and  ≥16.5 (S3b) among ever smokers66, and  ≥17 (S1a) among all patients159, again following the studies from which I derived their performance characteristics. In two scenarios (S1c and S3d), the CDQ and screening spirometry were implemented sequentially at the same primary care visit (Table 5-1).  84  Within each group of scenarios (S1a, S1b, S1c, S2a, S3a, S3b, S3c, S3d), I evaluated 3- and 5-year intervals between instances of case detection. In cases where an individual had no primary care visits at the scheduled interval, case detection was deferred to the next year in which they had a primary care visit.  5.2.2 Evaluating case detection scenarios The Evaluation Platform in COPD (EPIC) simulates the development and progression of COPD in the general population of Canadians ≥40 years old. From this simulated population, I created a cohort of individuals meeting the eligibility criteria for case detection at baseline (i.e. S1- all patients, S2- symptomatic patients, S3- ever smokers). To be eligible for cohort inclusion, individuals were also required to have undiagnosed COPD or to not have COPD (‘non-COPD’ individuals), and to have had at least one primary care visit in the previous 12 months as this was seen as an opportunity for study recruitment. Case detection occurred at routine primary care visits and resulted in one of four outcomes: a diagnosis of COPD in individuals with COPD (true positive), no diagnosis of COPD in individuals with COPD (false negative), a diagnosis of COPD in individuals with no COPD (false positive), or no diagnosis of COPD in individuals with no COPD (true negative). Individuals who tested negative after case detection (true negatives and false negatives) did not receive any further testing. Individuals who tested positive after case detection (true positives and false positives) underwent an additional primary care visit to assign a diagnosis of COPD using post-bronchodilator diagnostic spirometry. Diagnostic spirometry had 100% accuracy in ruling out a diagnosis of COPD in non-COPD individuals. Therefore, individuals with COPD (true positives) were definitively diagnosed, and non-COPD individuals were not diagnosed. The 85  cohort creation and case detection pathway are depicted in Figure 5-1. Details on the development and parameterization of the case detection module are provided in Section 5.2.5.6.  COPD could be diagnosed at primary care visits without the use of case detection, both among individuals with COPD (routine diagnosis) and non-COPD individuals (routine overdiagnosis). An additional primary care visit with diagnostic spirometry was required to correctly diagnose individuals with COPD, or to reverse the diagnosis among non-COPD individuals. Among individuals with COPD, a second pathway to diagnosis was during hospitalisation due to severe exacerbation. Details on the development and parameterization of the routine diagnosis module are provided in Section 5.2.5.4.      Figure 5-1 Framework for applying case detection strategies.  86  5.2.3 Parameter inputs 5.2.3.1 Treatment effectiveness All individuals diagnosed with COPD (through case detection or routine diagnosis) were treated with inhaled therapies, and among current smokers, treated for smoking cessation. Inhaled therapies were provided to all individuals with a diagnosis of COPD according to the GOLD ABCD criteria, which is based on exacerbation history in the previous 12 months and respiratory symptoms15. The full treatment algorithm is described in Section 5.2.5.5. Using evidence from landmark clinical trials160–162, the effectiveness of inhaled therapies was applied as a reduction in the rate of all exacerbations (parameter values are provided in Table 5-2). Following a previous modelling study156, symptom relief from treatment, outside an exacerbation episode, was assessed as a 0.0367 increase in quality adjusted life year (QALY) among diagnosed COPD patients with any respiratory symptoms. All newly diagnosed individuals who were current smokers received three months of nicotine replacement therapy (NRT). This was associated with an increase in the odds of successful smoking cessation in the year following diagnosis of 1.38. (This parameter was derived through a calibration process that is described in Section 5.2.5.5.) Nicotine replacement therapy had no effect on smoking cessation in subsequent years following diagnosis. Inhaled therapies and treatment for smoking cessation were assigned to individuals probabilistically based on an adherence of 70%. That is, in 30% of person-time in the model, individuals who met the criteria for inhaled therapy or NRT did not receive treatment, or for inhaled therapies, were not upgraded to the recommended treatment.  87  5.2.3.2 Costs Case detection, COPD diagnosis, and treatment were all associated with costs. These costs were converted to 2015 Canadian dollars using the healthcare component of the Consumer Price Index.  5.2.3.2.1 Case detection and diagnosis costs Case detection was administered during routine primary care visits. I assumed that case detection comprised the major topic of a primary care visit. According to a previous study, major topics require a median of 5.3 minutes out of a total visit length of 15.7 minutes, or 34% of the total visit time163,164. The time to administer case detection was costed at 34% of the total cost of a primary care visit, which was derived from the BC Medical Service Plan fee schedule165. The CDQ was assigned only this time-related cost. Screening spirometry was assigned the cost of the primary care physician’s time, plus the cost of administering spirometry with a handheld flow meter. Screening spirometry could be administered with or without a bronchodilator, and these had separate costs as determined from BC fee codes165. All costs related to case detection are shown in Table 5-2. The cost of an outpatient diagnosis was composed of the cost of post-bronchodilator diagnostic spirometry plus the duration of an additional primary care visit for interpreting the results, derived from BC fee codes165 (Table 5-2). An inpatient diagnosis of COPD had no cost in addition to that of hospitalisation for COPD exacerbation80.     88  5.2.3.2.2 Treatment costs The cost of inhaled pharmacotherapies was determined from medication dispensation records in BC health administrative data. I calculated the average annual cost of patients dispensed short-acting beta-agonists (SABA), long-acting muscarinic antagonists (LAMA), long-acting muscarinic antagonist and long-acting beta-agonist combination therapy (LAMA/LABA), and triple therapy with inhaled corticosteroids (ICS), LAMA, and LABA as a single inhaler or monotherapies at least once in 2014 (the last year for which full data was available)166. These costs are provided in Table 5-2. All newly diagnosed individuals who were current smokers received 3 months of NRT at a cost of C$350, which was based on the costs reported in a previous Canadian study167. The direct costs of maintenance COPD therapy, stratified by GOLD grade, were determined from Canadian studies84,168 and are shown in Table 5-2. To avoid double counting of treatment costs, I reduced the total maintenance costs by the proportion of total costs attributed to pharmacotherapy at each GOLD grade169. The direct costs of exacerbations were determined from the literature87,88; they have been described and provided previously80.  5.2.3.3 Health status The derivation and validation of background utilities by GOLD grade84,86, and the reduction in QALY due to exacerbations84, have been previously described80. A false positive diagnosis of COPD was not associated with a disutility following previous findings in another chronic disease170 (Table 5-2).   89  Table 5-2 Model parameters specific to the evaluation of case detection.  No COPD GOLD 1 GOLD 2 GOLD 3 GOLD 4 Global parameters    Annual discount for QALY 1.5%    Annual discount for costs 1.5%    Time horizon 20 years Health Outcomes    Background Utility84 0.8685 0.81 0.72 0.6886 0.5886    QALY reduction from mild-    moderate exacerbations*84  0.0225 0.0155 0.0488 0.0488    QALY reduction from severe-    very severe exacerbations*84  0.0728 0.0683 0.0655 0.0655    Disutility from false positive    diagnosis170 0     Costs†    Direct maintenance costs    without treatment‡ 84,169  $135 $330 $864 $1178168    Outpatient diagnosis165 $58.79    Screening spirometry without    bronchodilator165 $23.38    Screening spirometry with     bronchodilator165 $29.27    Administration of Case       Detection Questionnaire     (CDQ)165 $11.10 Treatment    Odds ratio of NRT for       successful smoking      cessation171,172 1.38    Annual costs of NRT167 $350.36    Adherence to NRT 70%    Inhaled therapies SABA LAMA LAMA/ LABA ICS/LAMA/LABA    Rate reduction for all    exacerbations§ 0 0.22160 0.23161 0.34162    Utility gain from    treatment§||156 0.0367    Annual cost166 $50.51 $335.55 $613.73 $1084.98    Medication adherence 70%  General model parameters have been reported previously80. 90  * The QALY reduction due to exacerbation was determined by dividing the 3-month disutility from an exacerbation84 by four. See the original publication for more details80. † The price year for all costs was 2015 Canadian dollars. ‡ Maintenance costs were composed of physician visits (generalist and specialist), rehabilitation programmes, laboratory tests and devices, and oxygen therapy84.  § Only individuals with COPD benefited from treatment (rate reduction in exacerbations or utility gain from treatment). Inhaled therapies had no effect on non-COPD individuals. || Only individuals with symptoms (the presence of cough, phlegm, wheeze, or dyspnea) received the additional utility benefit due to treatment.  5.2.4 Analysis 5.2.4.1 Reference case The main outcomes of this analysis were total costs and QALYs accumulated over a 20-year time horizon. The target population for this policy was the Canadian general population ≥40 years of age. In order to maintain the same total population across case detection scenarios with different eligibility criteria (S1- all patients, S2- symptomatic patients, S3- ever smokers), I standardised the analysis to the S1 (all patients) scenario. Total costs and QALYs of the S2 and S3 scenarios were adjusted to include the average costs and QALYs of individuals not meeting the selection criteria for that scenario. As a result, the expected per patient costs and QALYs were comparable across all case detection strategies and could be ranked. Unless otherwise specified, results are reported for a cohort of 1,000 individuals who were eligible for any case detection strategy. For each of the 16 case detection scenarios, I calculated incremental costs and QALYs relative to a comparator, no case detection scenario. I determined the preferred case detection 91  scenario using the efficiency frontier approach173. Using this approach, the scenario with the lowest incremental cost-effectiveness ratio (ICER) compared to the current standard of care (no case detection) was identified as the initial default scenario. Scenarios that were more costly and less effective than the default scenario were subject to ‘extended dominance’ and were not considered further. Of the remaining scenarios, the case detection scenario with the lowest ICER relative to the default became the new preferred scenario. This iterative process stopped when there were no remaining scenarios with an ICER below the willingness to pay (WTP) threshold of $50,000/QALY.  After determining the preferred scenario, I ranked the remaining scenarios using incremental net monetary benefit (INMB) at a WTP threshold of $50,000/QALY. Scenarios were considered cost-effective if they fell below this threshold. Costs and QALYs were discounted annually at a rate of 1.5%, and the analysis was conducted from the perspective of the Canadian healthcare system following guideline recommendations173.  5.2.4.2 Sensitivity analyses  I conducted one-way sensitivity analyses around the effectiveness of treatment for individuals diagnosed with COPD. Following previous assessments of medication adherence in the general population174, I reduced overall medication adherence to 0.5 and 0.3 in separate sensitivity analyses. Guidelines recommend providing treatment for smoking cessation to all current smokers regardless of COPD status6,175. Therefore, I conducted a sensitivity analysis in which treatment for smoking cessation at the time of COPD diagnosis was removed. Finally, I reduced the utility benefit due to symptom relief from treatment to 0.01, as this was the threshold for cost-effectiveness in a previous economic evaluation of case detection156. Following the same 92  study, I removed the utility benefit from the impact of treatment on symptoms. In each sensitivity analysis, I reassessed INMB and ranked the scenarios.   5.2.5 Model development 5.2.5.1 Evaluation Platform in COPD (EPIC) EPIC is a whole disease, discrete event microsimulation model of COPD in Canada. The development, internal and external validation of the model has been described in detail elsewhere80. In brief, EPIC is a dynamic population model that simulates Canadian adults ≥40 years old. It is composed of a series of modules, each of which addresses a different aspect of disease. EPIC has modules for: (1) the demography and risk factors of the general Canadian population 40 years and older; (2) COPD occurrence as defined by forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) less than the lower limit of normal40; (3) individualised lung function trajectories measured by FEV1 and the Global Initiative for chronic Obstructive Lung Disease (GOLD) COPD severity grades15,28; (4) the occurrence and severity of COPD exacerbations; (5) mortality related to COPD exacerbations and background mortality; and, (6) payoffs, including costs and quality adjusted life years (QALYs) for all individuals in the model. COPD occurrence, exacerbation start and end, change in smoking status, and death due to background mortality are modelled as events in EPIC. These modules performed well in tests of external validity80.  In order to assess the cost-effectiveness of case detection strategies for COPD, I constructed new modules for symptoms, primary care visits, COPD diagnosis, and treatment. The development of each new module is described below. All modules were validated to ensure 93  model results matched those of the input data source, and in some cases, external calibration targets. The results of validation are shown in the corresponding appendix for each module.  5.2.5.2 Symptom module Symptoms were added to EPIC based on the results of Chapter 2, which indicated they were a key factor in determining whether a patient with COPD was diagnosed. (Lung function, which was also strongly associated with a previous diagnosis, was already modelled in EPIC.) Simulated individuals in EPIC were assigned a binary value for the presence of cough, phlegm, wheeze, and dyspnea, annually. Separate linear mixed-effects logistic regression models were developed for each symptom using data from the population-based Canadian Cohort of Obstructive Lung Disease (CanCOLD)45. These equations were developed in Chapter 3; a detailed description of the dataset and equations can be found in Section 3.2. In brief, separate equations were fitted to participants with COPD (N=502) and those without COPD (N=1071). I used the subset of independent variables assessed in Chapter 3 that were modelled in EPIC. The log odds of experiencing a symptom was a function of age, sex, current smoking status, and for COPD patients, FEV1. I determined the covariance of the random effect term for participants from each symptom model, which I used to construct a multivariate normal distribution for symptoms. I sampled from this distribution to assign a unique random-effects term to individuals in EPIC for each symptom. The use of random effects captured additional heterogeneity in symptoms that could not be attributed to the independent variables in the model, as well as the tendency for symptoms to cluster within individuals and to differ in their stability over time150. Model specifications and the results of internal validation are provided in Appendix D.1.   94  5.2.5.3 Primary care visits module The results of Chapter 4 indicated that undiagnosed COPD patients frequently encountered primary care physicians before diagnosis, and that these types of visits could be used as opportunities for case detection. Individuals in EPIC were assigned an annual number of primary care visits based on their age, sex, smoking status, the presence of symptoms, and for COPD patients, FEV1. Because the health administrative data used in Chapter 4 did not have information on symptoms, smoking status, or FEV1, I used self-reported healthcare resource use data from CanCOLD to model primary care visits in EPIC. I developed separate regression models for COPD (N=502) and non-COPD (N=1070) patients. Participants in CanCOLD reported how many times they had seen their doctor in the past 12 months at in-person visits that occurred at 18 month intervals for a maximum of 3 years of follow-up45. I used generalised linear models with a negative binomial distribution and logarithmic link to model the rate of primary care visits for all individuals in EPIC. I used this to generate a count of annual primary care visits for all individuals. Model specifications and the results of internal validation are provided in Appendix D.2.  