Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Academic Managed Care Organizations and Adverse Selection Under Medicaid Managed Care in Tennessee

Academic Managed Care Organizations and Adverse Selection Under Medicaid Managed Care in Tennessee Abstract Context Health plans competing in a managed care system may face serious financial consequences if they are disproportionately selected by enrollees with expensive health conditions. Academic medical centers (AMCs) have traditionally provided medical care for the sickest patients and may be at particularly high risk for adverse selection, but whether this occurs is not known. Objective To determine whether managed care organizations (MCOs) representing AMCs are adversely selected by Medicaid managed care (MMC) enrollees with expensive chronic health conditions. Design and Setting Observational study using state Medicaid claims data from all of 1994 and January to August 1995 for Tennessee's statewide MMC program (TennCare). Participants All 12 capitated MCOs in Tennessee, which collectively provided services for 1.2 million Medicaid enrollees from January 1994 through August 1995 following the initiation of TennCare. Main Outcome Measures Prevalence of 6 state-specified high-cost chronic conditions—acquired immunodeficiency syndrome (AIDS), coagulation defects, cystic fibrosis, pregnancy, prematurity, and organ transplantation—and 27 additional high-cost conditions compared by academic, statewide, and regional MCOs. Results The prevalence of state-specified high-cost chronic conditions was generally higher for academic MCOs compared with other MCOs. Specifically, prevalence of AIDS was 14.1 times higher in academic MCOs than in statewide MCOs; coagulation defects, 6.4 times higher; transplantations, 4.4; pregnancy, 3.3; cystic fibrosis, 2.4; and prevalence of prematurity was equivalent. Prevalence was higher for academic than for statewide MCOs for 22 of the additional 27 high-cost conditions considered and similar for the remaining 5 conditions. Conclusions Our results suggest that academic MCOs in an MMC system are selected by a large percentage of the sickest patients. Adverse selection may present serious financial risks for AMCs participating in managed care. The term selection bias refers to the process whereby patients disproportionately select certain health plans, provider groups, or individual providers. If a plan is disproportionately selected by patients with expensive health conditions and payments to a health plan do not fully adjust for differences in health status or use of health services, the plan is said to experience adverse selection.1-3 It is well documented that Medicare and privately insured enrollees who select prepaid plans such as health maintenance organizations (HMOs) generally use fewer health services and cost less than their counterparts who remain in a fee-for-service arrangement.4 Health plans consider themselves to be adversely selected when a high percentage of their enrollees have high-cost, chronic, or acute illness, since they risk financial losses in a capitated environment.5 Technically speaking, adverse selection does not typically result from inappropriate consumer choice, since ill individuals appropriately choose providers whom they believe can best meet their health care needs. Rather, adverse selection results from rational selection by enrollees when pricing models fail to consider adequately health status in determining payment. This study primarily focuses on the first component of adverse selection, namely selection bias, by evaluating differences in case-mix experienced by competitive health plans in a Medicaid managed care (MMC) system.6 Academic medical centers (AMCs) may be at increased risk for adverse selection in a managed care system.7-11 Academic medical centers have a long history of providing a medical safety net for the poor and traditionally have provided medical care for the most severely ill patients on a primary care and referral basis.12,13 Whether AMCs own their own managed care organizations (MCOs) or if they are preferred providers for independent MCOs, they may be at particularly high risk for being adversely selected by MMC enrollees who have expensive chronic health conditions. Despite a common perception that AMCs participating in managed care may face high risk of adverse selection, this trend has not been clearly documented nor have its consequences been explored fully. Existing studies of adverse selection by AMCs provide only qualitative assessments and focus on the impact of private sector managed care developments.8-13 The evolution of MMC is more relevant to AMCs than these private sector developments, given the long-standing role of AMCs in the care of the poor, and the already lower payment rates for publicly insured persons who have greater morbidity than those with private insurance.14 As states move from traditional fee-for-service to managed care systems in the public sector, AMCs must continue to serve the publicly insured patient population to fulfill clinical, teaching, and research missions, but they may find that capitated payments are inadequate to cover costs of care for patients with high-cost conditions.8,9,14,15 Furthermore, third-party payers and health plans may increasingly avoid involvement with AMCs because of their higher costs and economically disadvantageous case-mix.8,9,16 Thus, adverse selection under MMC may contribute to increased risk by AMCs for financial insolvency and disruption of their ability to perform essential service, research, and teaching missions.8,9,17 Tennessee's experience with MMC provides an excellent opportunity to examine adverse selection in a managed care system. On January 1, 1994, Tennessee converted its entire Medicaid program to a capitated system. By the end of the first year, Tennessee's statewide experimental managed care program (TennCare) enrolled 1.2 million Medicaid-eligible and formerly uninsured persons who received services through 1 of 12 capitated MCOs (Table 1). Three of the original 12 MCOs were sponsored by AMCs (University of Tennessee, Memphis; Vanderbilt University, Nashville; and University of Tennessee, Knoxville).10,18,19 Capitation payments to the MCOs were initially adjusted only by age, sex, and disability status and averaged $98 per month, a rate considered to be quite low in comparison with actual pre-TennCare Medicaid costs and commercial managed care fees.10,14,15,20 Because of concerns that the low capitation rate might particularly hurt MCOs with an adverse selection, the state solicited MCO input into the development of a formula based on high-cost diagnoses for determining supplemental adverse selection payments to individual MCOs from a reserve fund pool. The amount budgeted