Association of Patient, Prescriber, and Region With the Initiation of First Prescription of Biologic Disease-Modifying Antirheumatic Drug Among Older Patients With Rheumatoid Arthritis and Identical Health Insurance Coverage | Geriatrics | JAMA Network Open | JAMA Network
[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Active Biologic Prescription Percentage as a Proportion of Active Users of All Disease-Modifying Antirheumatic Drugs (DMARDs) Among Older Adults With Rheumatoid Arthritis
Active Biologic Prescription Percentage as a Proportion of Active Users of All Disease-Modifying Antirheumatic Drugs (DMARDs) Among Older Adults With Rheumatoid Arthritis

Unadjusted proportion of all patients older than 65 receiving biologic DMARD for rheumatoid arthritis, by region.

Figure 2.  Time From First Conventional Synthetic Disease-Modifying Antirheumatic Drug (csDMARD) to First Biologic DMARD Among Older Adults With Rheumatoid Arthritis in Ontario, Canada, Stratified by Region
Time From First Conventional Synthetic Disease-Modifying Antirheumatic Drug (csDMARD) to First Biologic DMARD Among Older Adults With Rheumatoid Arthritis in Ontario, Canada, Stratified by Region

Unadjusted survival curves for patients older than 67 years who started their first csDMARD prescription between 2002 and 2015, inclusive, with a maximum follow-up to 2017. The shaded area around the aggregate curve represents the 95% CI.

Figure 3.  Quantity-Adjusted Supply of Rheumatologists by Region
Quantity-Adjusted Supply of Rheumatologists by Region

Vertical line represents the change in rheumatologist billing identifier code.

Figure 4.  Biologic Prescription Preferences Among 270 Physicians Who Prescribed Biologic Disease-Modifying Antirheumatic Drugs Over Time
Biologic Prescription Preferences Among 270 Physicians Who Prescribed Biologic Disease-Modifying Antirheumatic Drugs Over Time

Preference computed using all available patients. Dots represent individual physicians.

