ACSC indicates ambulatory care–sensitive condition; CKD, chronic kidney disease; and HbA1c, hemoglobin A1c.
Costs are reported in 2016 Canadian dollars (2016 exchange rate: $1.00 Canadian dollar = $0.75 US dollar). ACSC indicates ambulatory care–sensitive condition.
aIncludes other specialists and primary care physician costs.
bAmbulatory care–sensitive conditions include chronic kidney disease–specific ACSCs27 and the following Canadian Institutes for Health Information–defined conditions: chronic obstructive pulmonary disease, asthma, diabetes, heart failure and pulmonary edema, hypertension, and angina.26
cIncludes nephrology, cardiology, and diabetes clinics.
dChronic disease medications include antiarrhythmic drugs, nitrates and nitrites, statins, nonstatin cholesterol-lowering drugs, β-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, calcium channel blockers, diuretics, other blood pressure–lowering medications, anticoagulants, antidiabetes medications, antiplatelet agents, insulin, smoking cessation aids, erythropoietin, and darbepoetin alfa.
eIncludes the 25 of the most frequently ordered diagnostic tests at Canadian laboratories: complete blood cell count; prothrombin time (international normalized ratio); and creatinine, alanine aminotransferase, thyroid-stimulating hormone, hemoglobin A1c, low-density lipoprotein cholesterol, ferritin, alkaline phosphatase, albumin, random glucose, fasting glucose, calcium, urea, magnesium, iron and total iron-binding capacity, phosphate, total bilirubin, creatine kinase, free thyroxine, prostate-specific antigen, urate, lactate dehydrogenase, lipase, and albumin random urine levels.
fTotal is the sum of all categorical costs.
eTable 1. Comorbidity Characteristics of Patients With Diabetes or Chronic Kidney Disease by Physician Payment Model, Before and After Matching by Propensity Score
eTable 2. Unadjusted Overall Visit Rate for the Matched Cohort by Specialist Physician Characteristics and Payment Model
eTable 3. Unadjusted Mean Total and Categorical Costs per Patient (2016 CAD$) of Matched Cohort, One and Two Years After Index Visit by Primary Specialist Payment Model
Customize your JAMA Network experience by selecting one or more topics from the list below.
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.
Quinn AE, Hemmelgarn BR, Tonelli M, et al. Association of Specialist Physician Payment Model With Visit Frequency, Quality, and Costs of Care for People With Chronic Disease. JAMA Netw Open. 2019;2(11):e1914861. doi:https://doi.org/10.1001/jamanetworkopen.2019.14861
Is a specialist physician payment model associated with visit frequency, quality of care, and costs for people with chronic disease?
In this population-based cohort study that included a propensity-score matched cohort of 31 898 adults with diabetes or chronic kidney disease seen by 489 physicians, there was no statistical evidence of a difference in follow-up outpatient visit rates, quality, and costs between patients seeing salaried and fee-for-services physicians. The median association of physician clustering and the outcomes was greater than the association with the physician payment model.
Specialist physician payment does not appear to be associated with variation in use of chronic disease care, quality, and costs; however, these findings suggest large variation in outcomes between physicians.
Specialist physicians are key members of chronic care management teams; to date, however, little is known about the association between specialist payment models and outcomes for patients with chronic diseases.
To examine the association of payment model with visit frequency, quality of care, and costs for patients with chronic diseases seen by specialists.
Design, Setting, and Participants
A retrospective cohort study using propensity-score matching in patients seen by a specialist physician was conducted between April 1, 2011, and September 31, 2014. The study was completed on March 31, 2015, and data analysis was conducted from June 2017 to February 2018 and finalized in August 2019. In a population-based design, 109 839 adults with diabetes or chronic kidney disease newly referred to specialists were included. Because patients seen by independent salary-based and fee-for-service (FFS) specialists were significantly different in observed baseline characteristics, patients were matched 1:1 on demographic, illness, and physician characteristics.
Specialist physician payment model (salary-based or FFS).