for adverse selection payments was $40 million in 1994, which amounted to approximately 2.6% of the total capitation payments made to MCOs in that year.20 The academic MCOs jointly developed a list of 69 specific International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes based on similar methods used by New York, California, and Kentucky and suggested that adjustments be made for MCO size to protect low-enrollment MCOs. After considering many alternative case-mix adjustment methods, the state, in cooperation with the Health Care Financing Administration, adopted an approach that included 8 of the 69 conditions suggested by the AMCs and included pregnancy as an additional high-cost chronic condition and made no adjustments for MCO size. These conditions included acquired immunodeficiency syndrome (AIDS), coagulation defects, cystic fibrosis, pregnancy, prematurity, and heart, kidney, liver, and bone marrow transplantations. According to the adverse selection payment formula, payments for enrollees with high-cost chronic conditions were based on 1993 average Medicaid costs for each condition adjusted for inflation. High-cost chronic condition supplemental payments resulted in effective monthly capitation payments in the first quarter of 1994, which ranged from $153 per enrollee for pregnancy to $3759 for bone marrow transplantation (Table 2). However, the adopted mechanism for high-cost chronic condition payments was viewed unfavorably by representatives of the academic MCOs, who thought it would give larger MCOs a higher proportion of limited high-cost chronic condition payments and that remaining supplemental payment funds would not be sufficient to adjust adequately for the adverse selection experienced by the smaller academic MCOs. TennCare's approach for adverse selection payments was also considered to help adjust for selective marketing practices used by some MCOs during the enrollment period. Some MCOs reportedly offered enrollees inducements, such as secured credit cards, for joining and directed potential clients with chronic diseases or infirmities to other MCOs.10,18 Enrollees were generally assigned to MCOs based on their selection, but approximately 40% of the 714,000 Medicaid recipients who were sent MCO enrollment ballots did not initially respond.18 Many of these were subsequently assigned by the state based upon the proportion of enrollees choosing each MCO in the initial ballot. Therefore, any potential adverse selection experienced by individual MCOs could be attributed to patient selection, disproportionate assignment, or both. The AMC-sponsored MCOs in total received approximately 55,000 enrollees (4.5% of total) (Table 1), which was far fewer than expected and dampened expectations that potential adverse selection would be offset by large enrollment. The implications of the conversion to managed care for AMCs in Tennessee have been discussed by Meyer and Blumenthal,10 who estimated the impact of TennCare on volumes of clinical services and estimated clinical revenues for AMCs. However, their qualitative study relied on anecdotal reports and did not clearly demonstrate that academic MCOs experienced adverse selection or financial hardship in the MMC system. Summitt and colleagues21 have further demonstrated the impact of TennCare on graduate medical education funding. But the potential for AMCs participating in a managed care system to experience adverse selection has only been suggested. 8-10,13,17 This article documents the extent of adverse selection experienced by academic MCOs in Tennessee and the extent to which case-mix disparities may present financial risks for AMCs in an MMC system. Methods The Bureau of TennCare collected the primary data that serve as the basis for this cross-sectional study. The state required that TennCare MCOs submit a list of all enrollees with the 9 state-specified high-cost chronic conditions for all of 1994 and from January through August 1995. For each enrollee listed, the MCO was required to detail direct and indirect patient identifiers, the qualifying high-cost chronic condition diagnosis, and the days eligible for increased payment (based on the number of days of eligibility within the period following the date of diagnosis). The bureau then calculated an aggregate number of enrollees with each high-cost chronic condition eligible for additional payment for each month within each MCO. These data were analyzed further by calculating the prevalence of high-cost chronic conditions for academic, statewide, and regional MCOs. The high-cost chronic conditions analyzed included all 9 conditions specified by the state, with the 4 types of transplantations summarized in a single category (Table 3). For each condition and MCO, average prevalence per 10,000 enrollees was calculated as the period prevalence rate for the entire 20-month period. The numerator was the average number of enrollees eligible for additional payment and the denominator used the May 1995 enrollment figures to approximate the average enrollment for each MCO for the study period. May 1995 MCO enrollment figures were used to calculate prevalence because they were nearest to the middle of the study period of the figures available and enrollment varied less than 10% for all of the MCOs throughout the study period. A prevalence ratio was calculated for each high-cost chronic condition to compare the average prevalence between the academic and statewide MCOs. The period prevalence for additional high-cost conditions for each MCO was calculated using the state Medicaid/TennCare claims database. The conditions considered included those ICD-9-CM diagnoses identified as high cost or very high cost by Kronick and Dreyfus22 using disability payment system (DPS) categories in the Missouri Medicaid database. Twenty-seven diagnoses (listed in Table 4) were identified by the disability payment system, which were not already included as state-specified high-cost chronic conditions. In general, these diagnoses corresponded well with the 69 additional high-cost diagnoses in the list developed jointly by the academic MCOs. In addition, the period prevalence for all diabetes diagnoses was calculated as a summary measure since the high-cost diagnoses of type 1 diabetes (juvenile onset) and type 2 diabetes (adult onset) with complications were both considered. For each condition and MCO, period prevalence per 10,000 enrollees was calculated for the same 20-month period. The numerator was the number of enrollees with at least 1 occurrence of the specified diagnosis and the denominator used the May 1995 enrollment figures to approximate the average enrollment for each MCO for the study period. Data regarding MCO characteristics were obtained from the Bureau of TennCare and validated by external