Using prior utilization to determine payments for Medicare enrollees in health maintenance organizations by James Beebe, James Lubitz, and Paul Eggers The Tax Equity AndFiscal Responsibility Act of predict risk. Using statistical simulations, formulas 1982 is expected to make it more attractive for health incorporating prior use performed better for some maintenance organizations (HMO's) to participate in types ofbiased groups than a formula similar to the the Medicare program on an at-risk basis. Currently, one currently employed. Major concerns involve the payments to at-risk HMO's are based on a formula ability to "game the system.» The prior-use model is known as the adjusted average per capita cost now being tested in an HMO demonstration. This (AAPCC). This article describes the current formula article also outlines the limitations ofa prior-use and discusses a modification, based on prior use of model and areas for future research. Medicare services, that endeavors to more accurately Introduction (IPA) type ofHMO's in the studies. Persons may be attracted to IPA's because they can maintain their Health maintenance organizations (HMO's) have present ties to their physicians, and such persons may generally been considered to be more efficient be in poorer than average health (Lull, 1981). Third, systems for providing health care than the traditional in all the studies, only one fee--for-service alternative fee-for-service system. Many studies have found that was offered. A study ofthe Federal Employees health care costs and hospital use in HMO's, Health Benefits Program, where a range offee-for­ particularly in prepaid group practices (PGP), are service and HMO plans are available, did not find generally lower than in the fee-for-service sector selection oflow users into HMO's (Schuttinga, Fallik, (Manning et al., 1984; Wolinsky, 1980; Luft, 1978; and Steinwald, 1984). Instead, lower usen were and Roemer and Shonick, 1973). The studies often attracted into the low-premium, high-oost-sharing, point to lower hospital use. An example is the fee-for-service plans. The authors emphasize that, recently published study from the RAND Health ..... the extent ofadverse or beneficial selection into Insurance Experiment (Manning et al., 1984). The HMO's depends on the price and the authors found that an HMO achieved significant comprehensiveness ofbenefits ofeach available fee­ savings from lower hospitalization rates when for-service option." patients were randomly assigned to an HMO or to a The problem ofpotential overpayments to HMO's fee-for-service health care insurance package. bas been made more critical by the HMO provisions It has also been suggested that one reason for the ofthe Tax Equity and Fiscal Responsibility Act of lower costs may be biased selection ofhealthier 1982 (TEFRA). These provisions make it more persons into HMO's (Luft, 1978 and 1982). Over the attractive for HMO's to participate in Medicare on past few years, a number ofstudies have appeared an at-risk basis and are expected to increase the that show biased selection oflower-than-average number ofat-risk HMO's in Medicare. TEFRA users ofhealth care into HMO's, both by Medicare authorizes prospective reimbursement to HMO's enrollees (Eggers, 1980; and Eggers and Prihoda, under risk-sharing contracts at a rate equal to 95 1982), and by persons under 65 years ofage percent ofthe adjusted average per capita cost (Jackson-Beeck and Kleinman, 1983; Roghman, (AAPCC). TEFRA defmes the AAPCC to be the Sorensen, and Wells, 1980; Arthur D. Little, 1983). estimated average per capita amount that would be These studies examined the health care use ofHMO payable ifMedicare services for HMO members were enrollees before they joined an HMO, and found that furnished in the local fee--for-service market they bad used less services on the average than TEFRA requires that the AAPCC formula take persons notjoining. This suggests that the HMO into account those factors that are likely to be enrollees may have been in better-than-average associated with differences in health care use, such as health and would have been lower users even ifthey age and sex, and thereby adjust for differences in were not in HMO's. expected health service use between HMO enrollees The results ofthese studies should be qualified in and other Medicare beneficiaries living in the HMO's three ways. First, all but one study reported on cases service area. The current AAPCC formula uses age, in which an HMO option was offered for the first sex, welfare status, and institutional status as time. The findings might be different ifthe studies adjustment factors. 