5.2.5.4 Diagnosis module COPD patients in EPIC could be already diagnosed at the start of the simulation (baseline diagnosis), or during follow-up (incident diagnosis). The results of Chapter 2 informed the patient and disease characteristics that were modelled in EPIC in order to simulate diagnosis in current practice. However, I required patient-level data from CanCOLD (N=467) to develop logistic regression models for the odds of diagnosis based on age, sex, smoking status, the presence of symptoms, FEV1, and primary care visits. Diagnosis in CanCOLD was based on 95  self-report of having ever received a physician diagnosis of COPD, chronic bronchitis, or emphysema at each study visit. The regression equation for baseline diagnosis was fitted using data from the initial study visit only. Follow-up diagnosis was fitted using all study visits. In both cases, a diagnosis of COPD could only occur if the individual had ≥1 primary care visit in the preceding year. The regression equation for incident diagnosis is shown in Equation 5.1:  Equation 5-1     !"#$%(') = *! + *" × -#. + *# × /.0 + *$ × /1"2$3#	/%-%5/ +	*% × 6"5#ℎ 																																																					+*& × 'ℎ!.#1	 +	*' × 8ℎ..9.	 +	*( × :;/'3.-	 +	*) × <=>1 																																+	** × '@$1-@;	6-@.	A$/$%/	 +	*"! × 6-/.	:.%.6%$"3  where p is the probability of a patient with COPD being diagnosed during follow-up, and the */ are the input parameters for the probability of diagnosis.	*" through ** were obtained from the logistic regression model that was fitted to data from CanCOLD. *"! was not available in the data and was determined through a process described in Section 5.2.5.6.  Non-COPD individuals in EPIC could receive a false positive diagnosis of COPD (overdiagnosis) annually. I developed a logistic regression model to predict the odds of overdiagnosis among non-COPD subjects in CanCOLD (N=1,011). Model covariates were age, sex, smoking status, the presence of symptoms, and primary care visits. Overdiagnosed patients were re-assessed annually, and at each re-assessment there was a probability for the diagnosis to be corrected. This parameter was not reported in the literature and was determined through a calibration process. The calibration target was a stable proportion of false positive diagnoses at the 3% prevalence observed in CanCOLD, which is similar to previous observations131. This resulted in a probability of 0.50 for correcting overdiagnosis at any given year. The results of Chapter 4, as well as a previous study44, indicate that approximately one third of COPD patients are initially diagnosed in hospital. COPD patients in EPIC could be 96  diagnosed during instances of hospitalisation due to severe exacerbation. At each hospitalisation, COPD diagnosis occurred with a certain probability. This parameter was not reported in the literature and was determined through a calibration process. The probability of hospital diagnosis and the intercept of the incident diagnosis equation were calibrated simultaneously to maintain a stable proportion of undiagnosed COPD at the same prevalence observed in CanCOLD150. This corresponded to a probability of 0.50 for inpatient diagnosis, and 35% of diagnoses occurring in hospital, which is in line with previous reports. Model specifications and the results of internal validation are provided in Appendix D.3.  5.2.5.5 Treatment module Treatment was implemented in EPIC according to the GOLD ABCD criteria15. At outpatient diagnosis, I assumed that individuals were treated with a LAMA if they had dyspnea, or a SABA if they did not. All overdiagnosed individuals received a SABA. An individual’s medication could be changed annually or after an exacerbation. Individuals who experienced a moderate exacerbation could receive a LAMA or a SABA, depending on whether they had dyspnea or not, respectively. Individuals who were hospitalised due to an exacerbation, or who had ≥2 moderate exacerbations in the previous 12 months, received combination therapy. This could be either a LABA/LAMA, or an ICS/LAMA/LABA if they were also experiencing dyspnea. The full treatment algorithm is shown in Appendix D.4.1. The overall prevalence of inhaled therapies in EPIC are shown in Appendix Figure D.4.2. They are relatively reflective of overall trends in medication use among COPD patients in British Columbia (BC)166.   The effectiveness of 3-months of NRT on the odds of successful smoking cessation was determined through calibration. I used the one-year effectiveness of NRT from a large meta-97  analysis171 as the starting point for calibration. The calibration target was the marginal effect of the smoking cessation intervention on the proportion of non-smokers in the Lung Health Study, over 11-years172 (Appendix D.4.3). This resulted in an odds of successful smoking cessation of 1.38 due to the intervention. Background smoking cessation and incidence rates have been validated with overall trends in smoking in the general Canadian population, and have been previously described80.  5.2.5.6 Case detection module Case detection strategies increased the odds of true positive and false positive diagnoses according to their sensitivity and specificity, which I derived from previous case detection studies reported in the literature66,159.  In order to maintain accurate estimates of sensitivity and specificity, the selection characteristics for each strategy followed those used in the studies from which the performance characteristics were derived. The effect of case detection was added as a parameter to the regression equations for prevalent diagnosis, incident diagnosis (Equation 5-1) and false positive diagnosis. I used the sensitivity and specificity of each case detection method (shown in Table 5-1) to determine the coefficient for case detection (*"! in Equation 5-1). I solved for *"! such that the average probability of diagnosis equalled the sensitivity of the case detection method (for COPD patients), and the average probability of false positive diagnosis equalled the specificity of the case detection method (for non-COPD individuals). To accomplish this, I converted the sensitivity (for COPD patients) and the specificity (for non-COPD individuals) of the case detection method into odds, and divided by the odds of routine diagnosis for each patient to derive the average (log) odds ratio for the effect of case detection, as shown in Equation 5-2.  98   Equation 5-2   *"! = log E"+F ,--.	,0	-1234,.1.	5167	82.9	-969861,4,--.	,0	:,;6149	-1234,.1.!+<=" 	G 	  For a given case detection method, the ‘odds of diagnosis with case detection’ is the sensitivity (1 − sensitivity)⁄  for COPD patients, and the specificity (1 − specificity)⁄  for non-COPD individuals. The ‘odds of routine diagnosis1’ is the estimated probability of the ith patient being diagnosed without case detection from the logistic regression model (Equation 5-1 for COPD patients). The coefficient for each case detection method (*"!) and an example calculation are shown in Appendix D.5. Following guideline recommendations6, I assessed an additional scenario in which only patients with respiratory symptoms at the baseline visit were eligible for case detection (scenario S2a, Table 5-1). I was unable to find an example in the literature of the case detection methods of interest implemented in this subgroup of the population. Therefore, I performed an analysis of CanCOLD to determine the sensitivity and specificity of a theoretical screening spirometry intervention (based on pre-bronchodilator FEV1/FEV6<0.7) implemented among participants with any respiratory symptoms176. I repeated the process described above to convert these performance characteristics into coefficients for diagnosis.  5.3 Results 5.3.1 Reference case All case detection scenarios resulted in higher total costs and QALYs than no case detection. Scenarios with more inclusive eligibility criteria (S1- all patients) had higher total costs than more restrictive scenarios (S2- symptomatic patients and S3- smoking history). Scenarios using 99  the CDQ at its lowest threshold for referral to diagnostic spirometry produced the greatest number of total QALYs (S1a, S3b). Case detection methods with a high specificity, such as the CDQ and screening spirometry applied in combination (S1c and S3d), resulted in fewer total QALYs. Total costs and QALYs were higher when case detection was administered at 3-year intervals than at 5-years; however, these scenarios were not always more cost-effective (Table 5-3).  Table 5-3 Total costs and QALYs for case detection scenarios. Scenario Testing Interval Cost ($/ Patient) QALYs (QALYs/Patient) ICER ($/QALY) INMB* (Ranking) S0: No case detection NA $2,043 12.546 Reference Reference All patients (S1a) CDQ ≥17 points 3 years $2,318 12.560 18,791 457 (1) (S1a) CDQ ≥17 points 5 years $2,239 12.556 18,985 321 (2) (S1b) Screening spirometry 3 years $2,246 12.554 24,811 206 (6) (S1b) Screening spirometry 5 years $2,182 12.552 22,211 174 (8) (S1c) CDQ + Screening spirometry 3 years $2,268 12.551 45,047 25 (16) (S1c) CDQ + Screening spirometry 5 years $2,199 12.550 37,084 54 (15) Symptomatic patients (S2a) Screening spirometry 3 years $2,173 12.553 18,096 229 (4) (S2a) Screening spirometry 5 years $2,134 12.551 16,757 180 (7) Smoking history (S3a) CDQ ≥19.5 points 3 years $2,123 12.548 29,100 57 (14) (S3a) CDQ ≥19.5 points 5 years $2,097 12.548 21,688 71 (12) (S3b) CDQ ≥16.5 points 3 years $2,178 12.553 17,641 247 (3) (S3b) CDQ ≥16.5 points 5 years $2,138 12.552 15,542 210 (5) (S3c) Screening spirometry 3 years $2,143 12.550 22,946 118 (9) (S3c) Screening spirometry 5 years $2,113 12.549 19,404 110 (10) (S3d) CDQ + Screening spirometry 3 years $2,151 12.549 27,066 91 (11) (S3d) CDQ + Screening spirometry 5 years $2,116 12.548 26,435 65 (13)  The ‘All patients’ strategy encompassed the entire population of interest. 59% of the population was included in the ‘Symptomatic’ strategy, and 46% in the ‘Smoking history’ strategy. To maintain a constant reference population, 100  per patient costs and QALYs were adjusted to include individuals not selected by the strategy. For the symptomatic strategy, the costs (and QALYs) shown are the sum of the costs (QALYs) of individuals not included in the symptomatic strategy, and the costs (QALYs) of those included, weighted by the proportion in each group. The results of smoking history strategy were adjusted in the same fashion. * INMB is expressed incrementally in comparison to no case detection (scenario S0). Positive INMB indicates that the scenario is more cost-effective than no case detection at a willingness to pay threshold of $50,000/QALY.    The cost-effectiveness plane is shown in Figure 5-2. All scenarios were cost-effective compared to no case detection. A scenario in which the CDQ ≥16.5 points was delivered to ever smokers at 5-year intervals (scenario S3b) had the lowest ICER relative to no case detection. Therefore, the first step on the efficiency frontier was from the status quo (no case detection) to scenario S3b, which was the new default scenario. The CDQ ≥19.5 points among ever smokers (S3a), screening spirometry among ever smokers (S3c), the CDQ and screening spirometry among ever smokers (S3d), the CDQ and screening spirometry among all patients (S1c), and screening spirometry among symptomatic individuals (S2a) at 5-year intervals, were subject to extended dominance relative to scenario S3b.  Relative to the new default scenario (S3b), the remaining scenarios were: the CDQ ≥17 points among all patients (S1a), screening spirometry among all patients (S1b), screening spirometry among symptomatic individuals (S2a) at 3-year intervals, and the CDQ ≥16.5 points among ever smokers (S3b) at 3-year intervals. The CDQ ≥17 points delivered to all patients (S1a) at 3-year intervals had the lowest ICER, and because there were no remaining scenarios on the cost-effectiveness plane, this was the final step on the efficiency frontier. The CDQ ≥17 points delivered to all patients at 3-year intervals (scenario S1a) was associated with an incremental cost of $180 per eligible patient, and an incremental effectiveness of 0.009 QALYs 101  per eligible patient compared to the CDQ ≥16.5 points delivered to ever smokers at 5-year intervals (scenario S3b), resulting in an ICER of $21,108/QALY. The preferred case detection strategy in this analysis was the CDQ ≥17 points delivered at 3-year intervals to all patients ≥40 years of age (scenario S1a).   Figure 5-2 Cost-effectiveness plane for case detection scenarios.   Refer to Table 5-3 for a full legend of scenarios. 102  The solid blue line indicates the cost-effectiveness frontier. The grey dashed line indicates the WTP threshold ($50,000/QALY). The preferred scenario is highlighted in red.   For a cohort of 1,000 individuals eligible to receive case detection, the preferred case detection scenario (S1a) resulted in total costs of $2,318,121 and total QALYs of 12,560. The comparator scenario with no case detection had total costs of $2,043,113 and total QALYs of 12,546. Therefore, case detection with the CDQ ≥17 points delivered to all patients at 3-year intervals (S1a) resulted in an additional 15 QALYs at a cost of $275,008 relative to no case detection. In the same cohort of 1,000 individuals, the total patient years diagnosed with COPD increased from 356 diagnosed-years without case detection, to 890 diagnosed-years with case detection. Case detection increased the rate of treatment by 124 patient years for SABA, 46 patient years for LAMA, 304 patient years for LAMA/LABA, and 28 patient years for ICS/LAMA/LABA. This resulted in 13 fewer mild exacerbations, two fewer moderate, and three fewer severe exacerbations, and no change in the number of very severe exacerbations (<0.2 difference). Case detection resulted in an additional two patient years in GOLD grade II, and <1 patient year difference between all other GOLD grades. These results are depicted graphically in Appendix D.6.  5.3.2 Sensitivity analyses These results were sensitive to several key parameters related to the treatment of newly diagnosed patients (Figure 5-3). I removed 3-months of NRT for smoking cessation among newly diagnosed patients. As in the reference case, the preferred case detection scenario using 103  INMB was the CDQ ≥17 points delivered to all patients (S1a) at 3-year intervals. The INMB of most scenarios was unchanged or slightly lower, and the CDQ ≥19.5 points delivered to ever smokers (S3a) at 5-year intervals was no longer cost-effective. When medication adherence was reduced to 50%, results were similar to the reference case. The preferred scenario using INMB was again the CDQ ≥17 points delivered to all patients (S1a) at 3-year intervals, although its INMB was lower than in the reference case. The CDQ ≥19.5 points delivered to ever smokers (S3a) at 3-year intervals was not cost-effective. At 30% medication adherence, the preferred scenario using INMB was still the CDQ ≥17 points delivered to all patients (S1a), but at 5-year intervals. The INMB of all scenarios decreased; in addition to the CDQ ≥19.5 points delivered to ever smokers (S3a) at 3-year intervals, the CDQ and screening spirometry among all patients (S1c) was no longer cost-effective.  The cost-effectiveness of case detection was most sensitive to changes in the utility benefit due to symptom relief from treatment. This parameter was 0.0367 in the reference analysis, which was derived from a previous modelling study156. When the utility benefit was reduced to 0.01 in a sensitivity analysis, the INMB of all scenarios decreasing substantially. The preferred scenario was the CDQ ≥16.5 points delivered to ever smokers (S3b) at 5-year intervals, which had an ICER of $32,200/QALY compared to no case detection. The only other cost-effective scenarios were screening spirometry among symptomatic patients (S2a), the CDQ ≥17 points delivered to all patients (S1a) at 3-year intervals, screening spirometry among all patients (S1b), the CDQ ≥19.5 points among ever smokers (S3a), and screening spirometry among ever smokers (S3c), all delivered at 5-year intervals. Case detection was not cost-effective when the utility benefit due to symptom relief from treatment was removed (Figure 5-3). 104   Figure 5-3 Sensitivity analyses of the incremental net monetary benefit of case detection scenarios compared to no case detection.  105  5.4 Discussion 5.4.1 Summary of results I used a previously validated microsimulation model of COPD in the general Canadian population to assess the cost-effectiveness of early detection strategies for COPD. My results indicate that primary care-based case detection is highly cost-effective. Variations in the eligibility criteria for testing, the use of questionnaire- or spirometry-based case detection methods, and 3- or 5-year testing intervals were nearly all cost-effective compared with the status quo. The key feature in determining the value of case detection was treatment for newly diagnosed patients. In particular, the degree of symptom relief due to inhaled therapies. When this aspect of treatment was removed, case detection was no longer cost-effective.  5.4.2 Comparison with other research Previous guidelines have recommended against screening for COPD among asymptomatic adults in the general primary care population6. These recommendations apply to the use of diagnostic spirometry to test for COPD directly, and to methods such as screening spirometry and questionnaires, although the authors noted a lack of evidence on the value of screening tools in this population6. In contrast, I found that both questionnaires and screening spirometry, implemented at routine primary care visits, could be cost-effective methods for early detection. Similarly, a recent trial-based economic evaluation found that an active case finding approach to COPD detection using a mailed symptom assessment questionnaire was highly cost-effective155,156. My results extend this analysis and suggest that all primary care patients, regardless of symptoms, should routinely receive risk assessment questionnaires.   