Table.  Effect Estimates From Patient and Prescriber Variables
Effect Estimates From Patient and Prescriber Variables
1.
Widdifield  J, Bernatsky  S, Paterson  JM,  et al.  Accuracy of Canadian health administrative databases in identifying patients with rheumatoid arthritis: a validation study using the medical records of rheumatologists.  Arthritis Care Res (Hoboken). 2013;65(10):1582-1591. doi:10.1002/acr.22031PubMedGoogle Scholar
2.
Bykerk  VP, Akhavan  P, Hazlewood  GS,  et al; Canadian Rheumatology Association.  Canadian Rheumatology Association recommendations for pharmacological management of rheumatoid arthritis with traditional and biologic disease-modifying antirheumatic drugs.  J Rheumatol. 2012;39(8):1559-1582. doi:10.3899/jrheum.110207PubMedGoogle Scholar
3.
Smolen  JS, Landewé  R, Bijlsma  J,  et al.  EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2016 update.  Ann Rheum Dis. 2017;76(6):960-977. doi:10.1136/annrheumdis-2016-210715PubMedGoogle Scholar
4.
Singh  JA, Saag  KG, Bridges  SL  Jr,  et al; American College of Rheumatology.  2015 American College of Rheumatology guideline for the treatment of rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2016;68(1):1-25. doi:10.1002/acr.22783PubMedGoogle Scholar
5.
Pease  C, Pope  JE, Thorne  C,  et al.  Canadian variation by province in rheumatoid arthritis initiating anti-tumor necrosis factor therapy: results from the optimization of adalimumab trial.  J Rheumatol. 2010;37(12):2469-2474. doi:10.3899/jrheum.091447PubMedGoogle Scholar
6.
Pease  C, Pope  JE, Truong  D,  et al.  Comparison of anti-TNF treatment initiation in rheumatoid arthritis databases demonstrates wide country variability in patient parameters at initiation of anti-TNF therapy.  Semin Arthritis Rheum. 2011;41(1):81-89. doi:10.1016/j.semarthrit.2010.09.004PubMedGoogle Scholar
7.
Yelin  E, Tonner  C, Kim  SC,  et al.  Sociodemographic, disease, health system, and contextual factors affecting the initiation of biologic agents in rheumatoid arthritis: a longitudinal study.  Arthritis Care Res (Hoboken). 2014;66(7):980-989. doi:10.1002/acr.22244PubMedGoogle Scholar
8.
Yazdany  J, Tonner  C, Schmajuk  G.  Use and spending for biologic disease-modifying antirheumatic drugs for rheumatoid arthritis among US Medicare beneficiaries.  Arthritis Care Res (Hoboken). 2015;67(9):1210-1218. doi:10.1002/acr.22580PubMedGoogle Scholar
9.
Schmajuk  G, Trivedi  AN, Solomon  DH,  et al.  Receipt of disease-modifying antirheumatic drugs among patients with rheumatoid arthritis in Medicare managed care plans.  JAMA. 2011;305(5):480-486. doi:10.1001/jama.2011.67PubMedGoogle Scholar
10.
Sokka  T, Kautiainen  H, Pincus  T,  et al.  Disparities in rheumatoid arthritis disease activity according to gross domestic product in 25 countries in the QUEST-RA database.  Ann Rheum Dis. 2009;68(11):1666-1672. doi:10.1136/ard.2009.109983PubMedGoogle Scholar
11.
Neubauer  S, Cifaldi  M, Mittendorf  T, Ganguli  A, Wolff  M, Zeidler  J.  Biologic TNF inhibiting agents for treatment of rheumatoid arthritis: persistence and dosing patterns in Germany.  Health Econ Rev. 2014;4(1):32. doi:10.1186/s13561-014-0032-4PubMedGoogle Scholar
12.
Hoebert  JM, Mantel-Teeuwisse  AK, van Dijk  L, Bijlsma  JWJ, Leufkens  HGM.  Do rheumatoid arthritis patients have equal access to treatment with new medicines? tumour necrosis factor-alpha inhibitors use in four European countries.  Health Policy. 2012;104(1):76-83. doi:10.1016/j.healthpol.2011.10.011PubMedGoogle Scholar
13.
Souliotis  K, Papageorgiou  M, Politi  A, Ioakeimidis  D, Sidiropoulos  P.  Barriers to accessing biologic treatment for rheumatoid arthritis in Greece: the unseen impact of the fiscal crisis: the Health Outcomes Patient Environment (HOPE) study.  Rheumatol Int. 2014;34(1):25-33. doi:10.1007/s00296-013-2866-1PubMedGoogle Scholar
14.
Orlewska  E, Ancuta  I, Anic  B,  et al.  Access to biologic treatment for rheumatoid arthritis in Central and Eastern European (CEE) countries.  Med Sci Monit. 2011;17(4):SR1-SR13. doi:10.12659/MSM.881697PubMedGoogle Scholar
15.
Neovius  M, Simard  JF, Askling  J; ARTIS study group.  Nationwide prevalence of rheumatoid arthritis and penetration of disease-modifying drugs in Sweden.  Ann Rheum Dis. 2011;70(4):624-629. doi:10.1136/ard.2010.133371PubMedGoogle Scholar
16.
Neovius  M, Sundström  A, Simard  J,  et al; ARTIS Study Group.  Small-area variations in sales of TNF inhibitors in Sweden between 2000 and 2009.  Scand J Rheumatol. 2011;40(1):8-15. doi:10.3109/03009742.2010.493895PubMedGoogle Scholar
17.
Putrik  P, Sokka  T, Ramiro  S, Boonen  A.  Impact of socioeconomic gradients within and between countries on health of patients with rheumatoid arthritis (RA): lessons from QUEST RA.  Best Pract Res Clin Rheumatol. 2012;26(5):705-720. doi:10.1016/j.berh.2012.07.011PubMedGoogle Scholar
18.
Putrik  P, Ramiro  S, Kvien  TK,  et al; Working Group ‘Equity in access to treatment of rheumatoid arthritis in Europe’.  Inequities in access to biologic and synthetic DMARDs across 46 European countries.  Ann Rheum Dis. 2014;73(1):198-206. doi:10.1136/annrheumdis-2012-202603PubMedGoogle Scholar
19.
Putrik  P, Ramiro  S, Lie  E,  et al.  Less educated and older patients have reduced access to biologic DMARDs even in a country with highly developed social welfare (Norway): results from Norwegian cohort study NOR-DMARD.  Rheumatology (Oxford). 2016;55(7):1217-1224. doi:10.1093/rheumatology/kew048PubMedGoogle Scholar
20.
Putrik  P, Ramiro  S, Kvien  TK, Sokka  T, Uhlig  T, Boonen  A; Equity in Clinical Eligibility Criteria for RA treatment Working Group.  Variations in criteria regulating treatment with reimbursed biologic DMARDs across European countries: are differences related to country’s wealth?  Ann Rheum Dis. 2014;73(11):2010-2021. doi:10.1136/annrheumdis-2013-203819PubMedGoogle Scholar
21.
Brookhart  MA, Stürmer  T, Glynn  RJ, Rassen  J, Schneeweiss  S.  Confounding control in healthcare database research: challenges and potential approaches.  Med Care. 2010;48(6)(suppl):S114-S120. doi:10.1097/MLR.0b013e3181dbebe3PubMedGoogle Scholar
22.
Sommers  BD, Gawande  AA, Baicker  K.  Health insurance coverage and health: what the recent evidence tells us.  N Engl J Med. 2017;377(6):586-593. doi:10.1056/NEJMsb1706645PubMedGoogle Scholar
23.
Kessler  R, Glasgow  RE.  A proposal to speed translation of healthcare research into practice: dramatic change is needed.  Am J Prev Med. 2011;40(6):637-644. doi:10.1016/j.amepre.2011.02.023PubMedGoogle Scholar
24.
Rosenbaum  PR.  Design of Observational Studies. New York, NY: Springer; 2010.
25.
ICES. https://www.ices.on.ca/. Accessed November 5, 2019.
26.
von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  J Clin Epidemiol. 2008;61(4):344-349. doi:10.1016/j.jclinepi.2007.11.008PubMedGoogle Scholar
27.
The Johns Hopkins ACG Case Mix System.  Version 10 Release Notes. PC (DOS/WIN/NT) and Unix Version 10.0. Baltimore, MD: Johns Hopkins Bloomberg School of Public Health; 2014.
28.
Kralj  B.  Measuring rurality: RIO2008 BASIC: methodology and results. https://www.oma.org/wp-content/uploads/2008rio-fulltechnicalpaper.pdf. Accessed October 29, 2019.
29.
Statistics Canada.  Income Reference Guide, Census of Population, 2016. https://www150.statcan.gc.ca/n1/pub/82-221-x/2013001/quality-qualite/qua8-eng.htm. Accessed October 29, 2019.
30.
Matheson  FI, Dunn  JR, Smith  KLW, Moineddin  R, Glazier  RH.  Development of the Canadian Marginalization Index: a new tool for the study of inequality.  Can J Public Health. 2012;103(8)(suppl 2):S12-S16. doi:10.1007/BF03403823PubMedGoogle Scholar
31.
Evans  JL.  The dynamic behavior of alternative price adjustment mechanisms.  Manchester School. 1983;51(1):33-44. doi:10.1111/j.1467-9957.1983.tb00739.xGoogle Scholar
32.
Burke  N, Bowen  JM, Troyan  S,  et al.  Management of hospital formularies in Ontario: challenges within a local health integration network.  Can J Hosp Pharm. 2016;69(3):187-193. doi:10.4212/cjhp.v69i3.1554PubMedGoogle Scholar
33.
Austin  PC.  A tutorial on multilevel survival analysis: methods, models and applications.  Int Stat Rev. 2017;85(2):185-203. doi:10.1111/insr.12214PubMedGoogle Scholar
34.
Therneau  T.  Mixed effects Cox models. https://cran.r-project.org/web/packages/coxme/vignettes/coxme.pdf. Accessed March 17, 2018.
35.
Statistics Canada. 2016 Census topic: population and dwelling counts. https://www12.statcan.gc.ca/census-recensement/2016/rt-td/population-eng.cfm. Accessed October 29, 2019.
36.
Kalkan  A, Husberg  M, Hallert  E,  et al.  Physician preferences and variations in prescription of biologic drugs for rheumatoid arthritis: a register-based study of 4,010 patients in Sweden.  Arthritis Care Res (Hoboken). 2015;67(12):1679-1685. doi:10.1002/acr.22640PubMedGoogle Scholar
37.
Eriksson  JK, Johansson  K, Askling  J, Neovius  M.  Costs for hospital care, drugs and lost work days in incident and prevalent rheumatoid arthritis: how large, and how are they distributed?  Ann Rheum Dis. 2015;74(4):648-654. doi:10.1136/annrheumdis-2013-204080PubMedGoogle Scholar
38.
Ministry of Health and Long Term Care. Ontario public formulary. http://www.health.gov.on.ca/en/pro/programs/drugs/odbf_eformulary.aspx. Accessed October 29, 2019.
39.
Harrison  MJ, Tricker  KJ, Davies  L,  et al.  The relationship between social deprivation, disease outcome measures, and response to treatment in patients with stable, long-standing rheumatoid arthritis.  J Rheumatol. 2005;32(12):2330-2336.PubMedGoogle Scholar
40.
Brekke  M, Hjortdahl  P, Thelle  DS, Kvien  TK.  Disease activity and severity in patients with rheumatoid arthritis: relations to socioeconomic inequality.  Soc Sci Med. 1999;48(12):1743-1750. doi:10.1016/S0277-9536(99)00075-1PubMedGoogle Scholar
41.
Molina  E, Del Rincon  I, Restrepo  JF, Battafarano  DF, Escalante  A.  Association of socioeconomic status with treatment delays, disease activity, joint damage, and disability in rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2015;67(7):940-946. doi:10.1002/acr.22542PubMedGoogle Scholar
42.
Marra  CA, Lynd  LD, Esdaile  JM, Kopec  J, Anis  AH.  The impact of low family income on self-reported health outcomes in patients with rheumatoid arthritis within a publicly funded health-care.  Rheumatology (Oxford). 2004;43(11):1390-1397. doi:10.1093/rheumatology/keh334PubMedGoogle Scholar
43.
Widdifield  J, Bernatsky  S, Paterson  JM,  et al.  Serious infections in a population-based cohort of 86,039 seniors with rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2013;65(3):353-361. doi:10.1002/acr.21812PubMedGoogle Scholar
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Rheumatology
    December 6, 2019

