At the Intersection of Health, Health Care and Policy Cite this article as: Jessica Greene, Judith H. Hibbard and Valerie Overton Large Performance Incentives Had The Greatest Impact On Providers Whose Quality Metrics Were Lowest At Baseline Health Affairs, 34, no.4 (2015):673-680 doi: 10.1377/hlthaff.2014.0998
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/34/4/673.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2015 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
Primary Care By Jessica Greene, Judith H. Hibbard, and Valerie Overton 10.1377/hlthaff.2014.0998 HEALTH AFFAIRS 34, NO. 4 (2015): 673–680 ©2015 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Large Performance Incentives Had The Greatest Impact On Providers Whose Quality Metrics Were Lowest At Baseline
Jessica Greene (jessgreene@ gwu.edu) is a professor in the School of Nursing at the George Washington University, in Washington, D.C.
This study examined the impact of Fairview Health Services’ primary care provider compensation model, in which 40 percent of compensation was based on clinic-level quality outcomes. Fairview Health Services is a Pioneer accountable care organization in Minnesota. Using publicly reported performance data from 2010 and 2012, we found that Fairview’s improvement in quality metrics was not greater than the improvement in other comparable Minnesota medical groups. An analysis of Fairview’s administrative data found that the largest predictor of improvement over the first two years of the compensation model was primary care providers’ baseline quality performance. Providers whose baseline performance was in the lowest tertile improved three times more, on average, across the three quality metrics studied than those in the middle tertile, and almost six times more than those in the top tertile. As a result, there was a narrowing of variation in performance across all primary care providers at Fairview and a narrowing of the gap in quality between providers who treated the highest-income patient panels and those who treated the lowest-income panels. The large quality incentive fell short of its overall quality improvement aim. However, the results suggest that payment reform may help narrow variation in primary care provider performance, which can translate into narrowing socioeconomic disparities. ABSTRACT
M
ost current health care reform efforts in the United States include a focus on reforming payment to move away from paying for volume and toward paying for quality or value. There are a number of payment reform strategies being tested across the country. However, the evidence of these strategies’ impacts is based largely on the evaluation of pay-for-performance programs. These programs link part of clinicians’ compensation to their performance on designated quality metrics. The compensation can range from small bonuses for performance on a few quality indicators to as much as one-quarter of a provider’s
Judith H. Hibbard is a professor emerita and senior researcher in the Health Policy Research Group at the University of Oregon, in Eugene. Valerie Overton is vice president for quality and innovation at Fairview Medical Group, in Minneapolis, Minnesota.
income for performance on over 100 metrics.1,2 We examined the impacts on quality of a primary care provider compensation model implemented by Fairview Health Services that went much further than most pay-for-performance programs. During the time it was in use, Fairview’s model based 40 percent of physicians’ compensation on clinic-level quality metrics. We assessed which primary care providers improved the most under the payment model, and the degree to which the improvement narrowed the gap in patient outcomes for primary care providers who treated high- and low-income patient panels.
