|Home | About | Journals | Submit | Contact Us | Français|
Information on the impact of pay-for-performance programs is lacking in the nursing home setting. This literature review (1980-2007) identified 13 prior examples of pay-for-performance programs in the nursing home setting: 7 programs were active as of 2007, while 6 had been terminated. The programs were mostly short-lived, varied considerably in the choice of performance measures and pay incentives, and evaluations of the impact were rare.
More than 3 million frail and disabled individuals will rely on services provided by a nursing home at some point during the year, and among them 1.5 million will stay long enough to consider the nursing home their main residence (Doshi, Shaffer, and Briesacher, 2005). These individuals, their families, and their friends count on nursing homes to provide care that is of high quality. However, as of 2006, one in five nursing homes nationwide was cited for serious deficiencies—deficiencies that caused actual harm or placed residents in immediate jeopardy (Centers for Medicare & Medicaid Services 2006;U.S. Government Accountability Office, 2007).
One strategy for improving the quality of care in nursing homes is to link health care spending to quality and efficiency through pay-for-performance programs. Pay-for performance is a reimbursement approach designed to reward health care providers for achieving high levels of performance, or improvements in performance. This approach is in contrast to fee schedules of flat rates per service, where reimbursement is the highest when the most services are rendered, regardless of improved patient outcomes. In 2001, the Institute of Medicine identified payment systems based on only the quantity of care provided as a barrier to quality improvement and recommended that health care purchasers adopt reimbursement policies linked to quality improvement (Committee on Quality of Health Care in America and Institute of Medicine, 2001). Approximately one-half of all Medicaid Programs currently operate some type of pay-for-performance program, and 85 percent have plans to do so within 5 years (Kuhmerker and Hartman, 2007). Pay-for-performance programs in the nursing home setting have been rare; two recent surveys identified only four active programs (Georgia, Iowa, Vermont, and Utah) (White et al., 2006; Kuhmerker and Hartman, 2007). However, this will soon be changing as CMS, the largest purchaser of nursing home services (about $64 billion per year), will be sponsoring a pay-for-performance initiative through the Nursing Home Value-Based Purchasing Demonstration (Centers for Medicare & Medicaid, 2008,2006). Beginning in the summer of 2009, the pay-for-performance demonstration will include as many as 50 nursing homes per State in 4 host States.
Information on pay-for-performance programs in the nursing home setting is lacking, both on details of the programs as well as evaluations of any impacts. Research conducted in other health care settings find inconsistent improvements with pay-for-performance programs, both in terms of process measures of care (e.g., screening rates) and patient outcomes (Rosenthal and Frank, 2005; Petersen et al., 2006; Felt-Lisk, Gimm, and Petersen, 2007; Gilmore et al., 2007). For example, a recent analysis of pay-for-performance in the acute care setting found no significant differences in the quality of care or outcomes of patients with myocardial infarctions (Rosenthal and Frank 2005; Glickman et al., 2007). The lack of significant quality improvements in these different settings raises questions about the potential for similarly lackluster results in nursing homes. The objective of this study was to review the empirical literature to characterize current and former pay-for-performance programs in the nursing home setting and to compare the performance measures, pay incentives, and any evaluations of the impact of these programs.
Our conceptual framework for this review is an adaptation of the work by Rosenthal and Frank (2006). Within that framework, improving quality of care is addressed through a payment system that makes increasing quality in the best financial interest of the provider of care. However, a successful payment system requires good information to create a reward that is commensurate with the costs to the provider of increasing quality. Since information on the true costs of increasing quality is often lacking, a poorly designed payment system may have no effect on quality or even unintended negative effects. Thus, we probed the literature for valid evidence of the overall effectiveness of pay-for-performance in the nursing home setting. In addition, we were also interested in documenting the design elements of prior pay-for-performance programs in nursing homes to characterize key elements and current trends.