Main Outcomes and Measures
Follow-up outpatient visits, guideline-recommended care delivery, adverse events, and costs.
A total of 90 605 patients received care from FFS physicians and 19 234 received care from salary-based physicians. Before matching, the patients seen by salary-based physicians had more advanced chronic kidney disease (2630 of 14 414 [18.2%] vs 6627 of 54 489 [12.2%]), and a higher proportion had 5 or more comorbidities (5989 of 19 234 [31.3%] vs 23 326 of 90 605 [25.7%]). Propensity-score matching resulted in a cohort of 31 898 patients (15 949 FFS, 15 949 salary-based) seeing 489 specialists. In the matched cohort, patients were similar (mean [SD] age, 61.3 [18.2] years; 17 632 women [55.3%]; 29 251 residing in urban settings [91.7%]). Patients seen by salary-based specialists had a higher follow-up visit rate compared with those seen by FFS specialists (1.74 visits; 95% CI, 1.58-1.92 visits vs 1.54 visits; 95% CI, 1.41-1.68 visits), but the difference was not significant (rate ratio, 1.13; 95% CI, 0.99-1.28; P = .06). There was no statistical difference in guideline-recommended care delivery, hospital or emergency department visits for ambulatory care–sensitive conditions, or costs between patients seeing FFS and salary-based specialists. The median association of physician clustering with health care use and quality outcomes was consistently greater than the association with the physician payment, suggesting variation between physicians (eg, median rate ratio for follow-up outpatient visit rate was 1.74, which is greater than the rate ratio of 1.13).
Conclusions and Relevance
Specialist physician payment does not appear to be associated with variation in visits, quality, and costs for outpatients with chronic diseases; however, there is variation in outcomes between physicians. This finding suggests the need to consider other strategies to reduce physician variation to improve the value of care and outcomes for people with chronic diseases.
Noncommunicable chronic diseases pose a major challenge for health systems worldwide owing to rising prevalence and costs.1 Chronic disease management models have focused on the role of primary care2; however, specialists are also key members of the chronic care team providing additional support and care to patients with more complex needs.3 Outpatient care for chronic conditions is frequently suboptimal. There are effective interventions for chronic diseases,4-6 but less than half of eligible patients receive them.7 Physician payment has been identified as a barrier to the delivery of effective interventions for chronic disease.8
Fee-for-service (FFS) is the dominant physician compensation model in the United States and Canada: 95% of physician office visits in the United States9 and 72% of clinical payments in Canada10 were reimbursed under FFS. There is a robust literature on the association between payment mechanisms and physician behavior, but most of the empirical work addresses primary care payment and little is known about how specialists respond to payment models in general or, specifically, when caring for patients with chronic diseases. Studies in primary care found that FFS payment is associated with a higher number of primary care and specialty care visits compared with salary payments.11,12 Similarly, studies in specialty care have found FFS payment to be associated with increased health care use, including higher specialist visit rates13 and a higher volume of specialist physicians’ billable services, particularly for elective procedures.14 To date, few quasi-experimental studies of specialist payment have examined quality outcomes, and limited data exist on the outcomes associated with costs.15
Given the uncertainty around the association between the specialist physician payment model and care for patients with chronic diseases, we sought to examine the association of payment model with outpatient visit frequency, quality of care, and costs of care for patients with chronic diseases seen for the first time by a specialist by studying a payment reform in Alberta, Canada. The Academic Alternative Relationship Plan, offered to specialists in Calgary and Edmonton, Alberta, in 2004, compensated physicians with a salary-like payment that covered clinical, research, and teaching time. Physicians remained independent contractors and were not employees of Alberta’s health system. The goals were increased recruitment and retention, innovative care delivery, improved access to specialists, delivery of high-quality care, delivery of high-quality education, greater research capacity, and improved governance and accountability.16
We used the Interdisciplinary Chronic Disease Collaboration Data Repository.17-20 The database captures demographic, laboratory, and administrative health data (including vital statistics, prescription drug data, physician claims, hospitalizations, emergency department and outpatient visits, and all health care costs) for all individuals registered with Alberta Health (all residents of Alberta are eligible for insurance coverage; >99% participate17). Data are deidentified. This study was approved by University of Calgary’s Conjoint Health Research Ethics Board with waiver of informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
The cohort included newly referred adults with preexisting diabetes or nondialysis chronic kidney disease (CKD) seen by a specialist physician (endocrinologist, nephrologist, or internal medicine physician) in Alberta between April 1, 2011, and September 30, 2014, with no visit to the same physician or a physician from the same specialty in the 4 years before their index visit.21 In Canada, internal medicine physicians deliver specialty care. We selected and combined these conditions because they represent a large proportion of patients with chronic disease, frequently co-occur, and share common risk factors and treatment guidelines. Only specialists practicing in Calgary and Edmonton were included because salary-based programs are only available in those cities.