1 were repeated after an HMO option had been As previously noted, studies ofMedicare enrollees available for some years. Second, the finding oflower joining HMO demonstrations raise the issue of pre-enroJJment use only applied to prepaid group biased selection. After adjusting for age, sex, welfare practices, not to the individual practice association status, and institutional status, Eggers (1980) and Health CMe Fiaandq Re'lew/Spriag l98S/Voi111J!e 6, Number) 27 Eggers and Prihoda ( 1982) found in three out offour A prior-use adjustment might also be applicable if HMO's examined that Medicare enrollees still had a voucher system proposal is adopted. Such a system lower use in their preenrollment years. In the fourth, would restructure Medicare by giving vouchers to which was structured much like an IPA. Medicare Medicare enrollees to purchase private health plans enrollees had about average use. These findings (or to remain in Medicare). Although a voucher imply that a reimbursement formula that does not strategy may promote competition among health take health status into account may overpay HMO's plans, it can have the same undesirable effect found when they experience a favorable selection oflow in the Federal Employees Health Benefits Program of users or, conversely, may underpay when HMO's encouraging enrollment ofhealthy persons into low­ experience an unfavorable selection ofhigh users. As premium, high-cost-sharing plans, leaving sicker a result ofthe Eggers and Prihoda studies, the persons in high-premium, comprehensive-benefit Congressional Budget Office (1982) estimated that plans and raising overall program costs (Luft, 1982; under the present AAPCC, increased Medicare and Anderson and Knickman, 1984a). Adjusting the enrollment in HMO's would increase Medicare costs value ofvouchers to reflect health status. perhaps in the short run.2 through a prior-use adjustment, could reduce the Our purpose was to test, in a simulation, whether undue financial rewards to plans that attract healthy or not a regression model that uses available prior persons and reduce the undue losses to other plans utilization and demographic predictor variables that attract sicker ones. could produce more accurate predictions offuture For our study, we first developed an AAPCC Medicare reimbursement than the current AAPCC model incorporating prior-use variables in addition formula. Although many variables, such as self­ to most ofthe other facton used in the present reported health status or functional status, are AAPCC formula Then, we tested this prior-use associated with health care use, we restricted our AAPCC model by comparing its predictive accuracy examination to variables currently available from for groups ofenrollees with various kinds of administrative records. Thus, variables that might statistically simulated biased selection against the theoretically be more attractive but that would accuracy ofan AAPCC formula similar to the current require added data gathering were not investigated one. In this article, we also discuss the limitations of The utility ofprior use as a predictor offuture use a prior-use AAPCC formula as well as areas for is suggested by studies showing that a person's past future research. use ofhealth care is correlated positively with subsequent use (Roos and Shapiro, 1981; Densen, Methods Shapiro, and Einhorn, 1959; Eggers, 1980; Eggers and Prihoda, 1982; Anderson and Knickman, 1984b; and McCall and Wai, 1983). This is certainly Study design plausible. Many persons have chronic conditions, Before detailing the data sources and methods of such as cancer, stroke, or mental disorders, that our study, it is helpful to review the overall study require repeated treatments, often over a long time. design. The purpose ofthe study was to test how the In addition, some persons may have more ofa inclusion ofprior-use variables might improve the tendency to seek medical care than others. accuracy of the AAPCC formula. The actual Incorporating a prior-use variable into constraints ofdata availability that would be faced if reimbursement formulas for HMO's in Medicare has, an AAPCC formula with prior-use variables were in fact, been suggested by a number ofauthors really to be employed to pay HMO•s were duplicated (Trapnell, McKusick, and Genuardi, 1982; Anderson as far as possible. Thus, we used data ftles that could and Knickman, 1984a and b; and Thomas et al., actually be employed to implement an AAPCC 1983). These authors believe that a prior-use formula with a prior-use adjustment, rather than data adjustment would substantially remove the files that might, in theory, yield a better model. possibility offinancial losses to Medicare or the In the first step ofour study, models were built to HMO in case ofbiased selection, thus increasing predict an enrollee's 1976 total Medicare HMO participation. Adjustment for prior use could reimbursement (Figure 1). The models included ones also remove the current disincentive for HMO's to using variables similar to those in the present enroll sicker than average people, since premiums AAPCC formula, as well as ones incorporating would better reflect expected health care use. variables on prior hospital and physician use in 1975 and 1974. Then, the models were evaluated, using IWelfare status is measured by whether the Medicare enrollee is demographic data and 1978 and 1977 prior-use data also covered by Medicaid, and institutional status reflects whether theenrollee is living in an institution (e.g., a nursing home).