106  5.4.3 Policy implications The value of early detection for COPD is largely determined by the degree of adherence to best-practice recommendations for treatment, and the extent to which newly diagnosed patients benefit from symptom relief. Large clinical trials have demonstrated the effectiveness of inhaled therapies at reducing symptoms160–162,177, including among patients with mild COPD178. The impact of improved symptom control on health-related quality of life has also been well documented72,115. However, despite the importance of symptom relief to the value of early detection, the preferred case detection strategy in this analysis targeted all patients, rather than only patients with symptoms at the baseline primary care visit. In Chapter 3, I found substantial heterogeneity in the presence of respiratory symptoms between patients with COPD, and the stability of symptoms also varied within patients over time150. Similarly, I found that symptoms performed poorly as selection criteria for COPD screening176, and in this study, they had lower value when used as eligibility criteria for case detection.  The preferred scenario in this analysis targeted all patients ≥40 years of age to receive questionnaire-based case detection during their routine visits to a primary care physician. There are an estimated 1.8 million Canadians who currently have undiagnosed persistent airflow limitation46,179, meaning there are 15.6 million Canadians ≥40 years who do not have COPD180.  Extrapolating the per patient costs of the preferred case detection scenario to this population, the budget impact of this policy would be approximately $40.3 billion over 20 years (undiscounted). The budget impact of a more restrictive scenario, such as one targeting patients ≥50 years with a smoking history, would be much smaller ($27.4 billion). However, its effectiveness would also be lower; resulting in a total of 158.0 million QALYs, compared to 218.3 million QALYs with the preferred case detection scenario. 107  The case detection tools evaluated in this study differed mainly in terms of their performance characteristics (sensitivity and specificity) and mode of delivery. Both the questionnaire-based methods and screening spirometry were cost-effective compared to the status quo. In British Columbia, these case detection tools were relatively low in cost, although similar in cost to a questionnaire-based case detection tool used in a previous randomised controlled trial156. In all scenarios, case detection resulted in only small increases in per-patient costs. As a result, minimizing the use of diagnostic spirometry with a highly specific tool was less valuable. Scenarios in which the CDQ and screening spirometry were implemented sequentially had a specificity of 97%. These scenarios were typically associated with lower costs, but they also had lower value. The sensitivity of the case detection tool was the more important component of performance. The CDQ at a low threshold for referral to diagnostic spirometry was the preferred case detection tool. However, there was also support for the use of screening spirometry, particularly under conditions of poor adherence to inhaled therapies or diminished effectiveness of treatment. In these cases, the higher unit costs of screening spirometry were offset by their increased specificity. In addition, following previous evidence170, I did not assign a disutility to healthy patients who received a false positive diagnosis of COPD. These patients only incurred the additional costs of diagnostic spirometry. It is possible that I underestimated the costs of a false positive diagnosis, including unnecessary lung imaging, initiation of treatment without diagnostic spirometry, and additional physician visits63,64.   5.4.4 Clinical implications The effectiveness of inhaled therapies was almost entirely influenced by the additional utility due to symptom relief from treatment, rather than their effect on the rate of exacerbations. Earlier 108  initiation of treatment through case detection resulted in only a slight reduction in the overall rate of exacerbations. Because undiagnosed patients are more likely to have mild disease129, a lower rate of exacerbations, and exacerbations that are less severe80, this aspect of treatment was less influential. Case detection tools that specifically target patients with more severe COPD and a high risk of exacerbations181 might be more effective at preventing severe exacerbations and increase the value of this aspect of treatment. However, proper adherence to inhaled therapies, with the aim of relieving symptoms in patients with mild disease, is most critical to the benefit of an early diagnosis. Recent evidence indicates that inhaled therapies might be associated with improved lung function in patients with mild-moderate COPD178, and there is emerging evidence that early treatment can have a disease-modifying effect in these patients160. Treatment also has an indirect impact on lung function by reducing the rate of exacerbations, which are in turn associated with accelerated lung function decline57. Previous studies suggest this effect can be substantial, and greatest in patients with mild COPD56,57. However, I did not model the effect of treatment on lung function through either of these mechanisms. Therefore, my approach to assessing treatment effect can be regarded as conservative. My results address the long-standing debate on the value of early detection in patients with mild disease. Many authors have questioned the importance of detecting asymptomatic patients with limited therapeutic options6,63. This analysis indicates that even relatively small improvements in health outcomes due to early detection can be cost-effective over a long time horizon.   109  5.4.5 Strengths and limitations This study has several strengths. First, EPIC is an independently validated, open population model that simulates the demography and risk factors of the Canadian population, including their temporal trends during the period of evaluation. Given that the performance of case detection strategies is highly dependent on local settings and subsequent management decisions, the use of ‘whole disease’ platforms such as EPIC can result in more consistent evaluations. They can also be updated more easily to account for changes to ‘downstream’ technologies such as treatment. Second, my use of an individual-level simulation offered greater flexibility in the design of case detection strategies, including different permutations of eligibility criteria, case detection method, and treatment.  My study has several limitations. First, my approach to assessing case detection methods used a generic application of performance characteristics rather than a direct assessment of the case detection tool (e.g., in a randomised trial). As a result, I was highly reliant on previous studies that evaluated the effectiveness of case detection tools. I was also required to use similar selection criteria to what had been employed in those studies. This limited my ability to test even more subgroups of patients. Second, EPIC is not probabilistic, meaning that I could not explore parameter uncertainty in a probabilistic sensitivity analysis. Finally, with the exception of instances when case detection was delayed because a patient had no primary care visits, I modelled perfect uptake of the case detection tool at the scheduled intervals. My analysis should be extended to model gradual market penetration of case detection strategies in a population-level budget impact analysis. This type of analysis can also account for budget constraints, which might lead to a different case detection scenario being adopted than the one selected using the efficiency frontier approach. 110  5.4.6 Conclusions An early diagnosis of COPD, accompanied by timely and appropriate care, is very likely to be an efficient use of resources. Regular implementation of a screening questionnaire during the routine encounters of primary care patients ≥40 years of age is cost-effective compared to no case detection. Providing treatment to reduce the respiratory symptoms of newly diagnosed patients is the most important component of early detection. Future studies should evaluate the cost-effectiveness of the preferred case detection strategy in this analysis using a randomised control trial. Model-based evaluations such as this one should be extended to assess the impact of implementing case detection strategies in other healthcare settings, such as in pharmacist clinics. In summary, early detection strategies are a high value approach to improving the long-term health outcomes of patients with COPD.    111  Chapter 6: Conclusions  6.1 Contributions and implications In this thesis, I used a variety of methodological approaches to characterise undiagnosed COPD patients, quantify heterogeneity in their respiratory symptoms (which were a key factor in determining whether a patient with COPD had received a diagnosis), assess the healthcare encounters of patients before COPD diagnosis, and evaluate the cost-effectiveness of case detection strategies. My results indicate that undiagnosed COPD patients are identifiable, that they are in frequent contact with the healthcare system, and that strategies for improving early detection provide good value for money. These results can help practitioners recognize undiagnosed COPD patients during their routine healthcare encounters and inform strategies for improving their detection. Many authors have argued that an earlier diagnosis of COPD would have substantial public health benefits13, however, objective evaluations of early detection strategies were lacking. This thesis contains one of only two economic evaluations of case detection strategies for COPD156. It is also the only simulation study, which facilitated synthesizing evidence from diverse sources and more fully exploring the decision space created by early detection policies. My results provide an important message for policy-makers: early detection strategies for COPD are cost-effective and should be implemented, and for clinicians: patients with risk factors for COPD should receive case detection regularly. Given the large burden of COPD on patients and the healthcare system, better recognition of risk factors for COPD and the importance of early detection can have substantial public health benefits.  112  6.1.1 Improving recognition of risk factors for COPD The value of early detection for COPD is one of the most debated issues in the respiratory research community7,61. The results of this thesis suggest that clinicians and policy-makers should take a proactive approach towards diagnosing COPD. In a systematic review and meta-analysis (Chapter 2), I found that undiagnosed patients tended to have less severe airflow obstruction and fewer respiratory symptoms than diagnosed patients. However, common risk factors for COPD, including older age and current smoking, were not associated with a diagnosis. This suggests that diagnoses of COPD are currently being made based on lung function impairment and patient symptoms rather than risk factors for the disease, which may be contributing to the delay in diagnosis. The importance of respiratory symptoms in determining a previous diagnosis may be partly explained by current guidelines that recommend limiting screening or case detection for COPD to symptomatic individuals6.  In Chapter 3, I characterised heterogeneity in the presence of respiratory symptoms between patients with COPD using a longitudinal cohort of the general Canadian population. I found that the presence of cough, phlegm, dyspnea, and wheeze were highly heterogeneous between patients within COPD. Lung function and other commonly measured clinical characteristics explained a minority of this heterogeneity. Within patients, symptoms also differed in their stability over time. These results, and related work176, suggest that respiratory symptoms alone provide a poor signal for the presence of COPD. In order to improve the diagnosis of COPD in the community, risk factors for COPD should prompt an investigation of disease, rather than relying on patient impairment due to symptoms or airflow limitation. In my evaluation of the cost-effectiveness of case detection strategies (Chapter 5), symptom relief from inhaled therapies was critically important to the value of case detection. 113  However, using smoking history as eligibility criteria to receive case detection, or targeting all patients, was preferred to a strategy in which only patients with a history of symptoms were tested for COPD. Taken together, the results of this thesis indicate that reducing the burden of undiagnosed COPD will require more attention to risk factors for COPD, rather than exclusively evaluating patient symptoms. Guidelines for screening should be updated to reflect this focus6. These guidelines should also promote the use of case detection tools, such as the COPD diagnostic questionnaire, during routine healthcare encounters, instead of using diagnostic spirometry in the general population of primary care patients.  6.1.2 Early detection of COPD is cost-effective Case detection strategies rely on opportunities for diagnosis. Previous authors have suggested that the routine healthcare encounters of at-risk patients should be used to opportunistically test for COPD60–62. The results of this thesis support this conclusion. Patients with undiagnosed COPD frequently encounter the healthcare system, and missed opportunities for an earlier diagnosis are common. Using administrative health records for the province of British Columbia (Chapter 4), I determined that COPD patients had an average of 14 visits to pharmacists, and 10 visits to primary care physicians in each of the five years before diagnosis. Over the same time period, 85% of patients visited a primary care physician for a respiratory-related reason that did not result in a diagnosis of COPD. Due to the exclusion criteria used in this study, none of these patients received spirometry to test for COPD, and many were already receiving respiratory medications. These results indicate that case detection strategies relying on opportunistic encounters between undiagnosed COPD patients and the healthcare system have opportunities to 114  diagnose COPD earlier. There is a need for systematic strategies to increase the use of spirometry in practice. Regular administration of questionnaire-based assessments of COPD risk during routine primary care visits can be a high value, evidence-based approach to early detection. The use of screening spirometry with handheld flow meters (both with and without bronchodilators) was also supported in my evaluation of the cost-effectiveness of case detection strategies. However, in most cases, case detection strategies with a high sensitivity were preferred to highly specific tools such as screening spirometry. The use of case detection tools should be integrated into regular practice among primary care patients ≥40 years of age. The lack of treatment options for patients with mild disease is the most commonly cited argument against investing resources in early detection strategies7. The hallmarks of COPD therapy are risk factor modification and inhaled therapies15. There is mounting evidence that early administration of respiratory medications can reduce symptoms15, the risk of exacerbations178, and potentially, lung function decline57,58. In my evaluation of the cost-effectiveness of case detection strategies, symptom relief due to inhaled therapies was the most important component of treatment following diagnosis. These results indicate that current guideline recommendations for therapy, applied to undiagnosed patients in mild disease stages, can improve their long-term health outcomes. However, the cost-effectiveness of case detection is contingent upon high adherence to inhaled therapies following diagnosis. If a diagnosis does not result in treatment that reduces the symptoms of COPD patients, case detection is unlikely to be cost-effective. Therefore, in addition to increased testing for COPD, the importance of providing evidence-based treatment to newly diagnosed patients must be emphasised.  115  6.1.3 Disease heterogeneity should be quantified  The overarching themes of this thesis were evidence generation on the natural history of disease, followed by whole disease modelling to evaluate policy. Whenever possible, I generated evidence that can facilitate precision medicine, which requires information on heterogeneity in the disease characteristics of patients. In Chapter 2, I addressed a key knowledge gap on the heterogeneity of respiratory symptoms in patients with COPD, which are a core component of its natural history. I found that heterogeneity between individuals in the presence of symptoms was substantial, and that very little of this variation was explained by easily-measured patient and disease factors. Future developments in precision medicine, such as the discovery of biomarkers, might reduce this unexplained heterogeneity. This type of evidence can improve our understanding of the burden of symptoms, or identify disease phenotypes to inform management decisions. The equations describing between-patient variability in symptoms in this work became the key component of the symptom module of EPIC, which was used to evaluate case detection strategies. Because heterogeneity in the symptoms of simulated individuals was explicitly incorporated, future studies can use this model as a platform to evaluate technologies in precision medicine.   6.1.4 Whole disease modelling can inform policy development Whole disease models such as EPIC are a flexible platform for exploring the decision space created by a policy or intervention. These types of models are in contrast to de novo modelling, which are purpose-built decision models for evaluating a specific policy. The vast majority of previous COPD simulations have been de novo models182. As discussed in Chapter 1, I used a whole disease model because of the need to account for the downstream impacts of early 116  detection, especially treatment for newly diagnosed patients. The benefits and limitations of using a whole disease model to evaluate policy merits discussion.  