Patients with diabetes were identified using a validated algorithm.22,23 Chronic kidney disease was defined by serum creatinine level or—if serum creatinine level was not available—albuminuria or urine dipstick laboratory measurements before the index visit as in prior work.24 We excluded patients who (1) started dialysis within 90 days of their index visit, (2) had kidney or islet cell transplant before their index visit, (3) were seen for a preoperative visit (index visit within 60 days of an elective surgery25,26), and (4) saw a physician who switched payment models during the study period (Figure 1).
The primary outcome was the rate of outpatient visits with the primary specialist physician after the first visit. We included all patient visits (limited to 1 per day for estimation of rates) from physician claims data between the index date and the end of the study (March 31, 2015, unless a patient died, moved from Alberta, or started dialysis).
Secondary outcomes included the proportion of patients receiving guideline-recommended care and the rate of hospital admissions or emergency department visits for diabetes-specific, ambulatory care–sensitive conditions27 among patients with diabetes and for CKD-specific ambulatory care–sensitive conditions28 among those with CKD. Hospital visits were identified using hospitalization discharge data and emergency department visits were identified using ambulatory care data. Guideline-recommended care was defined using prescription drug data as the use of indicated medications 6 months after the index visit (angiotensin-converting enzyme inhibitors or angiotensin receptor blockers among people with CKD and albuminuria29,30 or diabetes and hypertension31; statins in patients aged >40 years with diabetes, those with diabetes and CKD, and in patients aged ≥50 years with CKD and no diabetes32,33) and using laboratory and claims data as the use of recommended testing following the index visit (urine albumin-creatinine ratio in those with CKD30 or diabetes; eye examinations and hemoglobin A1c [HbA1c (SI conversion factor: To covert HbA1c to proportion of total hemoglobin, multiply by 0.01)] level for those with diabetes34). Only patients with follow-up time required for each measure were eligible to be included.
Mean total and categorical costs per patient for services that could potentially be associated with specialist physicians’ behavior (primary specialist, other specialist, and primary care physician claims) for hospitalization and emergency department visits for ambulatory care–sensitive conditions (CKD,28 chronic obstructive pulmonary disease, asthma, diabetes, heart failure, pulmonary edema, hypertension, angina27), and use of ambulatory care (diagnostic imaging, laboratory tests, and prescription drug costs) were estimated for the first year and second year after the index visit for patients with 1 and 2 full years of follow-up data. Costs were censored if a patient died, moved from Alberta, or started dialysis. Primary specialist physician costs were estimated using actual amounts paid to FFS physicians and shadow billing (ie, claims submitted for administrative purposes that do not result in direct reimbursement) estimates for salary-based physicians. All costs are reported in 2016 Canadian dollars (2016 exchange rate: $1.00 Canadian dollar = $0.75 US dollar).35
The explanatory variable was the specialist physician payment model (salary-based vs FFS) at the patients’ index visit. A variable to consistently indicate payment model was added in 2011. Patient covariates included age, sex, neighborhood income quintile, rural vs urban residence status, prior use of health services, illness characteristics, and comorbidities defined using validated algorithms.36 Prior use of health services was defined as use of disease-related medications (ie, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, statins) in the 6 months before and hospitalizations and emergency department visits for ambulatory care–sensitive conditions in the year before the patient’s index visit. Illness characteristics included the presence of diabetes, CKD, or both conditions and details about the severity of their conditions based on laboratory measurements (ie, HbA1c, estimated glomerular filtration rate, albuminuria) and disease duration. Physician characteristics included physician type, defined as nephrologist, diabetes specialist (endocrinologists and internal medicine physicians with >50 different patients with diabetes each year and >30% of claims for outpatient diabetes care), and internal medicine physicians (excluding those in the diabetes specialist category); clinical workload (low, middle, and high based on the number of days billing per year); city of practice; and years practiced in Alberta since 1994.