    Association of Patient, Prescriber, and Region With the Initiation of First Prescription of Biologic Disease-Modifying Antirheumatic Drug Among Older Patients With Rheumatoid Arthritis and Identical Health Insurance Coverage

    Author Affiliations
    • 1University of Toronto, Toronto, Ontario, Canada
    • 2University Health Network, Toronto, Ontario, Canada
    • 3ICES, Toronto, Ontario, Canada
    • 4William Osler Health Care Centre, Brampton, Ontario, Canada
    • 5St Michael’s Hospital, Toronto, Ontario, Canada
    • 6University of British Columbia, Vancouver, British Columbia, Canada
    JAMA Netw Open. 2019;2(12):e1917053. doi:10.1001/jamanetworkopen.2019.17053
    Key Points español 中文 (chinese)

    Question  What patient, prescriber, and regional factors are associated with time to first prescription of biologic disease-modifying antirheumatic drug (DMARD) among patients 67 years or older with rheumatoid arthritis?

    Findings  In this cohort study of 17 672 older patients with rheumatoid arthritis, patients were more likely to receive biologic DMARDs earlier if they were younger, female, and living in urban areas closer to prescribers. Physician preference was strongly associated with differences in time from first conventional synthetic DMARD to first biologic DMARD.

    Meaning  In this study, pharmacologic care for rheumatoid arthritis was not uniform across patients and prescribers given similar disease characteristics in a population with identical health insurance coverage.

    Abstract

    Importance  Prescribing the first biologic treatment for rheumatoid arthritis (RA) is an important decision for patients, their physicians, and payers, with considerable costs and clinical implications. Conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) have known effectiveness and safety profiles and are less expensive; therefore, determining the variables contributing to csDMARD treatment duration is an essential question for patients, physicians, and payers.

    Objectives  To describe access to the first biologic DMARD prescription in a population of patients with RA and identical comprehensive health insurance coverage in Ontario, Canada, and to explore the associations of patient, prescriber, and geographic region with differences in time to first biologic prescription.

    Design, Setting, and Participants  This cohort study of incident patients with RA used administrative data with surveillance and patient-level data collected at yearly intervals. A total of 17 672 patients were included in the study; they were residents of Ontario, Canada, had an incident RA diagnosis at age 67 or older between 2002 and 2015, and received at least 1 csDMARD. Data were analyzed in November 2017.

    Exposure  Patient variables were age, sex, disease duration, socioeconomic status, distance to care, and supply of care in the patient’s area of residence. Prescriber covariates were year of graduation, specialty of practice, and supply of rheumatologic care in the patient’s geographic region.

    Main Outcomes and Measures  Time from first csDMARD prescription to receipt of first biologic medication.

    Results  Of 17 672 patients, 11 598 (65.6%) were women, and the mean (SD) age was 75.2 (5.8) years. Characteristics associated with longer time to receipt of a biologic prescription were older age (HR for every 5-year increase, 0.66; 95% CI, 0.62-0.71; P < .001), male sex (HR, 0.76; 95% CI, 0.66-0.89; P < .001), and distance to the nearest rheumatologist (HR per 10-km increase, 0.99; 95% CI, 0.98-0.99; P < .001). Prescribers were primarily rheumatologists (151 of 214 [70.6%]) and primary care physicians (26 of 214 [12.1%]). After adjusting for the number of patients eligible to receive biologic DMARDs, rheumatologists’ preferences (ie, yearly prescription rates) for using biologic DMARDs increased over time, from 1.7% in 2001 to 4.9% in 2015. After adjusting for calendar year and patient-, prescriber-, and region-level characteristics, substantial variation between prescribers in rates of prescribing a first biologic DMARD were found (65% variance).

    Conclusions and Relevance  This study found variation in time to receipt of first biologic DMARD after prescription of first csDMARD in a population with RA after adjustment for individual-level patient, prescriber, and geographic area covariates, despite identical universal health insurance coverage.

    Introduction

    In Canada, approximately 300 000 patients, or 1% of the population, are living with rheumatoid arthritis (RA).1 Treatment advances during the past 20 years have radically improved participation and quality-of-life outcomes for patients with access to evidence-informed care. These advances have included high-quality evidence to support the use of conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) at optimal dosing combinations and thresholds, and the development of biologic DMARDs, which specifically target the inflammatory mechanisms of RA.

    Switching to biologic therapy for RA is an expensive health care decision for patients, physicians, and payers. Optimizing dose and treatment time with first-line csDMARDs saves health care resources, with a difference in costs of more than CaD$15 000 (US$11 460) annually, while preserving the clinical choice to escalate treatment to biologic therapy.2-4 The ability to escalate to a biologic DMARD is critical; all clinical practice guidelines for RA recommend an adequate trial of csDMARDs before initiating a biologic medication.2-4 Despite these recommendations, limited evidence informs clinicians and policy makers on the factors associated with earlier receipt of a biologic DMARD. This limitation is troubling because wide variations in access to and time to initiation of biologic therapy have been observed across health care jurisdictions.5-20

    There are several challenges to research in this area. Research on the initiation of the first biologic medication is often limited by small nonrepresentative samples and short follow-up periods, reducing the interpretability and generalizability of findings.21,22 In addition, prescriber decisions to switch to biologic DMARDs are often influenced by nonmedical reasons associated with a patient’s insurance coverage.8,9 Variations in insurance coverage within patient populations confound clinical studies and limits causal attribution. Furthermore, patient, prescriber, and regional characteristics are often analyzed individually instead of simultaneously, a method that does not reflect the complexity of treatment decision-making in usual care.23,24 The objectives of this study were to describe access to the first biologic DMARD prescribed in a population of patients with RA and identical comprehensive health insurance coverage in Ontario, Canada, and to explore the associations of patient, prescriber, and geographic region with differences in time to first prescription of biologic DMARD.