sources where available.18,20 Managed care organization enrollment data were derived from TennCare enrollment and eligibility files and were obtained from state summaries. Results Academic MCOs experienced higher average prevalence for each of the state-specified high-cost chronic conditions compared with statewide MCOs (Table 3). Prevalence was similarly higher when comparing academic MCOs to regional MCOs or to the total Medicaid/TennCare population for all of the state-specified high-cost chronic conditions except prematurity. Although academic MCOs enrolled only 4.5% of the TennCare population, they cared for 38% of the patients with AIDS, 31% with coagulation defects, 10% with cystic fibrosis, 14% with pregnancy, and 26% with transplantations. Also, each of the 3 academic MCOs experienced higher prevalence than the average prevalence for the statewide MCOs or the regional MCOs for all of the state-specified high-cost chronic conditions except prematurity and transplantations. When additional high-cost conditions are considered, the academic MCOs were found to experience higher prevalence than the statewide MCOs for 22 of the additional 27 high-cost conditions considered; for the remaining 5 conditions, prevalence was similar (Table 4). Likewise, prevalence was higher when comparing academic MCOs with regional MCOs for 22 of 27 high-cost conditions. When considering only conditions for which the prevalence was very disparate (prevalence ratio ≥2.0 or ≤0.5), academic MCOs cared for 10.2% of the patients with decubitus ulcers, 8.2% with juvenile arthritis, 16.6% with quadriplegia, 28.4% with sickle cell disease in crisis, 27.7% with profound mental retardation, 13.1% needing tracheostomy, 10.8% needing tracheostomy attention, 16.9% with vena cava thrombosis, and 9.8% with hypertensive renal disease, even though they enrolled only 4.5% of the total TennCare population. The current status of the academic, regional, and statewide MCOs is indicated in Table 1, and their responses to their early experience in an MMC are described. All of the surviving TennCare MCOs have converted to an HMO structure. The academic MCOs have had significantly varied experiences. The state's largest academic MCO has responded to its ongoing financial difficulties by expanding to regional coverage and expanding its enrollment by 39%. Vanderbilt University's HMO has consolidated its TennCare participation by focusing its participation within a community-based practice staffed by nurse practitioners, and its enrollment has decreased by 23%. The University of Tennessee, Knoxville plan, following ongoing yearly losses, became financially insolvent and was subsumed by a new for-profit regional MCO formed by Blue Cross/Blue Shield in January 1996. Comment The prevalence of state-specified high-cost chronic conditions was much higher for academic MCOs than it was for either regional or statewide MCOs. The pattern of adverse selection was shared by each of the academic MCOs and persisted when a large number of additional high- and very high–cost conditions were considered. This study confirms that MCOs representing AMCs experience significant adverse selection by enrollees with many high-cost chronic conditions in an MMC system. Statistical modeling based on HMO claims and expenditure data has clearly demonstrated that smaller HMOs experience the greatest extremes of both favorable and adverse selection.23 Similarly, our data show that the small academic MCOs had the highest prevalence of high-cost conditions and that some of the small regional MCOs had the lowest prevalence seen among the MCOs. However, the case-mix of the statewide MCOs was most similar to that of the smaller MCOs with favorable selection, suggesting that MCO academic status is a stronger predictor of adverse selection than MCO size. The current data are subject to a number of limitations. First, although adverse selection by academic MCOs was demonstrated using the same methods as the State of Tennessee, it is not certain that all alternative methods for assessing case-mix disparities among MCOs would yield the same results. However, when we considered 27 additional high-cost conditions using a well-validated diagnosis-based method,22 the pattern of adverse selection by academic MCOs persisted. Second, because administrative data are known to be imprecise, it is possible that the differences in case-mix were due in part to differences in coding practices. However, there is little a priori reason to suspect systematic differences in coding practices between MCOs. Finally, it is unlikely that the methods used demonstrate the full extent of adverse selection since only a limited set of diagnostic categories was considered, no intermediate cost diagnostic categories were considered, and within each diagnostic category no adjustments were made for severity of illness or comorbidity. To fully delineate the extent of adverse selection by AMCs, it would be necessary to assess both the relative health of the entire population enrolled in the participating health plans and whether the payments provided fully adjust for differences in health status or use of health services. This study examined a limited set of high-cost conditions and did not evaluate the adequacy of payment for these conditions by comparing payments to costs. Several more complex risk adjustment systems are available for assessing case-mix disparities using administrative data, including the ambulatory care groups,6,24 diagnostic cost groups,25 and chronic disease score models.26 However, the 2 high-cost diagnosis methods used in the current study have several advantages over other models in that they are easy to implement, easily verifiable, and more resistant to differences in coding practices or exploitation through submission of exaggerated data.22 Further studies are needed to determine whether academic MCOs also experience adverse selection within intermediate cost diagnostic categories and greater severity of illness and comorbidity within each category. State, federal, and private payers must overcome major obstacles to set equitable payment schedules in managed care systems.27,28 Many alternative methods for risk-adjusting capitation payments exist, but all have limited ability to adjust for differences in case-mix. Risk adjustment by age and sex can account for at most 5% of the variability in annual costs depending on the patient population studied. The adjusted average per capita cost, using 4 demographic factors (age, sex, welfare status, nursing home status) and a geographic factor to adjust capitated payments to Medicare HMOs, only accounts for 1% of total variability in annual Medicare expenditures.25,28 The other more complex risk-adjustment systems using administrative data similarly account for at most 20% to 30% of the variation in health care costs6,24-26,28 and depend more on uniform individual and institutional coding practices.22 Because risk-adjustment methods are so limited, some experts advocate alternative payment schedules such as partial capitation whereby a portion of the payment would be based on actual use28 or additional specific payments to safety net providers. While further research continues to evaluate various payment methods, states must continue to make their best efforts in structuring their payment systems to provide sufficient risk adjustment in determining capitated payments to Medicaid MCOs and to pursue other approaches if necessary to ensure the viability of safety net providers caring for the sickest patients. This study demonstrates conclusively that the sickest patients are most heavily concentrated in academic MCOs in an MMC system. Many authors have noted that the low profitability of AMCs is not solely related to adverse selection but has multiple causes that AMCs must address to ensure their future economic viability.8,9,12,15,29 Clearly, AMCs must adapt traditional ways of doing business to maintain their service, training, and research missions in a managed care environment. However, because of managed care reforms in the public sector, a valuable public resource is increasingly endangered. A high proportion of AMC income comes from publicly funded health care, and academic MCOs may be less able than for-profit health care concerns to shift costs to private payers.13 The historic role of AMCs in providing a safety net for highly complex care is currently threatened by their very success in attracting and serving the sickest patients. Policymakers must explicitly recognize that AMC expertise in providing highly complex care, service to the poor, medical research, and physician training8,17 is dependent on both clinical revenues and government support. The public has a strong interest in ensuring their adequate funding. The public has a larger interest in developing public and private health care systems that do not discourage plans from enrolling and providing services to the sickest patients. References 1. Hellinger FJ. Selection bias in health maintenance organizations: analysis of recent evidence. Health Care Financ Rev.1987;9:55-63.Google Scholar 2. Diehr P, Madden CW, Martin DP. et al. Who enrolled in a state program for the uninsured: was there adverse selection? Med Care.1993;31:1093-1105.Google Scholar 3. Wilensky GR, Rossiter LF. Patient self-selection in HMOs. Health Aff (Millwood).1986;5:66-80.Google Scholar 4. Eggers P. Risk differential between Medicare beneficiaries enrolled and not enrolled in an HMO. Health Care Financ Rev.1980;1:91-99.Google Scholar 5. Kilbreth EH, Coburn AF, McGuire C. et al. State-sponsored programs for the uninsured: is there adverse selection? Inquiry.1998;35:250-265.Google Scholar 6. Fowles JB, Weiner JP, Knutson D. et al. Taking health status into account when setting capitation rates: a comparison of risk-adjustment methods. JAMA.1996;276:1316-1321.Google Scholar 7. Goldman L. The academic health care system. JAMA.1995;273:1549-1552.Google Scholar 8. Iglehart JK. Rapid changes for academic medical centers, I. N Engl J Med.1994;331:1391-1395.Google Scholar 9. Iglehart JK. Rapid changes for academic medical centers, II. N Engl J Med.1995;332:407-411.Google Scholar 10. Meyer GS, Blumenthal D. TennCare and academic medical centers: the lessons from Tennessee. JAMA.1996;276:672-676.Google Scholar 11. Fox P, Wasserman J. Academic medical centers and managed care: uneasy partners. Health Aff (Millwood).1993;12:85-93.Google Scholar 12. Epstein AM. US teaching hospitals in the evolving health care system. JAMA.1995;273:1203-1207.Google Scholar 13. Reuter JA.for the Task Force on Academic Health Centers. Patterns of Specialty Care: Academic Health Centers and the Patient Care Mission. New York, NY: The Commonwealth Fund; January 1999. 14. Holahan J, Rangarajan S, Schirmer M. Medicaid managed care payment rates in 1998. Health Aff (Millwood).1999;18:217-227.Google Scholar 15. Ku L, Hoag S. Medicaid managed care and the marketplace. Inquiry.1998;35:332-345.Google Scholar 16. Culbertson RA. Academic faculty practices: issues for viability in competitive managed care markets. J Health Polit Policy Law.1997;22:1359-1383.Google Scholar 17. Iglehart JK. Medicaid and managed care. N Engl J Med.1995;332:1727-1731.Google Scholar 18. Mirvis DM, Chang CF, Hall CJ, Zarr GT, Applegate WB. TennCare-health system reform for Tennessee. JAMA.1995;274:1235-1241.Google Scholar 19. Hughes R, Meyer DS, Jensen RN, Lennon KA, Kirksey BJ, Sloan K. Medicaid: Tennessee's Program Broadens Coverage but Faces Uncertain Future. Washington, DC: US General Accounting Office; 1995. Publication GAO/HEHS 95-186. 20. Hunt S, Staehlin M, Peters L, Stockard J. Actuarial Review of Capitation Rates in the TennCare Program. Nashville, Tenn: Office of the Comptroller; March 1999. 21. Summitt RL, Herrick RR, Martins M. Addressing a state's physician workforce priorities through the funding of graduate medical education. JAMA.1998;279:767-771.Google Scholar 22. Kronick R, Dreyfus T. The Challenge of Risk Adjustment for People With Disabilities: Health-Based Payment for Medicaid Programs. Princeton, NJ: The Robert Wood Johnson Foundation's Medicaid Managed Care Program for the Center for Health Care Strategies Inc; 1997:1-64. 23. Robinson JC, Gardner LB. Adverse selection among multiple competing health maintenance organizations. Med Care.1995;33:1161-1175.Google Scholar 24. Smith NS, Weiner JP. Applying population-based case mix adjustment in managed care: the Johns Hopkins ambulatory care group system. Managed Care Q.1994;2:21-34.Google Scholar 25. Ellis RP, Pope GC, Iezzoni LI. et al. Diagnosis-based risk adjustment for Medicare capitation payment. Health Care Financ Rev.1996;17:101-128.Google Scholar 26. Clark DO, Von Korff M, Saunders K, Baluch WM, Simon GE. A chronic disease score with empirically derived weights. Med Care.1995;33:783-795.Google Scholar 27. Rogal DL, Gauthier AK. Are health-based payments a feasible tool for addressing risk segmentation? Inquiry.1998;35:115-121.Google Scholar 28. Newhouse JP, Buntin MB, Chapman JD. Risk adjustment and Medicare: taking a closer look. Health Aff (Millwood).1997;15:27-43.Google Scholar 29. Hillman AL, Goldfarb N, Eisenberg JM, Kelley MA. An academic medical center's experience with mandatory managed care for Medicaid recipients. Acad Med.1991;3:134-138.Google Scholar http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JAMA American Medical Association

Academic Managed Care Organizations and Adverse Selection Under Medicaid Managed Care in Tennessee

Download PDF
6 pages

Loading next page...