We conducted the initial analysis from June 2017 to February 2018 and the final analysis in August 2019. We used propensity scores to adjust for baseline differences in patients and physicians. We first developed a multivariable logistic regression model including patient and physician characteristics to estimate the probability of being seen by a salary-based physician. Any missing data were defined as a category of variables included in the model (only 5 baseline characteristic variables had any missing data). We iteratively modified the patient and physician characteristics in the model if estimation and matching led to a sample with 1 or more characteristics with a standardized mean difference greater than or equal to 10. A standardized difference below 10% implied an acceptable balance.37 We matched 1 patient seen by an FFS physician with 1 patient seen by a salary-based physician (without replacement) on the closest matched propensity score (based on a caliper of 0.2 of the SD of the log odds of the propensity score).37,38 We assessed the balance in patient and physician characteristics between salary-based and FFS groups before and after matching using the standardized mean difference (expressed as a percentage). We further assessed the standardized differences in subgroup analyses (Figure 1).
Using the matched cohort, unadjusted overall visit rates were estimated using Poisson models for different physician characteristics by payment model. We examined the association between payment model and outcomes using generalized linear mixed models that included a fixed effect for payment model, a random intercept for a physician to address patients clustering within physicians, and a random residual for matched pairs to address correlation between propensity-matched pairs. We accounted for any imbalance (standardized difference ≥10%) in subgroups (Figure 1) by including unbalanced variables as covariates in mixed models. We assessed the association of payment model with visit rates and payment model with quality indicators by estimating rate ratios (RRs) using models of the Poisson family with a log link function, correcting for over-dispersion as necessary by using the negative binomial distribution.39 Follow-up visit rate and rate of hospital admissions and emergency department visits for diabetes- and CKD-specific ambulatory care–sensitive conditions mixed models included a denominator offset (ie, log number of days followed up) to account for patients’ variable follow-up times. Owing to the log-normal distribution of the random intercepts in the mixed models, FFS and salary-based rates and 95% CIs were transformed to population-averaged effects.40,41 The 95% CIs account for the variance of the fixed-effects estimates as well as the error in estimating the variance of the random intercepts. The association of mean total costs and payment model was estimated using the normal distribution.
For all use and quality outcomes, we also estimated median rate ratios (MRRs).42 The MRR is the median relative change in the outcome between a randomly selected patient seeing a physician with a higher rate of the outcome and a randomly selected patient with the same covariates seeing a physician with a lower rate of the outcome. The MRR quantifies the magnitude of the effect of clustering within physicians. An MRR greater than the RR for payment model suggests that the median association of physician clustering with an outcome was greater than the association with a payment model. An MRR lower than the RR suggests that the association of payment model with an outcome was greater than the median association with physician clustering.
For all outcomes, we addressed the issues of multiple comparisons by applying the Benjamini-Hochberg procedure43 that indicates which findings retain statistical significance after controlling for a prespecified false discovery rate. A false discovery rate of 10% was selected to balance our ability to detect the outcomes of interest against the consequence of false discoveries. Analysis was conducted using SAS statistical software, version 9.4 (SAS Institute Inc), and Stata statistical software, version 14.2 (StataCorp). The threshold for statistical significance was set at 2-tailed P < .05.