April 2 015
3 4:4
Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
Health Affa irs
673
Primary Care Background Rewarding clinicians for providing high-quality care or improving the quality of care they provide has strong intuitive appeal, and pay-for-performance programs have proliferated in recent years.3,4 Yet evaluations have found that these programs typically have modest impacts.5–9 In a recent article, Martin Roland and Stephen Campbell concluded: “As a summary of this increasing body of work, it is clear that pay for performance can be effective. However, the effects are sometimes only short-term and are often not as large as payers wish.”10(p1944) In response to the limited impact of pay-forperformance programs, many authors have highlighted the importance of programmatic design, including the size of the incentive, the metrics incentivized, and the entity that receives the incentive.10–16 Several articles have shown that some clinicians are more responsive than others to financial incentives.17–20 However, very few studies have explored the question of which providers respond to payfor-performance programs.18–20 Answering this question is key for better understanding the environments in which the programs may be more and less effective, and how much impact the programs can be expected to have on providers with varying characteristics. Several authors have argued that providers’ age, sex, and attitudes may affect quality improvement under pay-for-performance programs.6,19–22 There is some empirical evidence suggesting that providers with lower baseline quality and more positive attitudes toward the program improve more than those with higher baseline quality and negative attitudes.18–20 However, few provider characteristics have been investigated. Authors have also suggested that characteristics of primary care providers’ patient panels— such as patients’ age, socioeconomic status, engagement in their care, and comorbidities— may influence the providers’ efforts to improve quality in response to financial incentives.6,20,21 Providers have raised concerns about the possibility that providers who disproportionately treat patients with lower socioeconomic status may be less able to improve their patients’ quality metrics and thus perform worse in pay-forperformance programs, compared to similar providers who treat patients with higher socioeconomic status.23,24 The challenges in improving the outcomes of patients with lower socioeconomic status are related to the greater obstacles these patients face across areas such as transportation, job inflexibility, and limited access to laboratories and pharmacies.25 There is some evidence that 674
H ea lt h A f fai r s
April 2 015
34:4
providers treating patients with lower socioeconomic status perform worse in pay-forperformance programs, which supports providers’ concern about this point.25,26 This study examined primary care providers’ improvement in quality in response to financial incentives that made up 40 percent of their base compensation.We examined how much incentivized quality metrics improved over two years, and whether variation in performance narrowed across primary care providers. Many authors have speculated that the limited impact of pay-for-performance programs has been due, at least in part, to the small size of incentives.7,11–13,15,27,28 Some authors have recommended that incentives constitute at least 10 percent of compensation. Since Fairview’s payment model allocated four times that level to quality incentives, we examined whether improvement was substantially larger than that observed in previous research or compared to other delivery systems in Minnesota. Furthermore, since the different quality metrics were weighted differently in the compensation model, we were able to examine whether improvement was correlated with payment. We also examined which providers improved the most under the quality incentive. We investigated whether there were differential impacts based upon primary care providers’ characteristics or those of their patient panels. Finally, we explored whether improvements in quality had an impact on quality disparities among providers whose patient panels had higher and lower incomes. A number of scholars have argued that financial incentive programs have the potential to widen socioeconomic and racial disparities in quality.5,25,29,30 However, some evidence suggests that pay-for-performance programs can narrow gaps in health outcomes.5,30,31
The Study Setting Fairview Health Services is a Pioneer accountable care organization (ACO) in Minnesota. It has forty-four primary care clinics, which typically have five to fifteen primary care providers each. In April 2011 Fairview began implementing a team-based quality-focused compensation model. Before that time, Fairview paid primary care providers based upon fee-for-service productivity, with a small annual quality bonus. Forty percent of the new compensation model was based upon clinic-level performance on five Minnesota Community Measurement quality metrics (one of which, diabetes care, is used by the Centers for Medicare and Medicaid Services for ACOs).32,33 More specifically, metrics on
Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
diabetes care and vascular care each made up 12 percent of compensation at Fairview. Cancer screening and depression care metrics each made up 6 percent, and the asthma care metric made up 4 percent. Compensation for quality was pegged to the median market salary for a given provider’s specialty (for example, family medicine or internal medicine), and quality performance was compared to state-level quality benchmarks. If a clinic’s performance was at the state median for a given quality metric, the clinic’s primary care providers would receive their specialty’s median salary for the allocated percentage of their compensation. For example, if a clinic’s performance was at the state median for diabetes care (such as 32 percent of patients meeting the diabetes metric in 2010),34 primary care providers would receive the median market salary for 12 percent of their compensation. Performance above the median on a metric was compensated on a sliding scale of up to one and a half times