Whole disease models are time-consuming to develop and tend to be very complex. As a result, conventional methods for evaluating structural and parameter uncertainty are difficult to implement and computationally-intensive. This is an important limitation of whole disease models, as guidelines recommend that all health economic evaluations assess parameter uncertainty157. In this thesis, there was a trade-off between the granularity at which any given case detection method could be modelled, and fully accounting for the decision space created by a policy or intervention. As an alternative to whole disease modelling, I could have created a de novo model to evaluate the cost-effectiveness of the COPD diagnostic questionnaire in a particular subgroup of the population. This approach may have may have produced more consistent estimates of parameters measured from a single population. Whole disease modelling enables a broader approach to economic evaluation. The decision space for the evaluation of case detection strategies involves different permutations of eligibility criteria, case detection method, period of intervention, and disease management following diagnosis. A de novo model simulating a single case detection method in a particular population would have less ability to explore this decision space. It is also more likely to make simplifying assumptions about the distribution of risk factors, disease heterogeneity, and treatment effectiveness, all of which might impact the robustness of results. In contrast, whole disease models are a flexible platform for exploring this decision space, and can facilitate a wider range of economic evaluations. For example, in this research, I assessed different eligibility criteria for case detection, and identified the features of case detection methods (i.e. sensitivity versus specificity) and treatment that were most influential for the cost-effectiveness of early 117  detection. These results can inform the development of new case detection methods, or identify key targets for intervention. Although whole disease models are more time consuming to develop than de novo models, the contributions I made to the development of EPIC will have lasting benefit to future evaluations. The decision space that can be explored now includes three additional aspects of disease: diagnosis, respiratory symptoms, and healthcare encounters. Because EPIC is freely accessible to other researchers, this can reduce duplicate efforts within the community and increase the reproducibility of my analyses. It also makes whole disease modelling a more sustainable approach to evaluation. For example, the value of case detection is likely to change if new therapies for COPD are introduced or with new evidence on existing therapies. My evaluation of case detection strategies could be repeated to incorporate these changes to treatment effectiveness, which would provide valuable evidence to decision-makers.  Finally, by evaluating decisions at any point along the care pathway, whole disease modelling can promote system-level optimization. This enables new possibilities for resource management, such as disinvestment and finding optimal care bundles. In this research, I focused on the value of an early diagnosis of COPD. However, there are many other points of intervention along the COPD care pathway. For example, additional resources could be allocated to drug discovery, where new therapies to modify disease progression might be developed. Alternatively, decision-makers could invest resources in improving access to pulmonary rehabilitation programs183. Whole disease models can evaluate a variety of interventions using a single platform with consistent assumptions. This enables a more realistic quantification of the opportunity cost of investment decisions, and can improve overall disease management.  118  6.2 Strengths of this research There are several strengths to this work. I used a variety of methodological approaches to characterise undiagnosed COPD and evaluate strategies for improving its detection. I summarised existing evidence in a systematic review, conducted original analyses of a clinical cohort and administrative health records, and made significant contributions to a decision-analytic model of COPD.  The external validity of the data sources used in this research is one of its main strengths. My systematic review included studies from four continents and is likely to be representative of most high-income countries. The analyses in Chapter 3 used data from a population-based prospective cohort study, and administrative health records from the general population of COPD patients in Chapter 4. EPIC was developed to be representative of COPD in the Canadian general population. CanCOLD, a longitudinal cohort study, was the source of data for my analysis of heterogeneity in respiratory symptoms (Chapter 3), and for the additional modules I developed to simulate the case detection pathway in EPIC (Chapter 5). CanCOLD sampled adults ≥40 years from nine major urban sites in Canada with multi-level sampling to ensure representativeness of the general population. Because CanCOLD was developed to study the development and progression of COPD in the general population, the majority of subjects enrolled had mild to moderate COPD. This is in contrast to most clinical trials, which typically enrol patients in moderate to severe disease stages184. Therefore, CanCOLD is a unique data source for evaluating questions around undiagnosed COPD.   My use of a decision-analytic model for evaluating early detection strategies is another strength of this research. Using this approach, I incorporated multiple different sources of evidence, including from CanCOLD45, landmark trials in COPD therapy58,162,172,185, and 119  administrative health records from BC166.  This facilitated a fuller exploration of the decision space than any single empirical study could provide. The decision-analytic model also allowed evidence on intermediate outcomes from these studies to be extrapolated to a long time horizon. This is critical for chronic diseases such as COPD, in which the benefits of an early diagnosis may not be realised until much later in the disease progression.  6.3 Limitations of this research Decision-analytic models require many different assumptions and input data sources, and high-quality empirical data are not always available. There were two primary sources of uncertainty in the evidence used to populate parameters in EPIC. First, as discussed above, CanCOLD was the primary source of evidence for the case detection pathway simulated in EPIC. These data may not be representative of the general population of individuals without COPD. Non-COPD subjects in CanCOLD were selected because they were never smokers (healthy controls), or because they were ever smokers, and therefore at risk of developing COPD45. Ever smokers are likely to have more respiratory symptoms than the healthy population186. However, both groups of non-COPD subjects were used to simulate respiratory symptoms in the general, non-COPD population of EPIC. The frequency of respiratory symptoms may have been overestimated in this population as a result. The value of using symptoms as eligibility criteria for case detection may also have been underestimated. Other limitations of clinical studies such as healthy volunteer bias might also exist in CanCOLD. A second source of uncertainty was the effectiveness of inhaled therapies in patients with mild COPD. To date, most clinical trials of treatment effectiveness in COPD have enrolled patients in moderate to severe disease stages. With one notable exception58, early stage COPD 120  has been comparatively less studied. Given that most undiagnosed patients have mild COPD, applying the effectiveness of inhaled therapies to these patients might have overestimated the benefits of treatment. Finally, there was limited evidence on the effect of symptom relief due to treatment on health-related quality of life. This was the most influential parameter in determining the cost-effectiveness of case detection, and its value was based on a previous economic evaluation156. Quantifying the impact of treatment on health-related quality of life in patients with mild to moderate COPD would provide important evidence for future evaluations. To fill this knowledge gap, clinical trials or observational studies of inhaled therapies could administer the EQ-5D as a secondary outcome and provide a comparison of health-related quality of life between treatment groups.  Comorbidities are a key contributor to burden of COPD187. In Chapter 4, I found that circulatory disease was the most common reason for visits to a primary care physician among undiagnosed COPD patients. The presence of comorbidities is also likely to increase the probability that a patient with COPD has received a diagnosis, however, I was not able to investigate this factor in Chapter 2 due to the lack of existing literature. In part due to the lack of empirical data on comorbidities in patients with undiagnosed COPD, and also due to the complex interaction between COPD and other diseases, comorbidities are not currently modelled in EPIC. However, comorbidities influence the risk of COPD188 and the outcomes of COPD patients189. Therefore, the value of case detection is very likely to be affected by comorbid conditions. This is particularly true for cardiovascular comorbidities, which are highly prevalent among COPD patients23, and impact the management of COPD190. Primary care visits related to cardiovascular disease could also be used as opportunities for case detection. Future assessments 121  should explicitly incorporate the impact of cardiovascular comorbidities on the value of case detection. Another important limitation of this analysis is that I modelled an accelerated decline in FEV1 as the main disease process in COPD. Recent evidence has shown there are two primary biologic mechanisms for the development of COPD: an accelerated rate of lung function decline, or low maximum lung function due to abnormal lung development19. Despite evidence that disease progression and health outcomes differs between these groups, with important implications for treatment191, I did not consider this second pathway to disease in my evaluation. Further, other markers of disease such as blood eosinophil and history of asthma are also known to affect patient response to treatment149,191. Computed topography scans have the potential to identify markers of early disease before airflow obstruction develops192, and patients at risk of rapid progression could be identified and treated for smoking cessation and symptoms (if present)193. Incorporating these factors into a more nuanced treatment algorithm might affect the value of case detection. Finally, there are other approaches to early detection that were not evaluated in this thesis. In addition to questionnaires and screening spirometry, algorithms based on electronic medical records could be used to identify patients with undiagnosed COPD194. These algorithms could flag patients who are at high risk of having undiagnosed COPD, which would prompt a physician to investigate a diagnosis at their next visit. If this strategy reduced the number of patients eligible to receive case detection, it might increase the feasibility of early detection strategies. However, in developing these algorithms, it should be noted that the results of this thesis suggest that a more sensitive instrument is preferred to a highly specific one. Simulation 122  studies such as this one are well positioned to investigate the cost-effectiveness of strategies using these types of technologies.  6.4 Future research directions 6.4.1 Knowledge translation to policy-makers, knowledge users, and patients Expanding knowledge translation of this work is a key area for future research. Through peer-reviewed publications and conference presentations, I have communicated these results to care providers (e.g. at the Canadian and European Respiratory Congresses), and to technology assessment experts (e.g. at the Canadian Agency for Drugs and Technologies in Health Symposium). In addition, clinical experts and policy-makers were involved in the conceptualization and development of EPIC, and they remain active knowledge users. These stakeholders provided feedback on the relevant aspects of disease to model, the level of precision at which they needed to be represented, and assessed the model assumptions. However, further knowledge translation work remains. In particular, communicating the outputs of this research to stakeholders, and discussing the uncertainties and limitations of the model. The relevant stakeholders include other researchers, clinical guideline developers, government policy-makers, and patients with COPD. My results on the value of early detection for COPD should be accessible to these decision-makers and knowledge users. The case detection policies recommended by this research have a very large budget impact. As a result, policy adoption is likely to be highly dependent on factors specific to each jurisdiction. Allowing decision-makers to reassess the value of case detection under local conditions would produce more policy-relevant results.  123  Currently, EPIC is publicly available as an R package (http://resp.core.ubc.ca/research/Specific_Projects/EPIC). Existing functions allow other researchers to run the analysis described in Chapter 5 under different inputs and examine the results. However, this requires the user to be familiar with R Statistical Software92, which could be a substantial barrier to use. In future work, an existing model repository (http://resp.core.ubc.ca/ipress/prism) should be used to develop an interface in Excel for interacting with the model inputs and generating easily-understandable graphical outputs. This tool would facilitate policy-ready messages under conditions that are most relevant to decision-makers. It would also help communicate the uncertainties and variability in these results. Finally, other researchers in the scientific community could use this tool to replicate my analyses. This would increase model transparency and make my results more credible. Targeted knowledge translation such as this can increase the uptake of early detection strategies for COPD. Patients are a key stakeholder group who have thus far, not taken part in the development of EPIC, or my evaluation of case detection strategies. Communicating the results of this research to patients with COPD is a critical next step. Future researchers should elicit patient feedback on the values for certain parameters, and whether all outputs relevant to early detection have been collected. Given its influence on the cost-effectiveness of case detection, patient perspectives on the importance of symptom relief due to treatment would be particularly valuable.  Patients could also help prioritise future developments to EPIC as it is expanded to include other aspects of disease. For example, given that patients often view health economic models as simulating disease too narrowly195, they might emphasise the need to include comorbidities in future iterations of the model. Incorporating patient perspectives in this way 124  would increase the legitimacy of any subsequent results195. Finally, future researchers should discuss with patients the uncertainties and limitations of this research, and the degree to which this evidence should prompt changes in care. This information would be valuable to decision-makers and increase the acceptability of new policies related to the early detection of COPD.  6.4.2 Extending case detection to pharmacists Some of the resistance to implementing early detection strategies for COPD can be attributed to an overburdening of primary care physicians. Because many diseases have shared symptoms and risk factors, physicians must consider a multitude of potential diagnoses during very short visits with patients. A strategy in which all primary care patients ≥40 years were targeted for case detection would apply to an estimated 17.4 million Canadians180 and cost approximately $40.3 billion over 20 years (undiscounted). This is a major challenge to the feasibility of early detection strategies. Enlisting other healthcare providers to administer case detection is a promising solution.  In Chapter 4 of this thesis, I found that patients with COPD were in most frequent contact with pharmacists before diagnosis. In fact, the average annual visits to pharmacists reported in that chapter may have underestimated the frequency of encounters between undiagnosed COPD patients and pharmacists, as only visits in which patients were dispensed medications were recorded in the dataset. However, at these visits, patients with undiagnosed COPD were often dispensed respiratory medications. Pharmacists could target patients who are being dispensed respiratory medications without an existing respiratory disease diagnosis for case detection. These patients would be invited to complete the COPD diagnostic questionnaire. Patients who were found to be at risk of COPD using the questionnaire would be referred to their primary care 125  provider to investigate a diagnosis. Similar initiatives for improving detection of osteoarthritis in community pharmacy settings have been very successful196. In that study, community pharmacists recruited patients with symptoms, assessed their risk of osteoarthritis, and following examination, high risk patients were referred to their general practitioner to confirm a diagnosis196. There is general movement towards enabling pharmacist-provided care in Canada, which can increase the feasibility of this type of approach for COPD197. Financial incentives for collaboration between healthcare practitioners are becoming more widespread197, and can encourage innovative new approaches to improving patient care. New reimbursement models will be necessary to integrate case detection into the regular practice of pharmacists.  