To most stakeholders, the stated business value of employer health management strategies has largely been limited to reductions in healthcare costs. This article describes how employers and other stakeholders can gain a broader understanding of the relationship between workforce health status and business performance. This approach:

		Optimizes use of employer data sources that are readily available but have not been integrated.

		Yields a more comprehensive and compelling basis for assessing the impact of health benefits strategic planning.

		Can help employers and other stakeholders make more informed and value-based health benefits purchasing decisions.

Traditionally, employers have viewed health benefits as a cost of doing business. Benefits have served as a means to provide access to essential healthcare services, as well as an important consideration for employee recruitment and retention. With rising healthcare costs, employers have been forced to more closely scrutinize what has become a significant source of organizational expense. Some have reduced or eliminated healthcare benefits, or are considering paying the federal penalty and allowing their employees to join an exchange. Others are adopting progressive approaches such as account-based high deductible health plans, in an effort to control costs.

In light of the magnitude of the benefits expense, some employers have begun to more critically assess the value of business investments in workforce health. While workforce health status has been demonstrated to have a direct link to healthcare expenditures, absence, and presenteeism,

several recent reviews and commentaries have brought into question the degree to which corporate efforts to improve health produce positive health or financial outcomes.

 Illness increases business expense, but there may be limits on the extent to which businesses can achieve and sustain improvements in the prevalence and severity of illness.

At the juncture of continued healthcare cost inflation and increased scrutiny regarding the impact of healthrelated investments, employers have a fiduciary duty to carefully examine the link between health and business outcomes. Currently, measurement of the perceived effect of workforce health appears limited to 2 fundamental dimensions— healthcare costs and lost productivity (including absence)&mdash;both of which are in the human resources (HR) domain. Although workforce health status has a plausible and potentially comprehensive impact on business performance, measurement of this broader dimension of value of business investments in health has not been well articulated.

While a number of factors have contributed to the current scenario, 2 in particular stand out. First, HR departments have generally been viewed by senior leadership as cost centers, with cost management as their primary priority. This has effectively restricted measurement of the business impact of health management to HR metrics. Second, it makes intuitive sense that improvements in workforce health status can favorably influence healthcare costs, as well as the potential for greater employer engagement and reduced turnover, leading to improved work quality and greater customer satisfaction and sales revenue. Disappointingly, supporting data are woefully lacking beyond the impact of workforce health status changes on healthcare expenditures—and even that relationship has recently been called into question. It would therefore seem prudent for employers to realize a greater level of understanding of the business impact of health. Doing so may help to more appropriately prioritize employer human capital investments toward those likely to create the greatest net organizational value.

The intent of this commentary is to help employers, plan sponsors, and payers expand their approach to quantifying and therefore better understanding the broader business impact of workforce health status. It is important to note that the intent of this commentary is not to assess the extent to which employer investments in health impact healthcare costs and lost productivity. As applied, this expanded approach provides both HR personnel and their senior business leadership with a more comprehensive perspective of the impact and value of strategic investments in workforce health. This broader set of metrics more effectively aligns HR and business objectives around what can become a consensus framework, with organizational revenue and profitability as the common goal.

Employers have traditionally used health data metrics to identify causes of high healthcare costs. Used appropriately, these metrics can help to pinpoint specific sources of significant healthcare expenditures in order to develop risk mitigation strategies. To this end, benefits strategies have largely focused on cost management, with more recent approaches incorporating value-based benefit design, narrow or tiered provider networks, innovative pricing models, and population health management programs

Claims data warehouse technology enhancements have simultaneously advanced analytic capabilities and reduced the financial barriers to employer access. This has expanded employer understanding of population-level health metrics, including condition-specific healthcare compliance, utilization, and costs. Health risk and biometric screening data are now more commonly integrated with claims data analysis to quantify medical costs related to identified health risks and guide wellness program strategy.

 These advances have helped employers to refine their focus to more efficiently address the root causes of identified healthcare cost drivers. Publications regarding employer metrics for health management have also helped to broaden employer understanding and provide additional refinement for strategy development.

Human Resources as a Health and Productivity Management Silo 

During the past decade, efforts to gain a better understanding of the medical and productivity costs of poor health have yielded insightful information. Benchmark reports from the Health Enhancement Research Organization,

 Institute for Health and Productivity Management, 

 have broadened our understanding of how health risks and chronic conditions contribute to these organizational costs. However, it is important to note that these studies and others have limited their analysis to HR-related measures: healthcare costs and lost productivity. While healthcare expenditures are readily acknowledged at the C-suite level as a primary HR management responsibility, 

 lost productivity measures, particularly presenteeism, have not been as well received by business leaders as a meaningful business measure.

 And while presenteeism has been characterized as impacting work quantity and quality,

 little evidence exists to link these findings with business performance measures. Strengthening the connection between health and business performance has potential to increase business leader acceptance of the business value of a healthy workforce.

Why is it that no link between health and business performance been formalized? Several reasons may exist. First, these 2 areas rely on very different sources of information, with no precedent for data integration due to organizational silos. HR leaders have been significant contributors to our understanding of their departmental costs, limiting quantification of the costs of poor health to attributes under their direct control. Business performance is not a direct responsibility of HR function, which is typically viewed as a siloed cost management center, as shown in 

. Second, health absence and presenteeism measures are common across all industry types, facilitating data collection and analysis. In contrast, business performance metrics across different industries differ widely, substantially limiting generalization. Third, employer health and productivity data are typically collected at an individual level to reflect individual health and productivity costs, and are then aggregated at the business unit or organization level. In contrast, business performance metrics may or may not be tracked at the individual level, which can create challenges to meaningful data integration. As a result of these considerations, and likely others, there has been little focus on the connection between workforce health and the business value of a healthy workforce.

Existing Evidence for Health-Related Performance Impact 

A few examples of the relationship between employee health and job performance can be found. One area of recent research has targeted individual well-being as a more holistic approach to individual health, encompassing physical, behavioral, financial, social, community, and career dimensions. Researchers have demonstrated an association between individual employee well-being status and supervisor performance evaluation.

This association has also been demonstrated at an aggregate level, with business unit well-being scores correlating with team effectiveness and business performance for a large employer.