    Methods
    Study Design and Setting

    We conducted a retrospective cohort study using deidentified population-based administrative health care databases from ICES (formerly the Institute for Clinical Evaluative Sciences).25 All residents of Ontario aged 65 years and older are covered under the Ontario Health Insurance Plan, which includes physician visits, acute care hospital use, and all prescriptions. Research ethics committee approval was received from the University Health Network, Toronto, Canada, with individual consent waived in accordance with local privacy laws granted by section 45 of Ontario’s Personal Health Information Privacy Act. This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.26

    Data Sources

    Data sources included the Registered Persons Database for patient demographic data, Ontario Health Insurance Plan claims for physician visits, the Discharge Abstract Database for inpatient hospitalization records, and the Ontario Drug Benefit claims database for prescription medication dispensation. Prescriber variables were obtained from the ICES Physician Database. The Immigration Refugees and Citizenship Canada Database contained immigration status information for our study.

    Inclusion Criteria

    Patients were included if they were a resident of Ontario with a valid health card number and an RA diagnosis, which was identified using a validated algorithm. The incident date of RA diagnosis was determined as the date on which the final billing criteria was met (ie, 3 RA billings within 2 years, with ≥1 billing by a musculoskeletal specialist or ≥1 hospitalization for RA).1 To be eligible for analytic models, patients had to have incident onset RA at age 67 or older and be csDMARD naive. For unadjusted descriptive analysis, active users older than 65 years were included from 2001 to 2015, and unadjusted survival models used patients accrued from 2002 to 2015, with a maximum follow-up to 2017.

    Washout Period

    This population had a 2-year study washout period (ie, covering ages 65 and 66 years) to ensure the capture of all newly administered csDMARDs and biologic DMARDs since diagnosis of RA in the biologics era (ie, after 2001). An additional year of washout from 2001 to 2002 was used to remove potential bias from patients who may have immediately switched to biologic DMARDs when they received approval from regulators.

    Time to First Biologic DMARD

    To be eligible for the analysis of time from first csDMARD to first biologic DMARD, patients had to have received at least 1 csDMARD or biologic DMARD with an approved indication for RA. A full list of medications used in this analysis is provided in the eAppendix in the Supplement. The small-molecule selective inhibitor tofacitinib was excluded because fewer than 10 patients had an exposure during the observation window.

    Observation Window

    The study observation window was 2002 to 2015, with accrual ending in 2014, ensuring at least 1 complete year of follow-up for each patient (eFigure 1 in the Supplement). Data were analyzed in in November 2017, allowing a 2-year window to ensure all data were captured from each data source.

    Patient Characteristics

    Patient demographic variables were age, sex, disease duration, geographic area of residence, distance to prescriber of each medication, and distance to nearest rheumatologist. Health care resource use was operationalized using the Johns Hopkins Adjusted Comorbidity Group version 10,27 rurality index (range, 0-100; 0 indicates most urban; 100, most rural),28 neighborhood income quintile,29 and marginalization index.30 Neighborhood income quintile was calculated using national census data to set income quintile distribution with geographic assignment based on dissemination areas containing 500 to 800 people each. The marginalization index was also based on the characteristics of dissemination areas, focusing on 4 dimensions that contribute to the process of marginalization: residential instability, material deprivation, dependency, and ethnic concentration. Age and sex were considered fixed variables, with disease duration considered a time-varying (count) variable and calendar year considered a categorical variable. All patient variables except for age, sex, and disease duration were time-varying by calendar year.

    Prescriber Characteristics

    For each patient, prescriber variables were assigned by unique physician identifier number at the first new csDMARD or biologic DMARD prescription dispensation and carried forward to all subsequent prescriptions of the same medication. Therefore, the characteristics of the first prescriber of each new RA medication were assigned for each patient in the time-varying model. Biologic DMARD prescribers (eTable in the Supplement) were operationalized as physicians who had prescribed at least 10 biologic treatments during the study period. Prescriber characteristics were age, practice location, practice size, year of graduation, and specialty. Preference for biologic DMARD prescription for each physician was computed by dividing the number of times the clinician switched a biologic-naive patient to a biologic DMARD each year by the total number of biologic-naive patients in each year. This variable adjusts for physician-level preference for prescribing a biologic DMARD in a given year, while reflecting changing physician preferences over time.

    Regional Characteristics

    Access to physicians was estimated by computing the linear distance from patient to prescriber of each medication and from patient to the nearest available rheumatologist. To concurrently adjust for rheumatologist care in relation to regional demand for rheumatologists, we calculated the quantity-adjusted supply of rheumatologists,31 a new rheumatologist supply variable. This was calculated by dividing the annual number of rheumatologist visits in each region by the annual number of patients with RA who resided in each region. Rurality and socioeconomic status of physician practices were measured using rurality score28 and neighborhood income quintile.29 We operationalized 14 geographic regions using the Local Health Integration Network, an administrative geographic stratifier for health care delivery in Ontario32 (eFigure 2 in the Supplement).

    Outcome, Exposure, and Covariate Definitions

    The primary outcome for the explanatory analytic models was an intent-to-treat analysis from the time from the first csDMARD prescription to receipt of a biologic DMARD approved for RA, with censoring at death, loss of eligibility for public health care (ie, moved out of Ontario), or the end of follow-up (ie, December 31, 2015). Model explanatory exposure variables, including patient-level, physician-level, and region-level factors, were measured at yearly intervals, with medications recorded at each pharmacy dispensation and physician demographic characteristics updated every 90 days.

    Missing Data

    There were no missing patient characteristic variables or medication prescription variables; however, 18 571 of 385 612 physician characteristic variables (4.8%) attached to prescription data were missing because of pharmacy omission. Observations with a missing prescriber identification number value were assigned the identification number of the identical prescription closest in time for the patient.

    Statistical Analysis

    To illustrate the unadjusted association of region with receipt of first biologic DMARD, Kaplan-Meier curves stratified by region were calculated. Time from first csDMARD prescription to first biologic DMARD prescription was the outcome in mixed-effects Cox proportional hazard models, with random effects for region and physician nested within regions33 (eTable in the Supplement). This nested model choice reflects usual care; patients are prescribed medications by physicians who practice within regions. Time-dependent covariates were measured based on the covariates of the first prescriber of the csDMARD or biologic DMARD and the patient’s characteristics at time of prescription. Models also accommodated crossed random effects; patients could switch physicians, and physicians could move practice locations. Therefore, although models were nested, models were flexible to account for patients changing physicians or moving between regions, which reflects the realities of usual clinical care. To adjust for changing numbers of available biologic medications over time and year-specific factors, a categorical variable for calendar year was added to the model.

    Outcomes were measured at the patient level and reported as hazard ratios (HRs) with 95% CIs and P values. We used 2-tailed P < .05 as the threshold for statistical significance. From first csDMARD prescription, patient’s prescriptions were divided into a sequence of intervals based on the drug prescribed to accommodate time-varying patient, prescriber, and regional covariates. At the end of the follow-up period, each patient was either censored or had received a biologic DMARD, meaning trends in physician preference for biologic use were explicitly measured. A simple model showing the unadjusted association of the exposure with the outcome was contrasted with a full model, adjusting for all variables in the simple model as well as random effects, to calculate the association of physician variation with the time to receipt of first biologic DMARD. All analysis was performed using R statistical software version 3.1.2 (R Project for Statistical Computing).34

    Results

    A total of 17 672 patients met the study inclusion criteria, accruing 82 445 patient-years of follow-up (Table).35 A total of 719 patients (4.1%) received a first biologic prescription during the study.

    Differences in Biologic DMARD Prescription by Patient Characteristics
    Patient Demographic Characteristics

    Patients who received a csDMARD after RA diagnosis at 65 years or older had a mean (SD) age of 75.2 (5.8) years at baseline, were predominantly women (11 598 women [65.6%]), and tended to reside in more urban areas (mean [SD] rurality score, 11.86 [18.51]). Immigrants represented 1131 patients (6.4%) receiving csDMARDs at baseline (Table).

    In unadjusted descriptive analysis of all active medication users, disparities in biologic prescription percentage between regions increased over time in Ontario (Figure 1). In 2002, the difference between the highest and the lowest regional percentages of biologic DMARDs used per person was 1.8 percentage points (Toronto, 2.7% vs Central East, 0.9%; difference, 68.4%). In 2015, the difference between the highest and the lowest regional percentages was 8.7 percentage points (Hamilton Niagara, 21.3% vs Central East, 12.6%; difference, 66.3%). Unadjusted Kaplan-Meier curves stratified by region and as an aggregate of all regions illustrated significant differences in time to first biologic DMARD depending on region of patient residence (log-rank test, χ213 = 147; P < .001) (Figure 2).

    Calendar year was associated with time to first biologic DMARD in the fully adjusted Cox model (HR per 5-year increase in calendar year, 0.49; CI, 0.43-0.55; P < .001) (Table). For every 5-year increase in age at csDMARD initiation, patients were less likely to receive a biologic DMARD (HR, 0.66; 95% CI, 0.62-0.71; P < .001). Men (HR, 0.76; 95% CI, 0.66-0.89; P < .001) were also less likely than women to receive biologic medications. Patients who had a longer disease duration were more likely to receive a biologic DMARD (HR per 5-year increase in disease duration, 1.35; 95% CI, 1.19-1.54; P < .001). Health care resource use, represented by a higher Hopkins Adjusted Comorbidity Group score, had no observable association with time to receipt of first biologic DMARD (HR, 1.00; 95% CI, 0.99-1.01; P = .48) (Table).

    Socioeconomic Factors

    After measuring patient-level variables, including neighborhood income quintile, marginalization index composite scores, immigrant status, and supply of care variables (ie, distance to prescriber), we found mixed results for the association of socioeconomic factors with time to receipt of first biologic DMARD. Living in a neighborhood that belonged in income quintile 1 (HR, 0.92; 95% CI, 0.65-1.31; P = .16) and marginalization index composite score (HR, 0.99; 95% CI, 0.87-1.11; P = .81) were not statistically significant factors associated with receipt of first biologic prescription. Immigrants were statistically significantly less likely to receive biologic prescriptions (HR, 0.59; 95% CI, 0.42-0.84; P = .003) (Table). Greater distance to medication prescriber, irrespective of specialty, was associated with slight increase in likelihood of receiving a biologic prescription (HR per 10-km increase, 1.01; 95% CI, 1.00-1.02; P < .001), while patients with a longer distance to the nearest rheumatologist, who may not be the rheumatologist the patient was seeing, were less likely to receive a biologic DMARD (HR per 10-km increase, 0.99; 95% CI, 0.98-0.99; P < .001) (Table).

    Differences in Biologic DMARD Prescription by Prescriber Characteristics
    Prescriber Demographic Characteristics

    A total of 214 unique prescribers had written 10 or more biologic prescriptions during the study period. Prescribers were primarily rheumatologists (151 of 214 [70.6%]) and primary care physicians (26 of 214 [12.1%]). Among 405 321 prescriptions of biologic DMARDs and csDMARDs initiated or maintained during the study, 291 162 (71.8%) were by a rheumatologist, 56 470 (13.9%) by a primary care physician, 38 306 (9.40%) by an internist, and 16 461 (4.0%) by other specialists. Prescriptions of biologic DMARDs were predominantly initiated or maintained by rheumatologists (1629 of 1977 prescriptions [82.4%]), but primary care physicians (164 prescriptions [8.3%]), internal medicine physicians (55 prescriptions [2.8%]), and dermatologists (23 prescriptions [1.3%]) also prescribed biologic DMARDs in amounts greater than 1% of all biologic prescriptions. Having more recently graduated from medical school was associated with higher propensity to prescribe first biologic DMARD (HR per 5-year after graduation, 1.10; 95% CI, 1.04-1.17; P < .001), while women were not statistically different from men in propensity to initiate biologic treatment (HR, 0.97; 95% CI, 0.74-1.27; P = .83). Physicians practicing in more rural areas were negatively associated with prescription of biologic DMARDs (HR per 20-unit increase in rurality score, 0.67; 95% CI, 0.50-0.92; P = .01), while a greater quantity-adjusted supply of rheumatologists score was associated with a higher likelihood of receipt of biologic DMARDs (HR, 1.03; 95% CI, 1.00-1.06; P = .03) (Table and Figure 3).

    Rheumatologist Preference and Regional Variation

    Rheumatologist preferences for prescribing the first biologic DMARD among biologic-naive patients increased by a factor of 3 from 2001 through 2015, from 1.7% to 4.9%, adjusted for the number of patients eligible for receipt of first biologic DMARD (Figure 4). In models adjusted for age, sex, and calendar year, prescriber variance accounted for 74% of the variance in time from first csDMARD to first biologic DMARD (eTable in the Supplement). After adjustment for all patient and physician covariates, physician preference accounted for 65% of the between-region differences, while differences between the regions themselves contributed 4.6% to the overall prescription variation. The remaining 30.4% was owing to underlying variance (eTable in the Supplement).