the median income for performance at or above the ninetieth percentile (50 percent of patients meeting the diabetes metric). Performance below the median was compensated on a sliding scale of down to half of the median income for performance between the twentieth and twenty-ninth percentiles (21–25 percent of patients meeting the diabetes metric). Below that level, there was no compensation for that component of quality. Notably, there was sufficient range in quality performance across Fairview’s clinics that there were primary care providers receiving the full range of compensation for specific metrics (0–150 percent of the median market salary). The other components of the compensation model were clinic-level patient experience scores (10 percent), clinic-level panel size (10 percent), primary care provider–level panel size (15 percent), the provider’s number of billable and nonbillable interactions with patients (20 percent), and an assessment of provider “citizenship” or being a team player (5 percent). The model changed over time. The most significant change was implemented in mid-2012 to address the reduction in Fairview’s fee-forservice billing. Fairview added a productivity bonus based on relative value units and a limitation on how much primary care providers with low productivity could earn from team-based quality performance.35
Study Data And Methods This study principally used administrative data from Fairview Health Services to examine primary care providers’ change in quality performance
from before the compensation model was implemented to two years after the model’s implementation. It also used 2010 and 2012 quality performance data from Minnesota Community Measurement to assess the level of change at Fairview in comparison to changes at five other medical groups in Minnesota.32,34 Data And Sample We used Fairview administrative data on primary care provider–level quality from before the new model’s implementation (February and March 2011) and two years after its implementation (March and April 2013). Each given month’s data were made up of performance based upon the prior twelve months. We included only those providers with thirty or more patients who were eligible for a given metric at both baseline and follow-up. We computed the mean score for the two baseline months and the two follow-up months. This resulted in 207 primary care providers in the cancer screening analyses, 184 for diabetes, and 139 for vascular disease. Because the metrics used to assess asthma and depression care changed substantially during the study period, we did not include them in our analysis. Because of our interest in change in quality metrics, we did not include primary care providers who did not have both pre and post data. Notably, there was less than a 2-percentagepoint difference between performance on the three metrics at baseline and follow-up for providers in the sample and for all providers at Fairview. Data on primary care providers’ patient panel characteristics were computed from 2012 electronic health record data.36 Variables There were three dependent variables: optimal diabetes care, optimal vascular disease care, and cancer screening. The diabetes care metric was the percentage of a primary care provider’s patients ages 18–75 with diabetes who reached the following five goals: hemoglobin A1c
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/34/4/673.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2015 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
Primary Care By Jessica Greene, Judith H. Hibbard, and Valerie Overton 10.1377/hlthaff.2014.0998 HEALTH AFFAIRS 34, NO. 4 (2015): 673–680 ©2015 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Large Performance Incentives Had The Greatest Impact On Providers Whose Quality Metrics Were Lowest At Baseline
Jessica Greene (jessgreene@ gwu.edu) is a professor in the School of Nursing at the George Washington University, in Washington, D.C.
This study examined the impact of Fairview Health Services’ primary care provider compensation model, in which 40 percent of compensation was based on clinic-level quality outcomes. Fairview Health Services is a Pioneer accountable care organization in Minnesota. Using publicly reported performance data from 2010 and 2012, we found that Fairview’s improvement in quality metrics was not greater than the improvement in other comparable Minnesota medical groups. An analysis of Fairview’s administrative data found that the largest predictor of improvement over the first two years of the compensation model was primary care providers’ baseline quality performance. Providers whose baseline performance was in the lowest tertile improved three times more, on average, across the three quality metrics studied than those in the middle tertile, and almost six times more than those in the top tertile. As a result, there was a narrowing of variation in performance across all primary care providers at Fairview and a narrowing of the gap in quality between providers who treated the highest-income patient panels and those who treated the lowest-income panels. The large quality incentive fell short of its overall quality improvement aim. However, the results suggest that payment reform may help narrow variation in primary care provider performance, which can translate into narrowing socioeconomic disparities. ABSTRACT
M
ost current health care reform efforts in the United States include a focus on reforming payment to move away from paying for volume and toward paying for quality or value. There are a number of payment reform strategies being tested across the country. However, the evidence of these strategies’ impacts is based largely on the evaluation of pay-for-performance programs. These programs link part of clinicians’ compensation to their performance on designated quality metrics. The compensation can range from small bonuses for performance on a few quality indicators to as much as one-quarter of a provider’s
Judith H. Hibbard is a professor emerita and senior researcher in the Health Policy Research Group at the University of Oregon, in Eugene. Valerie Overton is vice president for quality and innovation at Fairview Medical Group, in Minneapolis, Minnesota.