 These recent results have provided a compelling basis for additional research in this area.

Generally, however, evidence supporting the impact of workforce health on organizational economic performance is scant. In support, some research has shown an association between employee health status and performance of individual tasks

 or occurrence rates of critical incidents, 

 but these results were not linked directly to business financial impact. It is easy to appreciate that absence and presenteeism are individual attributes, such that health status could affect individual productivity measures across diverse industries. In contrast, the link between health status and business financial performance is more difficult to generalize across industries, much less among employees within a single company.

In their “2010/2011 Staying At Work” report, Towers Watson and the National Business Group on Health showed that over a 5-year period, organizations substantially investing in workforce health had greater business success as measured by shareholder return.

 However, this analysis was cross-sectional in nature. While a causal effect was implied, an alternate interpretation of the findings is that employers with greater business success simply have more discretionary funds available to invest in workforce health.

Similar cross-sectional data linking workforce health with work quality have been reported in a manufacturing firm, where average employee healthcare costs (as a proxy for health status) correlated directly with waste production as a percentage of final stock value across 5 locations,

. At a location-specific level, using annual revenue per employee estimates, differences in healthcare costs of $1000 per employee were correlated with higher quality sufficient to increase stock value production (and therefore, revenue) by $2000 per employee. This finding, though not rigorously validated, has major economic implications for employers thinking beyond healthcare cost control.

The link between health and business performance has also been outlined in a recent industry white paper.

 The author presents a series of industry-specific business performance metrics for consideration for integration with health data, but unfortunately does not provide detailed quantitative examples. As technological advances lead to improved information management and tracking, it is likely only a matter of time before more substantial efforts at data aggregation help to demonstrate a causal relationship between workforce health and business performance.

Integrating Health and Business Performance Metrics—The Opportunity 

Why should employers—and particularly HR personnel &mdash; integrate these heretofore separate metrics? An understanding of the impact of workforce health and well-being on business performance has the potential to provide notable insights. Perhaps most importantly, this connection could more appropriately integrate health and well-being management, generally utilized solely as a benefits cost-control strategy, into the broader business investment strategy. Identifying the business impact of health and well-being management strategies can provide substantial justification for specific investments in these areas. For example, in a retail setting, if poor employee well-being is associated with reductions in customer satisfaction and lower sales volume, it may be worth investigating whether efforts to improve workforce well-being result in more contented employees, and, in turn, increased customer satisfaction and sales volume. If so, a comparatively small investment in well-being could have a substantial impact on business revenue.

Another important result of integration of workforce health, well-being, and business performance metrics is that HR leadership becomes a more involved strategic partner to business operations. Accordingly, with combined metrics, there is the potential for considerably greater alignment of organizational interests toward common business performance objectives, instead of the current perception of HR as solely a cost management function. Some organizations have achieved this effective transformation of HR from cost center to strategic business partner with measurable success.

Yet in a recent survey, the Integrated Benefits Institute found that only 20% of CFOs would consider using work output/labor costs as the best way to quantify health-related productivity.

Instead, health-related absence was reported by 46% as their preference for reporting this measure, which clearly fails to incorporate any at-work decrements in business performance. Further, in a global survey of CEOs, 33% identified benefits cost management as an important HR responsibility; workforce health and well-being as a business strategy was not identified as a business priority.

Linking Health Outcomes With Business Performance 

The process of integrating data to link workforce health and well-being measures with business performance does not have to be difficult. In many cases, the necessary information already exists, with the challenge being to integrate data from disparate sources. Both individual and group (facility or company)-level data may be accessible, and evaluations combining both may provide meaningful insights.

At the individual level, organizations are likely to have performance evaluations that may be in a quantitative or ordinal format. This information can be paired with individual health or well-being scores available from health risk assessment vendors or health plans to gain a better understanding of the association between health and worker performance. Depending on the industry, additional worker-specific performance data may be available, as may be the case for industries where specific measurable tasks are performed, such as call centers, banks, or pharmacies.

 provides a listing of representative individual business performance metrics for specific industries. These, and comparable metrics in other industry categories, may be used to demonstrate associations between individual health and well-being status and business performance.

At the group (facility, unit, or company) level, aggregated individual health measures and performance evaluations can be linked with team business performance data to better understand the link between health and organizational performance. 

 is also relevant to populations as well as individuals, with the caveat that other factors, notably market influences, will certainly impact aggregate results. An array of business performance measures are commonly tracked in business operations, including those related to quality, volume, customer satisfaction, revenue, and others that are more industry specific.

 provides a listing of representative aggregate metrics for various industries that are suitable for use in evaluation of healthrelated business impact. Employers should review their existing business reporting to identify the specific performance metrics that are central to business operations, and integrate those with health, well-being, and performance reporting. Integration of these key health and operations metrics can enhance organizational leadership interest in a broader perspective of the potential value of workforce health and well-being on business performance.

Limitations of Using This Measurement Approach 

There are several limitations to this approach of investigating the link between health, well-being, and business performance. Perhaps most importantly, at an aggregate level, fluctuations in business outcomes reflect a myriad of external economic forces that may obscure the impact of workforce health and well-being status on organizational performance. Accordingly, review of individual-level measures may be more meaningful indicators because they will be somewhat less dependent on market forces.

Second, while this approach represents a framework for connecting individual health to overall business performance, several assumptions in this approach must be acknowledged. The causal relationship between improved health and improved individual performance has not been fully characterized, though some existing evidence is certainly supportive. And while individual health and performance may be linked, little if any data exist which demonstrate that cumulative improvements in workforce health result in organization-level business performance improvement.

Notably, as only 1 of many components of human capital assets, individual health represents only 1 determinant of workforce performance. Other factors, including engagement

have also been shown to have an association with workforce performance. While it may be true that improvements in population health result in incremental improvements in workforce and business performance, it is possible that organizations can derive an even greater opportunity to increase performance by considering changes in other domains of HR management, including training, compensation design strategies, leave policies, andmanagement techniques to help maximize the value of workforce human capital.