6.4.3 Empirical evaluation of case detection strategies In this thesis, I determined that early detection strategies for COPD were cost-effective. These results should be confirmed in a clinical trial setting. The preferred case detection strategy was the COPD diagnostic questionnaire (CDQ) implemented at 3-year intervals during the routine primary care visits of all patients ≥40 years. This strategy should be investigated using a pragmatic randomised control trial following previous empirical evaluations of case detection155,156. To reduce the pool of eligible participants and increase the feasibility of this study, patients ≥40 years could be randomly selected using electronic medical records. These patients would have their records flagged, which would prompt their primary care physician to administer the CDQ at their next visit. Patients recruited into the opportunistic case detection arm would be compared to patients who were flagged but did not receive case detection (the usual care arm). Outcomes such as the proportion of COPD patients diagnosed, and the costs of case detection would be collected. Linking the administrative health records of trial participants 126  would provide valuable information on the long-term outcomes of early detection. However, the results of this thesis suggest that even small improvements in disease outcomes due to improved diagnosis can have important benefits over a long time horizon. Therefore, evidence of the short-term cost-effectiveness of a case detection strategy would provide strong support for its value. An observational study could also be used to evaluate case detection strategies. Existing variation in the use of case detection methods could be assessed in a longitudinal cohort of patients with COPD. Within this cohort, subjects who received case detection would be matched to control subjects who did not receive case detection. The rate of diagnosis, costs of case detection, and differences in the treatment patterns and healthcare resource use of diagnosed patients would be compared between groups. Although this type of study could provide valuable information on the associations between case detection and health outcomes, it would be at risk of confounding by disease severity and variation in physician or patient preferences for administering case detection. A rich dataset would be required to account for these factors.  6.5 Conclusions Reducing the burden of undiagnosed COPD requires a proactive approach to disease diagnosis. Risk factors for COPD, such as smoking, should prompt an investigation of disease, rather than only the presence of respiratory symptoms. Symptoms are highly heterogeneous between patients with mild to moderate COPD, and restricting early detection strategies to symptomatic patients reduces their value. Screening questionnaires are a systematic, evidence-based method for assessing the risk of COPD. They should be used routinely in primary care settings to identify patients who require further testing for COPD. This strategy can facilitate interventions to preserve lung function while the disease is early in its progression. Symptom relief through 127  inhaled therapies, and treatment for smoking cessation among current smokers, can improve the long-term health outcomes of patients with COPD.  In summary, strategies for improving the early detection of COPD can be cost-effective and represent a high value use of resources. Given the mounting evidence from this thesis and similar work, there is a need to shift the focus from evaluating strategies for the early detection of COPD, to investigating barriers to their implementation. More broadly, the framework used in this thesis of evidence generation to understand the natural history of disease, followed by whole disease modelling to evaluate policy, can be transferred to other chronic diseases and used to improve patient care and health outcomes.  128  References 1.  Wilt TJ, Harris RP, Qaseem A. Screening for cancer: advice for high-value care from the American College of Physicians. Ann Intern Med. 2015;162(10):718-725. doi:10.7326/M14-2326 2.  US Preventive Services Task Force (USPSTF). Screening for high blood pressure in adults: U.S. Preventive Services Task Force Recommendation Statement. Ann Intern Med. 2015;163(10):I-32. doi:10.7326/P15-9036 3.  Lindsay P, Connor Gorber S, Joffres M, Birtwhistle R, McKay D, Cloutier L. Recommendations on screening for high blood pressure in Canadian adults. Can Fam Physician. 2013;59(9):927-933. 4.  US Preventive Services Task Force (USPSTF), Curry SJ, Krist AH, et al. Screening for Cervical Cancer: US Preventive Services Task Force Recommendation Statement. JAMA. 2018;320(7):674-686. doi:10.1001/jama.2018.10897 5.  Wilson J, Jungner G. Principles and practice of screening for disease. WHO, Geneva. 1968. http://www.who.int/bulletin/volumes/86/4/07-050112BP.pdf. 6.  US Preventive Services Task Force (USPSTF), Siu AL, Bibbins-Domingo K, et al. Screening for Chronic Obstructive Pulmonary Disease: US Preventive Services Task Force Recommendation Statement. JAMA. 2016;315(13):1372-1377. doi:10.1001/jama.2016.2638 7.  Enright P, White P. Detecting mild COPD: don’t waste resources. Prim Care Respir J. 2011;20(1):6-8. doi:10.4104/pcrj.2011.00004 8.  Soriano JB, Zielinski J, Price D. Screening for and early detection of chronic obstructive pulmonary disease. The Lancet. 2009;374(9691):721-732. doi:10.1016/S0140-6736(09)61290-3 9.  Kahn R, Alperin P, Eddy D, et al. Age at initiation and frequency of screening to detect type 2 diabetes: a cost-effectiveness analysis. Lancet. 2010;375(9723):1365-1374. doi:10.1016/S0140-6736(09)62162-0 10.  Berry DA, Cronin KA, Plevritis SK, et al. Effect of screening and adjuvant therapy on mortality from breast cancer. N Engl J Med. 2005;353(17):1784-1792. doi:10.1056/NEJMoa050518 11.  Cobiac LJ, Magnus A, Barendregt JJ, Carter R, Vos T. Improving the cost-effectiveness of cardiovascular disease prevention in Australia: a modelling study. BMC Public Health. 2012;12(1):398. doi:10.1186/1471-2458-12-398 12.  Feldman AL, Griffin SJ, Fhärm E, et al. Screening for type 2 diabetes: do screen-detected cases fare better? Diabetologia. 2017;60(11):2200-2209. doi:10.1007/s00125-017-4402-4 129  13.  Steenbruggen I, Zielinski J, Lange P, Price D, Soriano JB. A BOLD statement on how to case-find moderate/severe COPD. Eur Respir J. 2013;41(3):503-504. doi:10.1183/09031936.00153512 14.  Jameson JL, Longo DL. Precision Medicine — Personalized, Problematic, and Promising. N Engl J Med. 2015;372(23):2229-2234. doi:10.1056/NEJMsb1503104 15.  From the Global Strategy for the Diagnosis, Management and Prevention of COPD. Global Initiative for Chronic Obstructive Lung Disease (GOLD). 2019. http://goldcopd.org. Accessed May 1, 2019. 16.  Vestbo J, Edwards LD, Scanlon PD, et al. Changes in forced expiratory volume in 1 second over time in COPD. N Engl J Med. 2011;365(13):1184-1192. doi:10.1056/NEJMoa1105482 17.  Sinden NJ, Stockley RA. Systemic inflammation and comorbidity in COPD: a result of “overspill” of inflammatory mediators from the lungs? Review of the evidence. Thorax. 2010;65(10):930-936. doi:10.1136/thx.2009.130260 18.  Mannino DM, Buist AS. Global burden of COPD: risk factors, prevalence, and future trends. Lancet. 2007;370(9589):765-773. doi:10.1016/S0140-6736(07)61380-4 19.  Agustí A, Hogg JC. Update on the Pathogenesis of Chronic Obstructive Pulmonary Disease. N Engl J Med. 2019;381(13):1248-1256. doi:10.1056/NEJMra1900475 20.  Canadian Institutes of Health Research. CIHI releases top 10 reasons for hospitalization and surgery. https://www.cihi.ca/en/types-of-care/hospital-care/acute-care/cihi-releases-top-10-reasons-for-hospitalization-and-surgery. Accessed February 20, 2017. 21.  Sidney S, Sorel M, Quesenberry CP, DeLuise C, Lanes S, Eisner MD. COPD and incident cardiovascular disease hospitalizations and mortality. Chest. 2005;128(4):2068-2075. doi:10.1378/chest.128.4.2068 22.  Schneider C, Jick SS, Bothner U, Meier CR. COPD and the risk of depression. Chest. 2010;137(2):341-347. doi:10.1378/chest.09-0614 23.  Chen W, Thomas J, Sadatsafavi M, FitzGerald JM. Risk of cardiovascular comorbidity in patients with chronic obstructive pulmonary disease: a systematic review and meta-analysis. Lancet Respir Med. 2015;3(8):631-639. doi:10.1016/S2213-2600(15)00241-6 24.  Patel AR, Hurst JR. Extrapulmonary comorbidities in chronic obstructive pulmonary disease: state of the art. Expert Rev Respir Med. 2011;5(5):647-662. doi:10.1586/ers.11.62 25.  Soriano JB, Lamprecht B. Chronic obstructive pulmonary disease: a worldwide problem. Med Clin North Am. 2012;96(4):671-680. doi:10.1016/j.mcna.2012.02.005 130  26.  Han MK, Agusti A, Calverley PM, et al. Chronic obstructive pulmonary disease phenotypes: the future of COPD. Am J Respir Crit Care Med. 2010;182(5):598-604. doi:10.1164/rccm.200912-1843CC 27.  Mannino DM. COPD: epidemiology, prevalence, morbidity and mortality, and disease heterogeneity. Chest. 2002;121:121S-126S. 28.  Zafari Z, Sin DD, Postma DS, et al. Individualized prediction of lung-function decline in chronic obstructive pulmonary disease. CMAJ. 2016;188(14):1004-1011. doi:10.1503/cmaj.151483 29.  Anthonisen NR, Connett JE, Kiley JP, et al. Effects of Smoking Intervention and the Use of an Inhaled Anticholinergic Bronchodilator on the Rate of Decline of FEV1: The Lung Health Study. JAMA. 1994;272(19):1497-1505. doi:10.1001/jama.1994.03520190043033 30.  Hurst JR, Vestbo J, Anzueto A, et al. Susceptibility to exacerbation in chronic obstructive pulmonary disease. N Engl J Med. 2010;363(12):1128-1138. doi:10.1056/NEJMoa0909883 31.  Han MK, Quibrera PM, Carretta EE, et al. Frequency of Exacerbations in COPD: An Analysis of the SPIROMICS Cohort. Lancet Respir Med. 2017;5(8):619-626. doi:10.1016/S2213-2600(17)30207-2 32.  Sadatsafavi M, Sin DD, Zafari Z, et al. The Association Between Rate and Severity of Exacerbations in Chronic Obstructive Pulmonary Disease: An Application of a Joint Frailty-Logistic Model. Am J Epidemiol. 2016;184(9):681-689. doi:10.1093/aje/kww085 33.  Statistics Canada. Chronic Obstructive Pulmonary Disease by Sex, by Province and Territory. CANSIM Table 105-0501. http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/health105a-eng.htm. 34.  Buist AS, McBurnie MA, Vollmer WM, et al. International variation in the prevalence of COPD (the BOLD Study): a population-based prevalence study. Lancet. 2007;370(9589):741-750. doi:10.1016/S0140-6736(07)61377-4 35.  Adeloye D, Chua S, Lee C, et al. Global and regional estimates of COPD prevalence: Systematic review and meta-analysis. J Glob Health. 2015;5(2):020415. doi:10.7189/jogh.05-020415 36.  Khakban A, Sin DD, FitzGerald JM, et al. Ten-Year Trends in Direct Costs of COPD: A Population-Based Study. Chest. 2015;148(3):640-646. doi:10.1378/chest.15-0721 37.  Najafzadeh M, Marra CA, Lynd LD, et al. Future impact of various interventions on the burden of COPD in Canada: a dynamic population model. PloS One. 2012;7(10):e46746. doi:10.1371/journal.pone.0046746 131  38.  Khakban A, Sin DD, FitzGerald JM, et al. The Projected Epidemic of COPD Hospitalizations Over the Next 15 Years: A Population Based Perspective. Am J Respir Crit Care Med. 2017;195(3):287-291. 39.  Ried JL, Harnmond D, Rynard VL, Burkhalter R. Tobacco Use in Canada: Patterns and Trends, 2015 Edition. Waterloo, ON: Propel Centre for Population Health Impact, University of Waterloo. 40.  Hansen JE, Sun X-G, Wasserman K. Spirometric criteria for airway obstruction: Use percentage of FEV1/FVC ratio below the fifth percentile, not < 70%. Chest. 2007;131(2):349-355. doi:10.1378/chest.06-1349 41.  Hankinson JL, Odencrantz JR, Fedan KB. Spirometric reference values from a sample of the general U.S. population. Am J Respir Crit Care Med. 1999;159(1):179-187. doi:10.1164/ajrccm.159.1.9712108 42.  Kessler R, Partridge MR, Miravitlles M, et al. Symptom variability in patients with severe COPD: a pan-European cross-sectional study. Eur Respir J. 2011;37(2):264-272. doi:10.1183/09031936.00051110 43.  Mannino DM, Gagnon RC, Petty TL, Lydick E. Obstructive lung disease and low lung function in adults in the United States: data from the National Health and Nutrition Examination Survey, 1988-1994. Arch Intern Med. 2000;160(11):1683-1689. 44.  Balcells E, Gimeno-Santos E, de Batlle J, et al. Characterisation and prognosis of undiagnosed chronic obstructive pulmonary disease patients at their first hospitalisation. BMC Pulm Med. 2015;15(4). doi:10.1186/1471-2466-15-4 45.  Bourbeau J, Tan WC, Benedetti A, et al. Canadian Cohort Obstructive Lung Disease (CanCOLD): Fulfilling the need for longitudinal observational studies in COPD. COPD. 2014;11(2):125-132. doi:10.3109/15412555.2012.665520 46.  Labonté LE, Tan WC, Li PZ, et al. Undiagnosed Chronic Obstructive Pulmonary Disease Contributes to the Burden of Health Care Use. Data from the CanCOLD Study. Am J Respir Crit Care Med. 2016;194(3):285-298. doi:10.1164/rccm.201509-1795OC 47.  Lamprecht B, Soriano JB, Studnicka M, et al. Determinants of underdiagnosis of COPD in national and international surveys. Chest. 2015;148(4):971-985. doi:10.1378/chest.14-2535 48.  Peto R, Speizer FE, Cochrane AL, et al. The relevance in adults of air-flow obstruction, but not of mucus hypersecretion, to mortality from chronic lung disease. Results from 20 years of prospective observation. Am Rev Respir Dis. 1983;128(3):491-500. 49.  Anthonisen NR, Connett JE, Murray RP. Smoking and lung function of Lung Health Study participants after 11 years. Am J Respir Crit Care Med. 2002;166(5):675-679. doi:10.1164/rccm.2112096 132  50.  McCarthy B, Casey D, Devane D, Murphy K, Murphy E, Lacasse Y. Pulmonary rehabilitation for chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2015;(2):CD003793. doi:10.1002/14651858.CD003793.pub3 51.  Raymakers AJN, Sadatsafavi M, Sin DD, FitzGerald JM, Marra CA, Lynd LD. Inhaled corticosteroids and the risk of lung cancer in chronic obstructive pulmonary disease (COPD): a population-based cohort study. Eur Respir J. 2019;53(6):1801257. doi:10.1183/13993003.01257-2018 52.  Fletcher CM, Elmes PC, Fairbairn AS, Wood CH. The significance of respiratory symptoms and the diagnosis of chronic bronchitis in a working population. BMJ. 1959;2(5147):257-266. 53.  Jones PW, Harding G, Berry P, Wiklund I, Chen W-H, Kline Leidy N. Development and first validation of the COPD Assessment Test. Eur Respir J. 2009;34(3):648-654. doi:10.1183/09031936.00102509 54.  Decramer M, Miravitlles M, Price D, et al. New horizons in early stage COPD--improving knowledge, detection and treatment. Respir Med. 2011;105(11):1576-1587. doi:10.1016/j.rmed.2010.12.015 55.  Parkes G, Greenhalgh T, Griffin M, Dent R. Effect on smoking quit rate of telling patients their lung age: the Step2quit randomised controlled trial. BMJ. 2008;336(7644):598-600. doi:10.1136/bmj.39503.582396.25 56.  Decramer M, Cooper CB. Treatment of COPD: the sooner the better? Thorax. 2010;65(9):837-841. doi:10.1136/thx.2009.133355 57.  Dransfield MT, Kunisaki KM, Strand MJ, et al. Acute exacerbations and lung function loss in smokers with and without chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2017;195(3):324-330. doi:10.1164/rccm.201605-1014OC 58.  Zhou Y, Zhong N, Li X, et al. Tiotropium in Early-Stage Chronic Obstructive Pulmonary Disease. N Engl J Med. 2017;377(10):923-935. doi:10.1056/NEJMoa1700228 59.  Haroon SM, Jordan RE, O’Beirne-Elliman J, Adab P. Effectiveness of case finding strategies for COPD in primary care: a systematic review and meta-analysis. NPJ Prim Care Respir Med. 2015;25:15056. 60.  Jithoo A, Enright PL, Burney P, et al. Case-finding options for COPD: results from the Burden of Obstructive Lung Disease study. Eur Respir J. 2013;41(3):548-555. doi:10.1183/09031936.00132011 61.  Kaplan A, Freeman D, Cleland J, Cerasoli F, Price D. Detecting mild COPD is not a waste of resources. Prim Care Respir J. 2011;20(3):238-239. doi:10.4104/pcrj.2011.00075 133  62.  Price D, Freeman D, Cleland J, Kaplan A, Cerasoli F. Earlier diagnosis and earlier treatment of COPD in primary care. Prim Care Respir J. 2011;20(1):15-22. doi:10.4104/pcrj.2010.00060 63.  Qaseem A, Wilt TJ, Weinberger SE, et al. Diagnosis and management of stable chronic obstructive pulmonary disease: a clinical practice guideline update from the American College of Physicians, American College of Chest Physicians, American Thoracic Society, and European Respiratory Society. Ann Intern Med. 2011;155(3):179-191. doi:10.7326/0003-4819-155-3-201108020-00008 64.  Krahn M, Chapman KR. Economic Issues in the Use of Office Spirometry for Lung Health Assessment. Can Respir J. 2003;10(6):320-326. doi:10.1155/2003/158736 65.  U.S. Preventive Services Task Force. Screening for Chronic Obstructive Pulmonary Disease Using Spirometry:  U.S. Preventive Services Task Force Recommendation Statement. Ann Intern Med. 2008;148(7):529-534. 66.  Haroon S, Jordan R, Takwoingi Y, Adab P. Diagnostic accuracy of screening tests for COPD: a systematic review and meta-analysis. BMJ Open. 2015;5(10):e008133. doi:10.1136/bmjopen-2015-008133 67.  Price DB, Tinkelman DG, Halbert RJ, et al. Symptom-based questionnaire for identifying COPD in smokers. Respiration. 2006;73(3):285-295. doi:10.1159/000090142 68.  Price DB, Tinkelman DG, Nordyke RJ, Isonaka S, Halbert RJ. Scoring system and clinical application of COPD diagnostic questionnaires. Chest. 2006;129(6):1531–1539. 69.  Yawn BP, Mapel DW, Mannino DM, et al. Development of the Lung Function Questionnaire (LFQ) to identify airflow obstruction. J Copd. February 2010:1-10. 70.  Stanley AJ, Hasan I, Crockett AJ, van Schayck OCP, Zwar NA. COPD Diagnostic Questionnaire (CDQ) for selecting at-risk patients for spirometry: a cross-sectional study in Australian general practice. NPJ Prim Care Respir Med. 2014;24:14024. doi:10.1038/npjpcrm.2014.24 71.  Hvidsten SC, Storesund L, Wentzel-Larsen T, Gulsvik A, Lehmann S. Prevalence and predictors of undiagnosed chronic obstructive pulmonary disease in a Norwegian adult general population: Undiagnosed COPD in a general population. Clin Respir J. 2010;4(1):13-21. doi:10.1111/j.1752-699X.2009.00137.x 72.  Miravitlles M, Ribera A. Understanding the impact of symptoms on the burden of COPD. Respir Res. 2017;18:67. doi:10.1186/s12931-017-0548-3 73.  Fletcher C, Peto R. The natural history of chronic airflow obstruction. Br Med J. 1977;1(6077):1645-1648. doi:10.1136/bmj.1.6077.1645 134  74.  Lopez-Campos JL, Calero C, Quintana-Gallego E. Symptom variability in COPD: a narrative review. Int J Chron Obstruct Pulmon Dis. 2013;8:231-238. doi:10.2147/COPD.S42866 75.  Mapel DW, Robinson SB, Dastani HB, Shah H, Phillips AL, Lydick E. The Direct Medical Costs of Undiagnosed Chronic Obstructive Pulmonary Disease. Value Health. 2008;11(4):628-636. doi:10.1111/j.1524-4733.2007.00305.x 76.  Jones RCM, Price D, Ryan D, et al. Opportunities to diagnose chronic obstructive pulmonary disease in routine care in the UK: a retrospective study of a clinical cohort. Lancet Respir Med. 2014;2(4):267-276. doi:10.1016/S2213-2600(14)70008-6 77.  Kotz D, Schayck OC van. Interpreting the diagnostic accuracy of tools for early detection of COPD. Prim Care Respir J. 2011;20(2):pcrj201150. doi:10.4104/pcrj.2011.00050 78.  Gewandter JS, McDermott MP, He H, et al. Demonstrating Heterogeneity of Treatment Effects Among Patients: An Overlooked but Important Step Toward Precision Medicine. Clin Pharmacol Ther. 2019;106(1):204-210. doi:10.1002/cpt.1372 79.  Brennan A, Chick SE, Davies R. A taxonomy of model structures for economic evaluation of health technologies. Health Econ. 2006;15(12):1295-1310. doi:10.1002/hec.1148 80.  Sadatsafavi M, Ghanbarian S, Adibi A, et al. Development and Validation of the Evaluation Platform in COPD (EPIC): A Population-Based Outcomes Model of COPD for Canada. Med Decis Making. 2019;39(2):152-167. doi:10.1177/0272989X18824098 81.  Hennessy DA, Flanagan WM, Tanuseputro P, et al. The Population Health Model (POHEM): an overview of rationale, methods and applications. Popul Health Metr. 2015;13(1):24. doi:10.1186/s12963-015-0057-x 82.  Hoogendoorn M, Feenstra TL, Hoogenveen RT, Al M, Mölken MR. Association between lung function and exacerbation frequency in patients with COPD. Int J Chron Obstruct Pulmon Dis. 2010;5:435-444. doi:10.2147/COPD.S13826 83.  Albert RK, Connett J, Bailey WC, et al. Azithromycin for Prevention of Exacerbations of COPD. N Engl J Med. 2011;365(8):689-698. doi:10.1056/NEJMoa1104623 84.  Spencer M, Briggs AH, Grossman RF, Rance L. Development of an economic model to assess the cost effectiveness of treatment interventions for chronic obstructive pulmonary disease. PharmacoEconomics. 2005;23(6):619-637. doi:10.2165/00019053-200523060-00008 85.  Guertin JR, Feeny D, Tarride J-E. Age- and sex-specific Canadian utility norms, based on the 2013–2014 Canadian Community Health Survey. CMAJ. 2018;190(6):E155. doi:10.1503/cmaj.170317 135  86.  Rutten-van Mölken M, Hoogendoorn M, Lamers LM. Holistic preferences for 1-year health profiles describing fluctuations in health: the case of chronic obstructive pulmonary disease. PharmacoEconomics. 2009;27(6):465-477. doi:10.2165/00019053-200927060-00003 87.  Mittmann N, Kuramoto L, Seung SJ, Haddon JM, Bradley-Kennedy C, Fitzgerald JM. The cost of moderate and severe COPD exacerbations to the Canadian healthcare system. Respir Med. 2008;102(3):413-421. doi:10.1016/j.rmed.2007.10.010 88.  Canadian Institute of Health Information (CIHI). Care in Canadian ICUs. November 2018. https://secure.cihi.ca/ free_products/ICU_Report_EN.pdf. 89.  PROSPERO: International prospective register of systematic reviews. 2017. http:// www.crd.york.ac.uk/PROSPERO/. 90.  National Heart, Lung, and Blood Institute. Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. 2014. https://www.nhlbi.nih.gov/health-pro/guidelines/in-develop/cardiovascular-risk-reduction/tools/cohort. Accessed January 8, 2017. 91.  Guido Schwarzer. meta: an R package for meta-analysis. R News. 2007;7(3):40-45. 92.  R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2019. http://www.R-project.org/. 93.  Cochran, WG. The combination of estimates from different experiments. Biometrics. 1954;10:101-129. 94.  Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-560. 95.  Güder G, Brenner S, Angermann CE, et al. GOLD or lower limit of normal definition? a comparison with expert-based diagnosis of chronic obstructive pulmonary disease in a prospective cohort-study. Respir Res. 2012;13:13. doi:10.1186/1465-9921-13-13 96.  Moreira GL, Manzano BM, Gazzotti MR, et al. PLATINO, a nine-year follow-up study of COPD in the city of Sao Paulo, Brazil: the problem of underdiagnosis. J Bras Pneumol. 2014;40(1):30-37. doi:10.1590/S1806-37132014000100005 97.  Nascimento OA, Camelier A, Rosa FW, Menezes AMB, Perez-Padilla R, Jardim JR. Chronic obstructive pulmonary disease is underdiagnosed and undertreated in Sao Paulo (Brazil): results of the PLATINO study. Braz J Med Biol Res. 2007;40(7):887-895. 98.  Talamo C, de Oca MM, Halbert R, et al. Diagnostic labeling of COPD in five Latin American cities. Chest. 2007;131(1):60-67. doi:10.1378/chest.06-1149 136  99.  Artyukhov IP, Arshukova IL, Dobretsova EA, Dugina TA, Shulmin AV, Demko IV. Epidemiology of chronic obstructive pulmonary disease: a population-based study in Krasnoyarsk region, Russia. J COPD. 2015;10:1781-1786. 100.  De Godoy I, Tanni S.E, Coelho L.S, et al. Smoking cessation program as a tool for the early diagnosis of chronic obstructive pulmonary disease. J Bras Pneumol. 2007;33(3):282-286. doi:10.1590/S1806-37132007000300009 101.  Ancochea J, Miravitlles M, Garcia-Rio F, et al. Underdiagnosis of chronic obstructive pulmonary disease in women: quantification of the problem, determinants and proposed actions. Arch Bronconeumol. 2013;49(6):223-229. doi:10.1016/j.arbres.2012.11.010 102.  Miravitlles M, Soriano JB, Garcia-Rio F, et al. Prevalence of COPD in Spain: impact of undiagnosed COPD on quality of life and daily life activities. Thorax. 2009;64(10):863-868. doi:10.1136/thx.2009.115725 103.  Schirnhofer L, Lamprecht B, Firlei N, et al. Using targeted spirometry to reduce non-diagnosed chronic obstructive pulmonary disease. Respiration. 2011;81(6):476-482. doi:10.1159/000320251 104.  Herrera A.C, De Oca M.M, Varela M.V.L, Aguirre C, Schiavi E, Jardim J.R. COPD underdiagnosis and misdiagnosis in a high-risk primary care population in four Latin American countries. A key to enhance disease diagnosis: The PUMA Study. PLoS ONE. 2016;11(4):DOI: 10.1371/journal.pone.0152266. doi:10.1371/journal.pone.0152266 105.  Hill K, Goldstein RS, Guyatt GH, et al. Prevalence and underdiagnosis of chronic obstructive pulmonary disease among patients at risk in primary care. CMAJ. 2010;182(7):673-678. doi:10.1503/cmaj.091784 106.  Llordes M, Jaen A, Almagro P, et al. Prevalence, Risk Factors and Diagnostic Accuracy of COPD Among Smokers in Primary Care. J COPD. 2015;12(4):404-412. doi:10.3109/15412555.2014.974736 107.  Mahishale V, Mahishale A, Angadi N, et al. Screening for chronic obstructive pulmonary disease in elderly subjects with dyspnoea and/or reduced exercise tolerance - A hospital based cross sectional study. J Chest Dis Tuberc. 2015;64:567-571. doi:10.1016/j.ejcdt.2015.03.010 108.  Queiroz MC de CAM de, Moreira MAC, Rabahi MF. Underdiagnosis of COPD at primary health care clinics in the city of Aparecida de Goiania, Brazil. J Bras Pneumol. 2012;38(6):692-699. 109.  Zhang J, Zhou J, Lin X, Wang Q, Bai C, Hong Q. Prevalence of undiagnosed and undertreated chronic obstructive pulmonary disease in lung cancer population. Respirology. 2013;18(2):297-302. doi:10.1111/j.1440-1843.2012.02282.x 137  110.  Roche N, Small M, Broomfield S, Higgins V, Pollard R. Real world COPD: association of morning symptoms with clinical and patient reported outcomes. COPD. 2013;10(6):679-686. doi:10.3109/15412555.2013.844784 111.  Bridevaux P-O, Gerbase MW, Probst-Hensch NM, Schindler C, Gaspoz J-M, Rochat T. Long-term decline in lung function, utilisation of care and quality of life in modified GOLD stage 1 COPD. Thorax. 2008;63(9):768-774. doi:10.1136/thx.2007.093724 112.  Rennard SI, Drummond MB. Early chronic obstructive pulmonary disease: definition, assessment, and prevention. Lancet. 2015;385(9979):1778-1788. doi:10.1016/S0140-6736(15)60647-X 113.  Miravitlles M, Worth H, Soler Cataluña JJ, et al. Observational study to characterise 24-hour COPD symptoms and their relationship with patient-reported outcomes: results from the ASSESS study. Respir Res. 2014;15:122. doi:10.1186/s12931-014-0122-1 114.  Ioanna T, Kocks J, Tzanakis N, Siafakas N, van der Molen T. Factors that influence disease-specific quality of life or health status in patients with COPD: a systematic review and meta-analysis of Pearson correlations. Prim Care Respir J. 2011;20(3):257-268. doi:10.4104/pcrj.2011.00029 115.  Monteagudo M, Rodríguez-Blanco T, Llagostera M, et al. Factors associated with changes in quality of life of COPD patients: a prospective study in primary care. Respir Med. 2013;107(10):1589-1597. doi:10.1016/j.rmed.2013.05.009 116.  Price D, Small M, Milligan G, Higgins V, Gil EG, Estruch J. Impact of night-time symptoms in COPD: a real-world study in five European countries. Int J Chron Obstruct Pulmon Dis. 2013;8:595-603. doi:10.2147/COPD.S48570 117.  Hurst JR, Vestbo J, Anzueto A, et al. Susceptibility to Exacerbation in Chronic Obstructive Pulmonary Disease. N Engl J Med. 2010;363(12):1128-1138. doi:10.1056/NEJMoa0909883 118.  Sadatsafavi M, Sin DD, Zafari Z, et al. The Association Between Rate and Severity of Exacerbations in Chronic Obstructive Pulmonary Disease: An Application of a Joint Frailty-Logistic Model. Am J Epidemiol. 2016;184(9):681-689. doi:10.1093/aje/kww085 119.  Miravitlles M, Menezes A, López Varela MV, et al. Prevalence and impact of respiratory symptoms in a population of patients with COPD in Latin America: The LASSYC observational study. Respir Med. 2018;134:62-69. doi:10.1016/j.rmed.2017.11.018 120.  Partridge MR, Karlsson N, Small IR. Patient insight into the impact of chronic obstructive pulmonary disease in the morning: an internet survey. Curr Med Res Opin. 2009;25(8):2043-2048. doi:10.1185/03007990903103006 138  121.  Espinosa de los Monteros MJ, Peña C, Soto Hurtado EJ, Jareño J, Miravitlles M. Variability of respiratory symptoms in severe COPD. Arch Bronconeumol. 2012;48(1):3-7. doi:10.1016/j.arbres.2011.07.004 122.  Donaldson GC, Goldring JJ, Wedzicha JA. Influence of season on exacerbation characteristics in patients with COPD. Chest. 2012;141(1):94-100. doi:10.1378/chest.11-0281 123.  Andrade SE, Kahler KH, Frech F, Chan KA. Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf. 2006;15(8):565-574. doi:10.1002/pds.1230 124.  Ten Have TR, Localio AR. Empirical Bayes estimation of random effects parameters in mixed effects logistic regression models. Biometrics. 1999;55(4):1022-1029. 125.  Kessler R, Partridge MR, Miravitlles M, et al. Symptom variability in patients with severe COPD: a pan-European cross-sectional study. Eur Respir J. 2011;37(2):264-272. doi:10.1183/09031936.00051110 126.  Sumner H, Woodcock A, Kolsum U, et al. Predictors of objective cough frequency in chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2013;187(9):943-949. doi:10.1164/rccm.201211-2000OC 127.  Agusti A, Calverley PMA, Celli B, et al. Characterisation of COPD heterogeneity in the ECLIPSE cohort. Respir Res. 2010;11:122. doi:10.1186/1465-9921-11-122 128.  Barsky AJ, Peekna HM, Borus JF. Somatic Symptom Reporting in Women and Men. J Gen Intern Med. 2001;16(4):266-275. doi:10.1046/j.1525-1497.2001.00229.x 129.  Johnson KM, Bryan S, Ghanbarian S, Sin DD, Sadatsafavi M. Characterizing undiagnosed chronic obstructive pulmonary disease: a systematic review and meta-analysis. Respir Res. 2018;19:26. doi:10.1186/s12931-018-0731-1 130.  GBD 2015 Chronic Respiratory Disease Collaborators. Global, regional, and national deaths, prevalence, disability-adjusted life years, and years lived with disability for chronic obstructive pulmonary disease and asthma, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet Respir Med. 2017;5(9):691-706. doi:10.1016/S2213-2600(17)30293-X 131.  Gershon AS, Thiruchelvam D, Chapman KR, et al. Health Services Burden of Undiagnosed and Overdiagnosed COPD. Chest. 2018;153(6):1336-1346. doi:10.1016/j.chest.2018.01.038 132.  BC Statistics. Population Estimates - Province of British Columbia. https://www2.gov.bc.ca/gov/content/data/statistics/people-population-community/population/population-estimates. Accessed February 15, 2019. 139  133.  Canadian Institute for Health Information. Discharge Abstract Database (Hospital Separations). V2. Data Extract. 2017. http://www.popdata.bc.ca/data. 134.  British Columbia Ministry of Health. Medical Services Plan (MSP) Payment Information File. V2. Data Extract. 2017. http://www.popdata.bc.ca/data. 135.  British Columbia Ministry of Health. PharmaNet. V2. Data Extract. Data Stewardship Committee. 2017. http://www.popdata.bc.ca/data. 136.  British Columbia Vital Statistics Agency. Vital Statistics Births. V2. Data Extract. (2017). 2017. http://www.popdata.bc.ca/data. 137.  British Columbia Vital Statistics Agency. Vital Statistics Deaths. V2. Data Extract. (2017). 2017. http://www.popdata.bc.ca/data. 138.  Canadian Institute for Health Information. Data Quality Documentation, Discharge Abstract Database—Multi-Year Information. 2012. https://www.cihi.ca/en/dad_multi-year_en.pdf. 139.  Williams JI, Young W. Inventory of studies on the accuracy of Canadian health administrative databases. Technical report. 1996. 140.  Gershon AS, Wang C, Guan J, Vasilevska-Ristovska J, Cicutto L, To T. Identifying individuals with physcian diagnosed COPD in health administrative databases. COPD. 2009;6(5):388-394. 141.  Richardson DB. An incidence density sampling program for nested case-control analyses. Occup Environ Med. 2004;61(12):e59. doi:10.1136/oem.2004.014472 142.  British Columbia Ministry of Health. Limited Coverage Drugs - Tiotropium. PharmaCare Prescriber Resources. https://www2.gov.bc.ca/gov/content/health/practitioner-professional-resources/pharmacare/prescribers/limited-coverage-drug-program/limited-coverage-drugs-tiotropium. 143.  Wouters EFM, Groenewegen KH, Dentener MA, Vernooy JHJ. Systemic Inflammation in Chronic Obstructive Pulmonary Disease. Proc Am Thorac Soc. 2007;4(8):626-634. doi:10.1513/pats.200706-071TH 144.  Cavaillès A, Brinchault-Rabin G, Dixmier A, et al. Comorbidities of COPD. Eur Respir Rev. 2013;22(130):454-475. doi:10.1183/09059180.00008612 145.  Roversi S, Fabbri LM, Sin DD, Hawkins NM, Agustí A. Chronic Obstructive Pulmonary Disease and Cardiac Diseases. An Urgent Need for Integrated Care. Am J Respir Crit Care Med. 2016;194(11):1319-1336. doi:10.1164/rccm.201604-0690SO 146.  Olson DE, Zhu M, Long Q, et al. Increased Cardiovascular Disease, Resource Use, and Costs Before the Clinical Diagnosis of Diabetes in Veterans in the Southeastern U.S. J Gen Intern Med. 2015;30(6):749-757. doi:10.1007/s11606-014-3075-7 140  147.  Desai U, Kirson NY, Ye W, Mehta NR, Wen J, Andrews JS. Trends in health service use and potentially avoidable hospitalizations before Alzheimer’s disease diagnosis: A matched, retrospective study of US Medicare beneficiaries. Alzheimers Dement. 2019;11:125-135. doi:10.1016/j.dadm.2018.12.005 148.  Yawn BP, Mapel DW, Mannino DM, et al. Development of the Lung Function Questionnaire (LFQ) to identify airflow obstruction. Int J Chron Obstruct Pulmon Dis. 2010;5:1-10. 149.  Postma DS, Bush A, van den Berge M. Risk factors and early origins of chronic obstructive pulmonary disease. Lancet. 2015;385(9971):899-909. doi:10.1016/S0140-6736(14)60446-3 150.  Johnson KM, Safari A, Tan WC, Bourbeau J, FitzGerald JM, Sadatsafavi M. Heterogeneity in the respiratory symptoms of patients with mild-to-moderate COPD. International Journal of Chronic Obstructive Pulmonary Disease. doi:10.2147/COPD.S184424 151.  GBD 2017 Mortality Collaborators. Global, regional, and national age-sex-specific mortality and life expectancy, 1950-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1684-1735. doi:10.1016/S0140-6736(18)31891-9 152.  Canadian Institute for Health Information. Hospital stays in Canada. April 2019. https://www.cihi.ca/en/hospital-stays-in-canada. Accessed October 8, 2019. 153.  Johnson KM, Khakban A, Bryan S, Sin DD, Sadatsafavi M. Healthcare system encounters before COPD diagnosis: a registry-based longitudinal cohort study. Thorax. November 2019. doi:10.1136/thoraxjnl-2019-213554 154.  Yawn BP, Duvall K, Peabody J, et al. The impact of screening tools on diagnosis of chronic obstructive pulmonary disease in primary care. Am J Prev Med. 2014;47(5):563-575. doi:10.1016/j.amepre.2014.07.030 155.  Jordan RE, Adab P, Sitch A, et al. Targeted case finding for chronic obstructive pulmonary disease versus routine practice in primary care (TargetCOPD): a cluster-randomised controlled trial. Lancet Respir Med. 2016;4(9):720-730. doi:10.1016/S2213-2600(16)30149-7 156.  Lambe T, Adab P, Jordan RE, et al. Model-based evaluation of the long-term cost-effectiveness of systematic case-finding for COPD in primary care. Thorax. 2019;74(8):730-739. doi:10.1136/thoraxjnl-2018-212148 157.  Husereau D, Drummond M, Petrou S, et al. Consolidated Health Economic Evaluation Reporting Standards (CHEERS)--explanation and elaboration: a report of the ISPOR Health Economic Evaluation Publication Guidelines Good Reporting Practices Task Force. Value Health. 2013;16(2):231-250. doi:10.1016/j.jval.2013.02.002 141  158.  Price DB, Tinkelman DG, Halbert RJ, et al. Symptom-based questionnaire for identifying COPD in smokers. Respiration. 2006;73(3):285-295. doi:10.1159/000090142 159.  Sichletidis L, Spyratos D, Papaioannou M, et al. A combination of the IPAG questionnaire and PiKo-6® flow meter is a valuable screening tool for COPD in the primary care setting. Prim Care Respir J. 2011;20(2):184-189. doi:10.4104/pcrj.2011.00038 160.  Zhou Y, Zhong N, Li X, et al. Tiotropium in Early-Stage Chronic Obstructive Pulmonary Disease. N Engl J Med. 2017;377(10):923-935. doi:10.1056/NEJMoa1700228 161.  Calverley PMA, Anderson JA, Celli B, et al. Salmeterol and Fluticasone Propionate and Survival in Chronic Obstructive Pulmonary Disease. N Engl J Med. 2007;356(8):775-789. doi:10.1056/NEJMoa063070 162.  Ferguson GT, Rabe KF, Martinez FJ, et al. Triple therapy with budesonide/glycopyrrolate/formoterol fumarate with co-suspension delivery technology versus dual therapies in chronic obstructive pulmonary disease (KRONOS): a double-blind, parallel-group, multicentre, phase 3 randomised controlled trial. Lancet Respir Med. 2018;6(10):747-758. doi:10.1016/S2213-2600(18)30327-8 163.  Tai-Seale M, McGuire TG, Zhang W. Time Allocation in Primary Care Office Visits. Health Serv Res. 2007;42(5):1871-1894. doi:10.1111/j.1475-6773.2006.00689.x 164.  Irving G, Neves AL, Dambha-Miller H, et al. International variations in primary care physician consultation time: a systematic review of 67 countries. BMJ Open. 2017;7(10). doi:10.1136/bmjopen-2017-017902 165.  British Columbia Ministry of Health. Medical Services Commission Payment Schedule. March 2019. https://www2.gov.bc.ca/gov/content/health/practitioner-professional-resources/msp/physicians/payment-schedules/msc-payment-schedule. 166.  Tavakoli H, Johnson KM, FitzGerald JM, et al. Trends in prescriptions and costs of inhaled medications in chronic obstructive pulmonary disease: a 19-year population-based study from Canada. Int J Chron Obstruct Pulmon Dis. 2019;2019(14):2003—2013. doi:https://doi.org/10.2147/COPD.S210897 167.  Mullen K-A, Coyle D, Manuel D, et al. Economic evaluation of a hospital-initiated intervention for smokers with chronic disease, in Ontario, Canada. Tob Control. 2015;24(5):489-496. doi:10.1136/tobaccocontrol-2013-051483 168.  Chapman KR, Bourbeau J, Rance L. The burden of COPD in Canada: results from the Confronting COPD survey. Respir Med. 2003;97:23-31. doi:10.1016/s0954-6111(03)80022-7 169.  Stafyla E, Geitona M, Kerenidi T, Economou A, Daniil Z, Gourgoulianis KI. The annual direct costs of stable COPD in Greece. Int J Chron Obstruct Pulmon Dis. 2018;13:309-315. doi:10.2147/COPD.S148051 142  170.  Tosteson ANA, Fryback DG, Hammond CS, et al. Consequences of false-positive screening mammograms. JAMA Intern Med. 2014;174(6):954-961. doi:10.1001/jamainternmed.2014.981 171.  Wu P, Wilson K, Dimoulas P, Mills EJ. Effectiveness of smoking cessation therapies: a systematic review and meta-analysis. BMC Public Health. 2006;6:300. doi:10.1186/1471-2458-6-300 172.  Murray RP, Connett JE, Rand CS, Pan W, Anthonisen NR. Persistence of the effect of the Lung Health Study (LHS) smoking intervention over eleven years. Prev Med. 2002;35(4):314-319. doi:10.1006/pmed.2002.1087 173.  CADTH. Guidelines for the economic evaluation of health technologies: Canada. 4th ed. Ottawa. March 2017. 174.  Oca MM de, Menezes A, Wehrmeister FC, et al. Adherence to inhaled therapies of COPD patients from seven Latin American countries: The LASSYC study. PLoS ONE. 2017;12(11):e0186777. doi:10.1371/journal.pone.0186777 175.  van Schayck OCP, Pinnock H, Ostrem A, et al. IPCRG Consensus statement: tackling the smoking epidemic - practical guidance for primary care. Prim Care Respir J. 2008;17(3):185-193. doi:10.3132/pcrj.2008.00060 176.  Johnson KM, Tan WC, Bourbeau J, et al. The diagnostic performance of patient symptoms in screening for COPD. Respir Res. 2018;19(1):147. doi:10.1186/s12931-018-0853-5 177.  Tashkin DP, Celli B, Senn S, et al. A 4-Year Trial of Tiotropium in Chronic Obstructive Pulmonary Disease. N Engl J Med. 2008;359(15):1543-1554. doi:10.1056/NEJMoa0805800 178.  Barrecheguren M, González C, Miravitlles M. What have we learned from observational studies and clinical trials of mild to moderate COPD? Respir Res. 2018;19(1):177. doi:10.1186/s12931-018-0882-0 179.  Government of Canada Statistics Canada. Estimating the prevalence of COPD in Canada: Reported diagnosis versus measured airflow obstruction. https://www150.statcan.gc.ca/n1/pub/82-003-x/2014003/article/11908-eng.pdf. Published March 1, 2014. Accessed October 25, 2019. 180.  Government of Canada Statistics Canada. Population estimates on July 1st, by age and sex. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1710000501. Published July 1, 2019. Accessed October 25, 2019. 181.  Martinez FJ, Mannino D, Leidy NK, et al. A New Approach for Identifying Patients with Undiagnosed Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med. 2016;195(6):748-756. doi:10.1164/rccm.201603-0622OC 143  182.  Zafari Z, Bryan S, Sin DD, Conte T, Khakban R, Sadatsafavi M. A systematic review of health economics simulation models of chronic obstructive pulmonary disease. Value Health. October 2016. doi:10.1016/j.jval.2016.08.003 183.  Camp PG, Hernandez P, Bourbeau J, et al. Pulmonary rehabilitation in Canada: A report from the Canadian Thoracic Society COPD Clinical Assembly. Can Respir J. 2015;22(3):147-152. 184.  Kruis AL, Ställberg B, Jones RCM, et al. Primary Care COPD Patients Compared with Large Pharmaceutically-Sponsored COPD Studies: An UNLOCK Validation Study. PLoS ONE. 2014;9(3):e90145. doi:10.1371/journal.pone.0090145 185.  Calverley PMA, Anderson JA, Celli B, et al. Salmeterol and Fluticasone Propionate and Survival in Chronic Obstructive Pulmonary Disease. N Engl J Med. 2007;356(8):775-789. doi:10.1056/NEJMoa063070 186.  Lambert AA, Bhatt SP. Respiratory symptoms in smokers with normal spirometry: clinical significance and management considerations. Curr Opin Pulm Med. 2019;25(2):138-143. doi:10.1097/MCP.0000000000000550 187.  Chen W, FitzGerald JM, Sin DD, Sadatsafavi M, Canadian Respiratory Research Network. Excess economic burden of comorbidities in COPD: a 15-year population-based study. Eur Respir J. 2017;50(1). doi:10.1183/13993003.00393-2017 188.  Silva GE, Sherrill DL, Guerra S, Barbee RA. Asthma as a risk factor for COPD in a longitudinal study. Chest. 2004;126(1):59-65. doi:10.1378/chest.126.1.59 189.  Sin DD, Anthonisen NR, Soriano JB, Agusti AG. Mortality in COPD: Role of comorbidities. Eur Respir J. 2006;28(6):1245-1257. doi:10.1183/09031936.00133805 190.  Cazzola M, Calzetta L, Rinaldi B, et al. Management of Chronic Obstructive Pulmonary Disease in Patients with Cardiovascular Diseases. Drugs. 2017;77(7):721-732. doi:10.1007/s40265-017-0731-3 191.  Celli BR, Wedzicha JA. Update on Clinical Aspects of Chronic Obstructive Pulmonary Disease. N Engl J Med. 2019;381(13):1257-1266. doi:10.1056/NEJMra1900500 192.  Labaki WW, Han MK. Improving Detection of Early Chronic Obstructive Pulmonary Disease. Ann Am Thorac Soc. 2018;15(Supplement 4):S243-S248. doi:10.1513/AnnalsATS.201808-529MG 193.  Sin DD, Leipsic J, Man SFP. CT in COPD: just a pretty picture or really worth a thousand words (or dollars)? Thorax. 2011;66(9):741-742. doi:10.1136/thx.2011.161430 194.  Mapel DW, Frost FJ, Hurley JS, et al. An algorithm for the identification of undiagnosed COPD cases using administrative claims data. J Manag Care Pharm. 2006;12(6):457-465. 144  195.  van Voorn GAK, Vemer P, Hamerlijnck D, et al. The Missing Stakeholder Group: Why Patients Should be Involved in Health Economic Modelling. Appl Health Econ Health Policy. 2016;14(2):129-133. doi:10.1007/s40258-015-0200-7 196.  Marra CA, Cibere J, Tsuyuki RT, et al. Improving osteoarthritis detection in the community: Pharmacist identification of new, diagnostically confirmed osteoarthritis. Arthritis Care Res. 2007;57(7):1238-1244. doi:10.1002/art.23019 197.  Tannenbaum C, Tsuyuki RT. The expanding scope of pharmacists’ practice: implications for physicians. CMAJ. 2013;185(14):1228-1232. doi:10.1503/cmaj.121990 145  Appendices  Appendix A    A.1 Search strategy MEDLINE (OVID) March 22, 2017  Database: Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations and Ovid MEDLINE(R) <1946 to Present> Search Strategy: -------------------------------------------------------------------------------- 1     *pulmonary disease, chronic obstructive/ or *bronchitis, chronic/ or *pulmonary emphysema/ (38797) 2     *airway obstruction/ (11509) 3     *bronchitis/ or *bronchiolitis/ or *bronchiolitis obliterans/ or *cryptogenic organizing pneumonia/ (18102) 4     *emphysema/ or *mediastinal emphysema/ or *subcutaneous emphysema/ or *alpha 1-antitrypsin deficiency/ (11041) 5     *Lung Diseases, Obstructive/ (13636) 6     limit 5 to yr="1980 -2001" (11016) 7     or/1-4,6 (84739) 8     di.fs. [Diagnosis] (2370855) 9     ep.fs. [Epidemiology] (1471514) 10     8 or 9 (3570350) 11     7 and 10 (23089) 12     Diagnostic Errors/ (35551) 13     Delayed Diagnosis/ (4539) 14     Early Diagnosis/ (22251) 15     Airway Obstruction/di [Diagnosis] (2772) 16     underdiagnos$.mp. (7477) 17     under diagnos$.mp. (2788) 18     undiagnos$.mp. (16208) 19     "Diagnostic Techniques and Procedures"/ (2914) 20     Diagnosis, Differential/ (432511) 21     "not diagnos$".mp. (5636) 22     misdiagnos$.mp. (26521) 23     or/12-22 (533328) 24     11 and 23 (4893) 25     "Risk Factors"/ (717586) 26     logistic models/ (119410) 27     risk assessment/ (224203) 146  28     risk factors/ (717586) 29     risk/ (115068) 30     protective factors/ (1844) 31     probability/ (54508) 32     odds ratio/ (79810) 33     risk factor$.mp. (954945) 34     risk assessment$.mp. (251708) 35     (characteri#tic? or characteri#e? or characteri#ation).mp. (2499735) 36     or/25-35 (3674000) 37     24 and 36 (992) 38     limit 37 to yr=1980 -current (981)   EMBASE (OVID) April 11, 2017  Database: Embase <1974 to 2017 August 14> Search Strategy: -------------------------------------------------------------------------------- 1     *chronic obstructive lung disease/ or *chronic bronchitis/ or *lung emphysema/ (68276) 2     *airway obstruction/ or *airflow limitation/ (10588) 3     *bronchitis/ or *chronic bronchitis/ or *bronchiolitis/ or *bronchiolitis obliterans/ or *bronchiolitis obliterans organizing pneumonia/ (24818) 4     *emphysema/ or *subcutaneous emphysema/ or *cigarette smoke-induced emphysema/ or *elastase-induced emphysema/ or *experimental emphysema/ or *alpha 1 antitrypsin deficiency/ (9667) 5     or/1-4 (103405) 6     di.fs. [Diagnosis] (2921008) 7     ep.fs. [Epidemiology] (976523) 8     6 or 7 (3661177) 9     5 and 8 (21021) 10     Diagnostic Error/ (50975) 11     Early Diagnosis/ (89985) 12     *Airway Obstruction/di [Diagnosis] (1452) 13     Diagnostic Procedure/ (80939) 14     Differential diagnosis/ (334345) 15     or/10-14 (535842) 16     9 and 15 (3600) 17     Risk Factor/ (828169) 18     regression analysis/ (116928) 19     multivariate analysis/ (141771) 20     risk assessment/ (424757) 21     risk factor/ (828169) 22     risk/ (496728) 23     probability/ (75672) 147  24     odds ratio/ (12341) 25     risk factor$.mp. (1080390) 26     risk assessment$.mp. (444895) 27     (characteri#tic? or characteri#e? or characteri#ation).mp. (3007687) 28     or/17-27 (4767936) 29     16 and 28 (717) 30     underdiagnos$.mp. (10792) 31     under diagnos$.mp. (5242) 32     undiagnos$.mp. (23402) 33     "not diagnos$".mp. (8488) 34     misdiagnos$.mp. (37341) 35     unrecogni$.mp. (33790) 36     Delayed Diagnosis/ (9134) 37     or/30-36 (123603) 38     9 and 37 (663) 39     29 or 38 (1299) 40     limit 39 to yr="1980 -current" (1296) 41     limit 40 to "english language" (1049) 42     40 not 41 (247)148  A.2 Unadjusted analysis in the general population  Associations between diagnosed (v. ‘undiagnosed’) COPD and sex, the presence of any respiratory symptoms, smoking status, smoking history, and COPD severity based on the contingency tables of studies using random sampling of the general population (‘unadjusted analysis’). Persistent airflow limitation was defined as post-bronchodilator FEV1/FVC<0.7. Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).  149  A.3 Unadjusted analysis using the lower limit of normal criterion  Associations between diagnosed (v. ‘undiagnosed’) COPD and sex, the presence of cough, wheeze, phlegm, dyspnea, and COPD severity based on contingency tables (‘unadjusted analysis’). Persistent airflow limitation was defined as post-bronchodilator FEV1/FVC< lower limit of normal (LLN). Squares represent individual study estimates with the size of the square corresponding to their weight in the pooled estimate (represented with diamonds).150  Appendix B   B.1 Symptom assessment questionnaire Cough Since your last visit, do you usually cough when you don’t have a cold? (Y/N) Phlegm Since your last visit, do you usually bring up phlegm from your chest, or do you usually have phlegm in your chest that is difficult to bring up when you don’t have a cold? (Y/N) Wheezing/Whistling Have you had wheezing or whistling in your chest at any time since your last visit? (Y/N) Breathlessness a) Are you unable to walk due to a condition other than shortness of breath? (Y/N) b) Dyspnea (MRC Dyspnea Scale) Level 1: Not troubled by breathlessness except with strenuous exercise. Level 2: Troubled by shortness of breath when hurrying on level ground or walking up a slight hill. Level 3: Walks slower than people of the same age on level ground because of breathlessness, or has to stop for breath when walking at own pace on level ground. Level 4: Stops for breath after walking about 100 yards (90 m) or after a few minutes on level ground. Level 5: Too breathless to leave the house or breathless when dressing or undressing.  Questionnaire used to measure symptoms at each study visit in CanCOLD. 151  B.2 Characteristics of study participants in the subset of the data used to assess dyspnea and any symptoms  Visit 1 (n=347) Visit 2 (n=241) Visit 3 (n=268) Age 65.0 (10.2) 66.7 (9.8) 67.9 (9.4) Male (vs. female) 54.5% 54.4% 54.1% BMI 27.0 (4.9) 27.1 (4.9) 27.0 (4.9) Caucasian (vs. non-Caucasian) 97.1% 97.1% 97.8% Comorbidities*       0 comorbidities 62.8% 56.4% 57.1%    1 comorbidity 30.5% 33.2% 24.3%    ≥2 comorbidities 6.6% 10.4% 18.7% Anxiety/Depression (vs. no) 18.2% 19.5% 19.4% Smoking between visits (vs. no) 74.1% 25.7% 24.6% Lifetime pack-years smoked 26.2 (25.8) 24.7 (25.4) 26.3 (25.6) Any exacerbations‡ (vs. no) 7.5% 10.0% 14.6% Medication Possession Ratio§ 76.9% (107.6) 75.7% (111.9) 76.5% (107.7) FEV1  (L) 2.2 (0.8) 2.1 (0.8) 2.1 (0.7) % Predicted FEV1 75.0 (17.8) 76.1 (19.0) 77.1 (18.5) Diagnosed COPD (vs. undiagnosed) 23.3% 27.0% 32.5% Asthma (vs. no) 21.0% 26.6% 29.9% Symptoms (present vs. absent)       Dyspnea 53.3% 46.5% 44.8%    Any symptoms 83.9% 77.2% 74.6%  Characteristics of study participants in the subset of data that was used to assess dyspnea and any symptoms because 34 participants were unable to walk due to a condition other than shortness of breath. Means (and standard deviations) are reported unless otherwise indicated * Participants reported whether they had ever been diagnosed with coronary artery disease, hypertension, diabetes, lung cancer, stroke, and tuberculosis at each study visit ‡ COPD exacerbations of any severity (mild, moderate, severe) over the past 12 months  § 12-month medication possession ratio for all respiratory-related medications152  Appendix C   C.1 List of medications classified as other respiratory medications DIN Category DIN Category 14923 Bronchodilators-Theophylline 1926608 Bronchodilators-Theophylline 156701 Bronchodilators-Theophylline 1926616 Bronchodilators-Theophylline 178497 Bronchodilators-Theophylline 1926640 Bronchodilators-Theophylline 261203 Bronchodilators-Theophylline 1966219 Bronchodilators-Theophylline 346071 Bronchodilators-Theophylline 1966227 Bronchodilators-Theophylline 405310 Bronchodilators-Theophylline 1966235 Bronchodilators-Theophylline 441724 Bronchodilators-Theophylline 1966243 Bronchodilators-Theophylline 441732 Bronchodilators-Theophylline 1966251 Bronchodilators-Theophylline 451282 Bronchodilators-Theophylline 1966278 Bronchodilators-Theophylline 458708 Bronchodilators-Theophylline 1966286 Bronchodilators-Theophylline 458716 Bronchodilators-Theophylline 2014165 Bronchodilators-Theophylline 460982 Bronchodilators-Theophylline 2014181 Bronchodilators-Theophylline 460990 Bronchodilators-Theophylline 2014270 Bronchodilators-Theophylline 461008 Bronchodilators-Theophylline 2014289 Bronchodilators-Theophylline 466409 Bronchodilators-Theophylline 2230085 Bronchodilators-Theophylline 476366 Bronchodilators-Theophylline 2230086 Bronchodilators-Theophylline 476390 Bronchodilators-Theophylline 2230087 Bronchodilators-Theophylline 476412 Bronchodilators-Theophylline 261238 Cromolyn Sodium 488070 Bronchodilators-Theophylline 534609 Cromolyn Sodium 497193 Bronchodilators-Theophylline 555649 Cromolyn Sodium 497207 Bronchodilators-Theophylline 638641 Cromolyn Sodium 503436 Bronchodilators-Theophylline 2046113 Cromolyn Sodium 511692 Bronchodilators-Theophylline 2049082 Cromolyn Sodium 532223 Bronchodilators-Theophylline 2219468 Cromolyn Sodium 536709 Bronchodilators-Theophylline 2231431 Cromolyn Sodium 556742 Bronchodilators-Theophylline 2231671 Cromolyn Sodium 565377 Bronchodilators-Theophylline 605255 Cromolyn Sodium Nasal Solution 575151 Bronchodilators-Theophylline 1950541 Cromolyn Sodium Nasal Solution 582654 Bronchodilators-Theophylline 2231326 Cromolyn Sodium Nasal Solution 582662 Bronchodilators-Theophylline 2231390 Cromolyn Sodium Nasal Solution 153  DIN Category DIN Category 589942 Bronchodilators-Theophylline 38121 Ephedrine 589950 Bronchodilators-Theophylline 438847 Ephedrine 599905 Bronchodilators-Theophylline 876534 Ephedrine 627410 Bronchodilators-Theophylline 893323 Ephedrine 631698 Bronchodilators-Theophylline 893331 Ephedrine 631701 Bronchodilators-Theophylline 2012111 Ephedrine 692689 Bronchodilators-Theophylline 2100231 Ephedrine 692697 Bronchodilators-Theophylline 2100258 Ephedrine 692700 Bronchodilators-Theophylline 2126400 Ephedrine 722065 Bronchodilators-Theophylline 2126419 Ephedrine 792934 Bronchodilators-Theophylline 2219743 Ephedrine 868450 Bronchodilators-Theophylline 2229678 Ephedrine 1926586 Bronchodilators-Theophylline 2229698 Ephedrine 1926594 Bronchodilators-Theophylline 2229711 Ephedrine 2236722 Ephedrine 30910 Oral Glucocorticoid Steroid 2237085 Ephedrine 30929 Oral Glucocorticoid Steroid 2242639 Ephedrine 30988 Oral Glucocorticoid Steroid 2242961 Ephedrine 36129 Oral Glucocorticoid Steroid 2243148 Ephedrine 36366 Oral Glucocorticoid Steroid 466417 Epinephrine 156876 Oral Glucocorticoid Steroid 525103 Epinephrine 176834 Oral Glucocorticoid Steroid 1927582 Epinephrine 210188 Oral Glucocorticoid Steroid 2017555 Epinephrine 232378 Oral Glucocorticoid Steroid 307548 Guaifenesin/ Theophylline 252417 Oral Glucocorticoid Steroid 317225 Guaifenesin/ Theophylline 271373 Oral Glucocorticoid Steroid 334510 Guaifenesin/ Theophylline 271381 Oral Glucocorticoid Steroid 356123 Guaifenesin/ Theophylline 280437 Oral Glucocorticoid Steroid 476374 Guaifenesin/ Theophylline 285471 Oral Glucocorticoid Steroid 545090 Guaifenesin/ Theophylline 295094 Oral Glucocorticoid Steroid 640093 Guaifenesin/ Theophylline 312770 Oral Glucocorticoid Steroid 721301 Guaifenesin/ Theophylline 354309 Oral Glucocorticoid Steroid 792942 Guaifenesin/ Theophylline 489158 Oral Glucocorticoid Steroid 828718 Guaifenesin/ Theophylline 508586 Oral Glucocorticoid Steroid 828726 Guaifenesin/ Theophylline 550957 Oral Glucocorticoid Steroid 828742 Guaifenesin/ Theophylline 598194 Oral Glucocorticoid Steroid 2449781 Humanised monoclonal antibody 607517 Oral Glucocorticoid Steroid 154  DIN Category DIN Category 2456419 Humanised monoclonal antibody 610623 Oral Glucocorticoid Steroid 577308 Ketotifen AL 868426 Oral Glucocorticoid Steroid 600784 Ketotifen AL 868434 Oral Glucocorticoid Steroid 2020017 Ketotifen AL 868442 Oral Glucocorticoid Steroid 2176084 Ketotifen AL 1946897 Oral Glucocorticoid Steroid 2218305 Ketotifen AL 1964070 Oral Glucocorticoid Steroid 2221330 Ketotifen AL 1964968 Oral Glucocorticoid Steroid 2230730 Ketotifen AL 1964976 Oral Glucocorticoid Steroid 2231679 Ketotifen AL 2063190 Oral Glucocorticoid Steroid 2231680 Ketotifen AL 2194082 Oral Glucocorticoid Steroid 2236606 Leukotriene receptor antagonist 2194090 Oral Glucocorticoid Steroid 2238216 Leukotriene receptor antagonist 2237044 Oral Glucocorticoid Steroid 2238217 Leukotriene receptor antagonist 2237045 Oral Glucocorticoid Steroid 2243602 Leukotriene receptor antagonist 2237046 Oral Glucocorticoid Steroid 2247997 Leukotriene receptor antagonist 2239534 Oral Glucocorticoid Steroid 2260565 Omalizumab 2240684 Oral Glucocorticoid Steroid 15016 Oral Glucocorticoid Steroid 2240685 Oral Glucocorticoid Steroid 15024 Oral Glucocorticoid Steroid 2240687 Oral Glucocorticoid Steroid 16438 Oral Glucocorticoid Steroid 2250055 Oral Glucocorticoid Steroid 16446 Oral Glucocorticoid Steroid 2260298 Oral Glucocorticoid Steroid 16462 Oral Glucocorticoid Steroid 2260301 Oral Glucocorticoid Steroid 21695 Oral Glucocorticoid Steroid 2261081 Oral Glucocorticoid Steroid 28185 Oral Glucocorticoid Steroid 2279363 Oral Glucocorticoid Steroid 2311267 Oral Glucocorticoid Steroid 2230210 Systemic Glucocorticoid Steroid 3859 Synthetic Derivative of Adrenaline 2230211 Systemic Glucocorticoid Steroid 254134 Synthetic Derivative of Adrenaline 2231893 Systemic Glucocorticoid Steroid 1923870 Synthetic Derivative of Adrenaline 2231894 Systemic Glucocorticoid Steroid 1928449 Synthetic Derivative of Adrenaline 2231895 Systemic Glucocorticoid Steroid 2017652 Synthetic Derivative of Adrenaline 2232748 Systemic Glucocorticoid Steroid 2017660 Synthetic Derivative of Adrenaline 2232750 Systemic Glucocorticoid Steroid 155  DIN Category DIN Category 16241 Systemic Glucocorticoid Steroid 2237835 Systemic Glucocorticoid Steroid 28096 Systemic Glucocorticoid Steroid 2241229 Systemic Glucocorticoid Steroid 30600 Systemic Glucocorticoid Steroid 2245400 Systemic Glucocorticoid Steroid 30619 Systemic Glucocorticoid Steroid 2245406 Systemic Glucocorticoid Steroid 30627 Systemic Glucocorticoid Steroid 2245407 Systemic Glucocorticoid Steroid 30635 Systemic Glucocorticoid Steroid 2245408 Systemic Glucocorticoid Steroid 30643 Systemic Glucocorticoid Steroid  30651 Systemic Glucocorticoid Steroid 30678 Systemic Glucocorticoid Steroid 30759 Systemic Glucocorticoid Steroid 30767 Systemic Glucocorticoid Steroid 36137 Systemic Glucocorticoid Steroid 213624 Systemic Glucocorticoid Steroid 664227 Systemic Glucocorticoid Steroid 716715 Systemic Glucocorticoid Steroid 732885 Systemic Glucocorticoid Steroid 732893 Systemic Glucocorticoid Steroid 751863 Systemic Glucocorticoid Steroid 783900 Systemic Glucocorticoid Steroid 872520 Systemic Glucocorticoid Steroid 872539 Systemic Glucocorticoid Steroid 156  DIN Category DIN Category 874582 Systemic Glucocorticoid Steroid 878618 Systemic Glucocorticoid Steroid 878626 Systemic Glucocorticoid Steroid 888206 Systemic Glucocorticoid Steroid 888214 Systemic Glucocorticoid Steroid 888222 Systemic Glucocorticoid Steroid 888230 Systemic Glucocorticoid Steroid 1934325 Systemic Glucocorticoid Steroid 1934333 Systemic Glucocorticoid Steroid 1934341 Systemic Glucocorticoid Steroid 1977547 Systemic Glucocorticoid Steroid 2063697 Systemic Glucocorticoid Steroid 2063700 Systemic Glucocorticoid Steroid 2063719 Systemic Glucocorticoid Steroid 2063727 Systemic Glucocorticoid Steroid 2204266 Systemic Glucocorticoid Steroid 2204274 Systemic Glucocorticoid Steroid 157  C.2 Parameter estimates from the sensitivity analysis of the subset of participants with spirometry-confirmed moderate to severe COPD near the time of diagnosis  All-cause primary care visits Respiratory-related primary care visits Specialist physician visits Pharmacist visits  Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI COPD (vs. comparator) 1.30 1.27-1.34 2.67 2.57-2.78 1.33 1.29-1.37 1.60 1.54-1.66 Period 1.07 1.06-1.08 1.18 1.17-1.19 1.08 1.08-1.09 1.07 1.06-1.08 Female (vs. male) 1.10 1.07-1.12 1.18 1.14-1.23 1.01 0.98-1.05 1.13 1.08-1.17 Age 1.02 1.02-1.02 1.00 1.00-1.00 1.02 1.02-1.03 1.01 1.01-1.01 SES     1 (vs. 5) 1.12 1.08-1.17 1.13 1.06-1.20 0.99 0.95-1.04 1.32 1.24-1.41    2 (vs. 5) 1.08 1.04-1.12 1.12 1.06-1.20 0.97 0.93-1.02 1.13 1.07-1.18    3 (vs. 5) 1.04 1.00-1.08 1.10 1.04-1.18 0.97 0.92-1.01 1.08 1.03-1.14    4 (vs. 5) 1.04 1.00-1.08 1.02 0.96-1.09 1.01 0.96-1.07 1.07 1.03-1.13 Calendar year 1.00 0.99-1.00 0.96 0.95-0.96 1.01 1.01-1.02 1.02 1.01-1.02  Parameter estimates and confidence intervals for the rate of outpatient visits among COPD patients who were prescribed long-acting muscarinic antagonist (LAMA) in the year before or after the index date (N=9,079). Patients in this subgroup had spirometry-confirmed moderate to severe COPD (FEV1/FVC<0.7 and FEV1 ≤65%). 158  C.3 Parameter estimates from the subset of participants with no missing data  All-cause primary care visits Respiratory-related primary care visits Specialist physician visits Pharmacist visits  Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI COPD (vs. comparator) 1.40 1.39-1.41 2.16 2.14-2.19 1.35 1.34-1.36 1.61 1.60-1.63 Period 1.06 1.06-1.06 1.14 1.14-1.14 1.07 1.07-1.08 1.06 1.06-1.06 Female (vs. male) 1.15 1.15-1.16 1.27 1.25-1.28 1.08 1.07-1.09 1.18 1.17-1.20 Age 1.01 1.01-1.01 1.00 1.00-1.00 1.02 1.02-1.02 1.01 1.01-1.01 SES     1 (vs. 5) 1.17 1.16-1.19 1.11 1.09-1.13 1.02 1.00-1.03 1.32 1.30-1.35    2 (vs. 5) 1.10 1.09-1.11 1.08 1.06-1.10 0.99 0.98-1.01 1.12 1.10-1.14    3 (vs. 5) 1.06 1.04-1.07 1.05 1.03-1.07 0.99 0.97-1.00 1.06 1.04-1.08    4 (vs. 5) 1.03 1.02-1.05 1.02 1.00-1.04 0.99 0.97-1.00 1.04 1.03-1.06 Calendar year 1.00 1.00-1.00 0.97 0.96-0.97 1.01 1.01-1.01 1.02 1.02-1.02  Parameter estimates and confidence intervals for the rate of outpatient visits in the subset of COPD patients and comparator subjects with no missing data (N=110,556).159  C.4 Parameters estimates from a new cohort of participants in which comparator subjects never developed COPD during follow-up  All-cause primary care visits Respiratory related primary care visits Specialist physician visits Pharmacist visits  Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI Rate Ratio 95% CI COPD (vs. comparator) 1.46 1.45-1.47 2.35 2.32-2.38 1.42 1.40-1.43 1.72 1.70-1.75 Period 1.06 1.06-1.06 1.13 1.13-1.14 1.07 1.07-1.08 1.06 1.05-1.06 Female (vs. male) 1.16 1.15-1.17 1.27 1.26-1.29 1.09 1.08-1.10 1.19 1.18-1.21 Age 1.01 1.01-1.01 1.00 1.00-1.00 1.02 1.02-1.02 1.01 1.01-1.01 SES     1 (vs. 5) 1.15 1.13-1.16 1.09 1.07-1.11 1.01 1.00-1.02 1.29 1.26-1.31    2 (vs. 5) 1.09 1.07-1.10 1.08 1.06-1.10 0.99 0.97-1.00 1.11 1.09-1.13    3 (vs. 5) 1.04 1.03-1.05 1.04 1.02-1.06 0.99 0.97-1.00 1.06 1.04-1.08    4 (vs. 5) 1.02 1.01-1.04 1.01 0.99-1.03 1.00 0.98-1.01 1.03 1.01-1.05 Calendar year 1.01 1.00-1.01 0.97 0.97-0.97 1.01 1.01-1.01 1.02 1.02-1.03  Parameter estimates and confidence intervals for the rate of outpatient visits in a cohort of 106,439 COPD/comparator pairs in which comparator subjects never developed COPD during follow-up. 160 Appendix D   D.1 Development and validation of the symptom module in EPIC D.1.1 Regression coefficients for the log odds of the presence of symptoms. Parameter Estimate (SE) p-value Estimate (SE) p-value Cough Wheeze COPD     Intercept 4.4006 (1.3129) <0.01 14.2686 (2.4553) <0.01 Age -0.0412 (0.0138) <0.01 -0.1408 (0.0255) <0.01 Female -0.9472 (0.2956) <0.01 -1.4995 (0.4489) <0.01 Smoking 0.6036 (0.2089) <0.01 0.1345 (0.2685) 0.62 FEV1 -0.9564 (0.2170) <0.01 -2.3122 (0.3931) <0.01 Non-COPD     Intercept -5.2872 (1.9482) 0.01 -6.6284 (1.5300) <0.01 Age 0.03429 (0.0158) 0.03 -0.02051 (0.0217) 0.34 Female 0.2178 (0.2743) 0.43 0.4671 (0.4208) 0.27 Smoking 0.3786 (0.2114) 0.07 0.2332 (0.3205) 0.47 Phlegm Dyspnea COPD     Intercept 3.5726 (1.845) 0.05 4.8358 (1.4826) <0.01 Age -0.02422 (0.0193) 0.21 -0.00891 (0.0150) 0.55 Female -2.089 (0.4924) <0.01 -0.6346 (0.3256) 0.05 Smoking 1.0754 (0.2826) <0.01 0.4177 (0.2281) 0.07 FEV1 -1.6443 (0.356) <0.01 -1.9942 (0.2824) <0.01 Non-COPD     Intercept -10.3164 (1.8041) <0.01 -7.1802 (0.9182) <0.01 Age 0.02771 (0.0252) 0.27 0.07002 (0.0119) <0.01 Female -0.3459 (0.4763) 0.47 0.9343 (0.2204) <0.01 Smoking 0.5865 (0.3413) 0.08 0.8388 (0.1856) <0.01  D.1.2 Covariance matrix of random effects.  Cough Phlegm Wheeze Dyspnea  COPD Cough 2.7934 2.11077 1.1737 0.64573 Phlegm 2.11077 7.0945 1.45258 0.89548 Wheeze 1.1737 1.45258 7.4793 1.48313 Dyspnea 0.64573 0.89548 1.48313 3.6357  Non-COPD Cough 10.4693 17.10328 9.39833 1.04549 Phlegm 17.10328 138.8000 36.8647 4.67073 Wheeze 9.39833 36.86470 107.09 5.78527  161  Cough Phlegm Wheeze Dyspnea Dyspnea 1.04549 4.67073 5.78527 5.1828  162 D.1.3 Results of internal validation between EPIC and CanCOLD for the prevalence of symptoms.  150,000 individuals were simulated in EPIC, and 1,573 participants were included from CanCOLD. GOLD 4 is not showing due to the small number of observations in CanCOLD (n=6). 163 D.2 Development and validation of the primary care visits module in EPIC D.2.1 Regression coefficients for the rate (log scale) of annual primary care visits. Parameter Estimate (SE) p-value COPD Intercept 0.4472 (0.5578) 0.42 Age 0.012 (0.0055) 0.03 Female -0.0725 (0.1143) 0.53 Smoking 0.0669 (0.0704) 0.34 Cough -0.0037 (0.0631) 0.95 Phlegm -0.0108 (0.0734) 0.88 Wheeze 0.0553 (0.0603) 0.36 Dyspnea 0.0947 (0.0737) 0.20 FEV1 0.1414 (0.0881) 0.11 Dispersion* 0.4310  Non-COPD Intercept -0.3596 (0.3337) 0.28 Age 0.0169 (0.0047) <0.01 Female 0.0095 (0.0668) 0.80 Smoking 0.0722 (0.0615) 0.24 Cough 0.181 (0.1186) 0.13 Phlegm -0.0275 (0.0763) 0.72 Wheeze 0.2262 (0.0637) <0.01 Dyspnea 0.0807 (0.0619) 0.19 Dispersion* 0.4093   * The reciprocal of the dispersion parameter is the alpha parameter in a gamma distribution, which was used to generate a negative binomial distribution. 164 D.2.2 Results of internal validation between EPIC and CanCOLD for the mean number of annual primary care visits.  150,000 individuals were simulated in EPIC, and 1,572 participants were included from CanCOLD. GOLD 4 is not showing due to the small number of observations in CanCOLD (n=6).   165 D.3 Development and validation of the diagnosis module in EPIC D.3.1 Regression coefficients for the log odds of diagnosis. Parameter Estimate (SE) p-value Estimate (SE) p-value Prevalent Diagnosis False Positive Diagnosis Intercept 1.0543 (1.2144) 0.39 -5.2169 (1.0711) <0.01 Age -0.0152 (0.0134) 0.26 0.0025 (0.0153) 0.87 Female -0.1638 (0.1395) 0.24 0.2597 (0.2906) 0.37 Smoking 0.1068 (0.1255) 0.39 0.6911 (0.2954) 0.02 FEV1 -0.6146 (0.2263) 0.01   Cough 0.075 (0.1302) 0.56 0.7264 (0.3722) 0.05 Phlegm 0.283 (0.1337) 0.03 0.7956 (0.375) 0.03 Wheeze -0.0275 (0.124) 0.82 0.66 (0.3468) 0.06 Dyspnea 0.5414 (0.1302) <0.01 0.8798 (0.3245) 0.01 Primary care visits*   0.0075 (0.0258) 0.77 Incident Diagnosis Intercept -2†  Age -0.0324 (0.0108) <0.01 Female -0.4873 (0.2273) 0.03 Smoking 0.3711 (0.1946) 0.06 FEV1 -0.8032 (0.1957) <0.01 Cough 0.208 (0.2164) 0.34 Phlegm 0.4088 (0.2356) 0.08 Wheeze 0.0321 (0.2081) 0.88 Dyspnea 0.722 (0.2048) <0.01 Primary care visits 0.0087 (0.0231) 0.71  * Primary care visits were not simulated at baseline; therefore, this parameter was not included in the regression equation for prevalent diagnosis. † Coefficient was determined through model calibration. (Original intercept 1.9614)   166 D.3.2 Results of internal validation between EPIC and CanCOLD for the prevalence of diagnosis among COPD patients.  150,000 individuals were simulated in EPIC, and 467 participants were included from CanCOLD. GOLD 4 is not showing due to the small number of observations in CanCOLD (n=2).   167 Results from EPIC were generated using the original regression equations for diagnosis that were developed using data from CanCOLD. Following internal validation, the intercept of the regression equation was calibrated to yield a stable prevalence of diagnosed patients at a proportion approximately equal to that observed in CanCOLD (29%150). 168 D.3.3 Number of COPD patients diagnosed over model time.  150,000 individuals were simulated in EPIC. This figure was used during calibration to assess the stability of diagnosis and the proportion of COPD patients diagnosed.   169 D.3.4 Results of internal validation between EPIC and CanCOLD for the prevalence of false positive diagnosis (‘overdiagnosis’) among non-COPD individuals.  150,000 individuals were simulated in EPIC. 1,011 non-COPD participants (n=53 overdiagnosed) were included from CanCOLD.  170 D.3.5 Proportion of non-COPD individuals with a false positive diagnosis (‘overdiagnosed’) over model time.  Results are shown for 101,579 non-COPD individuals simulated in EPIC at year one. This figure was used during calibration to assess the stability of overdiagnosis.  171 D.4 Development and validation of the treatment module in EPIC D.4.1 Treatment algorithm for individuals diagnosed with COPD. Exacerbation history* Symptoms  No dyspnea Dyspnea ≥2 moderate exacerbations† or 1 severe‡/very severe§ exacerbation LAMA/LABA ICS/LAMA/LABA No exacerbations or 1 moderate exacerbation SABA LAMA False positive diagnosis SABA  * Exacerbation history in the previous 12 months † Moderate exacerbations are defined as those requiring a visit to the emergency department ‡ Severe exacerbations require hospital admission § Very severe exacerbations require admission to an intensive care unit  172 D.4.2 Proportion of simulated individuals in EPIC on medications of each class over model time.  210,000 individuals were simulated in EPIC.   173 D.4.3 Effectiveness of smoking cessation among newly diagnosed patients in EPIC.  Cross-sectional proportion of non-smokers in the years since COPD diagnosis.  150,000 individuals were simulated in EPIC in both the no smoking cessation and smoking cessation scenarios. This figure was used during calibration to assess the marginal effect of the smoking cessation intervention. In year 1, 34.4% of individuals who received smoking cessation were non-smokers, compared to 7.4% under no smoking cessation. These figures were 70.2% and 61.8% after 11 years, respectively.   174 D.5 Performance characteristics of case detection methods D.5.1 Performance characteristics of case detection methods evaluated for each scenario. Case detection method Sensitivity β logit(Diagnosis)* Specificity β logit(Overdiagnosis) (S1) All patients S1a: CDQ ≥17 points 91 4.1013 49 4.3940 S1b: Screening spirometry (with bronchodilator) 80 3.1740 94 1.6025 S1c: CDQ + Screening spirometry (with bronchodilator) 72 2.7321 97 0.8779 (S2) Symptomatic patients (any respiratory symptoms) S2a: Screening spirometry (without bronchodilator) 81.5 3.2705 88.9 2.2735 (S3) Smoking history (ever smokers ≥ 50 years) S3a: CDQ ≥19.5 points 64.5 2.3848 65.2 3.7262 S3b: CDQ ≥16.5 points 87.5 3.7336 38.8 4.8098 S3c: Screening spirometry (without bronchodilator) 79.9 3.1677 84.4 2.6657 S3d: CDQ + Screening spirometry (with bronchodilator) 74.4 2.8545 97 0.8779  * This refers to !!" in Equation 5-1  D.5.2 Example calculation of the coefficient for case detection.  The sensitivity of the example case detection method is 75%, therefore the odds of diagnosis among COPD patients is 3. Subject Probability of routine diagnosis Odds of routine diagnosis Odds ratio for case detection 1 0.13 0.15 3/0.15 = 19.59 2 0.18 0.22 3/0.22 = 13.40 3 0.27 0.37 3/0.37 = 8.13   175 The average odds ratio for case detection among Subjects 1-3 is 13.71. Therefore, the coefficient for case detection (!!") in the regression equation for the logit of diagnosis (Equation 5-1) would be ln(13.71) = 2.62.    176 D.6 Results of the preferred case detection scenario  Results of the preferred case detection scenario (S1a: CDQ ≥17 points delivered at 3 years intervals to all patients ≥40 years) compared to no case detection. Results are shown for a cohort of 1,000 individuals eligible to receive case detection. 

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}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            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:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0389633/manifest

Comment

Related Items