    Discussion

    In a system in which all patients received identical comprehensive prescription coverage, differences in time to receipt of first biologic DMARD persisted after adjustment for individual patient, prescriber, and regional characteristics. Modest between-region variations for time to first biologic prescription were discovered concurrently with comparatively large between-prescriber variations. A 2015 study36 showed that several factors, including a complex disease course, multiple health insurers, patient preferences, and the prescribing behaviors of physicians, were associated with differences in time to prescription of the first biologic DMARD, highlighting the importance of nested analysis. The results of this study are similar to the biologic penetration rate of Sweden, where approximately 4% of older patients receive biologic DMARDs.37 In our study, the biologic penetration rate of 4.1% (719 of 17 692) was almost identical to the results from Sweden. Our nested adjustment for patients, prescribers, and region provided a complete evaluation of the relative association of a multitude of factors with time to first biologic prescription and demonstrated persistent variations between geographic regions. A 4.6% difference in time to receipt of biologic DMARDs between regions should be considered problematic in the absence of other explanatory factors. For example, every 1% increase in biologic prescriptions among a population of 72 000 funded patients with RA costs approximately CaD$10.8 million (US$8.25 million) per year (assuming 10% biologics penetration and CaD$15 000 annually per biologic DMARD prescription [US$11 460]).38 To bring the highest spending region in Ontario in line with the average spend would save approximately CaD$6 million to $8 million (US$4.6 million to $6.1 million) per year, accounting for 16% of the total RA biologic budget and 0.32% of the total all-cause drug formulary budget. Conversely, it would cost approximately CaD$6 million to $8 million (US$4.6 million to $6.1 million) to increase biologic penetration in the underserviced regions of Ontario to meet the population average of all regions.

    A higher quantity-adjusted supply of rheumatologists and a more urban rheumatology practice were associated with the receipt of biologic DMARDs. Patients residing in areas with lower socioeconomic status and in rural areas with less access to rheumatologists as well as patients who were immigrants were less likely to receive biologic prescriptions, suggesting a sociodemographic gradient of care for patients with more disadvantage. This study indicated that patients with lower socioeconomic status were the least likely to receive biologic DMARDs. This finding is surprising considering that other published studies show that socioeconomic status is a known risk factor for more severe disease.39-42 The findings of this study suggest that having local access to a rheumatologist may help extend the effective treatment course of csDMARDs. Our findings also suggest that patients may be more willing to travel longer distances to seek biologic treatments. The association between increasing age and greater time to receipt of a biologic DMARD suggested a continuing bias against providing biologic DMARDs for older patients, despite few safety signals indicating a higher risk of infection or other serious adverse events and infections.37,43

    Strengths and Limitations

    Strengths of this study were the combination of high-quality data and statistical methods. The administrative databases used for exposure, outcome, and covariate definition were validated across diseases and contained uniform data capture and follow-up. The statistical models used were innovative in assigning time-varying exposure to the prescribing physician instead of relying on proxy exposures, in which physician covariates are fixed based on the first prescriber of a csDMARD. In addition, the inclusion of nested random effects for clinicians within regions allowed us to estimate the variation associated with region and clinician. Furthermore, a new variable, the quantity-adjusted supply of rheumatologists, was created to reflect the available supply of rheumatologists in relation to the demand for their care. This variable has a strong theoretical basis in existing economic literature with potential applications to other studies of disparities in health care service delivery.23,24

    This study also had limitations, including restricting the analysis to patients 67 years or older, an age group in which the propensity to prescribe biologic DMARDs is known to be lower. Owing to higher biologic prescription percentage (ie, 30% estimate37) in the population of younger patients with RA, combined with lack of universal publicly funded insurance for patients younger than 65 years in Ontario, we expect that the resulting disparities in biologic prescriptions could be more severe among patients younger than 65 years. The study contained a large sample of patients with RA who live in Canada because Ontario is home to 38% of Canada’s population.35 With half of patients with RA being older than 65 years, approximately 19% of all patients with RA in Canada were included in the study. Inherent limitations of administrative data were also present. While validated, the exposure definition of RA could result in misclassified cases. This misclassification concern was alleviated somewhat by design because patients had to have been prescribed an RA-related csDMARD to be included in the study. Unavailable covariates, such as disease severity, functional disability questionnaires (eg, the Disease Activity Score, 28, or the Health Assessment Questionnaire), and other unobserved covariates, may be associated with systemic differential influence across regions, patients, and prescribers. Individual-level data for patients and prescribers were used to minimize the risks of systemic bias and ecological fallacy.

    Conclusions

    Prescription of the first biologic DMARD is a costly health care decision, with the choice to prescribe a biologic DMARD associated with large cost and clinical implications. Therefore, small changes in time to first biologic DMARD have major clinical and economic impacts. From a clinical perspective, the prescription of a biologic DMARD represents a transition to a more complex care plan, with less data to support the next prescription choice after the first biologic DMARD.

    Although disparities in access to biologic DMARDs increased in this study, the overall prescription percentage decreased relative to the population of active medication users with RA. This suggests that despite rheumatologists prescribing fewer biologic DMARDs on average per patient per unit of time, variations in the prescription of biologic DMARDs continue to grow. The random-effects term in our models indicated a high amount of variance associated with the practice of rheumatologists within regions. These differences in prescriber preferences have unclear implications for patient outcomes but show that between-prescriber differences exist in health care delivery for patients with RA, despite identical health insurance coverage.

    Back to top
    Article Information

    Accepted for Publication: October 16, 2019.

    Published: December 6, 2019. doi:10.1001/jamanetworkopen.2019.17053

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Tatangelo M et al. JAMA Network Open.

    Corresponding Author: Claire Bombardier, MD, FRCPC, Toronto General Hospital, 200 Elizabeth St, 13-Eaton North, Toronto, ON M5G 2C4, Canada (claire.bombardier@utoronto.ca).

    Author Contributions: Mr Tatangelo had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Tatangelo, Paterson, Ahluwalia, Bombardier.

    Acquisition, analysis, or interpretation of data: All authors.

    Drafting of the manuscript: Tatangelo, Ahluwalia, Bombardier.

    Critical revision of the manuscript for important intellectual content: All authors.

    Statistical analysis: Tatangelo, Tomlinson, Bansback.

    Obtained funding: Gomes.

    Administrative, technical, or material support: Paterson, Kopp.

    Supervision: Tomlinson, Paterson, Bombardier.

    Conflict of Interest Disclosures: Dr Ahluwalia reported receiving consulting fees from Amgen, Roche, Novaritis, Sanofi, Merck, and Eli Lilly and Co and receiving grants and consulting fees from Abbvie, Janssen Pharmaceuticals, Pfizer, and Hospira outside the submitted work. Dr Bombardier reported receiving grants and consulting fees from Abbvie, Janssen Pharmaceuticals, Hospira, Merck, Pfizer, Sanofi, and Novartis; receiving consulting fees from Covance; and receiving grants from Amgen, Eli Lilly and Co, Celgene, Medexus, Medreleaf, Roche, and Union Chimique Belge outside the submitted work. No other disclosures were reported.