There is a compelling need for employers to more comprehensively evaluate the business value of a healthy workforce. Commercial health plans can also benefit by more clearly measuring and articulating the value of provided services. Early research efforts are already under way and can be expected to shed more light on this developing area of evaluation. Organizations will benefit from thoughtful integration of selected relevant metrics from both HR and business operations, to better understand the link between workforce health, well-being, and business performance. Many organizations already have well-established business performance metrics, and are well positioned to integrate HR and business data for meaningful results. Importantly, employer ability to transition from cost-based to value-based purchasing can be accelerated by incorporating the business impact of workforce health when considering an array of benefits designs. Particularly considering the current environment of healthcare reform, when employers are considering large-scale changes in  benefits strategy, a clearer understanding of the business value of a healthy workforce cannot come at a better time.

 Employers Health Coalition, Inc, Canton, OH (BWS); Department of Medicine, Case Western Reserve University School of Medicine, Cleveland, OH (BWS); Lynch Consulting, Ltd, Steamboat Springs, CO (WDL); Altarum Institute, Ann Arbor, MI (WDL).

Dr Lynch reports receiving consulting and writing fees in the preparation of this paper, and also reports participation in consultancies or paid advisory boards with several companies with health management interests. Dr Sherman reports no relationships or financial interests with any entity that would pose a conflict of interest with the subject matter of this article.

Concept and design (BWS, WDL); acquisition of data (BWS); analysis and interpretation of data (BWS); drafting of the manuscript (BWS, WDL); critical revision of the manuscript for important intellectual content (BWS, WDL); provision of study materials or patients (BWS); administrative, technical, or logistic support (BWS); supervision (BWS).

Bruce W. Sherman, MD, 3175 Belvoir Blvd, Cleveland, OH 44122. E-mail: bws@case.edu.1. Burton W, Conti D, Chen CY, Schultz A, Edington D. The role of health risk factors and disease on worker productivity. J Occ Environ Med. 1999;41(10):863-877.

2. Burton W, Pransky G, Conti D, Chen C, et al. The association of medical conditions and presenteeism. J Occ Environ Med. 2004;46(6 suppl):S38-S45.

3. Goetzel R, Anderson D, Whitmer R, Ozminkowski R, Dunn R, Wasserman J. The relationship between modifiable health risks and health care expenditures: an analysis of the multi-employer HERO health risk and cost database. J Occup Environ Med. 1998;40(10):843-854.

4. Goetzel R, Long S, Ozminkowski R, Hawkins K, et al. Health, absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting U.S. employers. J Occ Environ Med. 2004;46(4):398-412.

5. Goetzel R, Pei X, Tabrizi M, Henke R, Kowlessar N, et al. Ten modifiable health risk factors are linked to more than one fifth of employeremployee health care spending. Health Aff. 2012;31(11):2474-2484.

6. Mattke S, Liu H, Caloyeras J, et al. Workplace Wellness Programs Study Final Report. Santa Monica, CA: RAND Corporation; 2013.

Articles

COMMENTARY

Organizations can better understand the link between workforce health and business performance by thoughtful integration of relevant metrics from both HR and business operations.

Connecting the Dots: Examining the Link Between Workforce Health and Business Performance

"This study describes positive patient attitudes toward the concept of telephone-based alternatives to specialty clinic follow-up.”

		Patients have positive attitudes toward using telephone-based follow-up in the specialty clinic setting.

		Using telephone-based follow-up may be an effective strategy to reduce discretionary follow-up visits and increase access to specialty care.

The number of specialty care visits in the United States is rising, contributing to increased costs and delayed access to specialty care.

 In primary care, telephone-based follow-up has decreased the number of outpatient visits.

However, telephone-based follow-up in specialty care has focused on clinical outcomes and quality of life, not on reduction of outpatient healthcare utilization.

Little is known about patient attitudes toward using telephone-based programs to reduce office visits in US medical specialty clinics. In the context of a clinical operations project aimed at reducing discretionary specialty follow-up visits, we explored patient attitudes about telephone-based follow-up.

We conducted semi-structured patient interviews over 2 months in an urban academic endocrinology clinic. Interviews were conducted in clinic with a convenience sample of patients immediately following their physician encounter.

Interviews started with questions on clinic experience and utilization followed by a description of a potential pilot program. The proposed program involved changing the type of scheduled follow-up for patients. Rather than scheduling a follow-up office visit at the end of the visit, an appointment for a telephone visit with the medical assistant (MA) would be scheduled. Information gathered during the MA telephone call would be conveyed to the physician, who would decide whether patients should have another telephone call by a health professional (MA or physician) or a scheduled physician office visit. Interviewers asked about patient attitudes toward the program. Participants were provided free parking as compensation.

Interview notes were iteratively reviewed individually by 5 evaluation team members, who then met and reached consensus on key findings and themes. This project did not meet criteria for review by the Internal Review Board, as it used existing data collected for a quality improvement project.

Team members interviewed 24 patients (71% female, aged 22 to 64 years, median age 44.5 years). Overall, patients were very satisfied with their clinic experiences and traveled a median of 1.5 hours one way to clinic visits. Patients expressed enthusiasm for alternatives to office-based follow-up. The major themes that emerged were patient convenience, communication facilitation, and altruism (

). Patient endorsement of telephone-based follow-up was predicated upon the expectation that physicians continue to have primary responsibility for clinical care. Two patients expressed reservations about MA phone followup, including concerns about MD-MA coordination.

This study describes positive patient attitudes towards the concept of telephone-based alternatives to specialty clinic follow-up. Unexpected findings include perceptions about the program’s potential to facilitate clinic communication and improve clinic access for others.

Previous studies have shown positive patient attitudes toward telephone-based alternatives to primary care clinic follow-up,

 and this study extends those findings to specialty clinic follow-up. New models of specialty care are needed to address patient and payer demands for better access and quality. These findings support pilot testing a program to introduce telephone follow-up visits as a strategy for reducing discretionary office visits and improving specialty care access in similar clinics. Further assessment of patient and provider attitudes about telephone-based follow-up care in other clinical settings and specialties is also warranted.

 San Francisco Veterans Affairs Medical Center, San Francisco, CA (JAE); University of California, San Francisco, School of Medicine, Division of Geriatrics (JAE); University of California, San Francisco, School of Medicine, Division of General Internal Medicine (CJH, MAH); University of California, San Francisco, School of Medicine, Department of Biostatistics and Epidemiology (MAH, CKB, RG); University of California, San Francisco, School of Nursing, Department of Social and Behavioral Sciences (SLA).