 
/lp/american-medical-association/academic-managed-care-organizations-and-adverse-selection-under-tHhqiuzeO2

References (28)

Publisher
American Medical Association
Copyright
Copyright © 1999 American Medical Association. All Rights Reserved.
ISSN
0098-7484
eISSN
1538-3598
DOI
10.1001/jama.282.11.1067
Publisher site
See Article on Publisher Site

Abstract

Abstract Context Health plans competing in a managed care system may face serious financial consequences if they are disproportionately selected by enrollees with expensive health conditions. Academic medical centers (AMCs) have traditionally provided medical care for the sickest patients and may be at particularly high risk for adverse selection, but whether this occurs is not known. Objective To determine whether managed care organizations (MCOs) representing AMCs are adversely selected by Medicaid managed care (MMC) enrollees with expensive chronic health conditions. Design and Setting Observational study using state Medicaid claims data from all of 1994 and January to August 1995 for Tennessee's statewide MMC program (TennCare). Participants All 12 capitated MCOs in Tennessee, which collectively provided services for 1.2 million Medicaid enrollees from January 1994 through August 1995 following the initiation of TennCare. Main Outcome Measures Prevalence of 6 state-specified high-cost chronic conditions—acquired immunodeficiency syndrome (AIDS), coagulation defects, cystic fibrosis, pregnancy, prematurity, and organ transplantation—and 27 additional high-cost conditions compared by academic, statewide, and regional MCOs. Results The prevalence of state-specified high-cost chronic conditions was generally higher for academic MCOs compared with other MCOs. Specifically, prevalence of AIDS was 14.1 times higher in academic MCOs than in statewide MCOs; coagulation defects, 6.4 times higher; transplantations, 4.4; pregnancy, 3.3; cystic fibrosis, 2.4; and prevalence of prematurity was equivalent. Prevalence was higher for academic than for statewide MCOs for 22 of the additional 27 high-cost conditions considered and similar for the remaining 5 conditions. Conclusions Our results suggest that academic MCOs in an MMC system are selected by a large percentage of the sickest patients. Adverse selection may present serious financial risks for AMCs participating in managed care. The term selection bias refers to the process whereby patients disproportionately select certain health plans, provider groups, or individual providers. If a plan is disproportionately selected by patients with expensive health conditions and payments to a health plan do not fully adjust for differences in health status or use of health services, the plan is said to experience adverse selection.1-3 It is well documented that Medicare and privately insured enrollees who select prepaid plans such as health maintenance organizations (HMOs) generally use fewer health services and cost less than their counterparts who remain in a fee-for-service arrangement.4 Health plans consider themselves to be adversely selected when a high percentage of their enrollees have high-cost, chronic, or acute illness, since they risk financial losses in a capitated environment.5 Technically speaking, adverse selection does not typically result from inappropriate consumer choice, since ill individuals appropriately choose providers whom they believe can best meet their health care needs. Rather, adverse selection results from rational selection by enrollees when pricing models fail to consider adequately health status in determining payment. This study primarily focuses on the first component of adverse selection, namely selection bias, by evaluating differences in case-mix experienced by competitive health plans in a Medicaid managed care (MMC) system.6 Academic medical centers (AMCs) may be at increased risk for adverse selection in a managed care system.7-11 Academic medical centers have a long history of providing a medical safety net for the poor and traditionally have provided medical care for the most severely ill patients on a primary care and referral basis.12,13 Whether AMCs own their own managed care organizations (MCOs) or if they are preferred providers for independent MCOs, they may be at particularly high risk for being adversely selected by MMC enrollees who have expensive chronic health conditions. Despite a common perception that AMCs participating in managed care may face high risk of adverse selection, this trend has not been clearly documented nor have its consequences been explored fully. Existing studies of adverse selection by AMCs provide only qualitative assessments and focus on the impact of private sector managed care developments.8-13 The evolution of MMC is more relevant to AMCs than these private sector developments, given the long-standing role of AMCs in the care of the poor, and the already lower payment rates for publicly insured persons who have greater morbidity than those with private insurance.14 As states move from traditional fee-for-service to managed care systems in the public sector, AMCs must continue to serve the publicly insured patient population to fulfill clinical, teaching, and research missions, but they may find that capitated payments are inadequate to cover costs of care for patients with high-cost conditions.8,9,14,15 Furthermore, third-party payers and health plans may increasingly avoid involvement with AMCs because of their higher costs and economically disadvantageous case-mix.8,9,16 Thus, adverse selection under MMC may contribute to increased risk by AMCs for financial insolvency and disruption of their ability to perform essential service, research, and teaching missions.8,9,17 Tennessee's experience with MMC provides an excellent opportunity to examine adverse selection in a managed care system. On January 1, 1994, Tennessee converted its entire Medicaid program to a capitated system. By the end of the first year, Tennessee's statewide experimental managed care program (TennCare) enrolled 1.2 million Medicaid-eligible and formerly uninsured persons who received services through 1 of 12 capitated MCOs (Table 1). Three of the original 12 MCOs were sponsored by AMCs (University of Tennessee, Memphis; Vanderbilt University, Nashville; and University of Tennessee, Knoxville).10,18,19 Capitation payments to the MCOs were initially adjusted only by age, sex, and