Abstract
Background: Financial incentives may improve primary care access for adults with schizophrenia or bipolar disorder (serious mental illness [SMI]). We studied the association between receipt of the SMI financial premium paid to primary care physicians and rostering of adults with SMI in different patient enrolment models (PEMs), including enhanced fee-for-service and capitation-based models with and without interdisciplinary team-based care.
Methods: We conducted a retrospective cohort study involving Ontario adults (≥18 yr) with SMI in PEM practices, in fiscal years 2016/17 and 2017/18. Using negative binomial models, we examined relations between rostering and the primary care model and the contribution of the incentive. Similar models were developed for adults with type 1 or 2 diabetes mellitus and the general population.
Results: Among 9730 physicians in PEM practices, 4866 (50.0%) received a premium and 448 319 (88.4%) people with SMI in PEMs were rostered. Compared with enhanced fee for service, the likelihood of rostering people with SMI was 3.0% higher for patients in capitation with team-based care (adjusted relative risk [RR] 1.03, 95% confidence interval [CI] 1.02–1.04), with similar results for capitation without team-based care (adjusted RR 1.00 95% CI 0.99–1.01). Rostering for people with diabetes was similar in team-based care (adjusted RR 1.02, 95% CI 1.02–1.03) but higher in capitation without team-based care (adjusted RR 1.03, 95% CI 1.02–1.03) and slightly higher for the Ontario population (team-based care 1.04, 95% CI 1.04–1.05, capitation without team-based care 1.03, 95% CI 1.03–1.04).
Interpretation: Rostering of people with SMI was lower than for the general population. Additional policy measures are needed to address persisting inequities and to promote rostering of this underserved population with complex needs.
Mental illnesses are prevalent, affecting 10%–20% of adults per year1,2 and up to 33% over their lifetime.1 They are responsible for an estimated 22.9% of years lived with a disability3 and a mortality gap estimated at 13–20 years,4 of which 60% of deaths are attributable to chronic conditions including cardiovascular and respiratory disease.4
Primary care physicians are the most frequently consulted health care professionals by adults with schizophrenia and bipolar disorder, collectively referred to as serious mental illnesses (SMI).5 However, adults with SMI are less likely to have an ongoing site of primary care6 and experience both difficulty accessing primary care6,7 and lower quality of care.8,9 Patient-reported barriers to accessing care occur at the patient level (socioeconomic and mental health or medication related), provider level (perceived stigma and lack of willingness to address mental health concerns) and the health system level (difficulty finding a family physician, inadequate time during appointments to meet their health needs and poor collaboration with other health care providers).7
Since 2000, Ontario has implemented a broad suite of voluntary reforms in the delivery and payment of primary care, aimed at improving access, quality of care and retention of primary care physicians.10 More than 75% of primary care physicians shifted from exclusive fee-for-service to new primary care models involving patient enrolment.10 Patient enrolment is voluntary; physicians can choose to provide care to some patients without rostering them (billing fee for service), and may be incented to do so if they anticipate that these patients may have more complex needs, requiring more frequent visits. Unrostered patients are not included in provincial quality improvement reporting to practices. Previous work has shown that fewer people with mental illnesses were enrolled in new models11 and that people with SMI who were enrolled in capitation models accessed fewer health services than those enrolled in enhanced fee-for-service models.12
Incentives to enrol patients with SMI were included in the reforms in 2003.13 We examined the effect of the SMI premium on primary care rostering in different primary care patient enrolment models (PEMs). We hypothesized that people with SMI would have lower rates of rostering than those with another complex chronic disease (diabetes mellitus) and the general population of Ontario. We also hypothesized that premium payment would be associated with the increased likelihood of rostering adults with SMI.
Methods
We conducted a retrospective observational cohort study using population-level administrative data provided by ICES. We examined the effect of the primary care SMI special premium incentives available to physicians practising in PEMs. The PEMs include the enhanced fee-for-service model (remunerated by fee-for-service payments with some bonuses for preventive care) and blended capitation models with and without integration of interdisciplinary team-based care (remunerated by capitation payments based on age and sex for in-basket services, and additional bonuses for comprehensive and preventive care). The primary care SMI special premium is an annual payment paid to physicians practising in PEMs for providing comprehensive primary care to a minimum of 5 enrolled patients with diagnoses of bipolar disorder or schizophrenia.14 There are 2 levels of payment: $1000 for the minimum first 5 enrolled patients and $1000 for an additional 5 or more enrolled patients (maximum $2000 annually).