We excluded 2882 patients undergoing dialysis, 670 with kidney or islet cell transplants, 10 378 who had preoperative index visits, and 2103 whose physicians switched payment models during the study, resulting in a cohort of 109 839 adults with diabetes or nondialysis CKD newly referred to specialist physicians. There were 90 605 patients seen by FFS physicians and 19 234 seen by salary-based physicians. A higher proportion of patients with diabetes was seen by FFS physicians (65% vs 52%, P < .001), while a higher proportion of patients with CKD was seen by salary-based physicians (75% vs 60%, P < .001) (Table 1). In general, patients seen by salary-based physicians were sicker, with a higher proportion with more advanced CKD (2630 of 14 414 [18.2%] vs 6627 of 54 489 [12.2%]; P < .001) and a higher prevalence of 5 or more comorbidities (5989 of 19 234 [31.3%] vs 23 326 of 90 605 [25.7%]; P < .001), as well as a higher prevalence of chronic heart failure and dementia (Table 1; eTable 1 in the Supplement). Salary-based physicians were more likely than FFS physicians to be kidney specialists (standardized mean difference, 63%) and have a medium clinical workload (standardized mean difference, 90%) (Table 1).
Propensity-score matching resulted in a well-balanced matched sample (31 898 patients; 15 949 FFS, 15 949 salary-based physicians), with all standardized differences less than 10% (Table 1). In the matched cohort, mean (SD) age of patients seeing both types of specialists was 61.3 (18.2) years, 17 632 were women (55.3%), and 29 251 lived in urban areas (91.7%). A total of 489 physicians were included in the analysis.
The unadjusted follow-up outpatient visit rates were 2.02 (95% CI, 1.69-2.40) for all patients seeing FFS specialists and 1.90 (95% CI, 1.67-2.15) for all patients seeing salary-based specialists (eTable 2 in the Supplement). Unadjusted follow-up visit rates varied by payment model and physician characteristics. The unadjusted follow-up visit rate was highest for patients seeing salary-based nephrologists (3.13; 95% CI, 3.07-319) and salary-based specialists with the highest clinical workload (3.16; 95% CI, 3.08-3.24), and the lowest rate was for patients seeing FFS specialists with the lowest clinical workload (0.98; 95% CI, 0.94-1.02).
After accounting for physician and matched pair clustering, patients seen by salary-based physicians had a higher rate of follow-up visits compared with those seen by FFS physicians (1.74; 95% CI, 1.58-1.92 vs 1.54; 95% CI, 1.41-1.68), but the difference was not significant (RR, 1.13; 95% CI, 0.99-1.28; P = .06) (Table 2). There was no statistical evidence of a difference in visit rates between specialist groups for any of the subgroups. The visit rate for patients at higher risk for adverse outcomes was greater compared with the rate for patients who were not at higher risk for both the FFS and salary-based specialist groups. For example, patients with sustained hemoglobin A1c levels greater than 9% seeing FFS specialists had a visit rate of 3.09 (95% CI, 2.17-4.39), while those in the low-risk group seeing FFS specialists had a visit rate of 1.56 (95% CI, 1.42-1.71). The MRR was higher than the RR for all subgroups.
There were no clinically meaningful or significant differences in delivery of guideline-recommended prescribing or use of recommended testing for patients seeing salary-based physicians compared with patients seeing FFS physicians (Table 3). The proportion of patients receiving recommended medications in the 6 months after their index visit (Table 3) was higher than the proportion receiving those medications in the 6 months before their visit (Table 1). For example, before seeing a specialist, 9622 of 12 867 patients (74.8%) with diabetes and hypertension used an angiotensin-converting enzyme inhibitor or angiotensin receptor blocker, while in the 6 months after seeing an FFS specialists, 6426 patients (89.3%) (95% CI, 85.2-93.6) used one of these drugs and, in the 6 months after seeing a salary-based specialist, 6441 patients (85.1%) (95% CI, 81.1-89.3) did. The MRR was higher than the RR for 6 of the 7 guideline-recommended care outcomes.
