How 10 leading health systems pay their doctors

This article was first published in Healthcare: The Journal of Delivery Science and Innovation. It is co-authored by Dhruv Khullar, Robert Kocher, Patrick Conway, and Rahul Rajkumar.

We conducted interviews with senior executives at 10 leading health systems to better understand how organizations use performance-based compensation. Of the organizations interviewed, five pay physicians using productivity-independent salaries, and five use productivity-adjusted salaries. Performance-based pay is more prevalent in primary care than in subspecialties, and the most consistently identified performance domains are quality, service, productivity, and citizenship. Most organizations have less than 10% of total compensation at risk, with payments distributed across three to five domains, each containing several metrics. Approaches with many metrics—and little at-risk compensation for each metric—may offer weak incentive to achieve any particular goal.

Provider organization; Performance-based compensation; Productivity; Incentives; Value-based healthcare delivery; Quality improvement

Large healthcare organizations that seek to create value by managing population health face a Principal-Agent problem, in which one actor (here, the physician—agent) makes decisions on behalf of another person or entity (the organization—principal). Organizations have a particular set of incentives, but it is ultimately frontline physicians—who at times face different incentives—that make clinical decisions. Problems can arise when the two parties׳ goals are in conflict, when they have different tolerances for risk, or when it is difficult for the principal to verify the agent׳s actions. In medicine, these challenges can result in the delivery of too much, too little, or inefficient care, depending on how procedures and services are reimbursed.

In an effort to align the interests of Agent and Principal, many organizations in other industries use performance-based compensation schemes, in which employees are paid based on how they perform on selected metrics that promote the organization׳s overarching goals. According to the human resources consulting group Aon Hewitt, more than 90% of US companies now offer employees performance-based pay, up from 78% in 2005 and only half in the early 1990s.1 Many healthcare organizations have also shown interest in using performance-based compensation to align provider incentives with organizational goals.

To better understand whether and how leading health systems use performance-based compensation, we conducted interviews using a standard questionnaire with senior executives at 10 leading provider organizations. We selected organizations that are either leaders in reputational surveys or that are established innovators in healthcare delivery. All 10 organizations cooperated with our survey. Key findings are presented in Table 1.

Table 1Physicians at these organizations are largely salaried with relatively minor adjustments for performance. Of the 10 organizations interviewed, five pay physicians using productivity-independent salaries, meaning that salaries are determined by a physician׳s specialty and experience and usually pegged to national or regional benchmarks—but not tied to measures of productivity like Relative Value Units (RVUs). Five systems pay physicians using a productivity-adjusted salary, in which salaries are based on a physician׳s recent RVU-productivity.

Our interviews suggest that compensation models vary greatly by medical specialty. Organizations emphasize performance-based pay more in primary care than in subspecialties. Compensation for procedural subspecialties is more closely linked to RVU-productivity, whereas non-procedural subspecialties tend to receive an unadjusted salary. This discrepancy may be due to the relative market power of subspecialists as compared to primary care physicians, which likely strengthens their bargaining position with provider organizations. Moreover, because performance measurement initially emerged in primary care, policymakers have a more robust set of validated quality metrics here than in the subspecialties. A relative dearth of quality metrics serves to maintain status quo volume-based compensation.

Interviewed organizations tie a relatively low percentage of total compensation to performance. At the Cleveland Clinic, Mayo Clinic, and Iora Health, for example, physicians are 100% salaried. At Group Health and Kaiser Permanente (Southern California) more than 90% of total physician compensation is salary. Importantly, even organizations that tie little or no compensation to performance attempt to track and encourage performance on a variety of metrics by conducting internal performance reviews. Furthermore, performance data for individual physicians is transparent in most systems; physicians are able to see their own performance and rank, as well as that of their colleagues.

At most organizations, senior leaders set overarching strategic aims, and then work closely with frontline physicians and department chiefs to develop fair and meaningful performance metrics. Group Health, for example, is undertaking a five-year process in which each year several departments meet with senior leadership to develop metrics to improve quality and efficiency in that specialty. Most organizations use a combination of group and individual metrics to make allocation decisions about compensation. Leaders note that some outcomes can only meaningfully be understood at the system-level (e.g., utilization patterns), while others are more appropriately assessed at the physician-level (e.g., patient satisfaction). Across systems, the most consistent performance domains were quality, service, productivity (generally measured by RVUs), and teamwork or citizenship.

Research has long shown that physicians respond to financial incentives. Productivity in medical groups tends to increase as a larger share of physician compensation is tied to his or her individual production, and this finding has been replicated in a number of specialties and organizational designs.2 and 3 But while studies suggest financial incentives can increase productivity in the traditional sense (number of patient encounters or hours worked), it is unclear if they do so when productivity is redefined as value or quality per unit time. Some have raised concerns about the ability of performance-based compensation to achieve desired outcomes, arguing that it creates unintended consequences such as avoiding high-risk patients, focusing on narrowly-defined metrics at the expense of holistic care, and “crowding out” internal motivation and professionalism.

A recent article by Mostashari et al. argues that primary care physicians should be thought of as CEOs responsible for $10 million in annual revenue and that physicians should be accountable for the overall quality and costs in health systems as senior executives are in other organizations.4 Our study of leading provider organizations, however, suggests that physicians generally have far less compensation at risk than senior executives in other industries.

Most organizations have less than 10% of total compensation at risk, and payments are typically distributed across 3–5 performance domains—each containing several metrics—so that physicians may be responsible for 10–20 metrics during any given measurement period. Approaches with many metrics—and very little compensation at risk for each metric—offer a relatively weak incentive to achieve any particular goal. Some physicians may hit performance targets just by chance, especially if thresholds are low, and even physicians who change nothing could potentially receive as much in payments as those who make significant changes. Furthermore, physicians cannot focus on everything at once and may be uncertain about where to devote their efforts.

This observation should be taken with some caveats. Firstly, even organizations with many performance domains may choose to focus on only two or three metrics in a given measurement period. Secondly, some researchers have found that small incentives can change physician behavior. For example, a recent Massachusetts General Physicians Organization (MGPO) quality-improvement program significantly increased hand hygiene compliance and electronic prescribing using incentives totaling less than 2% of the average physician׳s income.5 Thirdly, many leaders emphasized institutional culture, not financial incentives, as an important driver of quality and performance. Finally, performance-based compensation is certainly not the only method to address the Principle-Agent problem. Non-financial incentives such as public reporting and internal performance reviews and recognition may also be effective.

Despite calls for greater value-based purchasing, physician reimbursement continues to be driven primarily by volume of care delivered. In organizations with performance-based compensation, RVU productivity still dominates quality and service metrics by driving base salary, perhaps reflecting the continued prevalence of a fee-for-service payment environment. This is, however, at odds with new value-based purchasing goals, and may change as payment models continue to evolve. Indeed, several organizations reported that they plan to significantly increase performance-based pay for physicians in the near future.

Several conditions are necessary for organizations to move a greater proportion of payment into performance and away from volume. Researchers must continue to explore validated metrics that meaningfully impact patient outcomes. Organizations should encourage physician buy-in by giving them input into the metrics upon which they will be evaluated. And finally, to foster greater experimentation early on, payers should ensure it is in the financial interest of providers to explore novel reimbursement models and enter risk-based contracts.

There is likely no universally effective physician compensation model suitable for all organizations, and each system will need to incent physicians with its unique culture, goals, patient population, and financial situation in mind. But provider organizations may be missing an opportunity by linking too little compensation to too many metrics. Large changes in physician behavior and healthcare delivery—increasingly important in the post-reform era—may require larger amounts of at-risk compensation. Further research on physician compensation models and their effect on quality outcomes and costs is needed. Payers are moving toward paying provider organizations on quality and costs; but organizations themselves should do more to examine how they can better align financial incentives for frontline providers with those of the larger organization. By exploring payment models with more pay tied to performance, and less to productivity, provider organizations may be able to free physicians of traditional constraints and empower them to be not just aligned agents, but change agents.

The views herein represent the opinions of the authors and not necessarily policy or views of the Department of Health and Human Services, Centers for Medicare & Medicaid Services.

1 Aon Hewitt Salary Increase Survey. 〈〉; 2014.

2 M. Gaynor, M. Pauly
Compensation and productive efficiency in partnerships: evidence from medical groups practice
J Politic Econ, 98 (1990), pp. 544–573

3 D.A. Conrad, A. Sales, S.Y. Liang, et al.
The impact of financial incentives on physician productivity in medical groups
Health Serv Res, 37 (2002), pp. 885–906

4 F. Mostashari, D. Sanghavi, M. McClellan
Health reform and physician-led accountable care: the paradox of primary care physician leadership
J. Am Med Assoc (2014), pp. 1855–1856

5 D.F. Torchiana, D.G. Colton, S.K. Rao, S.K. Lenz, G.S. Meyer, TG. Ferris
Massachusetts General Physicians Organization׳s quality incentive program produces encouraging results
Health Aff, 32 (2013), pp. 1748–1756

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s