income for performance on over 100 metrics.1,2 We examined the impacts on quality of a primary care provider compensation model implemented by Fairview Health Services that went much further than most pay-for-performance programs. During the time it was in use, Fairview’s model based 40 percent of physicians’ compensation on clinic-level quality metrics. We assessed which primary care providers improved the most under the payment model, and the degree to which the improvement narrowed the gap in patient outcomes for primary care providers who treated high- and low-income patient panels.
April 2 015
3 4:4
Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
Health Affa irs
673
Primary Care Background Rewarding clinicians for providing high-quality care or improving the quality of care they provide has strong intuitive appeal, and pay-for-performance programs have proliferated in recent years.3,4 Yet evaluations have found that these programs typically have modest impacts.5–9 In a recent article, Martin Roland and Stephen Campbell concluded: “As a summary of this increasing body of work, it is clear that pay for performance can be effective. However, the effects are sometimes only short-term and are often not as large as payers wish.”10(p1944) In response to the limited impact of pay-forperformance programs, many authors have highlighted the importance of programmatic design, including the size of the incentive, the metrics incentivized, and the entity that receives the incentive.10–16 Several articles have shown that some clinicians are more responsive than others to financial incentives.17–20 However, very few studies have explored the question of which providers respond to payfor-performance programs.18–20 Answering this question is key for better understanding the environments in which the programs may be more and less effective, and how much impact the programs can be expected to have on providers with varying characteristics. Several authors have argued that providers’ age, sex, and attitudes may affect quality improvement under pay-for-performance programs.6,19–22 There is some empirical evidence suggesting that providers with lower baseline quality and more positive attitudes toward the program improve more than those with higher baseline quality and negative attitudes.18–20 However, few provider characteristics have been investigated. Authors have also suggested that characteristics of primary care providers’ patient panels— such as patients’ age, socioeconomic status, engagement in their care, and comorbidities— may influence the providers’ efforts to improve quality in response to financial incentives.6,20,21 Providers have raised concerns about the possibility that providers who disproportionately treat patients with lower socioeconomic status may be less able to improve their patients’ quality metrics and thus perform worse in pay-forperformance programs, compared to similar providers who treat patients with higher socioeconomic status.23,24 The challenges in improving the outcomes of patients with lower socioeconomic status are related to the greater obstacles these patients face across areas such as transportation, job inflexibility, and limited access to laboratories and pharmacies.25 There is some evidence that 674
H ea lt h A f fai r s
April 2 015
34:4
providers treating patients with lower socioeconomic status perform worse in pay-forperformance programs, which supports providers’ concern about this point.25,26 This study examined primary care providers’ improvement in quality in response to financial incentives that made up 40 percent of their base compensation.We examined how much incentivized quality metrics improved over two years, and whether variation in performance narrowed across primary care providers. Many authors have speculated that the limited impact of pay-for-performance programs has been due, at least in part, to the small size of incentives.7,11–13,15,27,28 Some authors have recommended that incentives constitute at least 10 percent of compensation. Since Fairview’s payment model allocated four times that level to quality incentives, we examined whether improvement was substantially larger than that observed in previous research or compared to other delivery systems in Minnesota. Furthermore, since the different quality metrics were weighted differently in the compensation model, we were able to examine whether improvement was correlated with payment. We also examined which providers improved the most under the quality incentive. We investigated whether there were differential impacts based upon primary care providers’ characteristics or those of their patient panels. Finally, we explored whether improvements in quality had an impact on quality disparities among providers whose patient panels had higher and lower incomes. A number of scholars have argued that financial incentive programs have the potential to widen socioeconomic and racial disparities in quality.5,25,29,30 However, some evidence suggests that pay-for-performance programs can narrow gaps in health outcomes.5,30,31
The Study Setting Fairview Health Services is a Pioneer accountable care organization (ACO) in Minnesota. It has forty-four primary care clinics, which typically have five to fifteen primary care providers each. In April 2011 Fairview began implementing a team-based quality-focused compensation model. Before that time, Fairview paid primary care providers based upon fee-for-service productivity, with a small annual quality bonus. Forty percent of the new compensation model was based upon clinic-level performance on five Minnesota Community Measurement quality metrics (one of which, diabetes care, is used by the Centers for Medicare and Medicaid Services for ACOs).32,33 More specifically, metrics on
Downloaded from content.healthaffairs.org by Health Affairs on June 28, 2015 at EAST CAROLINA UNIVERSITY
diabetes care and vascular care each made up 12 percent of compensation at Fairview. Cancer screening and depression care metrics each made up 6 percent, and the asthma care metric made up 4 percent. Compensation for quality was pegged to the median market salary for a given provider’s specialty (for example, family medicine or internal medicine), and quality performance was compared to state-level quality benchmarks. If a clinic’s performance was at the state median for a given quality metric, the clinic’s primary care providers would receive their specialty’s median salary for the allocated percentage of their compensation. For example, if a clinic’s performance was at the state median for diabetes care (such as 32 percent of patients meeting the diabetes metric in 2010),34 primary care providers would receive the median market salary for 12 percent of their compensation. Performance above the median on a metric was compensated on a sliding scale of up to one and a half times the median income for performance at or above the ninetieth percentile (50 percent of patients meeting the diabetes metric). Performance below the median was compensated on a sliding scale of down to half of the median income for performance between the twentieth and twenty-ninth percentiles (21–25 percent of patients meeting the diabetes metric). Below that level, there was no compensation for that component of quality. Notably, there was sufficient range in quality performance across Fairview’s clinics that there were primary care providers receiving the full range of compensation for specific metrics (0–150 percent of the median market salary). The other components of the compensation model were clinic-level patient experience scores (10 percent), clinic-level panel size (10 percent), primary care provider–level panel size (15 percent), the provider’s number of billable and nonbillable interactions with patients (20 percent), and an assessment of provider “citizenship” or being a team player (5 percent). The model changed over time. The most significant change was implemented in mid-2012 to address the reduction in Fairview’s fee-forservice billing. Fairview added a productivity bonus based on relative value units and a limitation on how much primary care providers with low productivity could earn from team-based quality performance.35
Study Data And Methods This study principally used administrative data from Fairview Health Services to examine primary care providers’ change in quality performance
from before the compensation model was implemented to two years after the model’s implementation. It also used 2010 and 2012 quality performance data from Minnesota Community Measurement to assess the level of change at Fairview in comparison to changes at five other medical groups in Minnesota.32,34 Data And Sample We used Fairview administrative data on primary care provider–level quality from before the new model’s implementation (February and March 2011) and two years after its implementation (March and April 2013). Each given month’s data were made up of performance based upon the prior twelve months. We included only those providers with thirty or more patients who were eligible for a given metric at both baseline and follow-up. We computed the mean score for the two baseline months and the two follow-up months. This resulted in 207 primary care providers in the cancer screening analyses, 184 for diabetes, and 139 for vascular disease. Because the metrics used to assess asthma and depression care changed substantially during the study period, we did not include them in our analysis. Because of our interest in change in quality metrics, we did not include primary care providers who did not have both pre and post data. Notably, there was less than a 2-percentagepoint difference between performance on the three metrics at baseline and follow-up for providers in the sample and for all providers at Fairview. Data on primary care providers’ patient panel characteristics were computed from 2012 electronic health record data.36 Variables There were three dependent variables: optimal diabetes care, optimal vascular disease care, and cancer screening. The diabetes care metric was the percentage of a primary care provider’s patients ages 18–75 with diabetes who reached the following five goals: hemoglobin A1c