    Funding/Support: This study was funded by grants from the Canadian Institute of Health Research, the Arthritis Society, the Ontario Drug Policy Research and Effectiveness Network, and the Canadian Institute of Health Research Drug Safety and Effectiveness Network. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care.

    Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Disclaimer: The analyses, conclusions, and statements expressed herein are solely those of the authors and do not reflect those of the funders or data providers; no endorsement is intended or should be inferred.

    Meeting Presentation: This article was presented at the 2018 Annual Meeting of the American College of Rheumatology; October 24, 2018; Chicago, Illinois.

    Additional Information: Parts of this article are based on data and information compiled and provided by the Ontario Ministry of Health and Long-Term Care, Immigration Refugees and Citizenship Canada, the Canadian Institute of Health Information, and IMS Brogan.

    References
    1.
    Widdifield  J, Bernatsky  S, Paterson  JM,  et al.  Accuracy of Canadian health administrative databases in identifying patients with rheumatoid arthritis: a validation study using the medical records of rheumatologists.  Arthritis Care Res (Hoboken). 2013;65(10):1582-1591. doi:10.1002/acr.22031PubMedGoogle Scholar
    2.
    Bykerk  VP, Akhavan  P, Hazlewood  GS,  et al; Canadian Rheumatology Association.  Canadian Rheumatology Association recommendations for pharmacological management of rheumatoid arthritis with traditional and biologic disease-modifying antirheumatic drugs.  J Rheumatol. 2012;39(8):1559-1582. doi:10.3899/jrheum.110207PubMedGoogle Scholar
    3.
    Smolen  JS, Landewé  R, Bijlsma  J,  et al.  EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2016 update.  Ann Rheum Dis. 2017;76(6):960-977. doi:10.1136/annrheumdis-2016-210715PubMedGoogle Scholar
    4.
    Singh  JA, Saag  KG, Bridges  SL  Jr,  et al; American College of Rheumatology.  2015 American College of Rheumatology guideline for the treatment of rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2016;68(1):1-25. doi:10.1002/acr.22783PubMedGoogle Scholar
    5.
    Pease  C, Pope  JE, Thorne  C,  et al.  Canadian variation by province in rheumatoid arthritis initiating anti-tumor necrosis factor therapy: results from the optimization of adalimumab trial.  J Rheumatol. 2010;37(12):2469-2474. doi:10.3899/jrheum.091447PubMedGoogle Scholar
    6.
    Pease  C, Pope  JE, Truong  D,  et al.  Comparison of anti-TNF treatment initiation in rheumatoid arthritis databases demonstrates wide country variability in patient parameters at initiation of anti-TNF therapy.  Semin Arthritis Rheum. 2011;41(1):81-89. doi:10.1016/j.semarthrit.2010.09.004PubMedGoogle Scholar
    7.
    Yelin  E, Tonner  C, Kim  SC,  et al.  Sociodemographic, disease, health system, and contextual factors affecting the initiation of biologic agents in rheumatoid arthritis: a longitudinal study.  Arthritis Care Res (Hoboken). 2014;66(7):980-989. doi:10.1002/acr.22244PubMedGoogle Scholar
    8.
    Yazdany  J, Tonner  C, Schmajuk  G.  Use and spending for biologic disease-modifying antirheumatic drugs for rheumatoid arthritis among US Medicare beneficiaries.  Arthritis Care Res (Hoboken). 2015;67(9):1210-1218. doi:10.1002/acr.22580PubMedGoogle Scholar
    9.
    Schmajuk  G, Trivedi  AN, Solomon  DH,  et al.  Receipt of disease-modifying antirheumatic drugs among patients with rheumatoid arthritis in Medicare managed care plans.  JAMA. 2011;305(5):480-486. doi:10.1001/jama.2011.67PubMedGoogle Scholar
    10.
    Sokka  T, Kautiainen  H, Pincus  T,  et al.  Disparities in rheumatoid arthritis disease activity according to gross domestic product in 25 countries in the QUEST-RA database.  Ann Rheum Dis. 2009;68(11):1666-1672. doi:10.1136/ard.2009.109983PubMedGoogle Scholar
    11.
    Neubauer  S, Cifaldi  M, Mittendorf  T, Ganguli  A, Wolff  M, Zeidler  J.  Biologic TNF inhibiting agents for treatment of rheumatoid arthritis: persistence and dosing patterns in Germany.  Health Econ Rev. 2014;4(1):32. doi:10.1186/s13561-014-0032-4PubMedGoogle Scholar
    12.
    Hoebert  JM, Mantel-Teeuwisse  AK, van Dijk  L, Bijlsma  JWJ, Leufkens  HGM.  Do rheumatoid arthritis patients have equal access to treatment with new medicines? tumour necrosis factor-alpha inhibitors use in four European countries.  Health Policy. 2012;104(1):76-83. doi:10.1016/j.healthpol.2011.10.011PubMedGoogle Scholar
    13.
    Souliotis  K, Papageorgiou  M, Politi  A, Ioakeimidis  D, Sidiropoulos  P.  Barriers to accessing biologic treatment for rheumatoid arthritis in Greece: the unseen impact of the fiscal crisis: the Health Outcomes Patient Environment (HOPE) study.  Rheumatol Int. 2014;34(1):25-33. doi:10.1007/s00296-013-2866-1PubMedGoogle Scholar
    14.
    Orlewska  E, Ancuta  I, Anic  B,  et al.  Access to biologic treatment for rheumatoid arthritis in Central and Eastern European (CEE) countries.  Med Sci Monit. 2011;17(4):SR1-SR13. doi:10.12659/MSM.881697PubMedGoogle Scholar
    15.
    Neovius  M, Simard  JF, Askling  J; ARTIS study group.  Nationwide prevalence of rheumatoid arthritis and penetration of disease-modifying drugs in Sweden.  Ann Rheum Dis. 2011;70(4):624-629. doi:10.1136/ard.2010.133371PubMedGoogle Scholar
    16.
    Neovius  M, Sundström  A, Simard  J,  et al; ARTIS Study Group.  Small-area variations in sales of TNF inhibitors in Sweden between 2000 and 2009.  Scand J Rheumatol. 2011;40(1):8-15. doi:10.3109/03009742.2010.493895PubMedGoogle Scholar
    17.
    Putrik  P, Sokka  T, Ramiro  S, Boonen  A.  Impact of socioeconomic gradients within and between countries on health of patients with rheumatoid arthritis (RA): lessons from QUEST RA.  Best Pract Res Clin Rheumatol. 2012;26(5):705-720. doi:10.1016/j.berh.2012.07.011PubMedGoogle Scholar
    18.