We reported data in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines.15
Study participants
Study participants included all adult (≥ 18 yr) Ontario residents eligible for publicly funded health insurance who were attributed (either rostered or virtually rostered) to primary care physicians practising in PEMs (Appendix 1, available at www.cmajopen.ca/content/11/1/E1/suppl/DC1). We created a cohort of people with schizophrenia or bipolar disorder and additional cohorts for people with type 1 or 2 diabetes mellitus and the general population of adults in Ontario (≥ 18 yr) for comparative purposes. Study inclusion dates were from Apr. 1, 2016, to Mar. 31, 2018. We identified people with schizophrenia (defined as a psychotic disorder characterized by disturbances in thinking, emotional responsiveness and behaviour)16 or bipolar disorder (defined as a group of brain disorders that cause extreme fluctuation in a person’s mood, energy and ability to function)16 if they had at least 1 outpatient visit at any time before the study period with a family physician or psychiatrist, or an emergency department visit or an inpatient hospital admission billing the diagnostic codes schizophrenia-schizoaffective disorder (International Classification of Diseases [ICD]-9: 295; ICD-10: F20, F25) or bipolar disorder (ICD-9: 296; ICD-10: F31). We identified people with diabetes using a validated administrative case definition.17 Exclusion criteria included children (< 18 yr), adults lacking a valid Ontario health card and who were therefore ineligible for Ontario health insurance and adults who died during the study period.
Primary care physicians were defined as those whose specialty was listed as general practitioner or family physician in the Corporate Provider Database.
Patients were attributed to a physician if they were formally enrolled (rostered) or had attended a minimum of 3 visits with the same primary care provider during the study period (virtually rostered). Previous work has virtually rostered patients to the physician who billed the largest dollar amount for primary care services in the preceding 2 years.18 We used a higher threshold of 3 visits with the same physician over the 2-year study period for virtual rostering in light of the high needs of this population, which is consistent with the approach previously used as a proxy measure for having a regular primary care physician for people with a chronic condition.19
For comparative purposes, we identified adult Ontario residents with type 1 or type 2 diabetes mellitus using the Ontario Diabetes Database, an ICES-derived cohort,17 who had a diabetes-related primary care visit in the 3 years before the study period (between Apr. 1, 2013, and Mar. 31, 2016) and an adult general population comparison sample.
Data sources
Several data sets were linked using unique encoded identifiers and analyzed at ICES (Appendix 2, available at www.cmajopen.ca/content/11/1/E1/suppl/DC1).20 To identify and describe the cohort, we used the Registered Persons Database (a registry of all Ontario residents eligible for the Ontario Health Insurance Plan [OHIP]); the National Ambulatory Care Reporting System (a registry of emergency department visits); the Discharge Abstracts Database (a registry of inpatient hospital admissions); and the Ontario Mental Health Reporting System (a registry of mental health care contacts including hospital admissions). We derived age, sex, rurality and recent migration status from the Registered Persons Database. We measured rurality using postal codes and the Rurality Index for Ontario, with categories of urban (score 0–9), suburban (score 10–39) and rural (score ≥ 40).21 We derived neighbourhood income quintile using postal codes linked to the census dissemination area. We identified recent migrants to Ontario as people who received an Ontario health card for the first time within the previous 10 years (about 75% of this group would be expected to be recent immigrants and the remainder would be expected to have migrated from other Canadian provinces).22 We used the Johns Hopkins Adjusted Clinical Groups System, Version 10, to capture comorbidity according to Aggregated Diagnosis Groups.23 We derived health service utilization from the OHIP, the National Ambulatory Care Reporting System, the Discharge Abstracts Database and the Ontario Mental Health Reporting System databases,20 including primary care attachment,24 number of core primary care visits25 and number of psychiatric visits. Health service utilization and comorbidity were also examined over the 3 years before the study period (Apr. 1, 2013–Mar. 20, 2016), referred to as the look-back period, for additional contextual data. Core primary care visits were defined by billing codes for which 80% or more of all billings were submitted by primary care physicians, and those for which total primary care billings for the code represented at least 0.1% of all billings by primary care physicians.25
We identified primary care physicians and utilization using the Corporate Provider Database (a registry of all providers and provider groups eligible to bill OHIP for their services), the Client Agency Program Enrolment database (which lists all patients enrolled with a primary care physician within a primary care group) and Primary Care Population Database (an ICES derived cohort which includes data on primary care rostering, models of care and health system utilization for Ontario residents eligible for health services). We derived physician characteristics (age, sex, panel size, years since medical school graduation) from the Corporate Provider Database. We derived payment of SMI premiums from the architected payments data set, which includes physician payments that do not pertain to individual patient level services, such as premiums and bonuses, that are summed across a physician’s entire practice.