We conducted searches of the electronic database MEDLINE and the Internet using Google U.S. Government Search for empirical data describing pay-for-performance programs in nursing homes. In addition, we interviewed key stakeholders to supplement the search. The review covered the period 1980-August 2007. The initial search strategy for MEDLINE included the subject heading nursing homes in combination with the term quality and the reimbursement terms of pay-for-performance, value, purchasing, incentive, or payment. As primary documents or review articles were identified, we scanned the text for mention of programs, and reviewed the reference lists for additional sources. A modified search was applied for the Web search, which was limited to Web sites with downloadable files (Adobe Acrobat® PDF, Microsoft Word® or Microsoft PowerPoint®). Lastly, five interviews were conducted with key stakeholders from academia, Minnesota's department of human services, CMS, MedPAC, and the Department of Veterans Affairs to learn about additional programs. All investigators developed the search strategy, retrieved the documents and extracted the data, conducted the interviews, and made the final selection of programs.
The pay-for-performance programs were selected if they met four criteria: (1) the program had been implemented as a functioning payment system, (2) the program was set in the nursing home setting, (3) the program used measures of performance related to quality rather than only efficiency, and (4) the performance incentive was financial, rather than recognition-based. Programs were excluded if the financial incentives were not tied to reimbursement rates, but instead awarded as grants. Those programs generally award the grants retroactively so the relationship between performance and financial incentives is less clear. For instance, in Vermont's Gold Star program, high performing nursing homes are eligible for quality awards of $25,000 but the award is not guaranteed. In addition, we limited the review to only programs conducted in the U.S.
The Medline search identified 75 articles related to pay-for-performance in nursing homes and the Web search identified 134 online documents. Of these, only 19 related to our study objective, and 18 met the inclusion criteria. A manual review of these documents and the interviews produced four more programs. Key characteristics of the final programs are summarized in Table 1.
We identified 13 pay-for-performance programs in the nursing home setting between 1980 and 2007: 7 programs were currently active as of 2007, while 6 had been terminated. All of the current programs were relatively new, with the oldest program dating back to 2002 (Iowa's Accountability Measures Incentive Program). Nearly all of the terminated programs were short-lived. For example, Texas implemented a performance-based add-on payment program for nursing homes in 2001 but 2 years later ended the program due to budget cuts (Carter, 2002). The reasons for the other terminations were not documented. Most (5 of 13) of the programs were set in the Central/Midwestern States, 4 in the South, 3 in the West, and only 1 in the Northeast. Participation was voluntary in all programs except for two, Georgia and Iowa. Figure 1 shows the distribution of active pay-for-performance programs in nursing homes from 1980 to 2007.
Only 4 of the 13 programs evaluated the impact of the program, and among them only one provided sufficiently robust evidence for drawing any clear conclusions. In the early 1980s, 32 San Diego nursing homes were randomly assigned into a control group or an experiment group with incentive payments for: (1) accepting patients needing the most functional assistance, (2) improved patient functional status within 90 days of admission, and (3) prompt discharges of patients who remained out of the facility for at least 90 days (Norton, 1992; Trisolini et al., 2006) The study tracked over 11,000 nursing home residents for 2-1/2 years, and found the incentive payments associated with beneficial effects. Nursing homes eligible for the incentive payment program were more likely to admit individuals with severe disabilities and more likely to send individuals home or to lower skilled facilities than nursing homes receiving only the per diems. In addition, residents in the experiment group nursing homes were less likely to be hospitalized or to die than those in the control group homes. However, this promising result came at a cost—the average daily cost to Medicaid rose by about 5 percent, due to the incentive payments and increased administrative burdens.
Studies with less rigorous research design found modest or no impact of the pay-for-performance program. In Illinois's Quality Incentive Program, the evaluation failed to find that the incentive payment influenced overall costs, Medicaid access, facility efficiencies, or improved resident care (Geron, 1991). Similarly, the evaluation of Massachusetts' pay-for-performance program also found no relationship between the incentive payment and improvement of quality (Willemain, 1983). However, in Iowa, the nursing homes demonstrated modest improvement in resident satisfaction, staffing and employee retention after implementation of the pay-for-performance program (Kane, Arling, and Mueller, 2007).