disability status and averaged $98 per month, a rate considered to be quite low in comparison with actual pre-TennCare Medicaid costs and commercial managed care fees.10,14,15,20 Because of concerns that the low capitation rate might particularly hurt MCOs with an adverse selection, the state solicited MCO input into the development of a formula based on high-cost diagnoses for determining supplemental adverse selection payments to individual MCOs from a reserve fund pool. The amount budgeted for adverse selection payments was $40 million in 1994, which amounted to approximately 2.6% of the total capitation payments made to MCOs in that year.20 The academic MCOs jointly developed a list of 69 specific International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes based on similar methods used by New York, California, and Kentucky and suggested that adjustments be made for MCO size to protect low-enrollment MCOs. After considering many alternative case-mix adjustment methods, the state, in cooperation with the Health Care Financing Administration, adopted an approach that included 8 of the 69 conditions suggested by the AMCs and included pregnancy as an additional high-cost chronic condition and made no adjustments for MCO size. These conditions included acquired immunodeficiency syndrome (AIDS), coagulation defects, cystic fibrosis, pregnancy, prematurity, and heart, kidney, liver, and bone marrow transplantations. According to the adverse selection payment formula, payments for enrollees with high-cost chronic conditions were based on 1993 average Medicaid costs for each condition adjusted for inflation. High-cost chronic condition supplemental payments resulted in effective monthly capitation payments in the first quarter of 1994, which ranged from $153 per enrollee for pregnancy to $3759 for bone marrow transplantation (Table 2). However, the adopted mechanism for high-cost chronic condition payments was viewed unfavorably by representatives of the academic MCOs, who thought it would give larger MCOs a higher proportion of limited high-cost chronic condition payments and that remaining supplemental payment funds would not be sufficient to adjust adequately for the adverse selection experienced by the smaller academic MCOs. TennCare's approach for adverse selection payments was also considered to help adjust for selective marketing practices used by some MCOs during the enrollment period. Some MCOs reportedly offered enrollees inducements, such as secured credit cards, for joining and directed potential clients with chronic diseases or infirmities to other MCOs.10,18 Enrollees were generally assigned to MCOs based on their selection, but approximately 40% of the 714,000 Medicaid recipients who were sent MCO enrollment ballots did not initially respond.18 Many of these were subsequently assigned by the state based upon the proportion of enrollees choosing each MCO in the initial ballot. Therefore, any potential adverse selection experienced by individual MCOs could be attributed to patient selection, disproportionate assignment, or both. The AMC-sponsored MCOs in total received approximately 55,000 enrollees (4.5% of total) (Table 1), which was far fewer than expected and dampened expectations that potential adverse selection would be offset by large enrollment. The implications of the conversion to managed care for AMCs in Tennessee have been discussed by Meyer and Blumenthal,10 who estimated the impact of TennCare on volumes of clinical services and estimated clinical revenues for AMCs. However, their qualitative study relied on anecdotal reports and did not clearly demonstrate that academic MCOs experienced adverse selection or financial hardship in the MMC system. Summitt and colleagues21 have further demonstrated the impact of TennCare on graduate medical education funding. But the potential for AMCs participating in a managed care system to experience adverse selection has only been suggested. 8-10,13,17 This article documents the extent of adverse selection experienced by academic MCOs in Tennessee and the extent to which case-mix disparities may present financial risks for AMCs in an MMC system. Methods The Bureau of TennCare collected the primary data that serve as the basis for this cross-sectional study. The state required that TennCare MCOs submit a list of all enrollees with the 9 state-specified high-cost chronic conditions for all of 1994 and from January through August 1995. For each enrollee listed, the MCO was required to detail direct and indirect patient identifiers, the qualifying high-cost chronic condition diagnosis, and the days eligible for increased payment (based on the number of days of eligibility within the period following the date of diagnosis). The bureau then calculated an aggregate number of enrollees with each high-cost chronic condition eligible for additional payment for each month within each MCO. These data were analyzed further by calculating the prevalence of high-cost chronic conditions for academic, statewide, and regional MCOs. The high-cost chronic conditions analyzed included all 9 conditions specified by the state, with the 4 types of transplantations summarized in a single category (Table 3). For each condition and MCO, average prevalence per 10,000 enrollees was calculated as the period prevalence rate for the entire 20-month period. The numerator was the average number of enrollees eligible for additional payment and the denominator used the May 1995 enrollment figures to approximate the average enrollment for each MCO for the study period. May 1995 MCO enrollment figures were used to calculate prevalence because they were nearest to the middle of the study period of the figures available and enrollment varied less than 10% for all of the MCOs throughout the study period. A prevalence ratio was calculated for each high-cost chronic condition to compare the average prevalence between the academic and statewide MCOs. The period prevalence for additional high-cost conditions for each MCO was calculated using the state Medicaid/TennCare claims database. The conditions considered included those ICD-9-CM diagnoses identified as high cost or very high cost by Kronick and Dreyfus22 using disability payment system (DPS) categories in the Missouri Medicaid database. Twenty-seven diagnoses (listed in Table 4) were identified by the disability payment