Variables
Outcome
The dependent variable was the percentage of adults with SMI, diabetes mellitus and in the general population who were rostered, defined at the physician level, during the study period. The percent rostered was calculated as the proportion of the number rostered, divided by the total rostered and virtually rostered.
Exposure
The primary independent variable was primary care physician model of care (enhanced fee for service, team-based care and capitation without team-based care). To assess the relative contribution of the SMI premium to rostering, we created models with and without SMI premiums to assess change in model estimates.
Covariates
Covariates were selected theoretically based on literature on factors associated with access to care and opportunities for rostering to address potential confounding and included patient age, sex, income quintile, newly arrived in Ontario, rurality, Aggregated Diagnosis Groups, health system utilization within 3 years before study dates (primary care attachment, mean number of primary care visits, psychiatric hospital admissions), primary care visits during the study period, continuity of care and physician age, sex, rurality of practice, location of training, panel size and amount of SMI premium paid.11,12
Statistical analysis
We compared the demographic characteristics of people with SMI, with those with diabetes mellitus and with the adult population in Ontario, including those who were rostered and virtually rostered using consistent approaches to rostering among all 3 populations. Next, we compared the characteristics of physicians receiving SMI premium payments with those who did not receive these payments during the study period. For the outcome of number of patients rostered, patient data were aggregated at the physician level and the unit of observation was the primary care physician.
We determined that the outcome (number of patients rostered) was overdispersed, and, therefore, developed negative binomial models to model the counts of the number of rostered patients, with an offset for the log of the number of patients in the practice. Using complete cases, we modelled the relations between the number of rostered patients in the practice (by condition or the Ontario population) and the model of primary care. To examine the relative contribution of SMI premium payment status, we added this variable into each model to assess change in model estimates. Physicians with fewer than 100 patients in total (rostered or virtually rostered) were excluded. The means for continuous variables and the frequencies in each category represented for categorical variables were calculated. We adjusted for a number of patient and physician characteristics as prespecified covariates. Patient characteristics were aggregated at the physician level. Patient characteristics included in the model were age, sex, rurality, recent migration, neighbourhood income, comorbidity using Aggregated Diagnosis Groups, continuity of care and health care utilization in the lookback period (primary care attachment,24 number of primary care visits and number of psychiatric hospital admissions). Continuity of care was determined at the practice level for patients with at least 3 primary care visits during the study period and was defined as the proportion of primary care visits with the patient’s own provider. Physician-related covariates were physician age, sex, rurality, panel size, model of care and primary care visits during the study period. We repeated the analyses weighting the observations by the sum of rostered and virtually rostered patients, both with and without panel size included as a covariate in the model, to address concerns about physicians with different practice sizes having the same weight in the analysis. Finally, we did a weighted analysis including panel size but excluding SMI premium in the model. Analyses were completed using SAS version 9.4 (SAS Institute).
Ethics approval
The study was approved by the Queen’s University Health Sciences Research Ethics Board. ICES is an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.
Results
We identified 592 431 adults in Ontario with an SMI (212 369 with schizophrenia and 380 062 with bipolar disorder) between Apr. 1, 2016, and Mar. 30, 2018, representing 5.7% of the general population in Ontario (Table 1). People with schizophrenia and bipolar disorder were more likely to live in lower income neighbourhoods (particularly those with schizophrenia) and in urban centres, and less likely to be recent immigrants to Ontario, than the general population. People with SMI were less likely to be attributed to a primary care provider (n = 559 505, 94.4% v. n = 990 193, 98.4%), and less likely to have accessed any primary care and to have lower continuity of care, in contrast to people with diabetes mellitus.