Table 1 summarizes the measures of performance that generally fall into eight major domains: staffing (e.g., low staff turnover); performance on certification survey (e.g., deficiency-free inspection); quality indicators from the Minimum Data Set (e.g., no physical restraints or no new pressure ulcers); facility efficiency (e.g., occupancy and operating costs); service to Medicaid enrollees; resident and family satisfaction; resident quality of life (e.g., privacy and comfort); and other outcomes (e.g., timely discharge and improved functioning). No program covers all of these domains, although Oklahoma's Focus on Excellence program and Minnesota's Quality-Add-on program assesses performance on six of the eight areas (Nursing Facility Rates and Policy Division, 2006; Oklahoma Health Care Authority, 2007). (It should be noted that Minnesota's program was initially implemented using only a portion of the complete list of quality measures.) More than one-half of these measures came from data already collected in the CMS surveying process, such as total compliance with State and Federal regulations for certified nursing homes. A few of the measures required new data collection, as is the case with the resident quality of life assessment in Minnesota's program, which costs the State about $700,000 each year. Also, Oklahoma's program uses its own measures of resident outcomes that are not derived from the Minimum Data Set. Figure 2 shows that the active programs most often define performance in terms of staffing, the certification survey, and resident satisfaction. In comparison, the older inactive programs showed more heterogeneity with few programs assessing performance using the same measures.
The incentive payments in the pay-for-performance programs varied modestly, ranging from a low of a $0.25 flat bonus per patient day to a high of a 5 percent increase in the daily per diem rate (Table 1). The rationale for the level of the incentive payment (whether there was any relationship to the costs of improving care) was poorly documented, and the proportion of providers receiving the incentive payment varied considerably. The proportion of eligible homes earning the incentive payment in any year ranged from 38 to 87 percent, and total expenditures generally varied according to the number of eligible homes. For example, at the end of the first year of Texas' Performance-based Add-On Payment program, 57.8 percent of 1,020 nursing homes in the State had earned the incentive payment, for a total cost of $4 million (Carter, 2002). Otherwise, the annual expenditures for the programs ranged from $500,000 (Utah's Nursing Home Quality Improvement Initiative) to $20 million (Illinois Quality Incentive Program). The source of the incentive payments was also poorly documented, such as whether they represented a new allocation of funds or a reallocation of other nursing home funds.
In our review, we found that few pay-for-performance programs have occurred in the nursing home setting since 1990, and even fewer have lasted more than a few years. Among the seven programs that are currently operational, three were initiated in 2007 and the oldest program is in its sixth year of operation.
We found little empirical evidence that pay-for-performance programs increase the quality of care of residents or the efficiency of that care in nursing homes. However, the program set in San Diego did find benefits, and it used the strongest of all evaluation designs, a randomized control design. This provocative finding deserves a more current replication. The disappointing results from the other programs are difficult to explain but may be related to barriers specific to this setting of care. According to our conceptual framework, the incentive payment must be sufficient to reward providers after accounting for the costs to improve care. For nursing homes already strained by high turnover among staff and administrators or ownership profits, the costs of improving care, such as implementing health information technology may far exceed the incentive payment.
Our review also revealed that nearly all of the programs were developed independently so there was modest agreement across the design elements of the programs and no clear state-of-the art trends. In most cases the performance measures are derived from existing data collection efforts related to State and Federal certification surveys. In addition to problems with accuracy and timeliness, these sources limit measurement options to a few components of quality with clear relationships to resident outcomes. Important quality issues such as the quality of life of residents and the satisfaction of residents and their families with nursing home care require additional data collection, increasing the cost of these programs.