For patients with diabetes, the rate of visits to the hospital or emergency department for diabetes-specific ambulatory care–sensitive conditions showed no statistical evidence of a difference among patients seeing salary-based compared with FFS physicians (RR, 1.12; 95% CI, 0.96-1.29) (Table 3). There did not appear to be a difference in the rate of visits to hospitals or emergency departments for CKD-specific ambulatory care–sensitive conditions between salary-based and FFS physicians for people with CKD (RR, 1.03; 95% CI, 0.87-1.22). The MRR was higher than the RR for both diabetes- and CKD-specific ambulatory care–sensitive condition visit rates.
There was variation in unadjusted categorical costs by payment model (Figure 2; eTable 3 in the Supplement). Chronic disease medication ($701; 95% CI, $674-$727 vs $628; 95% CI, $606-$727) and diagnostic imaging ($359; 95% CI, $351-$368 vs $302; 95% CI, $294-$309) costs in the first year after index visits were higher for patients seeing FFS specialists compared with those seeing salary-based specialists; all other categorical costs were lower. The adjusted mean total costs per patient in the year after the index visit were not significantly different for patients seeing FFS compared with salary-based specialists ($6916; 95% CI, $6482-$7352 vs $6925; 95% CI, $6469-$7380; P = .98). Mean total and categorical costs per patient were lower in the second year compared with the first year after the index visit. For example, among patients seeing salary-based specialists, diagnostic imaging was $302 (95% CI, $294-$309) in the first year and $177 (95% CI, $171-$183) in the second year after the index visit (eTable 3 in the Supplement).
In this study of specialist physician payment for chronic disease care, we found no statistical difference in follow-up outpatient visit rates, quality of care, or costs between patients seen by salary-based and FFS physicians. The MRR was larger than the RR for most follow-up visit rates and quality outcomes, suggesting that the median association with physician clustering was greater than the association with the physician payment model. This finding suggests variation in care among physicians.
Patients seeing salary-based physicians had a higher number of outpatient visits than patients seeing FFS physicians, which is not consistent with primary care research.11,12 However, previous research in specialty care payment also found no significant difference in the number of outpatient visits to salaried and FFS specialists.44 It is possible the salary-based payment model in Alberta may influence other types of health care use (eg, other patient groups or inpatient settings). Research comparing the association between specialist salary and FFS payment models with use in other health care settings found that salary payment led to a significant decrease in elective tubal ligation,44 but there was no statistical evidence of a difference in emergency department use,45 inpatient anesthesia use,46 or number of surgical procedures.44
In primary care settings, salary-based payment is associated with poorer quality of care in terms of continuity and adherence to the recommended number of visits compared with FFS payment; however, patients seeing salaried physicians report higher satisfaction.11,12 To our knowledge, no quasi-experimental studies comparing salary and FFS payment in specialty care have examined quality or cost outcomes. Although we found no statistical evidence of a difference in quality between FFS and salary-based specialists, the proportion of patients receiving recommended medications was higher after their first specialist visit compared with the year before the visit, which suggests that specialists are important partners in caring for people with chronic diseases regardless of the payment model.
Given that it is unlikely there will be randomized clinical trials of specialist payment models, we believe it is important to consider differences in patients selected by physicians and differences in the physicians who select different payment models when determining the outcome and implications of physician payment reforms. In contrast to the theoretical literature suggesting that FFS specialists select sicker patients,47 we found that patients seen by salary-based physicians were sicker, pointing to possible differences in the behavior of specialists selecting patients compared with primary care physicians. Results of a survey of 135 physicians participating in Alberta’s Academic Alternative Relationship Plan indicated that most believed that the payment reform had a positive association with their ability to spend more time with patients with complex needs.48 Salary-based physicians seeing sicker patients is an important outcome of the payment reform and points to the relevance of a salary-based model for specialists seeing patients with chronic diseases and participating in chronic disease management teams.