Among the 13 606 Ontario family physicians identified, 9730 (71.5%) practised in PEMs and would have been eligible to receive the SMI premium, and 4866 (50.0%) received a premium during the study period based on having at least 5 SMI patients on their roster (Table 2). Only 90 physicians were in a PEM and had at least 5 SMI patients in their roster but did not receive the premium. Compared with physicians practising in PEMs who were ineligible for the premium by having too few patients, those who received the highest premium payments were more likely to be male, had larger patient panel size and were more likely to work in capitation models (with and without team-based care). The patient panels of physicians practising in PEMs who were ineligible for the premiums did not differ by age and sex from the patient panels of physicians who received the premium or those who were not practising in PEMs, but included higher proportions of patients who were recent immigrants or living in urban settings (Table 3). Compared with practices of physicians practising in PEMs patients of physicians who were not practising in PEMs, were more likely to live in low-income neighbourhoods, be new immigrants, have higher morbidity and have more primary care visits. In total, $12 750 400 was paid in SMI premiums during the study period.
Among the 592 431 Ontario adults with SMI, 507 158 (85.6%) received primary care in PEMs, compared with 916 506 (91.0%) people with diabetes and 8 954 863 (85.6%) of the Ontario general population. Among the 507 158 people with SMI receiving primary care through PEMs, 88.4% were formally rostered, compared with 854 668 (93.3%) people with diabetes and 8 135 246 (90.8%) of the Ontario general population (Table 4). The proportion of adults with SMI rostered was consistently lower than those for either people with diabetes or in the Ontario general population across all patient and physician characteristics and all models of care. For people with SMI, rostering ranged from 145 252 (85.2%) for enhanced fee-for-service models, 147 487 (91.0%) for capitation models with team-based care and 149 674 (88.7%) for capitation models without team-based care, which were all less than rates observed for diabetes (90.6%–95.2%) and the Ontario general population (86.1%–94.1%) (Table 4).
Adjusted negative binomial models of the number of patients rostered, using panel size as an offset, determined that compared with enhanced fee for service, the likelihood of physicians rostering people with SMI was higher for those in capitation models with team-based care (adjusted relative risk [RR] 1.03 confidence interval [CI] 1.02–1.04) but not for capitation models without team-based care (adjusted RR 1.00 95% CI 0.99–1.01) (Table 5). In similar modelling for the population with diabetes, we found that the likelihood of rostering of patients with diabetes compared with enhanced fee for service was higher for capitation models with team-based care (adjusted RR 1.02, 95% CI 1.02–1.03) and for capitation models without team-based care (adjusted RR 1.02, 95% CI 1.02–1.03). Parameter estimates for similar modelling for the Ontario general population were capitation with team-based care (adjusted RR 1.04, 95% CI 1.04–1.05) and capitation without team-based care (adjusted RR 1.03, 95% CI 1.03–1.04). Probability testing for each of these models was conducted, testing against the null that there is no difference across enrolment models and was significant for all 3 models (p < 0.001). When the SMI premium was included in the model, the parameter estimates were unchanged.
Interpretation
Thirteen years after introduction of reforms into the payment and structure of primary care, including a financial incentive to promote enrolment of people with SMI, we found evidence of lower enrolment into new models for people with SMI compared with the Ontario general population. Including the SMI premium payment did not substantially change parameter estimates of the relation between enrolment model and rostering, suggesting that the SMI premium payment was not associated with rostering of SMI patients into PEM models.