Several limitations of this review should be noted. Pay-for-performance is an evolving concept, so standard terminology was sometimes lacking and descriptions of current activity were limited. For instance, the terms value-based purchasing and pay-for-performance were treated as synonymous, while quality recognition programs were handled on a case-by-case basis. Documentation of these programs was very limited. For instance, in the case of Utah, we could identify that a program had been initiated in 2004 but could find no subsequent information on the performance. In fact, this review raised more questions than answers. We could identify that many programs were short-lived but could rarely find the reasons for the terminations. Lastly, any assessment of current pay-for-performance activity will quickly become outdated. Our assessment can serve only as snapshot of the historical trends and evidence base for these programs before CMS invites participation in its nursing home demonstration.
Using financial incentives to encourage quality of healthcare is an increasingly popular approach to improving quality. That most of these programs in nursing homes have been terminated after a few years of operation sends an important warning that these approaches must be very carefully designed, although the reasons for these terminations are not clear and may be related more to lack of political viability rather than practical viability. At least one program had to be significantly scaled back in response to pressure from the State's nursing home industry an important constituency of State lawmakers (Kane, Arling, and Mueller, 2007). With the current lack of evidence to support the design and implementation of such programs, payers considering pay-for-performance in this setting should carefully consider the specific areas of performance and quality of most concern, review the capacity of the target nursing homes to actually accomplish the quality goals, involve the local nursing home industry in the design of the program from the beginning, and track all of the potential intended and unintended consequences of the program.
What can policymakers learn from this review, especially in terms of assessing the CMS Nursing Home Value-Based Purchasing Demonstration (NHVBP)? First, rigorous evaluations of the impact of pay-for-performance are almost nonexistent in the nursing home setting, yet they are critical for understanding the implications of this policy. It is reassuring to see that the NHVBP will employ an experimental design with randomized assignment of participating nursing homes into treatment and control groups and has contracted for an independent evaluation. Second, improvement in residents' outcomes and costs savings have been documented in only one program (in San Diego in the 1980s). Yet, the NHVBP demonstration requires cost savings expected to result from decreases in hospitalizations and skilled nursing facility stays and other savings, otherwise incentive payments are not awarded. The design of this payment system places a considerable burden on the provider to increase quality that ensures Medicare savings, and this is a burden unlike that of the other programs in our review. Third, the state of measuring quality in nursing homes has evolved to include informative domains such as resident quality of life and consumer satisfaction. The CMS demonstration is initially focused on the traditional measures of staffing (levels and stability), MDS based-quality measures, survey deficiencies, and inappropriate hospitalizations. The one exception in this list is the measure of inappropriate hospitalizations, although the validity and reliability of this measure in the nursing home setting is yet unclear. It should be noted, though, that resident satisfaction has been listed as a potential measure for future years of the demonstration (Centers for Medicare & Medicaid 2008). Lastly, the small number of prior programs, many of which were terminated, suggests that potentially serious barriers exist to implementing and sustaining pay-for-performance programs in this setting. It would be beneficial if the CMS demonstration would include a thoughtful assessment of these barriers, particularly which factors would be important to maintaining a pay-for-performance program after the demonstration is concluded.
The authors would like to thank Darlene O'Connor, Ph.D., Associate Professor of Family Medicine and Community Health and Director of Long-Term Care Policy, University of Massachusetts Medical School for her assistance.
The authors are with the University of Massachusetts Medical School and Meyers Primary Care Institute, Worcester, Massachusetts. The research in this article was prepared with support from Commonwealth Medicine, the public service healthcare consulting division of the University of Massachusetts Medical School. The statements expressed in this article are those of the authors and do not necessarily reflect the views or policies of the University of Massachusetts Medical School, Meyers Primary Care Institute, or the Centers for Medicare & Medicaid Services (CMS).
Reprint Requests: Becky A. Briesacher, Ph.D., University of Massachusetts Medical School, Division of Geriatric Medicine, Biotech Four, Suite 315, 377 Plantation Street, Worcester, MA 01605. E-Mail: email@example.com