 This project was supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through UCSF-CTSI Grant Number UL1 TR000004. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

The authors (JAE, CJH, MAH, CKB, RG, SLA) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Concept and design (JAE, CJH, MAH, CKB, RG, SLA); acquisition of data (JAE, CJH, CKB); analysis and interpretation of data (JAE, CJH, MAH, RG, SLA); drafting of the manuscript (JAE, MAH, CKB, RG, SLA); critical revision of the manuscript for important intellectual content (JAE, MAH, SLA); administrative, technical, or logistical support (CJH). Address correspondence to: Jessica A. Eng, MD, San Francisco VA Medical Center, 4150 Clement St, 181G, San Francisco, CA 94121. E-mail: jessica. eng@ucsf.edu.1. Mehrotra A, Forrest CB, Lin CY. Dropping the baton: specialty referrals in the United States. Milbank Q. 2011;89(1):39-68.

2. Wasson J, Gaudette C, Whaley F, Sauvigne A, Baribeau P, Welch HG. Telephone care as a substitute for routine clinic follow-up. JAMA. 1992; 267(13):1788-1793.

3. Weinberger M, Tierney WM, Cowper PA, Katz BP, Booher PA. Costeffectiveness of increased telephone contact for patients with osteoarthritis: a randomized, controlled trial. Arthritis Rheum. 1993;36(2): 243-246.

4. McKinstry B, Watson P, Pinnock H, Heaney D, Sheikh A. Telephone consulting in primary care: a triangulated qualitative study of patients and providers. Br J Gen Pract. 2009;59(563):e209-e218. 

LETTERS TO THE EDITORS

Patient Attitudes About Specialty Follow-up Care by Telephone

 To estimate: (1) the percentage of physicians whose compensation is variable; (2) the frequency at which performance incentives for productivity, care quality, patient satisfaction, and resource use were used to determine compensation; and (3) how much incentives differ for physicians who serve greater percentages of patients who are Medicaid-insured, racial/ethnic minorities, or who face language barriers, versus those who do not.

 Cross-sectional study of 3234 nationally representative physicians responding to the 2008 Center for Studying Health System Change’s Health Tracking Physician Survey (HTPS).

We examined the degree to which practices’ percentage of Medicaid revenues and physicians’ panel characteristics were associated with physicians’ financial incentives using x

 statistics and multivariate logistic regression (adjusting for physician specialty, practice type, and capitation levels, and area-based factors).

Compensation was variable for 69% of respondents, was most frequently tied to productivity (68%), and less often to care quality (19%), patient satisfaction (21%), or resource use (14%). Physicians were significantly less likely to report variable compensation if the percentage Medicaid revenues was 50% or more (adjusted odds ratio [OR] 0.73, 95% confidence interval [CI], 0.57-0.95) or if physician panels were at least 50% Hispanic (adjusted OR 0.74, 95% CI, 0.56-0.99). However, physicians were significantly more likely to report use of all 4 performance incentives if percentage of Medicaid revenues was 6% to 24%.

 Physicians report different types of financial incentives designed to alter care quality and quantity; incentive types differ by the degree that practices derive revenues from Medicaid or serve Hispanic patients. Further investigation is needed to understand how to align financial incentives with disparity-reduction efforts.

		Financial incentives alter the quality and quantity of care that physicians provide.

		Understanding physicians’ past experience with financial incentives can help shape future compensation strategies.

		There is a complex relationship between physicians’ experience of financial incentives and the degree to which practices derive revenue from Medicaid—pay based on performance is more prevalent when percent Medicaid is between 6% and 24%, but fixed salaries are more common when percent Medicaid is greater than or equal to 50%.

		Fixed salaries are also more common when physician panels are greater than or equal to 50% Hispanic.

US healthcare spending continues to rise faster than the gross domestic product (GDP) and significant gaps persist between the care people should and do receive, particularly for persons from lower socioeconomic status (SES) classes and racial/ethnic minorities.

 Since the Affordable Care Act of 2010 (ACA), payment and delivery system reform efforts have intensified, and there is substantial interest in healthcare delivery systems (eg, accountable care organizations [ACOs]) in which physicians increasingly serve as employees.

 This may further catalyze efforts to restructure the types of financial incentives that physicians face (eg, a greater emphasis on cost profiling).

Until now, payment and delivery system policy discussions have tended to focus on the incentives that flow between payers (eg, governmental and commercial health plans) and practices (which range in size from small solo practitioners to large multidisciplinary medical groups that may include a hospital system), and less attention has been paid to the types of financial incentives that physicians face as individuals through their compensation arrangements.

 Even less is known about how compensation may differ if practices derive a substantial proportion of their revenues from Medicaid-insured patients or if they care for greater percentages of patients with minority racial/ethnic backgrounds or non-English speaking patients, who are disproportionately medically and socially complex.

The types of incentives that physicians face affect healthcare costs, quality, and disparities.

 Physicians who are paid fixed salaries tend to see fewer patients and perform fewer procedures than physicians who are paid according to productivity.

 However, paying physicians according to visit or procedure volumes can cause physicians to focus on services that can be delivered quickly (eg, acute care for self-limited problems like colds) and spend less time on those that require more time (eg, teaching patients how to manage their diabetes), even though the latter has a better chance of improving patients’ health.

Tying financial bonuses to care quality (eg, pay for performance [P4P]) can be effective at improving care quality, but many are concerned that this strategy will yield negative unintended consequences (eg, unincentivized care will suffer).

There is even concern that P4P could cause physicians to avoid caring for patients of lower SES or patients of minority racial/ ethnic backgrounds because it may be more difficult to improve care quality for these patients.

 It is important to understand the quickly evolving payment environment, especially as certain types of incentives may recede (eg, those for productivity) while others increase (eg, those for quality, patient satisfaction, and resource use).