system, which were not already included as state-specified high-cost chronic conditions. In general, these diagnoses corresponded well with the 69 additional high-cost diagnoses in the list developed jointly by the academic MCOs. In addition, the period prevalence for all diabetes diagnoses was calculated as a summary measure since the high-cost diagnoses of type 1 diabetes (juvenile onset) and type 2 diabetes (adult onset) with complications were both considered. For each condition and MCO, period prevalence per 10,000 enrollees was calculated for the same 20-month period. The numerator was the number of enrollees with at least 1 occurrence of the specified diagnosis and the denominator used the May 1995 enrollment figures to approximate the average enrollment for each MCO for the study period. Data regarding MCO characteristics were obtained from the Bureau of TennCare and validated by external sources where available.18,20 Managed care organization enrollment data were derived from TennCare enrollment and eligibility files and were obtained from state summaries. Results Academic MCOs experienced higher average prevalence for each of the state-specified high-cost chronic conditions compared with statewide MCOs (Table 3). Prevalence was similarly higher when comparing academic MCOs to regional MCOs or to the total Medicaid/TennCare population for all of the state-specified high-cost chronic conditions except prematurity. Although academic MCOs enrolled only 4.5% of the TennCare population, they cared for 38% of the patients with AIDS, 31% with coagulation defects, 10% with cystic fibrosis, 14% with pregnancy, and 26% with transplantations. Also, each of the 3 academic MCOs experienced higher prevalence than the average prevalence for the statewide MCOs or the regional MCOs for all of the state-specified high-cost chronic conditions except prematurity and transplantations. When additional high-cost conditions are considered, the academic MCOs were found to experience higher prevalence than the statewide MCOs for 22 of the additional 27 high-cost conditions considered; for the remaining 5 conditions, prevalence was similar (Table 4). Likewise, prevalence was higher when comparing academic MCOs with regional MCOs for 22 of 27 high-cost conditions. When considering only conditions for which the prevalence was very disparate (prevalence ratio ≥2.0 or ≤0.5), academic MCOs cared for 10.2% of the patients with decubitus ulcers, 8.2% with juvenile arthritis, 16.6% with quadriplegia, 28.4% with sickle cell disease in crisis, 27.7% with profound mental retardation, 13.1% needing tracheostomy, 10.8% needing tracheostomy attention, 16.9% with vena cava thrombosis, and 9.8% with hypertensive renal disease, even though they enrolled only 4.5% of the total TennCare population. The current status of the academic, regional, and statewide MCOs is indicated in Table 1, and their responses to their early experience in an MMC are described. All of the surviving TennCare MCOs have converted to an HMO structure. The academic MCOs have had significantly varied experiences. The state's largest academic MCO has responded to its ongoing financial difficulties by expanding to regional coverage and expanding its enrollment by 39%. Vanderbilt University's HMO has consolidated its TennCare participation by focusing its participation within a community-based practice staffed by nurse practitioners, and its enrollment has decreased by 23%. The University of Tennessee, Knoxville plan, following ongoing yearly losses, became financially insolvent and was subsumed by a new for-profit regional MCO formed by Blue Cross/Blue Shield in January 1996. Comment The prevalence of state-specified high-cost chronic conditions was much higher for academic MCOs than it was for either regional or statewide MCOs. The pattern of adverse selection was shared by each of the academic MCOs and persisted when a large number of additional high- and very high–cost conditions were considered. This study confirms that MCOs representing AMCs experience significant adverse selection by enrollees with many high-cost chronic conditions in an MMC system. Statistical modeling based on HMO claims and expenditure data has clearly demonstrated that smaller HMOs experience the greatest extremes of both favorable and adverse selection.23 Similarly, our data show that the small academic MCOs had the highest prevalence of high-cost conditions and that some of the small regional MCOs had the lowest prevalence seen among the MCOs. However, the case-mix of the statewide MCOs was most similar to that of the smaller MCOs with favorable selection, suggesting that MCO academic status is a stronger predictor of adverse selection than MCO size. The current data are subject to a number of limitations. First, although adverse selection by academic MCOs was demonstrated using the same methods as the State of Tennessee, it is not certain that all alternative methods for assessing case-mix disparities among MCOs would yield the same results. However, when we considered 27 additional high-cost conditions using a well-validated diagnosis-based method,22 the pattern of adverse selection by academic MCOs persisted. Second, because administrative data are known to be imprecise, it is possible that the differences in case-mix were due in part to differences in coding practices. However, there is little a priori reason to suspect systematic differences in coding practices between MCOs. Finally, it is unlikely that the methods used demonstrate the full extent of adverse selection since only a limited set of diagnostic categories was considered, no intermediate cost diagnostic categories were considered, and within each diagnostic category no adjustments were made for severity of illness or comorbidity. To fully delineate the extent of adverse selection by AMCs, it would be necessary to assess both the relative health of the entire population enrolled in the participating health plans and whether the payments provided fully adjust for differences in health status or use of health services. This study examined a limited set of high-cost conditions and did not evaluate the adequacy of payment for these conditions by comparing payments to costs. Several more complex risk adjustment systems are available for assessing case-mix disparities using administrative data, including the ambulatory care groups,6,24 diagnostic