People with SMI have complex needs, and it is encouraging to observe that overall rostering was quite high. Nevertheless, inequitable access to new models (shown by lower enrolment than for the Ontario general population) was still observed. In Ontario, provincial quality improvement systems, including incentives and practice-level reporting, for preventive care (such as cancer screening and vaccination) apply only to rostered patients. Lower rostering of people with SMI may then translate into lower quality of preventive care and contribute to adverse outcomes in a high-needs population with elevated risks of chronic disease, including cancer.8,26 Furthermore, the incentive structure itself may limit its effect. Once a provider has enrolled 10 patients with SMI, there is no additional incentive to enrol additional patients. Modified capitation as implemented in Ontario includes adjustments for age and sex, but not for case mix, thereby embedding disincentives for enrolment of patients with complex needs. It is noteworthy that the intention to incorporate case-mix adjustment was outlined in the recent Physician Services Agreement, with specific details pending.27
Our findings are consistent with a substantial body of research demonstrating the limited effect of pay-for-performance measures. Pay for performance has been implemented in many countries and settings, and by using different structures and targets. A recent systematic review found that most pay-for-performance programs target chronic disease management in primary care, and found evidence of short-term improvements in process of care outcomes, but little or no effect was shown for improved health outcomes (intermediate or patient-important outcomes), or longer term improvements.28 Older systematic reviews drew similar conclusions.29,30 Few studies have examined pay for performance for mental health care. Rudoler and colleagues31 found no increased provision of follow-up care after psychiatric hospital admission or after suicide attempts after implementation of a financial incentive. In the United Kingdom, financial incentives were associated with improvements in screening and intervention on physical health (weight, blood pressure, lipid and glucose screening) in people with psychosis in secondary care.32 Gutacker and colleagues33 found that better performance on quality metrics of mental health care in the UK was associated with higher rates of psychiatric hospital admission. A pay-for-performance program in Taiwan was associated with reduction in unscheduled outpatient visits and compulsory admissions but no change in emergency department visits or acute psychiatric hospital admissions or readmissions.34 In British Columbia, incentives targeting primary mental health care for people with depression were associated with incremental improvements in the targeted domains but worsening continuity of care.35
Limitations
Study strengths include the inclusion of linked population level data, which limits potential selection bias. Our study has some limitations. The administrative data used were not designed for research purposes. Only people with valid health care coverage were included, which limited the sample to permanent residents of Ontario. The cross-sectional design precludes determination of a causal relation between premium payment with increased enrolment of people with SMI into new models. In addition, the results may be biased by residual confounding. For example, a substantial proportion of people experiencing homelessness or precarious housing are affected by SMI36 and have challenges accessing health care services.37,38 However, history of homelessness is not reported in administrative data. We also could not directly assess severity of illness through administrative data, though we used psychiatric hospital admissions as a proxy measure. We were unable to account for clustering at a clinic level. In addition, we could not account for people who did not access health services during the study period. Nonetheless, we expect the effect to be limited as we believe we have been thorough in identifying relevant confounders. The diagnostic code to select for bipolar disorder has not been validated. The extent to which it may include people with major depressive disorder is unclear.
Conclusion
This study found that incentives were not associated with rostering SMI patients. Though overall rostering for people with SMI was high, there were still inequities in the likelihood to be rostered. Additional policy measures are needed to promote rostering of this underserved population with complex needs.
Footnotes
Competing interests: Imaan Bayoumi reports the Centre for Studies in Primary Care Research Initiation Grant from Queen’s University funded this work. Imaan Bayoumi received funded grants from the Canadian Institutes of Health Research and a funded grant and salary award from PSI Foundation, unrelated to this work, and has received an honorarium for work on the Core Group of the Rourke Baby Record, from the Government of Ontario. No other competing interests were declared.
This article has been peer reviewed.
Contributors: All authors take responsibility for the integrity of the data and the accuracy of the data analysis. All authors conceptualized and designed the study, interpreted the data, critically revised and reviewed the manuscript for important intellectual content, and approved the final manuscript. Imaan Bayoumi prepared the initial draft of the manuscript. Marlo Whitehead and Wenbin Li performed the statistical analysis.
Funding: This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care. This study also received funding from the Centre for Studies in Primary Care Research Initiation Grant, Queen’s University. Parts of this material are based on data and information compiled and provided by the MOH, the Canadian Institute for Health Information, Cancer Care Ontario and The Johns Hopkins ACG® System Version 10. The analyses, conclusions, opinions and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.
Data sharing: The data set from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., health care organizations and government) prohibit ICES from making the data set publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at https://www.ices.on.ca/DAS (email: das{at}ices.on.ca). The full data set creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification.
Supplemental information: For reviewer comments and the original submission of this manuscript, please see www.cmajopen.ca/content/11/1/E1/suppl/DC1.
This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC-ND 4.0) licence, which permits use, distribution and reproduction in any medium, provided that the original publication is properly cited, the use is noncommercial (i.e., research or educational use), and no modifications or adaptations are made. See: https://creativecommons.org/licenses/by-nc-nd/4.0/
References
- © 2022 CMA Impact Inc. or its licensors