    Putrik  P, Ramiro  S, Kvien  TK,  et al; Working Group ‘Equity in access to treatment of rheumatoid arthritis in Europe’.  Inequities in access to biologic and synthetic DMARDs across 46 European countries.  Ann Rheum Dis. 2014;73(1):198-206. doi:10.1136/annrheumdis-2012-202603PubMedGoogle Scholar
    19.
    Putrik  P, Ramiro  S, Lie  E,  et al.  Less educated and older patients have reduced access to biologic DMARDs even in a country with highly developed social welfare (Norway): results from Norwegian cohort study NOR-DMARD.  Rheumatology (Oxford). 2016;55(7):1217-1224. doi:10.1093/rheumatology/kew048PubMedGoogle Scholar
    20.
    Putrik  P, Ramiro  S, Kvien  TK, Sokka  T, Uhlig  T, Boonen  A; Equity in Clinical Eligibility Criteria for RA treatment Working Group.  Variations in criteria regulating treatment with reimbursed biologic DMARDs across European countries: are differences related to country’s wealth?  Ann Rheum Dis. 2014;73(11):2010-2021. doi:10.1136/annrheumdis-2013-203819PubMedGoogle Scholar
    21.
    Brookhart  MA, Stürmer  T, Glynn  RJ, Rassen  J, Schneeweiss  S.  Confounding control in healthcare database research: challenges and potential approaches.  Med Care. 2010;48(6)(suppl):S114-S120. doi:10.1097/MLR.0b013e3181dbebe3PubMedGoogle Scholar
    22.
    Sommers  BD, Gawande  AA, Baicker  K.  Health insurance coverage and health: what the recent evidence tells us.  N Engl J Med. 2017;377(6):586-593. doi:10.1056/NEJMsb1706645PubMedGoogle Scholar
    23.
    Kessler  R, Glasgow  RE.  A proposal to speed translation of healthcare research into practice: dramatic change is needed.  Am J Prev Med. 2011;40(6):637-644. doi:10.1016/j.amepre.2011.02.023PubMedGoogle Scholar
    24.
    Rosenbaum  PR.  Design of Observational Studies. New York, NY: Springer; 2010.
    25.
    ICES. https://www.ices.on.ca/. Accessed November 5, 2019.
    26.
    von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  J Clin Epidemiol. 2008;61(4):344-349. doi:10.1016/j.jclinepi.2007.11.008PubMedGoogle Scholar
    27.
    The Johns Hopkins ACG Case Mix System.  Version 10 Release Notes. PC (DOS/WIN/NT) and Unix Version 10.0. Baltimore, MD: Johns Hopkins Bloomberg School of Public Health; 2014.
    28.
    Kralj  B.  Measuring rurality: RIO2008 BASIC: methodology and results. https://www.oma.org/wp-content/uploads/2008rio-fulltechnicalpaper.pdf. Accessed October 29, 2019.
    29.
    Statistics Canada.  Income Reference Guide, Census of Population, 2016. https://www150.statcan.gc.ca/n1/pub/82-221-x/2013001/quality-qualite/qua8-eng.htm. Accessed October 29, 2019.
    30.
    Matheson  FI, Dunn  JR, Smith  KLW, Moineddin  R, Glazier  RH.  Development of the Canadian Marginalization Index: a new tool for the study of inequality.  Can J Public Health. 2012;103(8)(suppl 2):S12-S16. doi:10.1007/BF03403823PubMedGoogle Scholar
    31.
    Evans  JL.  The dynamic behavior of alternative price adjustment mechanisms.  Manchester School. 1983;51(1):33-44. doi:10.1111/j.1467-9957.1983.tb00739.xGoogle Scholar
    32.
    Burke  N, Bowen  JM, Troyan  S,  et al.  Management of hospital formularies in Ontario: challenges within a local health integration network.  Can J Hosp Pharm. 2016;69(3):187-193. doi:10.4212/cjhp.v69i3.1554PubMedGoogle Scholar
    33.
    Austin  PC.  A tutorial on multilevel survival analysis: methods, models and applications.  Int Stat Rev. 2017;85(2):185-203. doi:10.1111/insr.12214PubMedGoogle Scholar
    34.
    Therneau  T.  Mixed effects Cox models. https://cran.r-project.org/web/packages/coxme/vignettes/coxme.pdf. Accessed March 17, 2018.
    35.
    Statistics Canada. 2016 Census topic: population and dwelling counts. https://www12.statcan.gc.ca/census-recensement/2016/rt-td/population-eng.cfm. Accessed October 29, 2019.
    36.
    Kalkan  A, Husberg  M, Hallert  E,  et al.  Physician preferences and variations in prescription of biologic drugs for rheumatoid arthritis: a register-based study of 4,010 patients in Sweden.  Arthritis Care Res (Hoboken). 2015;67(12):1679-1685. doi:10.1002/acr.22640PubMedGoogle Scholar
    37.
    Eriksson  JK, Johansson  K, Askling  J, Neovius  M.  Costs for hospital care, drugs and lost work days in incident and prevalent rheumatoid arthritis: how large, and how are they distributed?  Ann Rheum Dis. 2015;74(4):648-654. doi:10.1136/annrheumdis-2013-204080PubMedGoogle Scholar
    38.
    Ministry of Health and Long Term Care. Ontario public formulary. http://www.health.gov.on.ca/en/pro/programs/drugs/odbf_eformulary.aspx. Accessed October 29, 2019.
    39.
    Harrison  MJ, Tricker  KJ, Davies  L,  et al.  The relationship between social deprivation, disease outcome measures, and response to treatment in patients with stable, long-standing rheumatoid arthritis.  J Rheumatol. 2005;32(12):2330-2336.PubMedGoogle Scholar
    40.
    Brekke  M, Hjortdahl  P, Thelle  DS, Kvien  TK.  Disease activity and severity in patients with rheumatoid arthritis: relations to socioeconomic inequality.  Soc Sci Med. 1999;48(12):1743-1750. doi:10.1016/S0277-9536(99)00075-1PubMedGoogle Scholar
    41.
    Molina  E, Del Rincon  I, Restrepo  JF, Battafarano  DF, Escalante  A.  Association of socioeconomic status with treatment delays, disease activity, joint damage, and disability in rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2015;67(7):940-946. doi:10.1002/acr.22542PubMedGoogle Scholar
    42.
    Marra  CA, Lynd  LD, Esdaile  JM, Kopec  J, Anis  AH.  The impact of low family income on self-reported health outcomes in patients with rheumatoid arthritis within a publicly funded health-care.  Rheumatology (Oxford). 2004;43(11):1390-1397. doi:10.1093/rheumatology/keh334PubMedGoogle Scholar
    43.
    Widdifield  J, Bernatsky  S, Paterson  JM,  et al.  Serious infections in a population-based cohort of 86,039 seniors with rheumatoid arthritis.  Arthritis Care Res (Hoboken). 2013;65(3):353-361. doi:10.1002/acr.21812PubMedGoogle Scholar
    ×