Prior studies suggest that although the majority of physicians who receive variable compensation are paid for productivity, the use of productivity incentives may be declining, while performance incentives related to care quality, patient satisfaction, and resource use may be increasing (

 These studies also demonstrate that practice and regional factors (eg, private practice vs health maintenance organizations [HMOs], the degree to which practice revenue depends on capitated contracts, whether practices are located in low-SES areas, and region of the United States) influence the types of financial incentives that physicians face.

It is also unclear whether physicians who treat high proportions of underserved patients may be exposed to different financial incentives than those who do not. For example, there is anecdotal evidence that physicians serving vulnerable populations (eg, physicians affiliated with federally qualified health centers or academic medical centers) tend to receive fixed salaries.

 To our knowledge, there are no nationally representative data on whether the types of financial incentives that individual physicians face vary depending on the degree to which physicians care for Medicaid-insured patients, patients of minority racial/ethnic backgrounds, or patients who typically face language barriers when seeking medical care, after adjusting for other factors such as practice setting, for one, that may also influence the use of these incentives.

Thus, this study aims to: (1) estimate the percentage of physicians whose compensation is variable (as opposed to fixed); (2) determine, for physicians whose compensation is variable, the frequency at which physicians experience 4 common performance incentives—productivity, quality, patient satisfaction, and resource use; and (3) assess the degree to which these incentives may vary for physicians whose practices serve more Medicaid-insured patients, whose panels comprise larger proportions of minorities, or whose panels include more patients facing language barriers, adjusting for specialty, practice, and geographic characteristics.

This is a cross-sectional study of physicians responding to the Center for Health System Change’s 2008 Health Tracking Physician Survey (HTPS). The HTPS is a periodic national survey of nonfederal allopathic and osteopathic physicians engaged in patient care at least 20 hours per week. Residents, fellows, and specialty physicians with limited patient contact in ambulatory settings (radiology, anesthesiology, and pathology) were ineligible for the survey. The HTPS utilized a nationally representative sample that included information from more than 4700 physicians and was conducted by mail. To our knowledge, the HTPS (and its predecessor, the Community Tracking Study) is the only nationally representative survey of active, ambulatorybased physicians that gathers this level of detail on physician compensation arrangements; alternate physician surveys do not.

 The response rate for this survey was 61.9%.

Our first outcome of interest was the proportion of physicians that reported their compensation to be variable (as opposed to fixed). We defined physician compensation as “variable” if respondents answered “yes” to the survey item asking if a physician’s compensation was tied to any of the following measures: productivity, quality, patient satisfaction, or resource use; otherwise, physician compensation was considered to be fixed.

Our second set of outcomes was the performance measures to which compensation could be tied: productivity, care quality, patient satisfaction, and resource use. For example, we defined compensation as varying according to “productivity” if respondents answered “yes” to the question that asked if their practice took their personal productivity (eg, number of patients seen or procedures done per clinical session) into account when determining their pay. We did the same for questions related to physicians’ performance on clinical quality measures (eg, rates of preventive care services), satisfaction surveys (completed by their own patients), or pattern of using medical resources relative to their peers (ie, resource use or cost profiling). The 2008 HTPS did not ask these questions of solo practitioners, nor did it include questions about the size of the variable component of compensation.

We used 4 predictor variables to assess the degree to which physicians provided care to Medicaid-insured patients or patients with minority racial/ethnic backgrounds: (1) percentage of practice revenue derived from Medicaid, and the percentage of patients in each physician's practice who (2) were Hispanic, (3) were African American, or (4) faced language barriers when seeking healthcare (ie, the percentage of each physician's patient panel that had a “hard time speaking with or understanding because you speak different languages”). The validity of physician report of the racial/ ethnic composition of their patient panels has been previously established.

 In each of these cases, since the data were skewed, we defined categories based on those providing care to 5% or less, 6% to 24%, 25% to 49%, and 50% or more of the demographic group of interest. Alternative categorizations (eg, <50% vs 

50%) yielded substantively similar results. There was no co-linearity between these 4 variables (Variance Inflation Factor: 1.17) and less than 0.5% of the responses were missing.

In multivariate analyses, we adjusted for 5 major factors known to be associated with physician compensation: (1) physician specialty (primary care, medical sub-specialty, or surgery); (2) practice type (1-2 physicians, 

3 physicians, group model or staff model health maintenance organization [HMO], hospitalbased practice, medical school and other [eg, community health centers, municipal boards of health]); (3) the proportion of practice revenue that was capitated or prepaid (none, 1%-24%, or 

25%); (4) an area-based marker of SES (the median household income of the zip code in which the practice was located using the 2004 Area Resources File [ARF]); and (5) regional location (North, Midwest, South, West).

We used descriptive statistics to provide the percentages of our 2 outcomes of interest, stratifying by the degree to which physicians served patients insured by Medicaid, patients who were of minority background, or patients who faced language barriers. We then examined bivariate and multivariate associations between our financial incentives of interest and our covariates using x

statistics and logistic regression models. We used a separate multivariate logit model for each type of performance measure (ie, productivity, quality, satisfaction, and/or resource use incentives). In our multivariate models, we included covariates that may contribute to variation in financial incentives: physician specialty, practice type, proportion of practice revenues from capitation, area SES, and regional location.

We retained all covariates in the adjusted model because of their conceptual significance, regardless of the statistical significance of their bivariate association.

 We applied survey weights to account for the complex sampling strategy of the HPTS (ie, known differences between probability of selection, response, nonresponse, and location) in accordance with HTPS technical recommendations.

We conducted our analyses using STATA 9.0, and all reported P values are 2-sided. The Institutional Review Board (IRB) at the University of Chicago reviewed the study protocol and deemed it exempt from official review.

The funding sources had no role in the design or conduct of the study; collection, management, analysis or interpretation of the data; or preparation, review, or approval of the manuscript for publication. None of the authors have any conflicts of interest to declare.

National Estimates: Variable Compensation and Types of Performance Incentives 

In this nationally representative survey of physicians, 68.5% of respondents indicate that their compensation is variable (

). Among physicians reporting variable compensation, a majority (67.7%) report having their compensation tied to productivity; a minority indicate that their pay is tied to quality measures (18.7%), patient satisfaction (21.4%), or resource use (14.0%).

Relationship Between Variable Compensation and Degree to Which Physicians Serve Vulnerable Populations

In bivariate analysis, the degree to which compensation is variable depends significantly upon the degree to which practices derive revenues from Medicaid or physicians care for patients with Hispanic backgrounds; there is no association for physicians caring for varying proportions of African American patients or of patients facing language barriers (Table 2). These relationships remain significant in multivariate analysis (

). Physicians with 50% or more of their practice revenue from Medicaid have three-fourths the odds of receiving variable compensation compared with physicians with 5% or less of their practice revenues from Medicaid (adjusted OR: 0.73; 95% CI, 0.57-0.95, 

 <.05). Findings are similar for physicians with at least 50% Hispanic patients on their panels (adjusted OR: 0.74; 95% CI, 0.56-0.99, 

 <0.05) compared with those whose patient panels are 5% or less Hispanic.

Relationship Between Performance Incentives and Degree to Which Physicians Serve Vulnerable Populations 

Among physicians who indicate that their compensation depends on performance incentives, there is a significant bivariate association between physicians’ percentage of practice revenues from Medicaid and all 4 of our performance incentives of interest (

). This relationship remains significant after multivariate analysis adjusting for physician specialty, practice type, and area-level factors. Physicians who report deriving 6% to 24% of their practice revenues from Medicaid have significantly greater odds of reporting that their compensation is tied to productivity (adjusted OR 1.19, 95% CI, 1.01-1.40), care quality (adjusted OR 1.45, 95% CI, 1.19-1.76), patient satisfaction (adjusted OR 1.55, 95% CI, 1.28-1.87), and resource use (adjusted OR 1.43, 95% CI, 1.16-1.76) compared with those deriving 5% or less of practice revenues from Medicaid. Within the group of physicians, physicians at practices with 50% or more of their revenues from Medicaid have significantly lower odds of receiving incentives for productivity (adjusted OR 0.69, 95% CI, 0.53-0.88).

After adjusting for covariates, there is no consistently significant relationship between use of our 4 performance incentives of interest and the percentage of patient panels that are Hispanic, African American, or facing language barriers.

Practice Characteristics and Physician Performance Incentives 

The above multivariate analysis also confirms that the odds of physicians having variable compensation is significantly associated with practice type, level of capitation, and regional location, but not with physician specialty or area SES (Table 3). For example, physicians working in group/staff model HMOs are nearly 3 times as likely to report variable compensation as physicians working in solo/2-physician practices (adjusted OR 2.84, 95% CI, 1.79-4.51).

When evaluating the types of performance incentives; physician specialty, practice type, capitation level, and regional location are also—to varying degrees&mdash;significantly associated with incentives for productivity, care quality, patient satisfaction, and resource use in multivariate analysis (Table 4). For example, medicine subspecialists and surgeons have significantly lower odds of experiencing incentives for care quality, patient satisfaction, and resource use compared with primary care physicians, and surgeons alone have significantly greater odds of reporting incentives for productivity. Physicians in group/staff model HMOs have 2 to 8 times greater odds of reporting incentives for care quality, patient satisfaction, and resource use, but do not have different odds of exposure to productivity incentives.

First, the unadjusted results show that, in 2008, US physicians faced a variety of financial incentives to alter care quantity and quality, and that productivity incentives predominated. Although governmental and private payers implemented numerous large-scale interventions aimed at better aligning financial rewards with healthcare quality in the years immediately preceding the 2008 HTPS, the prevalence of productivity incentives remained similar to past waves of this survey’s predecessor, the Community Tracking Survey (Table 1).

Second, the adjusted multivariate results suggest how complex the relationship between physicians’ financial incentives and Medicaid were at that time. On the one hand, the more that practices derived revenue from Medicaid, the less likely physicians were to report having variable compensation. This finding corroborates anecdotal information about the prevalence of fixed salaries in underserved settings.

 On the other hand, physicians receiving variable compensation were significantly more likely to report all 4 types of performance incentives if their practices derived 6% to 24% of their revenues from Medicaid. The observation is consistent with the idea that physicians working in practices that are able to modulate the degree to which their practices care for Medicaid-insured patients may be better able to create and operate performance-based incentive programs.

 Practices may need to be of a certain size, level of infrastructure, or degree of technical sophistication in order to titrate Medicaid revenues or operate performance-based compensation programs.

CLINICAL

Financial incentives alter the quality and quantity of care that physicians provide. Understanding physicians' recent experience with incentives may help shape current payment reform efforts.

Physician Financial Incentives and Care for the Underserved in the United States

To examine trends in out-of-pocket spending and the financial burden of care for persons with diabetes between 2001 and 2009, and to examine whether these trends are consistent with trends in access to prescription drugs and utilization of hospital services.

 Data are from the 2001 to 2009 Medical Expenditure Panel Survey (MEPS). The sample includes persons aged 18 to 64 years with diagnosed diabetes. The primary outcome variable is the percent of people with out-of-pocket spending on insurance premiums and services that exceed 10% of family income. Secondary outcome measures include the percent with diabetes-related prescription drug use, perceived access to prescription drugs, hospital inpatient stays, and emergency department use in the past 12 months. Multiple regression analysis is used to control for changes in comorbid chronic conditions and other characteristics of persons with diabetes.

Both out-of-pocket spending and the percent with high financial burden decreased markedly for persons with diabetes between 2001 to 2003 and 2007 to 2009. The decrease in spending was driven primarily by a decrease in spending on prescription drugs, including diabetes-related prescriptions. The shift from brand name drugs to generics accounts for much of this decline, although decreases in out-of-pocket spending for both brand name and generic drugs also contributed. During the same period, utilization of and access to diabetes-related prescriptions increased, and hospital use decreased.

 Although the prevalence of diagnosed diabetes continues to increase, treatment is becoming more affordable, especially prescription drugs. This may offset some of the costs to the healthcare system of higher prevalence by reducing complications of uncontrolled diabetes that result in more costly hospital use.

		Patients receiving treatment for diabetes and related comorbidities often incur a substantial financial burden due to the out-of-pocket costs of care they receive. However, despite increased prevalence, the financial burden of diabetes care to patients decreased markedly over the past decade, due primarily to lower spending on prescription drugs.

		Increased use of generics explains much of the decrease in spending on prescription drugs, although other factors—possibly including increased use of mail order and valuebased health plans&mdash;may also be reducing costs for patients.

		Reduced financial burden was accompanied by fewer people with diabetes reporting that they had problems accessing prescription drugs, and fewer who had hospital inpatient stays and emergency department visits.

Diabetes has been widely recognized as a large and growing public health challenge in the United States. The number of adults with diagnosed diabetes increased 75% between 2000 and 2010 (to about 21 million by 2010), while the percentage of adults with diabetes increased from 6% to 9% during this same period.

The increasing prevalence of diabetes not only has profound consequences for the health of the US population, but also for healthcare costs. The direct medical costs of diabetes were estimated to be $116 billion in 2007.

 Average medical expenditures for persons with diabetes are about 2.3 times higher than for people without diabetes, in part because complications of the disease often lead to other medical conditions, including heart disease and stroke, hypertension, eye problems, kidney disease, and nervous system disease.

In addition, the high costs for treatment of diabetes and comorbid conditions often impose a high financial burden on individuals and their families, which itself can be a barrier to regular monitoring of the disease and adhering to treatment regimens.

 Development of treatment guidelines and new antidiabetic drugs has increased the intensity of treatment, which also increases the potential financial burden on people with diabetes.

On the other hand, the greater availability and use of generic medicines has reduced the out-of-pocket costs and financial burden of prescription medications for many patients. As the percent of prescription drugs filled using generic drugs has increased, out-of-pocket spending on prescription drugs declined sharply for the US population, from an annual average of $215 in 2001 to $162 in 2009.

 For persons with diabetes, the availability of frequently prescribed antidiabetic medications in generic form—such as metformin&mdash;is also likely to reduce the financial burden of medical care for people with diabetes and their families.

This paper has 2 objectives: (1) to examine trends in out-of-pocket spending and medical cost burdens for persons with diabetes between 2001 and 2009, focusing especially on trends in prescription drug spending; and (2) to examine trends in the percentage of people with diabetes using diabetes-related medications, overall access to prescription drugs, and use of hospital care for persons with diabetes. To the extent that diabetes treatment has become more affordable over the past decade, we would expect to see more people with diabetes using diabetes- related medications, and fewer reporting that they were unable to receive needed prescription drugs. As prior research has established that greater adherence to diabetes drugs is related to lower hospital emergency department (ED) use and inpatient stays, decreases in hospital use over the past decade among people with diabetes would also be consistent with greater affordability and access to treatment.

We used data from the 2001 to 2009 Medical Expenditure Panel Survey, Household Component (MEPS HC). The survey is based on a large nationally representative sample of the civilian noninstitutionalized population and is conducted annually by the federal Agency for Healthcare Research and Quality (2009 is the latest year for which data are publicly available at the time of this study). The MEPS sample is based on a subsample of the National Health Interview Survey (NHIS), conducted annually by the Centers for Disease Control and Prevention (CDC). The survey collects detailed information on healthcare expenditures, use of services, insurance coverage, health status, medical conditions, and other sociodemographic details of individuals and their families during 3 rounds of interviewing during the calendar year. Sample sizes range from between 33,000 to 37,000 for each year.

Cumulative response rates for the 9 years of the survey used in this analysis averaged around 60%. Cumulative response rates combine the initial response rate to the NHIS (the sample frame), as well as response rates in each of the 3 rounds of data collection for the MEPS sample. Survey weights used in this analysis calibrate population estimates to be consistent with the Current Population Survey conducted by the US Census Bureau, and adjust for differences in nonresponse by US census region, metro versus rural residence, income, age, gender, and family income.

The sample for this survey includes persons aged 18 to 64 years who, in response to a question in the survey, reported that a physician or other health provider had told them that they had diabetes, which is identical to the methods used by the CDC to assess diagnosed diabetes prevalence in the population. 8 Annual samples of nonelderly adults with diabetes ranged from about 900 to 1100. To increase the statistical precision of estimates, samples from multiple years were pooled. The study compares annual estimates for the years 2001 to 2003 (N = 2747); 2004 to 2006 (N = 3096); and 2007 to 2009 (N = 3439). Persons aged 65 years and older were excluded because most have Medicare coverage, and therefore financial burden is likely to differ considerably from that of the under- 65-years adult population.

Expenditures for each medical encounter for each sample respondent are collected during 3 rounds of survey interviews during the calendar year. Expenditures are reported separately for office-based medical provider visits, hospital inpatient, outpatient, and ED care, prescribed medicines, home healthcare, dental services, and vision aids. For each visit/ event, total expenditures are reported (ie, from all payer sources), as well as the amounts paid by third-party payers and out-of-pocket by the patient. To improve the quality and accuracy of expenditure reporting, the MEPS Medical Provider Component collects data from a sample of medical providers and pharmacies used by sample persons, which are used to either supplement or replace patient-reported data on expenditures.

 All expenditure data are inflated to reflect 2009 dollars, based on the Consumer Price Index.

The study also examines trends in the percent of persons with diabetes who live in families with “high financial burden.” This is defined similarly to previous studies using the MEPS, as the ratio of total out-of-pocket spending on health services and health insurance premiums to total family income.

 For this measure, out-of-pocket spending is defined at the family level (ie, summed across all members in the family, typically defined as the nuclear family). Each individual is assigned the family level burden measure. Individuals who live in families that spend more than 10% of family income on healthcare are defined as individuals with high financial burden.

More detailed analysis of spending on prescription drugs is also included. Respondents were asked to supply the name of any prescribed medicine they obtained. For each prescribed medicine, information was collected on the name, medical conditions (coded based on 

 classification), spending, and source of payment information. In the analysis, we distinguish between prescriptions related to diabetes versus prescriptions for other conditions based on the class of drug.

We also distinguish between “brand name” and “generic” prescriptions by linking the MEPS to Lexicomp, a database that identifies and distinguishes brand name and generic medications.

 National Drug Codes were used to link MEPS prescription drug records with Lexicomp data. However, data on generic versus brand name are available only for the years 2005 to 2009 because of changes in methods used to assign National Drug Codes to prescription drug records.