Our results suggest that individual physician practice behaviors have a greater association than payment model with use of outpatient services and quality outcomes. Because physicians in this study selected their payment model, it is possible that there is an association between payment model and physician characteristics in practice variation. For example, female specialist physicians are more likely to select alternative payment models12 and also more likely to deliver higher-quality diabetes care than male specialists.49 Research identifying significant associations between payment models and use, quality, and cost outcomes does not typically account for physician clustering. A randomized clinical trial of physician incentives that used mixed models to account for physician clustering also found no effect of physician incentives, although the trial identified an effect of combined physician and patient incentives.50 We believe it would be prudent to address physician clustering, physician characteristics that may be associated with practice variation, and other contextual factors, such as practice site, in future research on physician payment models in specialty and primary care to expand our understanding of the differences in physician behavior.
Our study has a number of limitations. First, while the inclusion of patients with diabetes and CKD is clinically relevant and the propensity matching increases the internal validity of the results, the generalizability outside of nephrology, endocrinology, and internal medicine physicians in Alberta is uncertain. Second, although we used propensity-score matching to control for measured patient differences, it is possible that there were unmeasured differences even in our matched sample. Third, salary-based physicians are more likely to practice at tertiary care hospitals and in major urban centers than are FFS physicians,51 which might influence case mix and illness severity. Fourth, it is possible that some of the differences that we observed related to the accuracy of shadow billing, but a past analysis of our data comparing shadow and FFS billing with medical record reviews showed similar accuracy.52
In this study, specialist physician payment is not associated with variation in chronic disease care outpatient visits, quality, and costs; however, we found a large variation in these outcomes among physicians. This variation suggests the need to consider other strategies, including incorporating principles of behavioral economics into payment models,53 to reduce variation between physicians to improve the value of care and outcomes for people with chronic diseases.
Accepted for Publication: September 8, 2019.
Published: November 8, 2019. doi:10.1001/jamanetworkopen.2019.14861
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Quinn AE et al. JAMA Network Open.
Corresponding Author: Braden J. Manns, MD, MSc, Cumming School of Medicine, Department of Medicine, University of Calgary, 3280 Hospital Dr NW, Calgary, Alberta T2N 4Z6, Canada (firstname.lastname@example.org).
Author Contributions: Drs Quinn and Manns had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Quinn, Hemmelgarn, McBrien, Edwards, Senior, Manns.
Acquisition, analysis, or interpretation of data: Quinn, Hemmelgarn, Tonelli, McBrien, Senior, Faris, Au, Ma, Weaver, Manns.
Drafting of the manuscript: Quinn, Faris, Manns.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Quinn, Hemmelgarn, Faris, Au, Ma, Weaver.
Obtained funding: Quinn, Hemmelgarn, Manns.
Administrative, technical, or material support: Quinn, Tonelli, Edwards, Manns.
Conflict of Interest Disclosures: Drs Manns, Hemmelgarn, Tonelli, McBrien, Edwards, and Senior are compensated under an Academic Alternative Relationship Plan (now called Academic Medicine and Health Services Program). Dr Hemmelgarn reported receiving grants from Alberta Innovates during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was funded by the Network of Alberta Health Economists Health Economics Scholar Award (Dr Quinn), a Banting Postdoctoral Fellowship (Dr Quinn), a Canadian Institutes of Health Research Foundation Grant (Dr Manns), and an Alberta Innovates Collaborative Research & Innovation Opportunity Team Grant.
Role of the Funder/Sponsor: The funding organizations 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: This study is based in part on data provided by Alberta Health. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the Government of Alberta or Alberta Health express any opinion in relation to this study.
Additional Contributions: Timothy B. Creedon, PhD (Health Equity Research Lab, Cambridge Health Alliance, Cambridge, Massachusetts), provided helpful, uncompensated consultation on the mixed-model analysis.
Create a personal account or sign in to: