AN EXAMINATION OF INFLUENCES ON PRESCRIPTION DRUG USE AMONGST OLDER ADULTS WITH AND WITHOUT CHRONIC ILLNESS IN TWO CANADIAN PROVINCES by FIONA KA IENG CHAN BSc (Pharm), University of British Columbia (Vancouver), 2012 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Population and Public Health) THE UNIVERSITY OF BRITISH COLUMBIA (VANCOUVER) December 2017 © Fiona Ka Ieng Chan, 2017 ii Abstract Background - Prescription medications, which are a critical component of modern medicine, are not covered under the Canada Health Act. Canadian prescriptions are financed through a combination of public financing, private insurance, and out-of-pocket payments. This leaves potential coverage gaps amongst different segments of the population. There are current discussions of a national pharmacare strategy to address this issue. It remains unclear how such a policy, particularly for seniors, should be financed. Methods – We first studied British Columbia’s publicly-funded pharmacare program to examine the impact of income-based deductibles on older adults with chronic obstructive pulmonary disease (COPD) in regards to prescription drug utilization and other health services. This was analyzed utilizing a regression discontinuity design with administrative datasets. The second study used logistic regression to examine the trend in employer-sponsored health insurance (EHI) availability amongst Ontario’s retirees from 2005 to 2014 using data from the Canadian Community Health Survey. Results - Deductibles had no effect on prescription utilization amongst a cohort of older adults living with COPD in BC. However, over 40% of the eligible person-years analyzed did not obtain a prescription for COPD treatment, suggesting severe under-treatment in this population. Results also suggest an increased use of inhaled corticosteroids, which may be due to a special authority process and may not be appropriate for COPD. A decline in EHI availability was apparent for Ontario’s retirees between 2005 and 2013-2014. EHI availability is strongly linked to household income, with those of a lower income-decile having the lowest odds of having EHI. Conclusion - Imposing a modest income-based deductible was not found to impact prescription utilization or utilization of other health services, even amongst a population with a chronic condition facing comparatively high prescription costs. In contrast, supplemental help in making prescriptions more affordable for the older adult population may be diminishing. A small but statistically significant decrease was observed in EHI – the main source of aid in prescription iii affordability apart from the public system. These results suggest that a comprehensive strategy to address medication adherence is warranted to minimize future health system burden. iv Lay Summary Prescription medications for Canadians are made more affordable by two financing systems: publicly-financed drug coverage, and employer-sponsored private health insurance. It remains unclear if current public policies using income-based deductibles are a viable funding strategy in ensuring proper access to prescription medications and preventing adverse health outcomes. Additionally, the scope and availability of employer-sponsored health insurance may have decreased over time, which may have decreased the affordability of non-publicly insured treatments such as prescription drugs. This thesis suggests that prescription utilization and health outcomes were not impacted by imposing an income-based deductible amongst a cohort of older adults living with chronic obstructive pulmonary disease – a disease where the cost per prescription may be comparatively high. The decrease in availability of employer-sponsored health insurance over the past decade amongst retirees is also presented. These results can help to inform current discussions around developing a national pharmacare strategy. v Preface This thesis is the result of work conducted and written by Fiona Chan (FC). The development of research questions, study design, and methodology was possible under the advisement of the thesis committee, consisting of Dr. Michael Law (UBC), Dr. J. Mark FitzGerald (UBC), Dr. Kimberlyn McGrail (UBC), and Dr. Sumit R. Majumdar (University of Alberta). The study of the impact of deductibles on prescription utilization amongst older adults with chronic obstructive pulmonary disease (COPD) in this thesis is a contribution to one of Dr. Law’s broader research projects entitled “The Impact of Out-of-Pocket Costs within the BC PharmaCare Program”. The study was approved by the Behavioural Research Ethics Board at the University of British Columbia (certificate number: H15-004). With the assistance of Dr. Law and Heather Worthington, FC completed the documentation required for data access and data extraction for this project from Population Data BC. Yan (Lucy) Cheng, a data analyst at the UBC Centre for Health Services and Policy Research (CHSPR) assisted with data extraction and preparing the final dataset for the regression discontinuity design analysis. Dr. Law assisted with the final analysis. FC designed the empirical approach, conducted all data analyses, compiled results, and drafted all materials related to this study. FC developed the research question, empirical approach, conducted all data analyses, and drafted all material related to the employer-health insurance analysis. The research was conducted in two phases, a preliminary phase using public-use microfiles and the final phase using microdata available through Statistics Canada’s Research Data Centre. Dr. Mieke Koehoorn (UBC) advised throughout the preliminary phase and provided input on how to proceed with the subsequent analysis. Dr. Law provided input in the final phase and presentation of results. This study is covered under the publicly available data clause (Item 7.10.3) of the University of British Columbia’s Policy #89: Research Involving Human Participants, which exempts researching involving the use of publicly available data protected by law from requiring study-specific ethics approval.1 vi The thesis committee provided guidance throughout the research process and critical feedback on earlier drafts prior to submission of this thesis to the Faculty of Graduate and Post-Graduate Studies. Versions of Chapter 2 and 3 will be submitted for publication in the coming months. Main findings from Chapter 3 were presented as poster presentations at the 14th Canadian Association for Health Services and Policy Research Conference on May 25, 2017 in Toronto, Canada, as well as the 29th Annual Centre for Health Services and Policy Research Conference on Mar 9-10, 2017 in Vancouver, Canada. vii Table of Contents Abstract.......................................................................................................................................... ii Lay Summary............................................................................................................................... iv Preface............................................................................................................................................ v Table of Contents........................................................................................................................ vii List of Tables ................................................................................................................................. x List of Figures ............................................................................................................................... xi List of Abbreviations ................................................................................................................. xiii Glossary ...................................................................................................................................... xiv Acknowledgements.................................................................................................................... xvi 1 Introduction ........................................................................................................................... 1 1.1 Background ..................................................................................................................... 1 1.2 Literature Review ............................................................................................................ 3 1.2.1 What is COPD? ....................................................................................................... 3 1.2.2 Current prescription coverage schemes in Canada ................................................. 6 1.3 Research Objectives ........................................................................................................ 7 1.4 Thesis Outline ................................................................................................................. 8 2 Impact of income-based deductibles amongst older adults living with chronic obstructive pulmonary disease in British Columbia ............................................................... 10 2.1 Background ................................................................................................................... 10 2.1.1 BC context ............................................................................................................ 11 2.2 Methods......................................................................................................................... 13 2.2.1 Data sources .......................................................................................................... 13 2.2.2 Study cohort .......................................................................................................... 14 2.2.3 Study period .......................................................................................................... 15 2.2.4 Outcomes of interest ............................................................................................. 15 2.2.5 Statistical analysis ................................................................................................. 17 2.3 Results ........................................................................................................................... 18 viii 2.3.1 Cohort characteristics ............................................................................................ 18 2.3.2 PharmaCare expenditure ....................................................................................... 19 2.3.3 Prescription drug use and costs ............................................................................. 19 2.3.4 Outpatient acute exacerbations medication use and costs .................................... 20 2.3.5 Physician visits and cost ....................................................................................... 20 2.3.6 Hospital admissions and length of stay ................................................................. 20 2.3.7 Sensitivity analysis................................................................................................ 21 2.4 Interpretation ................................................................................................................. 21 2.4.1 Overall use of COPD-related medications ............................................................ 22 2.4.2 Rescue vs. maintenance medications .................................................................... 23 2.4.3 Comparison to current literature ........................................................................... 24 2.4.4 Study limitations ................................................................................................... 25 2.5 Conclusions ................................................................................................................... 26 2.6 Tables and Figures ........................................................................................................ 28 3 Changes in employer-sponsored health insurance amongst retirees in Ontario between 2005 and 2013-2014 ..................................................................................................................... 47 3.1 Introduction ................................................................................................................... 47 3.2 Methods......................................................................................................................... 49 3.2.1 Survey data and study design ................................................................................ 49 3.2.2 Study samples ....................................................................................................... 49 3.2.3 Variables for analysis ............................................................................................ 49 3.2.4 Analysis plan ......................................................................................................... 50 3.3 Results ........................................................................................................................... 51 3.4 Interpretation ................................................................................................................. 52 3.5 Conclusion .................................................................................................................... 55 3.6 Tables and Figures ........................................................................................................ 56 4 Conclusion ........................................................................................................................... 61 4.1 Summary of Findings .................................................................................................... 61 ix 4.1.1 Impact of deductibles under British Columbia’s Fair PharmaCare amongst older adults living with chronic obstructive pulmonary disease (COPD) ...................................... 61 4.1.2 Changes in employer-sponsored health insurance amongst retirees in Ontario ... 62 4.2 Contribution to Current Literature ................................................................................ 63 4.3 Strengths and Limitations ............................................................................................. 65 4.3.1 Deductibles study .................................................................................................. 65 4.3.2 Employer-health sponsored insurance study ........................................................ 66 4.4 Recommendations for Future Research ........................................................................ 67 4.4.1 Factors underlying the low utilization rates of COPD-related medications ......... 67 4.4.2 Impact of current special authority program, particularly for COPD population . 68 4.4.3 Difference in characteristics of individuals not registered for Fair PharmaCare .. 68 4.4.4 More in-depth study of EHI changes for retirees across Canada .......................... 69 References.................................................................................................................................... 70 x List of Tables Table 2.1 – Classification of medications related to chronic obstructive pulmonary disease. .... 28 Table 2.2 - Respiratory antibiotics used in treatment of acute exacerbation of chronic obstructive pulmonary disease. ........................................................................................................................ 29 Table 2.3 – The number of person-years and observations used in analyzing each outcome variable in studying the impact of income-based deductibles amongst an older adult population living with chronic obstructive pulmonary disease (COPD). ....................................................... 30 Table 3.1 - Characteristics of study sample - investigating the relationship between availability of employer-sponsored health insurance (EHI) and survey year; data from the combined Cycle 3.1 (2005) and the 2013-2014 cycle of the Canadian Community Health Survey. ...................... 57 Table 3.2 - Results from Logistic Regression: The association between survey year (reference 2005 cycle) and the availability of employer-sponsored health insurance (yes/no). Unadjusted and adjusted odds ratios (aOR) with 95% confidence intervals. Bolded results denote statistical significance. .................................................................................................................................. 59 Table 3.3 - Predicted probability of receiving Employer-Sponsored Health Insurance in 2005 and 2013-2014 for individuals of certain characteristics; derived from estimates of logistic regression. ..................................................................................................................................... 60 xi List of Figures Figure 2.1 - Structure of British Columbia's Fair PharmaCare program for the Enhanced Assistance and Regular Assistance plans ..................................................................................... 31 Figure 2.2 - Average annual number of prescriptions for all medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth ...... 32 Figure 2.3 - Average annual drug expenditure for all medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth ........... 33 Figure 2.4 - Average annual drug expenditure for COPD-related medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth ....................................................................................................................................................... 34 Figure 2.5 - Average annual drug expenditure for “rescue” COPD medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease ....................... 35 Figure 2.6 - Average annual drug expenditure for “maintenance” COPD medications between 2004 to 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth .......................................................................................................................................... 36 Figure 2.7 - Average annual drug expenditure for all inhaled corticosteroids (ICS) between 2004 to 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth .............................................................................................................................................. 37 Figure 2.8 - Average annual number of prescriptions for all inhaled corticosteroids (ICS) between 2004 to for British Columbians living with chronic obstructive pulmonary disease, by year of birth .................................................................................................................................. 38 Figure 2.9 - Average annual number of prescriptions for prednisone between 2004 to for British Columbians living with chronic obstructive pulmonary disease, by year of birth ....................... 39 Figure 2.10 - Average annual number of prescriptions for respiratory antibiotics between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth .............................................................................................................................................. 40 Figure 2.11 - Average annual drug expenditure for respiratory antibiotics dispensed between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth . ......................................................................................................................................... 41 Figure 2.12 - Average annual physician expenditure on visits between 2004 and 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth ....................... 42 xii Figure 2.13 - Average annual number of physician visits made between 2004 and 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth ........... 43 Figure 2.14 - Average annual number of hospital admissions for any reason between 2004 and 2013 in British Columbians living with chronic obstructive pulmonary disease, by year of birth....................................................................................................................................................... 44 Figure 2.15 - Average annual number of days in hospital for any reason between 2004 and 2013 in British Columbians living with chronic obstructive pulmonary disease, by year of birth ....... 45 Figure 2.16 - Average annual number of hospital admissions due to chronic obstructive pulmonary disease (COPD) between 2004 and 2013 in British Columbians living with COPD, by year of birth .................................................................................................................................. 46 Figure 3.1 - Derivation of study sample from Cycle 3.1 (2005 cycle) and the 2013-2014 Cycle of the Canadian Community Health Survey, including exclusions due to missing/invalid responses....................................................................................................................................................... 56 xiii List of Abbreviations ATC Anatomical Therapeutic Chemical BC British Columbia CCHS Canadian Community Health Survey CHSPR UBC Centre for Health Services and Policy Research COPD Chronic obstructive pulmonary disease ED Emergency Department EHI Employer-sponsored health insurance GOLD Global Initiative for Chronic Obstructive Lung Disease ICD-9 International Classification of Diseases and Related Health Problems, 9th revision ICD-10CA International Classification of Diseases and Related Health Problems, 10th revision, Canada ICS Inhaled corticosteroids LABA Long-acting beta2-agonists LAMA Long-acting antimuscarinic antagonists MSP Medical Services Plan OECD Organisation for Economic Co-Operation and Development PDE-4 inhibitors Phosphodiesterase-4 inhibitor RDD Regression discontinuity design SABA Short-acting beta2-agonists xiv Glossary Acute exacerbation of COPD Acute worsening of COPD symptoms, often requiring medical attention. Copayment The cost payable by the user: the total cost less the cost paid by the insurer. Deductible Amount in eligible prescription cost to be paid in full prior to partial payment assistance on eligible prescriptions. Under Fair PharmaCare, this amount is shared amongst one’s household and is reset annually. Enhanced coverage The coverage scheme under Fair PharmaCare for those born in or before 1939. Deductibles, copayments, and family maximums are lower than those of Regular coverage in a comparable income bracket. Formulary A list of drug molecules which count towards the deductible and are eligible for coverage assistance by a particular insurance provider. Includes regular benefit and special authority drugs. Fair PharmaCare A voluntary, publicly-funded drug coverage program in British Columbia. Operates on a system of income-based deductibles and family maximums. Assistance commences after deductible (if applicable) is met through eligible prescription expense. Coinsurance period then starts until family maximum is reach, after which eligible prescriptions are fully covered. These are reset at the start of each calendar year. Special coverage schemes also available for particular populations. Family maximum Amount in eligible prescription cost to be paid prior to complete coverage of eligible prescriptions. This amount is shared amongst one’s household and is reset annually under Fair PharmaCare. Maintenance medications Medications used to manage symptoms of COPD and reduce risk of acute exacerbations. Costs ~$40-150 per month. Premium Cost to buy into a particular insurance plan. xv Regular benefit Coverage status of a drug that is automatically recognized by the Fair PharmaCare program as eligible expense. Additional documentation of clinical need is not required. Regular coverage The coverage scheme under Fair PharmaCare, which applies to those born in or after 1940. Rescue medications Medications intended for the relief of acute symptoms of COPD or symptom relief during COPD exacerbations. Costs ~$7-31 per month. Special authority Coverage status for a drug not considered as a Regular benefit under BC’s Fair Pharmacare, but may be recognized as eligible expense after a physician applies and demonstrates that a patient meets the “limited coverage criteria” set by the Ministry of Health for the specific drug. Some physician specialties are automatically exempted from this process for particular drugs (e.g. respirologists for some COPD maintenance medications). Threshold In regression discontinuity design (RDD), the point of change in allocation of a particular intervention. The change in allocation is not related to the outcome of the intervention and is often an arbitrary cut off for where a treatment or policy is applied. For a successful RDD application, subjects on either side of the threshold are comparable in every other way apart from the intervention. xvi Acknowledgements I would like to thank my supervisor, Dr. Michael Law, for his guidance and providing me with many opportunities to pursue a passion which I have held for many years. This experience would not have been possible without the generosity of Mike in allowing me to participate in so much of his research. I am grateful to Mike for this experience. I would also like to thank my thesis committee members, Dr. J. Mark FitzGerald, Dr. Kimberlyn McGrail, and Dr. Sumit R. Majumdar, for their input and patience with this new researcher. Thank you for valuing my insights. Your support has been most empowering. I would like to thank Mike’s staff (Lucy Cheng, Heather Worthington, and Ashra Kolhatkar) as well as the staff at CHSPR for making me feel welcomed and enabling me to do this work. You were all always available and kin, particularly in times of need. I also extend my thanks to faculty members at the School of Population of Public Health, particularly Drs. Mieke Koehoorn, Craig Mitton, Nick Bansback, and Larry Lynd, whose course material and inspiring ideas have been incorporated into this thesis, enrichening it in one form or another. Last but not least, I would like to thank my family and friends for their support throughout my degree. Most of all, I would like to thank my parents for seeing the value of education and believing in me and my choices. Love you both more than words can describe. To my sister Karen, for always challenging me and my ideas. To Mohamed, my best friend and partner who had encouraged me to pursue this passion. Thank you also for keeping me on track when I was struggling the most and listen to me ramble on about the same things every day. I definitely could not have done it without you by my side every step of the way. 1 1 Introduction 1.1 Background More than 50 years after the creation of universal insurance for physician and hospital services, many Canadians still lack insurance coverage for prescription drugs. Canadians rely on a combination of publicly-financed prescription drug programs, private health insurance, and out-of-pocket payments to afford prescription medications.2 While there have been recent conversations around the development of a national pharmacare strategy to address this issue, there is much disagreement as to how such a policy should be financed – some advocate for an age-based strategy where individuals of certain age groups (e.g. seniors) receive enhanced benefits compared to other parts of the population, while others advocate for an income-based strategy with a system of copayments and/or deductibles.2,3 More recently, proponents of national pharmacare in Canada suggest adopting a universal, first-dollar coverage model for everyone, regardless of age or income.4 Evidence suggests that out-of-pocket costs are associated with cost-related nonadherence (CRNA).5–7 In British Columbia (BC), in particular, drug affordability may be of concern, as one in six residents report difficulty paying for prescription medications – the highest nationally.8 While some authors have suggested that BC’s income-based deductible for public pharmaceutical coverage is in part to blame for this high rate,8 a recent study found that the use of income-based deductibles in an older adult population in BC did not impact prescription drug use or health care utilization.9 This is in contrast to previous studies which had suggested that the use of deductibles – income-based and otherwise – may be detrimental to health status in individuals with specific medical conditions.5–7 It also remains unclear whether all segments of the population are affected by income-based deductibles, or if particular subpopulations (e.g. those with a specific disease) are affected differently. Further study of the impact of income-based deductibles amongst specific subpopulations can help inform current discussions on sustainable and equitable financing strategies for Canada’s national pharmacare program. 2 Out-of-pocket costs may be particularly concerning amongst patients living with chronic obstructive pulmonary disease, or COPD. This chronic illness is a major public health burden in Canada and internationally.10 There is strong evidence for the regular use of COPD maintenance medications to prevent acute exacerbations and hospitalizations.10 However, many COPD patients express concerns with these inhalers due to their high cost.11 Inhalers are particularly unique due to their upfront cost: inhalers cannot be split the same way that it is possible to split a 3-month prescription for tablets into 1-month instalments to make costs more manageable.10,12,13 Compared to other types of medications, the addition of a deductible may make these inhalers a larger financial burden, as an individual would be forced to pay a larger sum per prescription up front with limited reimbursement. While decreased adherence may be a common mechanism to defer cost for any type of medications (e.g. taking one dose once daily, rather than the prescribed twice daily dosing), the imposition of deductibles on costly inhalers may present a unique financial challenge. This may increase the foregoing of these medications in addition to decreased adherence, both of which may lead to poorer health outcomes, increased risk of acute exacerbations, and increased need for other health services.10 Thus, the impact of out-of-pocket costs, such as through deductibles, may not only impact prescription utilization amongst this population, but also the number of physician visits and hospitalizations. While studying the impact of BC’s income-based deductible system will allow us to understand one important factor which influences the use of medications, other factors outside the public system play an important role. Assistance with payment for prescription drugs by private insurance plans is another prominent factor likely affecting the use of prescription medications by Canadians. Most Canadians receive private insurance through their employer, the cost of which is subsidized through tax exemptions in all provinces except Quebec.14 This is true even for retired individuals, who may receive benefits from their previous employer as part of their retirement package alongside items such as their pension.14 Private insurance may enhance one’s ability to afford prescription medications. However, there has been some speculation, mostly from anecdotal Canadian evidence and American data, that employers have been increasingly less generous in providing insurance coverage to employees, both current and retired.15–17 Any decrease in employer-sponsored health insurance – 3 either in the availability of coverage or in scope and level of coverage in terms of premiums, deductibles, and copayments – may decrease one’s ability to pay, particularly if the public safety net is inadequate. Thus, it is critical to examine if such a trend is indeed occurring. 1.2 Literature Review This literature review is in two parts: the first part provides a review of current treatment recommendations for COPD and examines the current literature on cost-related non-adherence to COPD medications and the impact this has on patients and the health system. The purpose of this literature review was to help motivate and refine the research questions and hypotheses for the analysis presented in Chapter 2. The second part of this review examines the current coverage schemes for prescription drugs across Canada. This is meant to provide some context as to how prescription medications are financed in Canada in order to better understand the impact that changes in employer-sponsored health insurance may have in the context of the entire health care system. This will help inform the discussion for the analysis presented in Chapter 3 and the final conclusions for this thesis presented in Chapter 4. 1.2.1 What is COPD? The Global Initiative for Chronic Obstructive Lung Disease (GOLD) strategy, providing annual updates since 2001, is the most widely adopted clinical strategy for the diagnosis and treatment of COPD.10 GOLD defines COPD as a progressive disease associated with chronic airway inflammation, resulting in structural damage of the lungs and narrowing of airways.10 These contribute to airflow obstruction and, with more severe disease, hypoxemia.10 The main symptoms of COPD include dyspnea, chronic cough, sputum production, wheezing, and chest tightness.10 These characteristics of COPD are coupled with episodes of acute worsening of symptoms, or acute exacerbations, usually lasting about 7 to 10 days.10 In contrast with asthma, which is characterized by reversible airway constriction due to hyper-responsiveness, COPD is irreversible and is associated with older age.10 COPD is a leading and growing cause of morbidity and mortality globally.10 In Canada, COPD is the most common medical reason for hospitalization.18 It is associated with various risk 4 factors, the most prominent of which is smoking, but also includes biomass exposure, occupational exposures such as through rubber and plastic manufacturing, or construction work19, and air pollution.10 Lower socioeconomic status as well as older individuals are also at increased risk of developing COPD, possibly due to increased durations of exposure to risk factors and the effects of aging.10 Comorbid conditions, such as ischemic heart disease, heart failure, diabetes mellitus, and metabolic syndrome, can contribute to the severity of COPD.10,20 Conversely, they can also worsen the outcomes of these diseases, which can ultimately impact survival.10,20 In a study by Curkendall et al., the prevalence of cardiovascular disease was significantly higher amongst COPD patients compared to otherwise similar individuals without COPD.21 Correspondingly, Kinnunen et al. found that individuals with comorbidities hospitalized for COPD had longer lengths of stay compared to those without comorbidities.22 It has thus been suggested that better management of COPD in patients with comorbidities may also improve these conditions and overall health outcomes.20 Why Treat COPD? There is a well-established evidence base for the treatment of COPD which has been shown to reduce the risk of exacerbations while also improving quality of life when symptoms are controlled. Acute exacerbations are common in COPD and are often triggered by infection or environmental exposure.10 Comorbidities, such as pneumonia, congestive heart failure, and cardiac arrhythmias, may sometimes mimic and/or aggravate exacerbations.10 These exacerbations are usually treated with oral corticosteroids (OCs), bronchodilators (including short-acting inhaled beta2-agonists and/or antimuscarinics), and, for those with signs of bacterial infections, antibiotics.10 Exacerbations often result in unscheduled visits to physicians, emergency departments, or hospitalizations.10 In addition to increased system costs, they are associated with high mortality rates the risk of which further increases with each additional exacerbation10.The mortality rate following an exacerbation requiring hospitalization is estimated to be up to 40% 1-year post-discharge, and the all-cause mortality rate after 3 years can be as high as 49%.10 Additionally, the risk of myocardial infarction or stroke may be increased up to two-fold following a COPD exacerbation.23 Their prevention is thus key in reducing the burden of disease10 and the cost of current and future care. 5 Medications to Treat COPD The goals of COPD treatment are to relieve symptoms, increase quality of life, prevent disease progression, decrease the risk of exacerbations, and to reduce mortality.10 Strong evidence supports the use of “maintenance medications” to manage daily symptoms and to reduce risk of acute exacerbations.10,12 These medications include inhaled long-acting beta2-agonists (LABA), inhaled long-acting antimuscarinics (LAMA), and inhaled corticosteroids (ICS).10 LABAs and LAMAs are considered the mainstay of COPD therapy, as they have been shown to reduce exacerbations and hospitalizations significantly.10 In accordance with current guidelines, patients (particularly those with moderate to very severe COPD) often use these medications in combination.10 While the risk of exacerbation is also dependent on disease severity and history of prior exacerbations,10 a combination LABA/ICS therapy has been shown to reduce one exacerbation every 2 to 4 years - about a 27% risk reduction.24 Though no treatments to date have been shown to reduce mortality, a recent Cochrane review estimated that number needed to treat to prevent 1 death over 3 years is only 42 for combination LABA/ICS therapy.24 These therapies, however, are quite costly as each medication, excluding pharmacy markup and dispensing fee, ranges from approximately CAD $40 to $150 per month.10,12,25,13 These large lump sums can be particularly difficult to afford for lower-income individuals who are more likely to report cost-related non-adherence.8 Consequently, they may be replaced by the patient with cheaper, short-acting bronchodilators – intended for the relief of acute symptoms or acute exacerbations – that provide some symptom relief but have no effect on the risk of acute exacerbations.10,12,26 As well, individuals who have been prescribed combination maintenance therapy may exhibit partial or poor adherence to one or more of their concurrent treatments. This may be the case particularly for higher cost therapies. This type of non-adherence may decrease overall treatment efficacy and increase risk of acute exacerbations. Under-utilization and poor adherence to COPD therapy is a well-documented phenomenon.10,11,27 While cost is an important factor in influencing adherence to COPD therapy11,12, there are other factors which may impact adherence and subsequent success of treatment. As many individuals with COPD are older, the dexterity and coordination required to use some inhaler devices may make it more difficult for this older population to use them 6 properly.11 In addition to this, individuals may have multiple medications as part of their COPD therapy, along with other comorbidities with their own treatment regimens. As treatment complexity increases, adherence may also worsen.11,27 This, coupled with fears of side effects, may decrease use of these medications.27 In contrast, patient understanding and having realistic expectations of therapy may help increase adherence to COPD medications.11,27 Thus, adherence to COPD therapy is influenced by multiple factors. While cost concerns are important, they may not be solely responsible in explaining low utilization and poor adherence to COPD treatment. 1.2.2 Current prescription coverage schemes in Canada Prescription medications in Canada are financed in a multi-payer system, involving a mix of public and private insurance payments, as well as direct out-of-pocket payments. While most provinces offer some coverage to individuals based on age, income, employment, or disability status, the Canadian Institute for Health Information estimates that, in 2014, only 42.6% of prescription expenditure was publicly-financed.28 Public coverage varies from province to province. Most provinces offer special coverage for very low-income populations, and some provinces have adopted an age-based approach where individuals of a certain age (usually seniors) qualify for public prescription benefits (e.g. Ontario, Alberta, and the Maritime provinces).29 In such provinces, copayments are common.2 In contrast, other provinces have adopted an income-based approach to prescription drug coverage (e.g. BC, Saskatchewan, Manitoba).29 In these provinces, annual deductibles – out-of-pocket payments in prescription costs based upon a percentage of household income – are required to be paid prior to public coverage for prescriptions.2,29 The proportion of income that is set as the deductible ranges from 0% to 13% depending on the province.2,29 In BC, this percentage is set as 0% to 3% depending on year of birth and income.30 Other forms of payment are required for medications which are not listed on the provincial formulary (and therefore not covered by public plans), for copayments and deductibles, and for individuals who are not covered under public plans.2 These payments are either made through private insurance or directly out-of-pocket. Private insurance is obtained most commonly through employers, which may also include family members as beneficiaries.2 7 Approximately 10% of private insurance policies are taken out independently.31 Together, it is estimated that two-thirds of Canadians have some access to a private insurance policy.2 Of the 57.4% of total drug expenditure paid for privately, 35.2% is financed through private insurance and 22.2% through direct out-of-pocket payments.28 Private insurance plans are not equally available to all segments of the Canadian population. Higher income and healthier individuals are more likely to hold private insurance coverage, while those who are older, of lower income, and use less prescription medications are more likely to be uninsured or depend on public insurance.14,32,33 Retirees may also hold employer-sponsored health insurance as part of a retirement package;34 Allin et al. estimated that, in 2005, approximately 27% of Ontario residents over the age of 65 receive private insurance from their current or previous employer.14 However, there is much speculation that this benefit is on the decline due to the increased cost of running these plans for employers.34 1.3 Research Objectives The first study in this thesis examining the impact of deductibles on older adults was part of a larger project to examine BC’s current PharmaCare program. Through this project, a study was published recently that examined the impact of the income-based deductible imposed on older adults at the population level.9 This would have included individuals with a clear need (i.e. requiring regular medication use) and those who may not have a need (i.e. those who do not require regular medication use). In contrast, this thesis is focused on the impact of these deductibles on a select group of individuals – those with the chronic illness COPD. The motivation behind choosing a COPD population is two-fold: a) to study the effects of income-based deductibles on older adults with COPD due to their often higher prescription costs and risk of making trade-offs in their treatment; and b) to assess whether cost is the sole motivator in this population given the chronic nature of symptoms, the periodic occurrence of exacerbations, and its relationship with an individual’s morbidity and comorbidities. 8 The main focus of the first study was to examine differences in prescription use patterns and health outcomes for seniors with COPD covered by two different deductible schemes under Fair PharmaCare. The two research objectives for this portion of the thesis were to: 1. Describe the impact of deductibles on prescription drug use amongst a population of older adults living with COPD; and 2. Describe the impact of deductibles on health outcomes – that is, frequency of acute COPD exacerbations, physician visits, and hospitalizations – amongst this population. For the second study, the focus shifts to private financing for non-publicly covered medical services and treatments, including for prescriptions. An examination of trends in employer-sponsored health insurance (EHI) is warranted given a lack of current Canadian evidence on the topic. Older adults, who may require these services and treatments more frequently compared to younger individuals, are particularly vulnerable to changes to this supplemental help. Previous Canadian data only provided a snapshot of how employers have been changing EHI policies and it remains unclear how individuals are affected by the actions of employers.17 This study examined the impact over time at the individual-level. These results can also help inform current policies governing the coverage of services and treatments. Using Canadian Community Health Survey data amongst the Ontario population, the research objective for this study were to: 1. Describe the prevalence of EHI amongst retirees and the characteristics of those who do or do not have EHI; and 2. Describe the change in availability of EHI amongst retirees over the period of 2005 to 2014. 1.4 Thesis Outline This thesis is presented in four chapters. Chapters 2 and 3 include the full background and rationale, methodology, results, and interpretations of the two studies described above. Chapter 2 presents the study of deductibles imposed on older adults living with COPD in British Columbia, Canada, made possible using administrative datasets. Chapter 3 describes the trends in EHI in Ontario, Canada, using data from the Canadian Community Health Survey. These two chapters aim to describe the impact of an existing public drug coverage scheme and the change 9 in availability of supplemental help through EHI to fill in the coverage gaps left by public insurance. Observations from investigating these important influencers, in both the public and private insurance systems, on prescription drug use are summarized in Chapter 4. Potential implications of these results on prescription drug use in older adults with or without a chronic illness and suggestions for future research and policy direction are presented. 10 2 Impact of income-based deductibles amongst older adults living with chronic obstructive pulmonary disease in British Columbia 2.1 Background Chronic obstructive pulmonary disease (COPD) is a progressive disease primarily affecting adults over the age of 40.10 It is associated with chronic airway inflammation resulting in structural damage of the lungs and narrowing of airways.10 The most prominent risk factor is smoking, but is also associated with air pollution and occupational exposure.10 These contribute to airflow obstruction and, with more severe disease, hypoxemia.10 The main symptoms of COPD include progressive dyspnea, chronic cough, sputum production, wheezing, and chest tightness.10 It is a leading and growing cause of morbidity and mortality globally.10 In Canada, COPD is a major public health burden as it is the most common medical reason for hospitalization,18 costing $445M in 2012-2013 in hospital care alone.35 Despite long-standing, evidence-based guidelines to manage the condition and reduce acute exacerbations, the under-treatment of COPD with drug therapies remains a major global problem.10,11 COPD medication can largely be categorized as rescue or maintenance therapies, especially for patients with moderate to severe disease. Rescue medications are those which are intended to provide relief of acute symptoms and are also used, at an increased dose, to treat acute exacerbations. Maintenance medications are prescribed to manage ongoing symptoms and also to reduce the risk of acute exacerbations.10,12 These maintenance therapies, however, can be quite costly as each medication, excluding pharmacy markup and dispensing fee, ranges in cost from approximately $40 to $150 per month.10,12,25,13 These large upfront payments can be particularly difficult to afford for lower-income individuals, who are more likely to report cost-related non-adherence, and may lead to inappropriate use of the cheaper, rescue medications.10,12,26 11 Adherence to COPD medications has multiple causes, including knowledge of disease, inhaler technique, comorbidities, and patient preferences.11 It has been suggested as well that out-of-pocket costs to patients are an important factor influencing adherence to preventative COPD therapies. This view is supported by substantial national and international evidence suggesting that out-of-pocket drug costs are barriers to access for patients, resulting in reduced medication use and negative impacts on health.5,7,36 Non-adherence to preventative COPD therapies may help explain the considerable costs incurred and health resources utilized. In summary, out-of-pocket cost to patients is one of the most clinically important factors influencing adherence to COPD therapies11 and can potentially worsen outcomes and increase health service utilization. Thus, any changes to drug benefits, which may decrease public drug expenditures, requires consideration of potential issues with regards to access to medication and the subsequent long-term economic impacts across the health care system. This is particularly important for low-income seniors who do not qualify for social assistance and are still required to pay out-of-pocket for medications. While a study was published recently examining the impact of deductibles on older adults at the population level,9 it remains unclear what the impact is of deductibles on individuals with a chronic condition facing relatively high drug costs, such as individuals living with COPD. Thus, the purpose of this study was to determine the impact of deductibles on elderly COPD patients in British Columbia (BC) with regards to drug usage, health outcomes, and health care system burden. 2.1.1 BC context Prior to 2002, low-income seniors in BC received full coverage for prescriptions.25 During the transition period in 2002, these individuals paid a fixed copayment of $10 per prescription until total out-of-pocket expenditure reached $200, after which they received full coverage.25 Starting on May 1, 2003, BC introduced an income-based catastrophic public drug benefits program, Fair PharmaCare.25 Facing rising drug costs, this new program was created to help curb provincial public drug expenditures.37 This plan offers coverage to individuals who have had a high level of out-of-pocket spending. Under the new program, annual deductibles – the amount in eligible prescription cost to be paid in full prior to partial payment assistance – and “family maximums” – the amount to be paid within the household prior to complete coverage of 12 eligible prescriptions – are determined based on predefined income categories (See Figure 2.1).25 Those in the lowest income category have no deductibles while the deductible for others are set between 1-3% of household income; family maximums are set between 1.25-4% of household income.25 Additionally, households with a member who was 65 years of age or older when the program was implemented (i.e. born in or before 1939) also qualify for “Enhanced coverage”, which has lower deductibles, coinsurance during the period of partial coverage, and family maximums compared to those born after 1939, who are covered under “Regular Coverage” (See Figure 2.1).25 For example, an individual born in 1939 with a household income of $35,000 would face an annual deductible of $350, while a similar individual born in 1940 would have to pay $1000.30 This has created a natural experiment to study the impact of higher out-of-pocket costs on health outcomes in two otherwise very similar populations. Patients in the province of BC might be at particularly high risk for cost-related non-adherence to medications for COPD. For example, a study using national data from the 2007 Canadian Community Health Survey observed the highest rates of cost-related non-adherence in BC.8 This was suggested to be attributable to either the high deductibles set out by Fair PharmaCare or the high debt burden of residents.8 Dormuth et al., in the year following Fair PharmaCare’s implementation, studied the impact of this change of coverage in low-income seniors from full payment to a fixed copayment per prescription, then to the current income-based deductible scheme a year later. In the study, patients under Enhanced coverage with a new diagnosis of COPD or asthma (a disease which uses the same inhaled medications) were 25% less likely to initiate treatment compared to the two previous coverage regimes,6 suggesting that cost is an important factor in the uptake of these medications in seniors. However, firstly, the authors studied the effect of two subsequent policies which increased cost-sharing, rather than the use of income-based deductibles specifically. Secondly, this study considered the impact on all patients covered under Enhanced coverage – that is, regardless of whether the individual actually had a deductible. As such, the impact which income-based deductibles have on medication use and health service utilization in this population remains unclear. 13 Many maintenance COPD medications also require a so-called “special authority” in order for the medication to be covered. To be approved for these limited-coverage medications to contribute towards the annual deductible and be eligible for coverage, physicians must make a one-time application to the Ministry of Health. For maintenance COPD medications, prescriptions from respirologists are automatically exempt and no special authority is required. However, this may mean that subsequent prescriptions from other types of prescribers will require this application prior to coverage. The usual criteria for approval requires that patients have a confirmed diagnosis of COPD and have failed on previous therapy (generally short-acting inhaled beta2-agonists or antimuscarinics). For many first-line medications which are effective and indicated for use in COPD (especially long-acting inhaled beta2-agonists, or LABAs), however, COPD is excluded as a required diagnosis for coverage. This requirement presents as an additional barrier to access for patients, particularly if patients or their physicians are not familiar with the coverage criteria. Additionally inhaled corticosteroids (ICS) – first-line treatment for asthma but adjunctive therapy for COPD10 – are covered automatically without special authority. This may result in unintentional overprescribing and overuse of this class of medication, which has been associated with an increased risk of pneumonia and a general increased risk of health care utilization.10,38 2.2 Methods 2.2.1 Data sources This research linked three existing, de-identified, population-based administrative databases available through Population Data BC, which includes individual-level data on virtually all British Columbians who have utilized any of the services captured.39 The following databases were linked together: the BC Medical Services Plan (MSP) database, which captures all fee-for-services outpatient physician encounters, including reasons for encounters and expenditures;40 the Discharge Abstract Database, which provides information on hospital admissions, reason for stay, and length of stay;41 and the BC PharmaNet system, which records all prescription dispensations in the province, including amount paid for by the province.42 Consolidation files were also used to obtain demographic characteristics and establish household 14 members.43 Those who receive drug benefits through federally administered drug benefits plans (e.g. First Nations and veterans) were excluded from this analysis as we did not have access to their data (less than 5% of the population).44 2.2.2 Study cohort All individuals born between 1923 and 1956, who were registered with MSP and Fair Pharmacare for at least one full calendar year between January 1, 2004 and March 31, 2015 were eligible. All individuals who were born in or before 1939 with Enhanced Coverage were eligible for inclusion into the pre-1939, no-deductible group while those who were born after 1939 and covered under Regular Coverage are included in the post-1939, deductible group. To account for changes in income and level of coverage, person-years where an individual changed income groups mid-year were excluded. The unit of observation was the person-year. The study cohort was limited to person-years with a household income between $14,000 to $30,000 for individuals covered under Enhanced coverage (no deductible) and between $15,000 to $30,000 for individuals covered under Regular coverage (deductible of 2% of household income). As it is possible for an individual born after 1939 to receive Enhanced coverage by virtue of living with an individual born in or before 1939, the analysis was limited to the oldest individual in each household. Outliers – defined as individuals in the top 0.1% in terms of overall drug expenditure – were removed from the analysis. The study cohort was further restricted to individuals with COPD, defined over a 12-month period as those with two or more physician billings with COPD being the main reason for visitation (ICD-9 code of 491, 492, 493.2x and/or 496) and/or one or more hospitalizations with COPD indicated as a diagnosis, regardless of type (ICD-10CA code of J41, J42, J43, or J44).45 This case definition was based upon a definition previously validated by Gershon et al., which used the same clinical criteria over a 24-month period.45 While the 24-month definition has a specificity of 91.5%,45 this study attempts to capture individuals with moderate-severe disease, who would require use of the higher-cost, maintenance inhalers, thus enabling an examination of whether there were differences attributable to the deductible in rates of rescue inhaler use compared to maintenance inhaler use. Sensitivity analysis was also performed using the original definition as developed by Gershon et al. of two COPD-related outpatient encounters or one 15 hospitalization over a 24-month period.45 To determine the appropriateness of the changed definition, we compared the study sample which was derived from each definition in terms of the sample size obtained and the proportion of person-years utilizing any medications, as well as COPD-related medications. 2.2.3 Study period The study period was from Jan 1, 2004 to Dec 31, 2013. Individuals who met the case definition between Jan 1, 2003 to Dec 31, 2012 were included in the analysis. This allowed individuals who have had enough COPD-related health service use to be included in the first year of analysis. This also resulted in the exclusion of individuals meeting the definition in the last year (2013), as this would not have yielded a complete year for analysis. We identified the first occurrence of two COPD-related health service records within a 365-day period and used the earlier of the two dates as the “entry date”. We subsequently identified all eligible person-years by identifying all years following the year of entry where the individual was also registered for both MSP and Fair PharmaCare. 2.2.4 Outcomes of interest We studied the impact of income-based deductibles on our COPD cohort, in sequential order from a patient receiving a prescription to changes in health status, by examining outcomes within the following five categories: 1) PharmaCare expenditure: We calculated the average expenditure per person-year by PharmaCare. These averages were calculated by overall expenditure, expenditure on COPD-related medications, and by specific categories of COPD medications (See Table 2.1 for classification of COPD medications). We then estimated the impact of Enhanced coverage on the availability and extent of public drug payments for each drug category. The accepted drug cost, excluding dispensing fees, was used to calculate PharmaCare expenditure. The Statistics Canada Consumer Price Index was used to adjust for inflation. Results are presented in 2015 dollars. 16 2) Prescription drug use and cost: The average expenditure and number of prescriptions dispensed per person-year were calculated in terms of overall drug usage, usage of COPD-related medications, and usage of specific categories of COPD medications. We estimated the impact of having a 2% deductible on prescription utilization using these categories. The submitted drug cost, excluding dispensing fees, were used to calculate average expenditure. All expenditures were adjusted to 2015 dollars. We also estimated the average number of unique drugs dispensed per person-year based on level-seven Anatomical Therapeutic Chemical Classification (ATC-7) codes to estimate the impact of deductibles on the number of unique medications used. Drugs were categorized as follows for analysis: all drugs, COPD-related drugs, acute medications, maintenance medications, long-acting antimuscarinics (LAMA), ICS, and LABA/ICS combination (See Table 2.1). 3) Special authority availability: The proportion of prescriptions which had special authority was calculated for the drug categories LAMA and LABA/ICS combination. The proportion of maintenance medication prescriptions with special authority was also calculated. This ensured that the differences in total expenditure and PharmaCare expenditure were not due to lack of special authority. 4) Outpatient exacerbation medication usage: We calculated the average expenditure and number of prescriptions for prednisone and respiratory antibiotics commonly used for acute exacerbations of COPD (AECOPD) (See Table 2.2). This was to estimate the impact of deductibles on the frequency of AECOPD. 5) Physician visits and expenditure: We calculated the average number of visits and physician expenditures per person-year, adjusted to 2015 dollars. These were calculated for COPD-specific visits, as well as visits for any other indication. Multiple visits to the same practitioner on the same day were considered to be one unique visit. 6) Hospital admissions and length of stay: The average number of hospital admissions and the average length of stay of visits were calculated. Hospitalizations where an individual was 17 transferred from one hospital to another were considered as one stay. Hospitalizations were categorized as any cause and COPD-related hospitalization. 2.2.5 Statistical analysis Regression discontinuity analysis, one of the strongest quasi-experimental research designs, was used to observe the impact of deductibles on the outcomes of interest.46,47 It utilizes the quasi-random nature of assignment at the 1939 birth year into one of two coverage schemes to derive causal estimates of the impact of the program. A major assumption of this method is that individuals on either side of this threshold at the 1939 birth year are similar in terms of other characteristics (sex, comorbidities, smoking rate, and socioeconomic factors). If these assumptions are true, any abrupt differences in the outcomes of interest observed at the threshold can be attributed to the deductible scheme prescribed by the Fair PharmaCare program. To test this assumption, estimates at the threshold were also produced for the proportion of females and average household size. The proportion of individuals in “low”, “middle”, and “high” income categories were also examined to test if there are changes at the threshold within each category, the presence of which may bias the analysis. As the income groups do not align across the threshold, household incomes for the no-deductibles group were categorized into $14,000 to $17,999 for “low” income; $18,000 to $25,999 for “middle” income, $26,000 to $30,000 for “high” income. Those in the deductibles group were categorized as $15,000 to $18,749 for “low”; $18,750 to $26,249 for “middle”; and $26,250 to $30,000 for “high”. We also tested our assumption that the presence of comorbidities was smooth across the threshold by using the Charlson Comorbidity Index (CCI), a commonly used method to describe and control for patient clinical characteristics in clinical research.48 We calculated the CCI for each person-year studied by examining the MSP billings and discharge abstracts for the person-year in question. Each of the 17 categories of comorbidity which contributed to the index was flagged using the ICD-9 and ICD-10 coding algorithms developed by Quan et al. to be used in administrative datasets.49 To test our assumption, we produced estimates at the threshold for the proportion of individuals with a CCI in the following categories: 0, 1, 2, and 3+. 18 To produce statistical models, averages for each of the outcomes of interest were produced for each birth year cohort using data from each person-year identified above. Using these averages, linear regression models were fitted for each outcome with six terms: 1) an intercept term, 2) the slope, which represents the incremental effect of each year from 1923 onwards, 3) the square of this term to capture non-linear trends in the outcome, 4) an indicator variable to indicate “post-1939” observations, 5) the slope for post-1939 observations, which represents the incremental effect of each year from 1940 onwards, and 6) the square of this term to capture non-linear trends. 2.3 Results 2.3.1 Cohort characteristics After exclusions, our cohort consisted of 39,320 unique individuals contributing a total of 131,331 person-years for analysis. Individuals contributed a mean of 3.34 years [standard deviation (SD)=2.36]. The mean age of the cohort was 75.7 years old (SD=7.51), while the mean age of when individuals first entered the cohort was 71.9 years old (SD=7.54). In testing our assumption that there were no significant differences in the makeup of the cohort across the 1939 threshold, we found no significant differences in terms of proportion of females nor average household size over the threshold (p=0.22 and 0.87, respectively). There was a slight increase in the proportion of individuals in the “low” income category across the 1939 threshold (3.5%; 95% confidence interval [CI] 0.3% to 6.8%), which is likely due to the difference in design of income intervals between Enhanced and Regular Coverage. The proportions across the threshold remain similar for the “middle” and “high” income categories (p=0.30 and 0.26, respectively), where the majority of the study population was included. We also found no statistically significant differences across the threshold in the proportion of individuals with CCI scores in each of the following categories: 0, 1, 2, or 3+ (p=0.78, 0.49, 0.56, and 0.98, respectively). See Table 2.3 for the number of observations used for each of the outcome variables below. 19 2.3.2 PharmaCare expenditure We found that the income-based deductible imposed on those enrollees of Fair PharmaCare born after 1939 living with COPD resulted in a large drop in both the number of prescriptions covered by PharmaCare and the amount paid by PharmaCare. Our estimates at the 1939 threshold showed that there was a drop of 9.44 prescriptions paid for by PharmaCare (95% CI -14.66 to -4.22; p=0.001), a decrease of 19.7% (See Figure 2.2). Average annual expenditure by PharmaCare was decreased by $335.21 (95% CI -$502.77 to -$167.64; p=0.0003), a decrease of 18.4% (See Figure 2.3). For COPD-specific medications, there was a decrease of 1.04 prescriptions paid by PharmaCare (95% CI -1.52 to -0.55; p=0.0001) with a drop in expenditure of $68.83 (95% CI -$110.62 to -$27.04; p=0.002), representing a decrease of 15.1% in overall expenditure on COPD-related medications (See Figure 2.4). Similar decreases were seen in both the rescue and maintenance medication categories: the decrease in expenditure by Pharmacare across the threshold was 18.2% for rescue medications and 14.4% for maintenance medications (See Figures 2.5 and 2.6); and the decrease in the average number of prescriptions paid were 22.6% and 17.6%, respectively (See Figures 2.5 and 2.6). In examining the individual drug categories, there were no statistically significant differences, with the exception of ICS, where the decreases in expenditure and number of prescriptions covered for the deductibles group were 32.0% and 30.9% respectively – the largest decrease in all categories examined. 2.3.3 Prescription drug use and costs While the difference in the extent of public coverage was apparent, this did not affect overall prescription usage, as measured by total prescription expenditure (including both public and private spending), average annual number of prescriptions used, and average annual number of unique molecules. Figures 2.2 and 2.3 represents the results of total drug expenditure and the average number of prescriptions dispensed, respectively. These results were similar for the COPD-related drugs category and COPD drugs categorized as rescue or maintenance drugs (See Figures 2.4, 2.5, and 2.6). In examining individual COPD drug classes, we also found no statistically significant change in any of the classes. However, it is worth noting that in the ICS category, the decreases at the threshold in overall expenditure and number of prescriptions was near statistical significance (0.09 and 0.20, respectively; See Figures 2.7 and 2.8). 20 2.3.4 Outpatient acute exacerbations medication use and costs In examining the impact of deductibles on health status, we first examined use of outpatient exacerbation medications. We found a statistically significant decrease at the 1939 threshold in average PharmaCare expenditure and average number of prescriptions covered for both prednisone and respiratory antibiotics. Figure 2.9 shows that there were no statistically significant differences across the threshold in the average number of prescriptions for prednisone (estimate -0.03; 95% CI -0.20 to 0.15; p=0.76). For antibiotics, we found no difference in the average number of prescriptions across the threshold (See Figure 2.10). Interestingly, there was a small, though not statistically significant increase in the average total expenditure on respiratory antibiotics at the threshold, as shown in Figure 2.11 (estimate $6.49; 95% CI -0.18 to 13.16; p=0.06). This suggests that while rates of antibiotic use are similar across the threshold, those in the deductibles group may be using more expensive molecules. 2.3.5 Physician visits and cost We found no differences at the 1939 threshold in either the average number of physician visits nor in average physician expenditure (p=0.69 and 0.98, respectively). The results are presented in Figures 2.12 and 2.13. These results persisted for visits categorized as COPD-related or non-COPD related visits (results not shown). 2.3.6 Hospital admissions and length of stay Similar to the indicators for physician utilization, we did not find meaningful differences in either average number of hospital stays nor the average number of days in hospital amongst our cohort. Figure 2.14 shows the average annual number of admission while Figure 2.15 shows the average annual number of days in hospital, with no change apparent at the 1939 threshold (p=0.67 and 0.85, respectively). It is also worth noting that there was a very small, near statistically significant decrease at the threshold for the average annual number of hospitalization where COPD was the primary reason for hospitalization (estimate -0.01; 95% CI -0.02 to 0.00; p=0.06). This effect, presented in Figure 2.16, is in the opposite direction of what would be expected if indeed deductibles on prescription drugs were a barrier to accessing prescription drugs and negatively impacted health status. 21 2.3.7 Sensitivity analysis In the sensitivity analysis to determine the appropriateness of our altered definition of COPD, using the original Gershon et al. definition yielded a slightly larger sample size of 138,274 person-years (5.3% increase). The original definition produced a study sample with a smaller proportion of person-years utilizing COPD-related medications compared to the definition used in this study (56.6% vs 57.6% in our altered definition) while the proportion of person-years using any prescription medications was 88.7%, the same as in our altered definition. Given the small difference in proportion of person-years using any prescriptions, we feel confident that our altered definition identified less false positives and yielded a population with more severe disease, who are more likely to require use of higher cost inhalers. 2.4 Interpretation The current study utilized ten years of administrative data from British Columbia’s public health care system to determine the impact of a modest deductible of 2% of household income on older adults with household incomes between $15,000 and $30,000, living with chronic obstructive pulmonary disease. While public drug expenditure and the average number of prescriptions which had received public subsidy decreased, we found that the deductible did not impact overall prescription drug use, either in terms of overall expenditure or number of prescriptions dispensed. Additionally, we did not find evidence of medication usage patterns differing between the no-deductible and deductible groups. Specifically, regarding COPD treatment, we did not find evidence that the deductible caused lower rates of use of “maintenance” medications and higher rate of use “rescue” medications – a common cost-related non-adherence coping mechanism amongst patients with asthma and suspected in patients with COPD.11,26 Moreover, it did not appear to affect the rates of use of exacerbation medications, use of physician services (either in terms of expenditure or number of visits), or hospitalization. The lack of statistically significant differences on these outcomes suggests that income-based deductibles did not produce clinically significant differences in these outcomes. 22 COPD is a high-burden chronic illness where the role of costly “maintenance” medications to alleviate daily symptoms, slow progression of disease, and decreased risk of acute exacerbations is well-established.10,11 This finding confirms that, even in patients with a definitive need (i.e. a diagnosis of COPD) and in face of potentially high costs for each prescription, imposing this modest deductible produces cost-savings in the public system, while not appearing to impact prescription in utilization or health service utilization in this income group. The potential confounders of sex, income, and comorbidities were not found to be significantly different across the threshold, and were unlikely to have biased our results. This finding is in line with previous research done by our research group, which found that this deductible did not affect prescription use and health service utilization in the general population for the same income group.9 The results suggest that adding a modest deductible of 2% of household income in addition to copayments did not have any obvious unintended consequences. 2.4.1 Overall use of COPD-related medications It must be noted that the overall use of COPD-related medications was very low. While 88.7% of person-years in the study cohort had filled a prescription for any medication in the eligible year, only 57.6% of person-years (or 65.0% of unique individuals) had filled a prescription for COPD-related medications. 46.7% of person-years had at least one claim for rescue medications while 45.7% had at least one claim for maintenance medications. Though it is possible that some individuals may have disease of lesser severity and thus are controlled with regular use of a rescue inhaler, it seems unlikely to be the case here for a large majority of person-years. Stricter criteria than the one that had been previously validated by Gershon et al. was used to define our COPD cohort in an attempt to minimize false positives and identify individuals with more severe disease (warranting COPD-related physician consult twice in a 365-day span or one hospitalization with COPD listed as a medical condition). It is unclear what impact this low utilization rate had on the results of our study. Given that the proportion of person-years with COPD-related prescription did not change across the 1939 threshold, the impact of deductibles would likely have been similar if the utilization rates were higher. While most studies measure utilization rates using traditional methods of measuring adherence (e.g. proportion of days covered),50 Stuart et al. observed that, among Medicare-23 covered individuals with COPD in the United States, only 40% of observed person-years had filled a prescription for maintenance medications.51 This is similar to our observation of 45.7% person-years. A study by Haupt et al. found that, amongst Swedish patients with asthma or COPD, only 49% of patients had rescue or maintenance medications filled more than once a year.52 This suggests that the issues which influenced utilization were similar in this population compared to other studies of COPD populations and deductibles likely do not contribute to low utilization. Noting this, however, results from a 2012 study showed that British Columbians experienced the highest rate of cost-related non-adherence in Canada, with the authors suggesting the high out-of-pocket costs associated with Fair PharmaCare as a possible culprit.8 Due to the coverage structure of Fair PharmaCare, we were not able to determine whether the mere fact of having copayments – regardless of size - was prohibitive to obtaining a prescription for COPD-related drugs in this population. The fact that utilization rates for medications generally were not as low as that for COPD-related medications may be attributable to the high cost of many COPD-related medications. If cost is indeed a concern for these medications, our results would suggest that making the 25%-30% copayment for these drugs is still too costly for this population, even in the absence of a deductible. It is also feasible that the special authority structure inhibits obtaining COPD treatment. However, special authority is only imposed on maintenance medications and, given that rates of use of rescue medications and maintenance medications are similar, the special authority process alone likely does not explain the low utilization rate. 2.4.2 Rescue vs. maintenance medications We hypothesized that the deductible may make maintenance medications, which are comparatively more expensive, less affordable, leading to decreased utilization. As a consequence, the use of rescue medications may increase at the threshold. While we observed decreases in expenditure by PharmaCare and the average number of prescription paid across the threshold, this did not translate into increased utilization of rescue medications and decreased utilization of maintenance medications at the threshold. This trade-off is a widely cited phenomenon amongst asthma patients,53 though it has not been well-studied amongst the COPD patients. A recently published study by Fan et al. found evidence of overuse of rescue 24 medication in their small sample of 32 COPD patients but it is not clear if this is to compensate for underutilization of maintenance medications.26 The results of our study suggest that the imposition of an income-based deductible did not influence either the overuse of rescue medications, or the underutilization of maintenance ones. The increased use of ICS amongst the no-deductible group is worth noting, particularly since it has the least evidence of benefit amongst COPD patients.10,38 This increase may be an artifact of how maintenance medications are covered under Fair PharmaCare: all classes of maintenance medications for COPD, except for ICS, require so-called “special authority” from physicians prior to being eligible for coverage. This is contradictory to current COPD guidelines, which recommend for other long-acting bronchodilators to be used before using ICS as adjuvant therapy.10 Overuse of ICS amongst individual living with COPD is a well-documented phenomenon, with implications on health status and health care costs.38 Though the result was not statistically significant, the 17.5% increase in expenditure of ICS amongst the no-deductibles group may be clinically significant, particularly as ICS use is associated with an increased risk of pneumonia and other side effects.54 While studying the effects of special authority is outside of the scope of this analysis, further research in this area is warranted to determine the effects of a special authority program in conjunction with an income-based deductible system on the patterns of use of different types of medications under this program. 2.4.3 Comparison to current literature The findings from the current study are in contrast to studies from the early years of Fair PharmaCare, which compared the current program to two previous public programs in BC.6,7 The findings from Dormuth et al. suggested that the implementation of Fair Pharmacare’s deductibles resulted in decreased utilization of maintenance medications in patients with asthma or COPD,6 which led to increased emergency department visits.7 Because of the design of these studies by Dormuth et al., the results are an artefact of increased cost-sharing (i.e. policy changes from full coverage to some coverage, then to deductibles), rather than the income-based deductibles themselves. The results of these previous studies may not be surprising, given that over 40% of individuals in those studies had household incomes below $22,000. Older individuals have been found in various studies to be sensitive to prescription prices.14,55 For 25 example, Allin et al. found that further reduction of a few dollars in face of low copayments appear to incentivize seniors in using certain types of medications.14 These results echoed the body of literature which suggests that out-of-pocket cost adversely impact prescription use and subsequent health status.5,36 While our analysis showed that deductibles did not adversely affect prescription utilization and health status, the low overall utilization rate may indicate that incomes in this population are too low for individuals to afford more costly prescriptions, even when only a copayment needs to be satisfied and a deductible does not need to be met. Given the similar utilization rates as cited in COPD literature, however, there may be factors other than cost influencing utilization, such as patient education and preferences, treatment complexity, and ease of use of inhaler devices.10,11 2.4.4 Study limitations While this study does benefit from a strong quasi-experimental design, there are a few limitations which are noteworthy. Firstly, while we were able to estimate the impact of income-based deductibles at the threshold (i.e. individuals born in 1939 and 1940), the extrapolation and generalizability of these results may be limited to older adults with a current household income between $15,000 and $30,000. These results may not apply to younger adults, children, or vulnerable populations. Secondly, while PharmaNet captures the total drug cost and PharmaCare expenditure, it does not contain information on the availability of private insurance plans and the extent of coverage. These plans are important and may be protecting individuals from the deductibles and out-of-pocket payments. It is unlikely, however, that the trend in the availability or the extent of coverage would differ at the 1939 threshold, given that there were no changes in private insurance regulation which would affect the no-deductibles and the deductibles group disproportionately. Thirdly, data on Emergency Department (ED) visits which did not result in hospitalization were not available for analysis. A mild to moderate COPD exacerbation may result in an ED visit. These exacerbations may be treated with a short course of oral treatment(s) and hospitalization may not be necessary.10 Given that there were no differences in the rates of 26 use of medications used to treat acute exacerbations of COPD (a likely reason for a COPD-related ED visit), it is unlikely the case that rates of emergency department visits differed. Lastly, we did not link use of exacerbation medications to COPD-related physician visits. Particularly, prednisone may be used for multiple indications (e.g. arthritis, inflammatory bowel disease, dermatitis, etc.)56 and may also be prescribed to patients in anticipation of a potential AECOPD. Thus, our examination of these medications as a proxy for COPD exacerbation rate may be imperfect. While there may be an increasing trend in utilization of these medications with age, there is unlikely to be an abrupt change at the 1939 threshold, the presence of which would have biased our estimates. 2.5 Conclusions The results of this research confirm that a modest, income-based deductible resulted in substantial decreases in public expenditure across all drug categories for this population of older adults with COPD. Despite having a specific medical need and in the face of high out-of-pocket cost per prescription, the deductible did not result in changes in access to medications or change in utilization of physician services or hospitalizations in this specific population. As Canada is currently debating implementing a national strategy to increase the affordability and the accessibility to prescription medications, the results of this study present an alternate financing scheme for a public drug program if copayments are to remain a part of such a system. While the use of income-based deductibles and copayments did not have any apparent adverse outcomes on our cohort, the Fair PharmaCare policy does shift the cost burden from the public coffers to private payments through private insurance and out-of-pocket payments. This shift in cost may not be appropriate for vulnerable populations, but these considerations are outside of the scope of this study. As noted above and in international research,10,11 other interventions, such as education on COPD management and medication reminders, may be more important in this population of older adults living with COPD to ensure appropriate use of medications to decrease the risk of acute exacerbations. Future research could examine the appropriateness of copayments for COPD-related medications, as well as the effects of a special authority program 27 on access to medications for COPD and other chronic diseases, and whether this is of benefit to the population. 28 2.6 Tables and Figures Table 2.1 – Classification of medications related to chronic obstructive pulmonary disease. Classification for analysis Medication Class Molecules in Class Anatomical Therapeutic Chemical (ATC-7) Acute medications Short-acting beta2-agonist (SABA) salbutamol orciprenaline terbutaline R03AC02 R03AB03, R03CB03 R03AC03 Short-acting antimuscarinics (SAMA) ipratropium R03BB01 SABA/SAMA combination salbutamol/ipratropium R03AL02 Maintenance medications Long-acting antimuscarinics (LAMA) tiotropium bromide aclidinium bromide glycopyrronium bromide R03BB04 R03BB05 R03BB06 Long-acting beta2-agonists (LABA) salmeterol formoterol fumarate dehydrate indacaterol R03AC12 R03AC13 R03AC18 Inhaled corticosteroid (ICS) beclometasone budesonide ciclesonide fluticasone mometasone R03BA01 R03BA02 R03BA08 R03BA09 R03BA07 LABA/ICS combination budesonide/formoterol fumarate dehydrate fluticasone/salmeterol fluticasone/vilanterol mometasone/formoterol R03AK07 R03AK06 R03AK10 R03AK09 R03AK09 LAMA/LABA combination tiotropium/olodaterol umeclidinium/vilanterol R03AL06 R03AL03 PDE-4 inhibitor montelukast roflumilast zafirlukast R03DC03 R03DX07 R03DC01 Biologic omalizumab R03DX05 Other medications Methylxanthines aminophylline choline theophyllinate theophylline R03DA05 R03DA02 R03DA04 29 Table 2.2 - Respiratory antibiotics used in treatment of acute exacerbation of chronic obstructive pulmonary disease. Antibiotics Anatomical Therapeutic Chemical (ATC-7) amoxicillin & clavulanate J01CR02 cefprozil J01DC10 cefuroxime J01DC02 levofloxacin J01MA12 moxifloxacin J01MA14 clarithromycin J01FA09 azithromycin J01FA10 erythromycin J01FA01 sulfamethoxazole & trimethoprim J01EE01 doxycycline J01AA02 30 Table 2.3 – The number of person-years and observations used in analyzing each outcome variable in studying the impact of income-based deductibles amongst an older adult population living with chronic obstructive pulmonary disease (COPD). Number of unique persons (%) Number of Person-years (%) Number of observations Total 39,320 (100) 131,331 (100) - Prescription drugs All drugs 35,508 (90.3) 116,533 (88.7) 6,114,348 COPD-related drugs 25,570 (65.0) 75,643 (57.6) 819,523 Rescue drugs 22,419 (57.0) 61,292 (46.7) 383,548 Maintenance drugs 20,777 (52.8) 59,968 (45.7) 399,110 Inhaled corticosteroids (ICS) 10,246 (26.1) 22,923 (17.5) 85,698 Long-acting antimuscarinics (LAMA) 7,406 (18.8) 17,920 (13.6) 97,988 With special authority 4,960 (12.6) 11,923 (9.1) 64,347 With special authority (excluding respirologists) 4,313 (11.0) 10,136 (7.7) 52,222 Combination ICS/LABA 13,968 (35.5) 37,732 (28.7) 185,513 With special authority 11,088 (28.2) 31,474 (24.0) 161,329 With special authority (excluding respirologists) 10,365 (26.4) 28,598 (21.8) 139,937 Outpatient exacerbation medications Prednisone 12,923 (32.9) 25,029 (19.1) 87,523 Respiratory antibiotics 21,633 (55.0) 46,668 (35.5) 103,750 Physician utilization All visits 38,822 (98.7) 128,622 (97.9) 4,062,293 COPD-related visits 28,406 (72.2) 69,867 (53.2) 248,528 Hospitalizations All hospitalizations 24,928 (63.4) 45,595 (34.7) 77,617 COPD-related hospitalizations 5,850 (14.9) 8,186 (6.2) 11,265 31 Figure 2.1 - Structure of British Columbia's Fair PharmaCare program for the Enhanced Assistance and Regular Assistance plans. Within each income band and year of birth group, the pre-defined household deductible needs to be met prior to paying the coinsurance amount. The combined out-of-pocket expenditure from the deductible and coinsurance period contribute to the family maximum, after which eligible prescriptions would be fully covered. The highlighted groups is the study population for this analysis. Born in 1939 or earlier“Enhanced Assistance” Deductible: 0%Coinsurance: 25%Maximum: 1.25%Born in 1940 or later“Regular Assistance”$0$15,000$30,000$14,000Deductible: 0%Coinsurance: 30%Maximum: 2%Deductible: 0%Coinsurance: 25%Maximum: 1.25%Deductible: 2%Coinsurance: 30%Maximum: 3%Deductible: 1%Coinsurance: 25%Maximum: 2%Deductible: 3%Coinsurance: 30%Maximum: 4%Household Income (2 years prior)$30,000$032 Figure 2.2 - Average annual number of prescriptions for all medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of 9.44 less prescriptions assisted by PharmaCare (95% confidence interval -14.66 to -4.22; p=0.001), but a nonsignficant change in the annual average number of prescriptions dispensed (decrease of 2.88; 95% confidence interval -8.85 to 3.08; p=0.33). 33 Figure 2.3 - Average annual drug expenditure for all medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of $335.21 of assistance by PharmaCare (95% confidence interval -$502.77 to -$167.64; p=0.0003), but a nonsignficant change in annual average total expenditure of all dispensed prescriptions (decrease of $101.47; 95% confidence interval -$352.30 to $149.35; p=0.41). 34 Figure 2.4 - Average annual drug expenditure for COPD-related medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of $68.83 of assistance by PharmaCare (95% confidence interval -$110.62 to -$27.04; p=0.002), but a nonsignficant change in annual average total expenditure of all dispensed prescriptions (increase of $9.04; 95% confidence interval -$54.22 to $72.30; p=0.77). 35 Figure 2.5 - Average annual drug expenditure for “rescue” COPD medications between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of $16.50 of assistance by PharmaCare (95% confidence interval -$27.44 to -$5.56; p=0.005), but a nonsignficant change in annual average total expenditure of all dispensed prescriptions (increase of $4.78; 95% confidence interval -$9.41 to $18.97; p=0.50). 36 Figure 2.6 - Average annual drug expenditure for “maintenance” COPD medications between 2004 to 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of $50.06 of assistance by PharmaCare (95% confidence interval -$85.97 to -$14.15; p=0.008), but a nonsignficant change in annual average total expenditure of all dispensed prescriptions (decrease of $5.63; 95% confidence interval -$50.38 to $61.64; p=0.84). 37 Figure 2.7 - Average annual drug expenditure for all inhaled corticosteroids (ICS) between 2004 to 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of $23.61 of assistance by PharmaCare (95% confidence interval -$37.38 to -$9.84; p=0.002), but a nearly signficant decrease in annual average total expenditure of all dispensed prescriptions (decrease of $15.97; 95% confidence interval -$33.97 to $2.38; p=0.09). This represents a 17.5% decrease. 38 Figure 2.8 - Average annual number of prescriptions for all inhaled corticosteroids (ICS) between 2004 to for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of 0.21 less prescriptions assisted by PharmaCare (95% confidence interval -0.35 to -0.07; p=0.004), but a small, nonsignficant decrease the annual average number of prescriptions dispensed (decrease of 0.11; 95% confidence interval -0.28 to 0.06; p=0.20). This represents at 16.2% decrease. 39 Figure 2.9 - Average annual number of prescriptions for prednisone between 2004 to for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of 0.18 less prescriptions assisted by PharmaCare (95% confidence interval -0.33 to -0.03; p=0.02), but a nonsignficant change in the annual average number of prescriptions dispensed (decrease of 0.03; 95% confidence interval -0.20 to 0.15; p=0.76). 40 Figure 2.10 - Average annual number of prescriptions for respiratory antibiotics between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a significant decrease of 0.22 less prescriptions assisted by PharmaCare (95% confidence interval -0.32 to -0.13; p=0.00003), but a nonsignficant change in the annual average number of prescriptions dispensed (increase of 0.03; 95% confidence interval -0.07 to 0.14; p=0.50). 41 Figure 2.11 - Average annual drug expenditure for respiratory antibiotics dispensed between 2004 to 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in deductibles at the 1939 threshold led to a nonsignificant decrease of $3.67 of assistance by PharmaCare (95% confidence interval -$8.87 to $1.53; p=0.16) and a nearly signficant increase in annual average total expenditure of all dispensed prescriptions (increase of $6.49; 95% confidence interval -$0.18 to $13.15; p=0.06). 42 Figure 2.12 - Average annual physician expenditure on visits between 2004 and 2013 for British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in Fair PharmaCare deductible at the 1939 threshold led to nonsignificant decrease in expenditure of $1.40 (95% confidence interval -137.09 to 134.28; p=0.98). 43 Figure 2.13 - Average annual number of physician visits made between 2004 and 2013 by British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in Fair PharmaCare deductible at the 1939 threshold led to nonsignificant decrease of 0.38 in annual number of visits (95% confidence interval -2.32 to 1.56; p=0.69). 44 Figure 2.14 - Average annual number of hospital admissions for any reason between 2004 and 2013 in British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in Fair PharmaCare deductible at the 1939 threshold led to nonsignificant decrease of 0.01 admissions in annual number of admissions (95% confidence interval -0.05 to 0.08; p=0.67) 45 Figure 2.15 - Average annual number of days in hospital for any reason between 2004 and 2013 in British Columbians living with chronic obstructive pulmonary disease, by year of birth. The discontinuity at the change in Fair PharmaCare deductible at the 1939 threshold led to nonsignificant increase of 0.08 days in hospital (95% confidence interval -0.72 to 0.87; p=0.85). 46 Figure 2.16 - Average annual number of hospital admissions due to chronic obstructive pulmonary disease (COPD) between 2004 and 2013 in British Columbians living with COPD, by year of birth. The discontinuity at the change in Fair PharmaCare deductible at the 1939 threshold led to a nearly significant decrease of 0.01 admissions (95% confidence interval -0.02 to 0.00; p=0.06). 47 3 Changes in employer-sponsored health insurance amongst retirees in Ontario between 2005 and 2013-2014 3.1 Introduction In contrast to other universal systems, Canada has a health care system where physician and hospital services are fully publicly funded, while other services, including pharmaceuticals, dental, and vision care, have a mix of public and private payment.57 Private payments come from a mix of private insurance and out-of-pocket contributions and, as a result, private health insurance plays an important role in Canada in making costs manageable and enhancing access to these essential treatments.14,55,58 Approximately 60% of Canadians hold private health insurance.14 These are provided mostly by employers, which are subsidized through tax exemptions (except in Quebec) to encourage uptake of these plans.14 The availability of employer-based health insurance, or EHI, represents an important determinant for access to these other types of health care. As dental care is almost entirely privately-funded, retirees are particularly vulnerable, since costs significantly increase with loss of EHI and incomes may be limited.59 Even for prescription medications, which are at least partially publicly-funded in most provinces, the availability of extra coverage to lower copayments has been shown to increase use of essential medications and may also increase physician utilization for individuals who might otherwise self-medicate.14,55,60 Thus, it is important to observe any prevailing trends in EHI coverage for Canadians in order to adapt policies and protect those most vulnerable. Prior studies have found substantial increases in out-of-pocket expenses from 1998 to 2009, with private health insurance (including EHI) premiums being prominent expenses.61 Additionally, a growing proportion of Canadian households are spending more than 10% of their income on health expenses.61 It is plausible that these increases may be a manifestation of changes in both the scope and availability of EHI, especially if changes made by employers 48 affect payment of services and treatments sought by plan beneficiaries. Out-of-pocket costs at the point of service can present as significant barriers in accessing these treatments, potentially resulting in poorer health outcomes and increased expenditure elsewhere in the health care system.5,7,36,55,58,59,62 The landscape of EHI may be changing for Canadians. While Canadian EHI data are limited, data from the United States (US) consistently show that EHI availability has been declining for over a decade.15,16 Though the role of EHI for working Americans is different, its role for Americans over the age of 65 is similar to that for Canadians. These plans provide supplemental coverage for services, copayments, and deductibles not covered by US Medicare – a publicly administered insurance which covers medical services and pharmaceuticals for individuals aged 65 and over.15 Stuart et al. observed the greatest decline amongst new American retirees where, between 1996 to 2000 alone, the proportion of EHI-covered individuals aged 65 to 69 decreased by 7%.15 Surveys of employers on coverage for current and retired employees confirm that they are becoming less generous over time.15 Similar evidence was found in Ontario by Mercer in 2011, with employers citing the aging population and program costs as main reasons for limiting coverage.17 However, it is unclear what the study methodologies were or the number of employers surveyed, thus bringing to question the integrity of these results. There have been no incentives or policy changes in Canada in recent decades for either employers or employees to encourage EHI provision or uptake. If the trend in Canada is similar to that of the US, individuals retiring in the near future may experience a growing gap in coverage compared to their older counterparts, potentially making treatments too costly and inaccessible.15,16 The current study aims to address this knowledge gap of EHI trends in Canada by using data from a national survey to investigate the change in availability of EHI for retirees from 2005 to 2014. 49 3.2 Methods 3.2.1 Survey data and study design This study used data from the Canadian Community Health Survey (CCHS) – a cross-sectional survey conducted by Statistics Canada. The survey sample is derived from a multistage stratified cluster sampling design and is intended to be representative of approximately 97% of the population aged 12 and older. Exclusions include individuals who reside on Reserves and in certain remote regions of the country, are institutionalized, or are in the Armed Forces full-time. Additional information on the sampling and interviewing methods are available from Statistics Canada.63 The response rates were 76% for the 2005 cycle and 66% for 2013-2014.64,65 3.2.2 Study samples The survey samples used for the current analysis were obtained from the 2005 cycle (or Cycle 3.1) and the 2013-2014 cycle. The study sample was restricted to respondents who resided in Ontario at the time of interview, as it was the only province which included the optional survey module on the health insurance in more than one survey cycle, enabling a comparison over time. To capture retirees, respondents were included if they were aged 65 to 75 years old and responded that they had not worked at a job or business at any time in the past 12 months, or if they were 75 or older. The analysis excluded individuals who immigrated to Canada less than 10 years ago, to limit the number of individuals who arrived in Canada after retirement. Respondents who did not provide valid responses to the questions on job status or immigration status were excluded. 3.2.3 Variables for analysis The dependent variable of interest is a binary variable to describe whether the individual reported having EHI or not. This variable is derived from the optional health insurance module* in the CCHS that asked about coverage in four different areas: prescription medication, dental * While most responses regarding health insurance were provided in the insurance module, for some respondents in the 2013-2014 cycle, dental insurance information was provided in the dental module. For these individuals, this study used their responses regarding dental insurance from the dental module. Responses from all respondents in the 2005 cycle were all provided the insurance module. 50 care, eye glasses, and private/semi-private hospital room cost. An individual was flagged as having EHI if they reported having employer-sponsored insurance in any of these areas. Individuals who reported having either a source of coverage other than an employer or no coverage for all four areas were flagged as having no EHI. Individuals who provided invalid responses about employer-sponsored coverage in all four areas were excluded. In order to analyze the relationship over time, the explanatory variable was survey cycle – a binary variable differentiating the 2005 and 2013-2014 cycles. Potential confounders for the relationship between time and EHI coverage among retirees include categorical variables for age, sex, marital status, urban/rural setting, household income, highest level of education within the household, self-reported health status, and number of chronic illness(es). Age was categorized as four groups: 65-69, 70-74, 75-79, or 80+ years. Previous research has found that individuals with higher income are more likely than those with lower income to have private insurance policies.14 Household income was chosen over personal income for the analysis, as coverage may be available through a spouse. Marital status was also included for this reason, categorized as single, common-law, married, and widowed/separated. In order to compare between the two survey cycles, information from the income decile variable – derived based on all respondents in the province – was used and re-categorized as quintiles for analysis. Highest level of education within the household was categorized as less than high school, secondary school graduation, or at least some post-secondary. Self-reported health status was categorized in the survey into five levels, ranging from “poor” to “excellent”. The number of self-reported chronic illnesses was counted for each individual and categorized into four groups: no illnesses, 1-2 illnesses, 3-4 illnesses, or 5+ illnesses. Of the diseases which were assessed in the survey, the diseases that contributed to the count were asthma, arthritis, hypertension, chronic obstructive pulmonary disease, diabetes, heart disease, previous stroke, bowel disease, and mood disorder. 3.2.4 Analysis plan We calculated descriptive statistics to characterize the retiree sample and EHI in the two survey cycles. Population estimates and their variances were possible by applying the probability and bootstrap weights, respectively, in all statistical analyses. These weights were provided by Statistics Canada that account for the non-proportional sampling technique used.63 The 51 probability weights from the individual survey cycles were adjusted using the pooled approach in order to produce a single dataset to be analyzed.66 Combining survey cycles is feasible here, as the questions from which the variables for analysis were derived, survey coverage, and mode of collection had not changed.66 A multivariable logistic regression model was used to estimate the association between survey cycle as a surrogate measure for time and the proportion of individuals with or without EHI, adjusting for the confounders.67 Adjusted odds ratios (aOR) were calculated with 95% confidence intervals (CI). Predicted probabilities given individual characteristics were calculated using the point estimates from the multivariate analysis. To understand the significance of the change if it is present, we used Statistics Canada’s population estimates from 2014 to estimate the number of individuals who were potentially affected.68 3.3 Results 6,234 individuals from 2005 and 6,509 from 2013-2014 met the study criteria. This yielded a weighted study sample representing 479,192 individuals in 2005 and 455,072 in 2013-2014, for a total study sample of 934,265 individuals across the two survey cycles (See Figure 3.1). While respondents were equally distributed across the four age groups, 55.0% of respondents were female (See Table 3.1). About 61.0% of respondents reported at least some post-secondary education as the highest level of education attained by a member of the household. Respondents in 2013-2014 were in slightly better educated households compared to those in the 2005 cycle, with a slight increase in the proportion of individuals having graduated from secondary school from those who had not graduated. The majority of respondents reported having one or two chronic illnesses (58.6%), with 62.5% of respondents self-reporting as being of very good or good health. A higher proportion of respondents in the 2013-2014 cycle also self-report being in a better health state compared to those in the 2005 cycle, though this better self-rated health is not reflected in the number of chronic diseases reported. Approximately 57.1% of individuals reported a total household income within the first two quintiles. The majority of respondents resided in an urban community setting. About one-third of respondents reported having some form of EHI (32.8%; See Table 3.1), with 32.6 % and 33.1% of respondents receiving EHI in the 2005 cycle and the 2013-2014 cycle, respectively. In the final, pre-specified multivariate model adjusting for all potential 52 confounders available, the aOR estimate was 0.87 (95% CI 0.77-0.99, p-value=0.03; See Table 3.2). These results suggest that, all else being equal, the odds of having EHI decreased substantially in 2013-2014 compared to 2005. While other variables statistically significantly confound the relationship under investigation (See Table 3.2), the decrease in odds ratio estimates after adjusting for confounding is almost solely attributable to the household income variable. Individuals earning in the 2nd quintile have 2.71 times the odds of having EHI, compared to individuals in the 1st quintile in the adjusted analysis. Those earning in the 4th quintile have the highest odds of having EHI. Using estimates from the multivariate logistic regression (See Table 3.3), the absolute decrease in predicted probability of receiving EHI from 2005 to 2013-2014 ranged from 0.60% to 3.35%. To extrapolate using 2014 population estimates from Statistics Canada, this affects approximately 13,000 to 72,000 of Ontario’s 2,139,666 residents over the age of 65.68 The segment of the population with the lowest predicted probability of receiving EHI are older individuals, with lower levels of education and income in the household. In contrast, the population with the highest predicted probability of receiving EHI are those with 3 to 4 chronic conditions, with a household income in the 4th quintile. The larger relative decreases over time are amongst those with the lowest predicted probabilities of having EHI. 3.4 Interpretation Employer-sponsored health insurance remains an important mechanism through which many Canadians access important forms of health care. We utilized two cycles of the Canadian Community Health Survey and found that the adjusted rates of EHI coverage for retirees have declined over time. This study provides evidence that, at least in Ontario, that there has been a decrease in the odds of a retiree receiving EHI in 2013-2014 compared to in 2005, after controlling for confounders. These findings confirm that, much like the United States,15,16 the odds of a retired employee receiving EHI has decreased by 13% in the past decade in Ontario. Depending on personal characteristics, this represents a decrease of up to 3.35% in absolute probability of having EHI. From population estimates, up to 72,000 Ontario residents over the age of 65 were potentially affected by this trend. The public health implications of this finding 53 are important, as Canadians often rely on private insurance provided by employers in order to afford health treatments not covered under the Canadian Health Act.14,55,58 The results of this study provide an explanation to some of the observations described by previous research. Our results corroborate previous industry surveys conducted in the province that found that employers have or will increasingly limit EHI coverage.17 As these plans help offset the out-of-pocket costs for treatments which are not publicly funded,14,55,58 the observed decrease in coverage availability may be linked to evidence of individuals having increased out-of-pocket payments in order to obtain items such as dental services and prescription drugs.61 EHI is available to retirees often as part of a benefits package offered to the individual at the time of retirement.34 While availability may be one way of measuring changes in EHI, there are a variety of EHI options available to employers should they wish to decrease their cost in providing retiree EHI benefits. Options such as increasing employee-paid premiums and cost-sharing for services can help curb employer cost while still providing some EHI benefits. This may explain in part why prior studies have observed larger out-of-pocket cost increases on insurance premiums than out-of-pocket costs for treatments.61 The decision for an employer to cease providing EHI for retirees is likely a drastic one. This change is further deterred by precedent cases where small changes to the extent of coverage resulted in instances of legal disputes between retirees,34 making any changes or complete cessation of plans previously offered unlikely. As such, changes are more likely to occur with newer retirees without previously agreed upon retirement benefits. Coupled with the fact that employers providing EHI may claim their contributions as tax-deductible expenses,69 employers may instead choose to shift the cost onto employees while maintaining some coverage for newly retired individuals. Prior studies have found that overall (both for non-retirees and retirees), there had been an increase in the use of cost-controlling mechanisms amongst private, Canadian prescription drug plans, particularly in the form of increased deductibles and limited reimbursement on prescription dispensing fees.70 However, the population-level impact of these cost-saving mechanisms was unclear. 54 Ontario seniors receive generous public subsidy on prescription drugs under the Ontario Drug Benefits program.71 This program, with its low copayments likely decreases the burden of cost on employers and EHI plans. Prescription drug costs constitutes approximately one-third of expenses by private insurance plans available nationally – one of the largest portions compared to other health services.31 As such, the decrease in EHI availability may be more profound in other provinces where drug coverage for seniors is not as generous. As previously discussed, much of our finding is attributable to changes in the household income structure of retirees. Indeed, in examining the composition of the population under study in the two time periods, those in 2013-2014 reported earning in a higher quintile (relative to the entire province) compared to those in 2005. It has been found previously that private insurance availability (through an employer or otherwise) is associated with one’s income.14,33,59,62 Thus, with more individuals reporting higher household incomes, it may appear that EHI availability was maintained in the latter period. However, as the results suggest in the adjusted analysis, the odds of having EHI in fact decreased over this period, after taking into account income and other confounders. There are some limitations with the current study worth noting. First, the data is derived from two cross-sectional surveys and may be subject to recall bias. However, it seems unlikely that knowledge about EHI would have been different amongst the two cohorts. Second, we were only able to examine the association between EHI insurance and time by using two survey cycles and limiting the outcome variable to a binary variable. The results may oversimplify how EHI for retirees has changed over time. Given that discontinuing EHI is perhaps the most severe form of cost control, we feel these results provide a potentially important body of preliminary evidence that warrants further investigation. Future studies should investigate the proportion of retirees facing increased policy premiums or increased copayments for treatments.70 Third, the sample was restricted to examining Ontarians only. While we were only able to study the effects on the largest province in Canada, as incentivizing policies for employers to 55 provide EHI are applied nationally,69 the results observed in the current study are likely generalizable to the rest of Canada. Last, although data from the CCHS is cross-sectional in nature, it is still valuable in many regards. Firstly, it is a national, population-based survey conducted by Statistics Canada, which is renowned for the high quality of their sampling and data collection techniques. Second, the survey provides important information on many aspects of respondents’ health – from access, to health status, to important social determinants of health – while also allowing for confounders to be adjusted for in an analysis. Lastly, and most relevant to this study, information on private insurance is not readily available to the public. This previously cited issue has made it difficult to assess the current landscape of health care coverage for many essential services in Canada.14,55,57,61 3.5 Conclusion The current study provides valuable information regarding population-level EHI availability changes over time in Canada. The decrease in EHI availability presents as a significant public health issue, as cost-related non-adherence to medically necessary treatments may subsequently increase adverse health outcomes and hospital and/or physician use. This potential burden on the public system may provide impetus on policymakers to further study other important EHI trends in Canada such that appropriate policy action may be taken to protect the health of the population and the health care system. 56 3.6 Tables and Figures Figure 3.1 - Derivation of study sample from Cycle 3.1 (2005 cycle) and the 2013-2014 Cycle of the Canadian Community Health Survey, including exclusions due to missing/invalid responses. 2005 Cyclen=132,947n=8,998n=7,239n=6,474n=6,434n=6,2342013-2014 Cyclen=128,310n=13,122n=8,129n=6,815n=6,723n=6,509Total Study Sample n = 12,743 Weighted Study Sample n = 934,265 Respondents from Ontario, ≥65 years old Those with valid income responses/income not imputed Not worked in past 12 months (for those age 65-75) Non-immigrant or immigrated ≥10 years ago Valid responses to variables of interest 57 Table 3.1 - Characteristics of study sample - investigating the relationship between availability of employer-sponsored health insurance (EHI) and survey year; data from the combined Cycle 3.1 (2005) and the 2013-2014 cycle of the Canadian Community Health Survey. Study Sample by Survey Year Total 2005 (Cycle 3.1) 2013-2014 Cycle Weighted frequency Percentage (standard error) Weighted frequency Percentage (Standard error) Weighted frequency Percentage (Standard error) Total Study Sample 934,265 100 479,192 51.3 (0.73) 255,072 48.7 (0.73) Insurance availability No employer health insurance 627,455 67.2 (0,65) 323,043 67.9 (0.61) 304,412 66.9 (0.74) Have employer health insurance 306,810 32.8 (0.65) 156,150 32.6 (0.50) 150,660 33.1 (0.50) Have prescription coverage 257,584 27.6 (0.61) 129,195 13.8 (0.44) 128,388 13.7 (0.47) Have dental coverage 236,215 25.3 (0.59) 119,371 12.8 (0.44) 116,844 12.5 (0.46) Have eyeglasses coverage 233,992 23.0 (0.59) 117,961 12.6 (0.44) 116,031 12.4 (0.45) Have hospital room coverage 247,031 26.4 (0.60) 133,887 14.3 (0.47) 113,143 12.1 (0.44) Age 65 to 69 years 258,626 27.7 (0.68) 128,429 26.8 (0.48) 130,187 28.6 (0.58) 70 to 74 years 231,253 24.8 (0.63) 122,302 25.5 (0.43) 108,951 23.9 (0.52) 75 to 79 years 218,646 23.4 (0.59) 114,066 23.8 (0.43) 104,579 23.0 (0.46) 80 years or more 225,740 24.2 (0.60) 114,386 23.9 (0.45) 111,354 24.5 (0.46) Sex Male 420,238 45.0 (0.72) 207,369 43.3 (0.57) 212,869 46.8 (0.66) Female 514,026 55.0 (0.72) 271,823 56.7 (0.62) 242,203 53.2 (0.66) Urban/rural dwelling Rural 153,579 16.4 (0.43) 68,660 14.3 (0.27) 84,919 18.7 (0.33) Urban 780,686 83.6 (0.43) 410,533 85.7 (0.71) 370,153 81.3 (0.74) Total household income – provincial quintiles Quintile 1 283,233 30.3 (0.70) 168,447 35.2 (0.51) 114,786 25.2 (0.61) Quintile 2 250,340 26.8 (0.63) 131,512 27.4 (0.45) 118,828 26.1 (0.50) Quintile 3 186,714 20.0 (0.56) 86,534 18.1 (0.40) 100,180 22.0 (0.43) Quintile 4 138,111 14.8 (0.51) 60,824 12.7 (0.35) 77,287 17.0 (0.40) Quintile 5 75,866 8.1 (0.36) 31,875 6.7 (0.25) 43,991 9.7 (0.26) 58 Table 3.1 (continued) - Characteristics of study sample - investigating the relationship between availability of employer-sponsored health insurance (EHI) and survey year; data from the combined Cycle 3.1 (2005) and the 2013-2014 cycle of the Canadian Community Health Survey. Total Study Sample by Survey Year 2005 (Cycle 3.1) 2013-2014 Cycle Weighted frequency Percentage (Standard error) Weighted frequency Percentage (Standard error) Weighted frequency Percentage (Standard error) Highest level of education within household < Than secondary 204,336 21.9 (0.53) 118,025 24.6 (0.40) 86,312 19.0 (0.37) Secondary grad. 160,176 17.1 (0.62) 74,614 15.6 (0.34) 85,562 18.8 (0.56) At least some post-secondary 569,752 61.0 (0.70) 286,553 59.8 (0.66) 283,199 62.2 (0.68) Number of chronic illness(es) None 169,573 18.2 (0.55) 88,705 18.5 (0.41) 80,868 17.8 (0.42) 1 to 2 547,896 58.6 (0.71) 285,002 59.5 (0.64) 262,895 57.8 (0.70) 3 to 4 192,687 20.6 (0.56) 94,151 19.6 (0.38) 98,536 21.7 (0.46) 5 or more 24,108 2.6 (0.23) 11,335 2.4 (0.13) 12,774 2.8 (0.19) Marital Status Single/Never married 41,097 4.4 (0.26) 20,871 4.4 (0.17) 20,226 4.4 (0.20) Common-law 21,913 2.3 (0.25) 6,364 1.3 (0.09) 15,550 3.4 (0.23) Married 558,578 59.8 (0.69) 290,260 60.6 (0.66) 268,317 59.0 (0.70) Widow/Separated/Divorced 312,677 33.5 (0.66) 161,697 33.7 (0.47) 150,980 33.2 (0.54) Self-reported health status Excellent 120,993 13.0 (0.48) 55,032 11.5 (0.30) 65,961 14.5 (0.40) Very good 278,512 29.8 (0.64) 139,569 29.1 (0.47) 139,942 30.5 (0.52) Good 305,299 32.7 (0.70) 158,345 33.0 (0.51) 146,954 32.3 (0.60) Fair 162,263 17.4 (0.54) 89,677 18.7 (0.41) 72,586 16.0 (0.39) Poor 67,199 7.2 (0.38) 36,570 7.6 (0.25) 30,629 6.7 (0.30) 59 Table 3.2 - Results from Logistic Regression: The association between survey year (reference 2005 cycle) and the availability of employer-sponsored health insurance (yes/no). Unadjusted and adjusted odds ratios (aOR) with 95% confidence intervals. Bolded results denote statistical significance. Unadjusted Adjusted† Odds Ratios 95% Confidence Interval Odds Ratios 95% Confidence Interval Survey year 2005 cycle (Cycle 3.1) 1 1 1 1 2013-2014 cycle 1.02 0.91 - 1.15 0.87 0.77 - 0.99 Age 65 to 69 years 1 1 1 1 70 to 74 years 0.87 0.74 - 1.02 0.87 0.74 - 1.02 75 to 79 years 0.76 0.65 - 0.89 0.80 0.67 - 0.94 80 years or more 0.72 0.60 - 0.85 0.84 0.70 - 1.00 Sex Female 1 1 1 1 Male 1.23 1.09 - 1.38 1.01 0.89 - 1.15 Urban/rural dwelling Rural 1 1 1 1 Urban 1.04 0.91 - 1.19 1.36 1.18 - 1.56 Total household income provincial quintile Quintile 1 1 1 1 1 Quintile 2 2.88 2.41 - 3.44 2.70 2.26 - 3.25 Quintile 3 4.36 3.65 - 5.20 4.01 3.31 - 4.86 Quintile 4 5.73 4.65 - 7.07 5.20 4.18 - 6.48 Quintile 5 4.99 3.91 - 6.37 4.46 3.45 - 5.76 Highest level of education within household < Than secondary 1 1 1 1 Secondary grad. 1.69 1.42 - 2.02 1.31 1.08 - 1.58 At least some post-secondary 1.98 1.72 - 2.29 1.13 0.96 - 1.32 Number of chronic illness(as) None 1 1 1 1 1 to 2 0.92 0.79 - 1.07 1.01 0.86 - 1.19 3 to 4 0.86 0.72 - 1.03 1.18 0.96 - 1.45 5 or more 0.49 0.34 - 0.69 0.73 0.49 - 1.10 Marital Status Single/Never married 1 1 1 2 Common-law 1.82 1.09 - 3.03 1.42 0.83 - 2.43 Married 1.79 1.38 - 2.32 1.58 1.19 - 2.10 Widow/Separated/Divorced 0.89 0.68 - 1.16 1.04 0.78 - 1.39 Self-reported health status Excellent 1 1 1 1 Very good 0.88 0.74 - 1.05 0.93 0.77 - 1.12 Good 0.77 0.64 - 0.93 0.89 0.74 - 1.08 Fair 0.61 0.50 - 0.74 0.81 0.65 - 1.01 Poor 0.48 0.37 - 0.62 0.71 0.54 - 0.93 † Adjusted for age, sex, urban/rural dwelling, household income, highest level of education within household, number of chronic illnesses, marital status, and self-reported health status. 60 Table 3.3 – Predicted probability of receiving Employer-Sponsored Health Insurance in 2005 and 2013-2014 for individuals of certain characteristics; derived from estimates of logistic regression. Characteristics 2005 2013-2014 Absolute change Relative change 65-69, married, urban dwelling, 1-2 chronic illnesses, 2nd income quintile, some post-secondary, very good health Male 52.61% 49.26% -3.35% -6.37% Female 52.35% 49.00% -3.35% -6.40% 65-69, married, urban dwelling, 1-2 chronic illnesses, 4th income quintile, some post-secondary, very good health Male 59.02% 55.74% -3.28% -5.55% Female 58.77% 55.48% -3.28% -5.59% 65-69, married, urban dwelling, 1-2 chronic illnesses, 2nd income quintile, some post-secondary, very good health Male 42.84% 39.60% -3.25% -7.58% Female 42.59% 39.35% -3.24% -7.61% 65-69, married, urban dwelling, 1-2 chronic illnesses, 1st income quintile, some post-secondary, very good health Male 21.68% 19.49% -2.19% -10.10% Female 21.50% 19.33% -2.18% -10.12% 70-74, widowed, urban dwelling, 1-2 chronic illnesses, 1st income quintile, secondary school grad, very good health Male 14.39% 12.81% -1.57% -10.94% Female 14.26% 12.70% -1.56% -10.96% 75-79, widowed, rural dwelling, 1-2 chronic illnesses, 1st income quintile, secondary school grad, very good health Male 11.03% 9.78% -1.25% -11.32% Female 10.92% 9.68% -1.24% -11.33% 80+, never married, rural dwelling, 5+ chronic illnesses, 1st income quintile, < secondary school, poor health Male 5.02% 4.42% -0.60% -12.00% Female 4.97% 4.38% -0.60% -12.00% 61 4 Conclusion 4.1 Summary of Findings This thesis investigated the factors which may influence prescription drug use amongst older adults in Canada. The first study investigated the impact of deductibles in a publicly-funded system, specifically the deductibles imposed under British Columbia’s Fair PharmaCare program amongst older adults living with chronic obstructive pulmonary disease. The second study investigated the availability of employer-sponsored health insurance (EHI) for retirees over 65 years of age. 4.1.1 Impact of deductibles under British Columbia’s Fair PharmaCare amongst older adults living with chronic obstructive pulmonary disease (COPD) In chapter 2, “Impact of income-based deductibles amongst older adults living with chronic obstructive pulmonary disease in British Columbia”, we studied the impact of having to satisfy a deductible that is 2% of the household income versus having no deductible prior to public assistance of prescription medications using a regression discontinuity design. Similar to the previously published results from this project on all eligible older adults,9 we found that even in individuals with an identifiable need (i.e. having COPD, a chronic illness requiring long-term treatment), imposing this modest deductible did not impact prescription expenditure, average number of prescriptions, nor the use of particular classes of medications. While the previous study found a small increase in physician expenditure at the 1939 threshold,9 the current study did not find the deductible to influence physician visits, physician expenditure, or hospitalizations. We had hypothesized that the deductible would influence the types of medications used – specifically that the deductible would compel the post-1939 population to decrease use of expensive, maintenance medications to prevent acute exacerbations of COPD in favour of using less expensive rescue medications. We found no evidence of this. We did observe that use of ICS 62 was slightly higher in the zero deductibles group compared to the deductibles group. This may be related to the fact that ICS, unlike other maintenance medications, do not require “special authority” prior to them being eligible for coverage. While the increase was not statistically significant, the 17.5% increase in expenditure of ICS amongst the no-deductibles may be of clinical significance and warrants further investigation. It must be noted that the use of COPD-related medications in this population was very low – only 57.6% of eligible person-years filled a prescription for COPD-related medications – rescue or maintenance. Though medication utilization is more commonly measured by calculating medication possession ratios or proportion of days covered,50 this low annual utilization rate is similar to ones found amongst Swedish patients and US Medicare beneficiaries.51,52 The results of the current study suggest that adding a modest deductible of 2% of household income in addition to copayments did not impact non-pharmaceutical health care utilization in the population studied here. The low utilization rate of COPD-related medications suggests that there are other factors influencing whether or not such a prescription is filled by the patient. Given the importance of drug treatment for COPD, suggestions for future research will be offered in section 4.4.1 below. 4.1.2 Changes in employer-sponsored health insurance amongst retirees in Ontario Chapter 3, “Changes in employer-sponsored health insurance amongst retirees in Ontario between 2005 and 2013-2014”, is the first Canadian population-based evidence on employer-sponsored health insurance (EHI), which showed a decline in their availability for retirees. We used data available from two cycles of the Canadian Community Health Survey (CCHS) and used multiple logistic regression to estimate the odds ratio of a retired Ontario resident having EHI in 2013-2014 compared to 2005. Adjusting for a priori confounders, the odds ratio of receiving EHI in 2013-2014 compared to 2005 was 0.87. The shift in estimates was mainly attributable to household income, here measured in quintiles relative to the entire Ontario population surveyed. This may be due to the fact that respondents in 2013-2014 included in the analysis reported earning in a higher quintile relative to those in 2005. Thus, while the overall rates of EHI are similar over the two 63 time periods, this was driven by the fact that retirees have become wealthier relative to the entire province in the latter period. The decrease in odds of receiving EHI in the latter period was more profound amongst individuals who had lower odds of receiving EHI – that is, individuals with lower income, who are older, have never been married, and of poorer health status. While the analysis was limited to Ontario – the only province with data available over time – the results suggest that the much-speculated decrease in coverage availability in this population may be true. 4.2 Contribution to Current Literature There has been much debate in Canada on how to finance public prescription drug coverage.2,3 Most academics and policymakers agree that, since pharmaceuticals are considered a corner-stone of modern medicine, some level of coverage should be provided to minimize cost-related non-adherence, thereby also decreasing any downstream increase in non-pharmaceutical health care utilization, which is often more costly.1 Income-based deductible schemes, such as the one imposed in British Columbia under Fair PharmaCare, have been considered as a viable option, though the evidence on their impact is mixed.5–7 While a previous study using survey data found that British Columbians experience the highest rate of cost-related non-adherence,8 our study found that having to satisfy a deductible was not associated with changes in prescription utilization patterns amongst older adults living with the chronic condition COPD. Further, it was also not associated with changes in non-pharmaceutical health care utilization. This is similar to findings from the larger study published by Law et al., examining the same population regardless of diagnosis.9 The results of this study add to growing evidence and current debate on alternate financing options for proposed provincial and national pharmacare reforms. This study, as part of a larger study, helps to inform where deductibles may be appropriately utilized, allowing resources to be reallocated elsewhere to provide optimal health outcomes for the population while ensuring long-term sustainability of the insurance system. 64 This study of deductibles also found that medication utilization rates amongst this population living with COPD is very low. This may be interpreted in two ways: 1) if cost-related non-adherence is present, the fact that the deductible makes no impact on prescription utilization may be that cost-related non-adherence is experienced the same in both the no deductibles and the deductibles group; and 2) there are factors other than cost which drive prescription utilization in this elderly population living with COPD. While investigating the factors which drive this behaviour is beyond the scope of this thesis, the results of this study does seem to suggest that factors other than out-of-pocket cost are stronger predictors of medication use. The results of the second study on EHI may help inform the urgency to ensure seniors retain the ability to afford their prescription medications. Approximately one-third of Ontario’s retirees had access to EHI during the study period. While the decrease observed in the study is relatively small, it does impact a significant portion of the population and this decrease is expected to continue more drastically, with larger, national firms recently announcing their plans to cease offering retiree benefits.34 This would place a large burden on Canadian seniors, many of whom rely on these plans to make prescription drugs and other non-publicly funded treatments more affordable.59–11 Collectively, this thesis tells a tale of what has passed and what is to come: income-based deductibles have been used in British Columbia for the past decade, with no observable changes in prescription drug use or health care utilization. However, policymakers should be aware that EHI availability, which has played a large role in making prescription medications more affordable, has been declining and is expected to continue. This may create a coverage gap in older, lower income seniors. These results may be of interest to policymakers as well as health care professionals to further investigate this matter and determine whether policy action and interventions are necessary to prevent unnecessary increases in costlier, non-pharmaceutical health care utilization. Amidst current discussions around implementation of a national pharmacare strategy in Canada, the results presented in this thesis suggest that the strategy should keep in mind the decrease of supplemental help from EHI for retired individuals and the potential vulnerabilities this may create. As well, this thesis highlights the need for such a 65 strategy to be comprehensive in addressing, apart from cost, other important challenges in achieving optimal medication adherence in chronic disease management. 4.3 Strengths and Limitations 4.3.1 Deductibles study While the structure of Fair Pharmacare has allowed for us to use one of the most rigorous quasi-experimental designs, there are some limitations to this study. Firstly, the portion of the prescription cost not paid by PharmaCare may not be entirely out-of-pocket. It is possible that at least a portion of this cost is paid for by a private insurance plan, information which we did not have. However, in order to bias the estimates from the regression discontinuity design, individuals on either side of the threshold would have to have vastly dissimilar rates of holding a private insurance policy and/or the reimbursement scheme of these policies to be very different. This is unlikely to be the case. Secondly, while we had information on physician visits and hospitalizations, information on emergency department (ED) visits which did not result in hospitalization was not available. As there were no differences in the rate of use of medications used to treat acute exacerbations of COPD (a likely reason for a COPD-related ED visit10), it is improbable that rates of ED visits differed. Thus, it is unlikely that deductibles on prescription drugs had an impact on ED visits. Thirdly, the analysis excluded individuals who were not registered for Fair PharmaCare. Of the 1,764,635 households in BC in 2011,68 1,211,692, or 68.7% were registered for Fair PharmaCare.72 It is unclear how individuals registered for Fair PharmaCare and those not registered differ and whether their prescription drug utilization patterns also differ. A plausibility is that individuals with little assistance to prescription drug costs (i.e. those with no private insurance) may rely on public coverage more than those who have more assistance, and thus more inclined to register. If this is the case, this would minimize the influence of private insurance on our results. 66 Despite these limitations, the strong study design along with use of a rigorous definition of identifying individuals with COPD has allowed us to determine the effect of imposing deductibles on individuals with an identifiable need for chronic medications, which to our knowledge is the first study of its kind. The definition used in the current study more accurately accounts for individuals diagnosed with COPD, in comparison to the definition used in the study by Dormuth et al, which found that the increase in cost-sharing moving from previous pharmacare programs to Fair PharmaCare, decreased initiation of treatment in patients with asthma and COPD.6 4.3.2 Employer-health sponsored insurance study A major limitation of this study is the credibility of responses regarding insurance availability. While it was not studied specifically, there is a sizeable proportion of individuals in both survey cycles who reported that they did not have publicly-funded insurance for prescription drugs. However, all Ontario residents over the age of 65 are automatically provided with coverage assistance on prescription medications under the Ontario Drug Benefits program.71 This brings to question whether or not responses to EHI are reliable. However, as these benefits are likely negotiated and outlined in a retirement benefits package,34 it is likely more memorable than being automatically enrolled into the Ontario Drug Benefits program. Moreover, the health status of individuals – measured as number of chronic illnesses and self-reported health status – were similar between the two survey cycles, suggesting ability to recall is likely similar between the cycles. Unless ability to recall is affected disproportionately between the cycles, the estimates are unlikely to be biased. While this study does provide the first empirical evidence of changes in EHI for retirees, the data available did not allow me to study the other ways in which employers may be decreasing generosity with EHI. Other cost-saving mechanisms which have been hypothesized include limiting annual or lifetime maximums, increased deductibles, limited formularies, and limited reimbursement on prescription dispensing fees.70 These were beyond the scope of this study. 67 The use of national, population-based survey data has the potential to illustrate current trends while having a variety of variables to ensure a rigorous statistical analysis. This study demonstrates of the type of work which may be possible with the rich data which comes from Statistics Canada’s annual CCHS, particularly when reliable data is lacking from traditional sources. This type of data may allow for detection of suspected epidemiological phenomena and, as with this study, the suspected trends in health-related matters. The results yielded by studies such as ours will hopefully provide impetus for researchers to pursue more rigorous analysis by gathering data via other means. 4.4 Recommendations for Future Research The findings of this thesis are relevant to ongoing discussions on pharmacare in Canada. Together, these findings provide evidence for an alternate financing system for public drug coverage programs while warning about the potential inadequacies in traditional aid in the coverage of pharmaceuticals through EHI. However, in order to present the complete picture, below are a few recommendations for future research. 4.4.1 Factors underlying the low utilization rates of COPD-related medications In the deductibles study, only 57.6% of all eligible person-years utilized any COPD-related medications. While the special authority structure for most maintenance medications may contribute to the lower utilization rate of maintenance medications,73 the fact that use of rescue medications (which are not subject to special authority73) is similarly low may indicate that active use of COPD medications is generally low. Even though cost is an often cited factor for poor adherence to COPD treatment in literature74 – this was not seen as the case here, as least not with increased out-of-pocket payments due to imposition of a deductible. If cost-related non-adherence is present, such a result would mean that, even with assistance, copayments remain too high due to the high cost of these treatments to begin with. Dosing frequency, non-adherence to treatment for other comorbidities, and other patient factors such as health beliefs and understanding of the disease, patient experiences, and behaviour, are all influential in the uptake of COPD-related treatments.11,75,76 It would be important to understand the most influential 68 factors in order to design programs to increase uptake of COPD medications and subsequently improve the health status of individuals whose treatment for COPD is not optimized. 4.4.2 Impact of current special authority program, particularly for COPD population In the deductibles study, we found that use of ICS was slightly higher in the no deductibles group compared to the deductibles group. This near statistical significant result may be related to the fact that ICS, unlike other classes of maintenance medications for COPD, are a regular benefit eligible for coverage without having to go through the special authority process. This process requires a physician to submit an application, demonstrating that a patient satisfy a clinical need for the medication. The listing of ICS as a regular benefits allows them to be eligible for coverage immediately for the no deductibles group. While ICS are generally considered the first-line of treatment in asthma over other types of maintenance inhalers, their role in COPD management is debatable.54 Current COPD guidelines consider ICS to be adjunctive therapy to other classes of maintenance medications and it is recommended that they not be used alone.10 As such, this increase in ICS use in the no deductibles group may potentially be inappropriate and the structure of special authority around inhaler coverage may be contributing to this phenomenon. It is unclear from our study the implications of this, but given that ICS use may increase risk of pneumonia – a trigger for COPD exacerbations, this increase may be of clinical significance and warrants further investigation. 4.4.3 Difference in characteristics of individuals not registered for Fair PharmaCare According to 2011 estimates, only 68.7% of families in BC were registered for Fair PharmaCare. As we were not able to obtain verified income information for the proportion of individuals not registered, these individuals were not included in our analysis. It would be helpful to understand, likely through qualitative methods, the characteristics which influence whether or not a family registers or not. These factors may include income status, availability of private insurance, and perceived need to receive pharmaceutical care. In line with the deductibles analysis, it would be valuable to understand whether utilization patterns of pharmaceutical care are the same amongst individuals not registered in the program. A potential research question could be to examine the rate of uptake of COPD-related treatment and whether utilization rates 69 are as low as the Fair PharmaCare population. As individuals would not be subject to the special authority requirements, studying the rate of ICS use amongst this population may be helpful in understanding the increased rate observed in the no deductibles group. 4.4.4 More in-depth study of EHI changes for retirees across Canada As alluded to in other parts of this thesis, more data on insurance availability and coverage policies is required to study the trends in EHI. This is particularly important for older adults, who may have limited income and diminished ability to afford prescription drugs and other non-publicly insured treatments. Specifically, more information is needed on how coverage policies have changed over the last decade in regards to out-of-pocket premiums and deductibles, and formulary changes. The extent of which these strategies are being employed and the characteristics of individuals affected need to be investigated in order to understand the potential impact of these changes on the population. For many, EHI is the last resort in making treatments more available. As decreasing generosity of EHI would likely increase out-of-pocket costs for individuals affected, it follows that the uptake of treatments may also be affected. This may have potential implications on the publicly-financed health care system. This topic should be studied nationally, as the experience of residents in each province may be different. While this thesis has focused on implications on prescription drug utilization, these results have implications for other health care services and treatments frequently utilized by this older adult population, such as dental treatment, home care, and long-term care. The impact of this trend on these service areas are less well-studied and deserve our attention to minimize impact on individuals and the health care system. 70 References 1. The University of British Columbia Board of Governors. Policy #89: Research involving human participants. http://universitycounsel.ubc.ca/files/2012/06/policy89.pdf. Published 2012. Accessed October 18, 2016. 2. Morgan S, Daw J, Law M. Rethinking Pharmacare in Canada (June 13, 2013) C.D. Howe Institute Commentary 384. Toronto, Ont; 2013. 3. Busby C, Pedde J. Should Public Drug Plans Be Based on Age or Income? (December 3, 2014) C.D. Howe Institute Commentary 417. Toronto, Ont; 2014. 4. Morgan SG, Li W, Yau B, Persaud N. Estimated effects of adding universal public coverage of an essential medicines list to existing public drug plans in Canada. CMAJ. 2017;189(8):E295-E302. doi:10.1503/cmaj.161082. 5. Tamblyn R, Laprise R, Hanley J, et al. Adverse events associated with prescription drug cost-sharing among poor and elderly persons. JAMA. 2001;285(4):421–429. 6. Dormuth CRC, Glynn RRJ, Neumann P, Maclure M, Brookhart AM, Schneeweiss S. Impact of two sequential drug cost-sharing policies on the use of inhaled medications in older patients with chronic obstructive pulmonary disease or asthma. Clin Ther. 2006;28(6):963-964. doi:10.1016/j.clinthera.2006.06.007. 7. Dormuth CR, Maclure M, Glynn RJ, Neumann P, Brookhart AM, Schneeweiss S. Emergency hospital admissions after income-based deductibles and prescription copayments in older users of inhaled medications. Clin Ther. 2008;30 Spec No:1038-1050. doi:10.1016/j.clinthera.2008.06.003. 8. Law MR, Cheng L, Dhalla IA, Basc DH, Morgan SG. The effect of cost on adherence to prescription medications in Canada. CMAJ. 2012;184(3):297-302. 9. Law MR, Cheng L, Worthington H, et al. Impact of income-based deductibles on drug use and health care utilization among older adults. CMAJ. 2017;189(19):690-696. doi:10.1503/cmaj.161119. 10. Global Initiative for Chronic Obstructive Pulmonary Disease (GOLD). Global Strategy for the Diagnosis, Management and Prevention of COPD (GOLD 2017).; 2017. http://goldcopd.org/gold-2017-global-strategy-diagnosis-management-prevention-copd/. Accessed February 16, 2017. 11. Restrepo R, Alvarez M, Wittnebel L, et al. Medication adherence issues in patients treated for COPD. Int J Chron Obstruct Pulmon Dis. 2008;3(3):371-384. 12. Criner G, Bourbeau J, Diekemper R, Al. E. Prevention of acute exacerbations of chronic obstructive pulmonary disease: American College of Chest Physicians and Canadian Thoracic Society Guideline. Chest. 2015;147(4):894–942. 13. Therapeutic Research Centre. PL Detail-Document #310107. Inhalers for COPD (Last modified January 2016). Pharmacist’s Letter. http://canadianpharmacistsletter.therapeuticresearch.com/pl/ArticlePDF.aspx?cs=&s=PLC&DocumentFileID=0&DetailID=290924&SegmentID=0. Published 2015. Accessed January 15, 2016. 14. Allin S, Law MR, Laporte A. How does complementary private prescription drug insurance coverage affect seniors’ use of publicly funded medications? Health Policy (New York). 2013;110(2-3):147-155. doi:10.1016/j.healthpol.2013.02.010. 15. Stuart B, Singhal PK, Fahlman C, Doshi J, Briesacher B. Employer-sponsored health 71 insurance and prescription drug coverage for new retirees: Dramatic declines in five years. Heal Aff. 2003;Suppl Web:W3-334-341. 16. Buchmueller TC, Monheit AC. Employer-sponsored health insurance and the promise of health insurance reform. Inquiry. 2009;46(2):187-202. 17. West D, Mercer (Canada) Limited. Cost Trends in Health Benefits for Ontario Businesses: Analysis for Discussion. Toronto, Ont; 2011. 18. Canadian Institute for Health Information. A Snapshot of Health Care in Canada as Demonstrated by Top 10 Lists, 2011.; 2012. 19. Hnizdo E, Sullivan P, Bang KM, Wagner G. Airflow Obstruction Attributable to Work in Industry and Occupation Among U.S. Race / Ethnic Groups : A Study of NHANES III Data. Am J Ind Med. 2004;46:126-135. doi:10.1002/ajim.20042. 20. Chatila WM, Thomashow BM, Minai OA, Criner GJ, Make BJ. Comorbidities in chronic obstructive pulmonary disease. Proc Am Thorac Soc. 2008;5(4):549-555. doi:10.1513/pats.200709-148ET. 21. Curkendall S, DeLuise C, Jones J, et al. Cardiovascular disease in patients with chronic obstructive pulmonary disease, Saskatchewan Canada cardiovascular disease in COPD patients. Ann Epidemiol. 2006;16:63-70. 22. Kinnunen T, Saynajakangas O, Tuuponen T, Keistinen T. Impact of comorbidities on the duration of COPD patients’ hospital episodes. Respir Med. 2003;97:143-146. 23. Donaldson GC, Hurst JR, Smith CJ, Hubbard RB, Wedzicha JA. Increased risk of myocardial infarction and stroke following exacerbation of COPD. Chest. 2010;137(5):1091-1097. doi:10.1378/chest.09-2029. 24. Nannini LJ, Poole P, Milan S, Holmes R, Normansell R. Combined corticosteroid and long-acting beta 2-agonist in one inhaler versus long-acting beta 2-agonists for chronic obstructive pulmonary disease (Review). Cochrane Database Syst Rev. 2013;(11):1-166. doi:10.1002/14651858.CD006829.pub2.Copyright. 25. Morgan SG, Coombes M. Income-based drug coverage in British Columbia: Towards an understanding of the policy. Healthc Policy. 2006;2(2):92-108. 26. Fan V, Glylys-Colwell I, Locke E, et al. Overuse of short-acting beta-agonist bronchodilators in COPD during periods of clinical stability. Respir Med. 2016;116:100-106. 27. Mäkelä M, Backer V, Hedegaard M, Larsson K. Adherence to inhaled therapies, health outcomes and costs in patients with asthma and COPD. Respir Med. 2013;107:1481-1490. doi:10.1016/j.rmed.2013.04.005. 28. Canadian Institute for Health Information. Prescribed Drug Spending in Canada, 2016: A Focus on Public Drug Programs. Ottawa, ON; 2016. 29. Morgan SG, Daw J. Canadian pharmacare: Looking back, looking forward. Healthc Policy. 2012;8(1):14-23. 30. Government of British Columbia. PharmaCare for B.C. residents: How your coverage is calculated. PharmaCare for B.C. residents: How your coverage is calculated. http://www2.gov.bc.ca/gov/content/health/health-drug-coverage/pharmacare-for-bc-residents/who-we-cover/fair-pharmacare-plan/how-your-coverage-is-calculated. Published 2015. Accessed October 10, 2015. 31. Canadian Life and Health Insurance Association. Canadian Life and Health Insurance Facts. Toronto; 2016. 32. Kapur V, Basu K. Drug coverage in Canada: who is at risk? Health Policy (New York). 72 2005;71(2):181-193. doi:10.1016/j.healthpol.2004.08.006. 33. Dewa C, Hoch J, Steele L. Prescription drug benefits and Canada’s uninsured. Int J Law Psychiatry. 2005;28(5):496-513. 34. Bull D. The end of retiree benefits? Benefits Canada. October 2015:15-18. http://www.benefitscanada.com/benefits/health-benefits/the-end-of-retiree-benefits-72327. Accessed August 10, 2017. 35. Canadian Institute for Health Information. Inpatient hospitalizations, surgeries and childbirth indicators in 2012-2013. https://secure.cihi.ca/estore/productFamily.htm?locale=en&pf=PFC2624. Published 2014. Accessed October 17, 2015. 36. Goldman D, Joyce G, Zheng Y. Prescription drug cost sharing: associations with medication and medical utilization and spending and health. JAMA. 2007;298(1):161-169. 37. Morgan SG, Evans R, Hanley GE, Caetano P. Income-based drug coverage in British Columbia: Lessons for BC and the rest of Canada. Healthc Policy. 2006;2(2):115-127. 38. De Miguel-Díez J, Carrasco-Garrido P, Rejas-Gutierrez J, et al. Inappropriate overuse of inhaled corticosteroids for COPD patients: Impact on health costs and health status. Lung. 2011;189(3):199-206. doi:10.1007/s00408-011-9289-0. 39. Population Data BC. About PopData. https://www.popdata.bc.ca/aboutus. Published 2016. Accessed February 2, 2016. 40. BC Ministry of Health: Medical Services Plan (MSP) payment information file. Data extract. Vancouver: Population Data BC; 2014. www.popdata.%0Abc.ca/data/internal/health/msp. Accessed February 18, 2016. 41. Canadian Institute for Health Information. Discharge Abstract Database (hospital separations). Data Extract. Victoria: BC Ministry of Health; 2014. www.popdata.bc.ca/data/internal/health/dad. Accessed February 18, 2016. 42. PharmaNet. Data extract. Victoria: BC Ministry of Health, Data Stewardship Committee; 2014. www.popdata.bc.ca/data/external/PharmaNet. Accessed February 18, 2016. 43. BC Ministry of Health: Consolidation file (MSP registration & premium billing). Data Extract. Vancouver: Population Data BC; 2014. 44. Morgan SG, Cunningham C. Population aging and the determinants of healthcare expenditures : The case of hospital, medical and pharmaceutical care in British Columbia, 1996 to 2006. Healthc Policy. 2006;7(1):68-79. 45. 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. doi:10.1080/15412550903140865. 46. O’Keeffe A, Geneletti S, Baio G, Sharples L, Nazareth I, Petersen I. Regression discontinuity designs: an approach to the evaluation of treatment efficacy in primary care using observational data. BMJ. 349:g5293–g5293. 47. Moscoe E, Bor J, Bärnighausen T. Regression discontinuity designs are underutilized in medicine, epidemiology, and public health: a review of current and best practice. J Clin Epidemiol. 68(2):122–133. 48. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis. 1987;40:373-383. 49. Quan H, Sundararajan V, Halfon P, et al. Data Coding Algorithms for Defining Comorbidities in ICD-9-CM and ICD-1 0 Administrative Data. Med Care. 73 2005;43(11):1130-1139. 50. Boland MRS, Van Boven JFM, Kruis AL, et al. Investigating the association between medication adherence and health-related quality of life in COPD: Methodological challenges when using a proxy measure of adherence. Respir Med. 2016;110:34-45. doi:10.1016/j.rmed.2015.11.008. 51. Stuart BC, Simoni-wastila L, Zuckerman IH, et al. Impact of maintenance therapy on hospitalization and expenditures for Medicare beneficiaries with chronic obstructive pulmonary disease. Am J Geriatr Pharmacother. 2010;8(5):441-453. doi:10.1016/j.amjopharm.2010.10.002. 52. Haupt D, Krigsman K, Nilsson JL. Medication persistence among patients with asthma / COPD drugs. Pharm World Sci. 2008;30:509-514. doi:10.1007/s11096-008-9197-4. 53. Anis AH, Lynd LD, Wang X, et al. Double trouble: impact of inappropriate use of asthma medication on the use of health care resources. 2001;164(5):625-631. 54. Yang I, Clarke M, Sim E, Fong K. Inhaled corticosteroids for stable chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2012;7:CD002991. 55. Kratzer J, Cheng L, Allin S, Law MR. The impact of private insurance coverage on prescription drug use in Ontario, Canada. Healthc Policy. 2015;10(4):62-74. 56. Buttgereit F. A fresh look at glucocorticoids how to use an old ally more effectively. Bull NYU Hosp Jt Dis. 2012;70(S1):S26-29. 57. Law MR, Kratzer J, Dhalla IA. The increasing inefficiency of private health insurance in Canada. CMAJ. 2014;186(12):E470-E474. 58. Quiñonez C, Figueiredo R. Sorry doctor, I can’t afford the root canal, I have a job: Canadian dental care policy and the working poor. Can J Public Heal. 2010;101(6):481-485. 59. Ramraj C, Sadeghi L, Lawrence HP, Dempster L, Quiñonez C. Is accessing dental care becoming More Difficult? Evidence from Canada’s Middle-Income Population. PLoS One. 2013;8(2):e57377-e57377. doi:10.1371/journal.pone.0057377. 60. Devlin RA, Sarma S, Zhang Q. The role of supplemental coverage in a universal health insurance system: Some Canadian evidence. Health Policy (New York). 2011;100(1):81-90. doi:10.1016/j.healthpol.2010.08.011. 61. Law MR, Daw JR, Cheng L, Morgan SG. Growth in private payments for health care by Canadian households. Health Policy (New York). 2013;110(2-3):141-146. doi:10.1016/j.healthpol.2013.01.014. 62. Locker D, Maggirias J, Quiñonez C. Income, dental insurance coverage, and financial barriers to dental care among Canadian adults. J Public Heal Dent. 2011;71(3):327-334. doi:10.1111/j.1752-7325.2011.00277.x. 63. Statistics Canada. Canadian Community Health Survey (CCHS). Definitions, data sources and methods. http://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3226. Published 2016. Accessed October 14, 2016. 64. Statistics Canada. Canadian Community Health Survey (CCHS) Cycle 3.1 (2005): Public Use Microdata File (PUMF) User Guide. Ottawa, ON; 2006. 65. Statistics Canada. Canadian Community Health Survey (CCHS) Annual Component: User Guide 2014 and 2013-2014 Microdata Files. Ottawa, ON; 2015. 66. Thomas S, Wannell B. Combining cycles of the Canadian Community Health Survey. Heal Reports. 2009;20(1):53-58. 74 67. Vittinghoff E, Glidden D, Shiboski S, McCulloch C. Chapter 5: Logistic regression. In: Regression Methods in Biostatistics. 2nd ed. New York: Springer New York; 2012:139-202. 68. Statistics Canada. CANSIM - 051-0001 - Estimates of population, by age group and sex for July 1, Canada, provinces and territories. CANSIM (database). http://www5.statcan.gc.ca/cansim/a47. Published 2016. Accessed August 9, 2017. 69. Agency G of CCR. Line 9060 - Salaries, wages, and benefits (including employer’s contributions). Business Expenses. http://www.cra-arc.gc.ca/tx/bsnss/tpcs/slprtnr/rprtng/t2125/ln9060-eng.html. Published 2016. Accessed December 4, 2016. 70. Kratzer J, McGrail KM, Strumpf E, Law MR. Cost-control mechanisms in Canadian private drug plans. Healthc Policy. 2013;9(1):35-43. 71. Government of Ontario. Get coverage for prescription drugs | Ontario.ca. https://www.ontario.ca/page/get-coverage-prescription-drugs. Published 2017. Accessed August 15, 2017. 72. Medical Beneficiary Pharmaceutical Division Services. PharmaCare Trends 2011/12. Victoria, BC; 2013. 73. BC Ministry of Health. Special authority - Province of British Columbia. Practitioner & professional resources. http://www2.gov.bc.ca/gov/content/health/practitioner-professional-resources/pharmacare/prescribers/special-authority. Published 2016. Accessed May 3, 2016. 74. Castaldi PJ, Rogers WH, Safran DG, Wilson IB. Inhaler costs and medication nonadherence. Chest. 2010;138(3):614-620. doi:10.1378/chest.09-3031. 75. Toy EL, Beaulieu NU, Mchale JM, et al. Treatment of COPD: Relationships between daily dosing frequency, adherence, resource use, and costs. Respir Med. 2011;105(3):435-441. doi:10.1016/j.rmed.2010.09.006. 76. Schwab P, Hopson S, Moretz C, Burslem K, Renda A, Kaila S. Association between adherence to medications for COPD and medications for other chronic conditions in COPD patients. Int J COPD. 2017;12:115-122.