To assess perceived access to prescription drugs, MEPS also asks respondents whether there was a time in the previous year when they were unable to obtain prescription drugs they or a doctor thought necessary. For hospital use, MEPS obtains information on all visits for inpatient stays and ED visits made in the prior year.

For the analysis of trends in spending and medical cost burdens, the analysis compares means and percentages for pooled samples of persons aged 18 to 64 years for the years 2001 to 2003, 2004 to 2006, and 2007 to 2009. All spending estimates reflect annual averages. Tests of statistical significance are computed for the change between 2001 to 2003 and 2007 to 2009 and between 2004 to 2006 and 2007 to 2009. All estimates are weighted to be nationally representative of the US civilian noninstitutionalized population with diagnosed diabetes.

For measures of the likelihood of having high financial burden, any use of diabetes-related prescriptions in the past year, access to prescription drugs, and use of hospital care, we use logistic regression analysis to examine whether changes between the three time periods are statistically significant, even after accounting for other changes in the aggregate characteristics of diabetics, including their age, gender, race/ ethnicity, number of comorbid chronic conditions, body mass index, limitations in daily activities, family income relative to the US poverty level, educational attainment, and health insurance coverage.

The regressions are estimated by pooling all 9 years of data into a single model. The models include the covariates mentioned above, as well as binary indicators for the 2001 to 2003 sample and the 2004 to 2006 sample. Coefficients for the 2001 to 2003 and 2004 to 2006 indicators show the difference with the 2007 to 2009 sample (the omitted group) on the dependent variable after controlling for changes in all other characteristics in the model. Standard errors for these coefficients are used to assess whether the change between 2007 and 2009 and prior years is statistically significant.

Given the number of regressions, we show the results only for the 2001 to 2003 and 2004 to 2006 variables—presented as odds ratios&mdash;which are the main interest in this study. Fullregression results are available in 

. Also, we use the results of the regression to compute adjusted means and percentages for the dependent variables, based on predicted values for the 3 time periods.

Characteristics of Persons With Diabetes 

The percentage of the US population aged 18 to 64 years with diagnosed diabetes increased from 4.6% in the 2001 to 2003 period to 6.2% in 2007 to 2009 (findings not shown). Among persons with diabetes, the percentage with 3 or more comorbid conditions increased from 53.1% in 2001 to 2003 to 62.7% in 2007 to 2009 (

). A higher percentage were obese in 2007 to 2009 compared with 2001 to 2003, although there were fewer who reported being limited in daily activities in 2007 to 2009 compared with 2001 to 2003. There were no changes in family income (relative to the US poverty level) during the 3 periods, while the percent uninsured increased from 11% in 2001 to 2003 to 13% in 2007 to 2009 (

In 2007 to 2009, total personal spending on healthcare for persons with diabetes was almost 3 times higher than spending for all persons aged 18 to 64 years. Total spending on healthcare increased for persons with diabetes between 2001 to 2003 and 2007 to 2009, as it did for all persons aged 18 to 64 years (

). For persons with diabetes, total spending (from all sources and for all health conditions) increased from an average of $8746 in 2001 to 2003 to $10,327 in 2007 to 2009 (

 <.01), representing an 18% increase over this period. The rate of increase in spending is similar for all persons aged 18 to 64 years.

Out-of-pocket spending on healthcare for persons with diabetes was more than double that for all persons aged 18 to 64 years. Average out-of-pocket spending on services decreased for persons with diabetes, mostly between 2004 to 2006 and 2007 to 2009. On average, persons with diabetes spent $1373 out-of-pocket on medical care in 2007 to 2009 compared to $1624 in 2004 to 2006 (

 <.001), a 15% decrease. Out-ofpocket spending did not decrease significantly for the general population during the study period.

Persons with diabetes are more likely to be in families that spend more than 10% of family income on healthcare compared with all persons aged 18 to 64 years (18.6% vs 7.8%, 

<.001). Despite an increase in comorbid conditions and no change in income, the percentage of persons with diabetes with high financial burden decreased, from 23.9% in 2001 to 2003 to 18.6% in 2007 to 2009 (

 <.001). There was no stmtistically significant decrease in high financial burden for all persons aged 18 to 64 years during the same period.

Overall, spending on prescription drugs accounted for more than half of all out-of-pocket spending for persons with diabetes in 2007 to 2009 (Table 3). However, outof- pocket spending on prescription drugs decreased sharply, from $1095 in 2001 to 2003 to $763 in 2007 to 2009 (

<.001). For diabetes-related prescriptions, out-of-pocket spending began to decrease in the 2004 to 2006 period, while for non—diabetes-related prescriptions, the decrease occurred in the 2007 to 2009 period. Out-of-pocket spending also decreased for other medical expenses (eg, eyeglasses, hearing aids, other medical equipment), but either did not change or increased (dental care) for all other types of services.

Generic drugs accounted for 52.9% of all diabetes-related prescriptions in 2007 to 2009, up from 46.9% in the 2005 to 2006 period (data on brand name vs generic that could be linked to the MEPS were not available prior to 2005). Both total spending and out-of-pocket spending per prescription are considerably lower for generic medications than they are for brand name prescriptions. More than two-thirds of generic prescriptions involved a copayment of $10 or less, compared with about one-third of brand name prescriptions (

Decreased out-of-pocket spending on prescription drugs accounts for reduced financial burden of care to patients, and is consistent with improved access to care.

February 2014

Connecting the Dots: Examining the Link Between Workforce Health and Business Performance

Patient Attitudes About Specialty Follow-up Care by Telephone

Physician Financial Incentives and Care for the Underserved in the United States

Trends in the Financial Burden of Medical Care for Nonelderly Adults with Diabetes, 2001 to 2009

The Impact of Patient Assistance Programs and the 340B Drug Pricing Program on Medication Cost

Complying With State and Federal Regulations on Essential Drug Benefits: Implementing the Affordable Care Act

Formulary Restrictions on Atypical Antipsychotics: Impact on Costs for Patients With Schizophrenia and Bipolar Disorder in Medicaid

Impact of a Medicare MTM Program: Evaluating Clinical and Economic Outcomes

Dialing In: Effect of Telephonic Wellness Coaching on Weight Loss

Positive Predictive Values of ICD-9 Codes to Identify Patients With Stroke or TIA