cost groups,25 and chronic disease score models.26 However, the 2 high-cost diagnosis methods used in the current study have several advantages over other models in that they are easy to implement, easily verifiable, and more resistant to differences in coding practices or exploitation through submission of exaggerated data.22 Further studies are needed to determine whether academic MCOs also experience adverse selection within intermediate cost diagnostic categories and greater severity of illness and comorbidity within each category. State, federal, and private payers must overcome major obstacles to set equitable payment schedules in managed care systems.27,28 Many alternative methods for risk-adjusting capitation payments exist, but all have limited ability to adjust for differences in case-mix. Risk adjustment by age and sex can account for at most 5% of the variability in annual costs depending on the patient population studied. The adjusted average per capita cost, using 4 demographic factors (age, sex, welfare status, nursing home status) and a geographic factor to adjust capitated payments to Medicare HMOs, only accounts for 1% of total variability in annual Medicare expenditures.25,28 The other more complex risk-adjustment systems using administrative data similarly account for at most 20% to 30% of the variation in health care costs6,24-26,28 and depend more on uniform individual and institutional coding practices.22 Because risk-adjustment methods are so limited, some experts advocate alternative payment schedules such as partial capitation whereby a portion of the payment would be based on actual use28 or additional specific payments to safety net providers. While further research continues to evaluate various payment methods, states must continue to make their best efforts in structuring their payment systems to provide sufficient risk adjustment in determining capitated payments to Medicaid MCOs and to pursue other approaches if necessary to ensure the viability of safety net providers caring for the sickest patients. This study demonstrates conclusively that the sickest patients are most heavily concentrated in academic MCOs in an MMC system. Many authors have noted that the low profitability of AMCs is not solely related to adverse selection but has multiple causes that AMCs must address to ensure their future economic viability.8,9,12,15,29 Clearly, AMCs must adapt traditional ways of doing business to maintain their service, training, and research missions in a managed care environment. However, because of managed care reforms in the public sector, a valuable public resource is increasingly endangered. A high proportion of AMC income comes from publicly funded health care, and academic MCOs may be less able than for-profit health care concerns to shift costs to private payers.13 The historic role of AMCs in providing a safety net for highly complex care is currently threatened by their very success in attracting and serving the sickest patients. Policymakers must explicitly recognize that AMC expertise in providing highly complex care, service to the poor, medical research, and physician training8,17 is dependent on both clinical revenues and government support. The public has a strong interest in ensuring their adequate funding. The public has a larger interest in developing public and private health care systems that do not discourage plans from enrolling and providing services to the sickest patients. References 1. Hellinger FJ. Selection bias in health maintenance organizations: analysis of recent evidence. Health Care Financ Rev.1987;9:55-63.Google Scholar 2. Diehr P, Madden CW, Martin DP. et al. Who enrolled in a state program for the uninsured: was there adverse selection? Med Care.1993;31:1093-1105.Google Scholar 3. Wilensky GR, Rossiter LF. Patient self-selection in HMOs. Health Aff (Millwood).1986;5:66-80.Google Scholar 4. Eggers P. Risk differential between Medicare beneficiaries enrolled and not enrolled in an HMO. Health Care Financ Rev.1980;1:91-99.Google Scholar 5. Kilbreth EH, Coburn AF, McGuire C. et al. State-sponsored programs for the uninsured: is there adverse selection? Inquiry.1998;35:250-265.Google Scholar 6. Fowles JB, Weiner JP, Knutson D. et al. Taking health status into account when setting capitation rates: a comparison of risk-adjustment methods. JAMA.1996;276:1316-1321.Google Scholar 7. Goldman L. The academic health care system. JAMA.1995;273:1549-1552.Google Scholar 8. Iglehart JK. Rapid changes for academic medical centers, I. N Engl J Med.1994;331:1391-1395.Google Scholar 9. Iglehart JK. Rapid changes for academic medical centers, II. N Engl J Med.1995;332:407-411.Google Scholar 10. Meyer GS, Blumenthal D. TennCare and academic medical centers: the lessons from Tennessee. JAMA.1996;276:672-676.Google Scholar 11. Fox P, Wasserman J. Academic medical centers and managed care: uneasy partners. Health Aff (Millwood).1993;12:85-93.Google Scholar 12. Epstein AM. US teaching hospitals in the evolving health care system. JAMA.1995;273:1203-1207.Google Scholar 13. Reuter JA.for the Task Force on Academic Health Centers. Patterns of Specialty Care: Academic Health Centers and the Patient Care Mission. New York, NY: The Commonwealth Fund; January 1999. 14. Holahan J, Rangarajan S, Schirmer M. Medicaid managed care payment rates in 1998. Health Aff (Millwood).1999;18:217-227.Google Scholar 15. Ku L, Hoag S. Medicaid managed care and the marketplace. Inquiry.1998;35:332-345.Google Scholar 16. Culbertson RA. Academic faculty practices: issues for viability in competitive managed care markets. J Health Polit Policy Law.1997;22:1359-1383.Google Scholar 17. Iglehart JK. Medicaid and managed care. N Engl J Med.1995;332:1727-1731.Google Scholar 18. Mirvis DM, Chang CF, Hall CJ, Zarr GT, Applegate WB. TennCare-health system reform for Tennessee. JAMA.1995;274:1235-1241.Google Scholar 19. Hughes R, Meyer DS, Jensen RN, Lennon KA, Kirksey BJ, Sloan K. Medicaid: Tennessee's Program Broadens Coverage but Faces Uncertain Future. Washington, DC: US General Accounting Office; 1995. Publication GAO/HEHS 95-186. 20. Hunt S, Staehlin M, Peters L, Stockard J. Actuarial Review of Capitation Rates in the TennCare Program. Nashville, Tenn: Office of the Comptroller; March 1999. 21. Summitt RL, Herrick RR, Martins M. Addressing a state's physician workforce priorities through the funding of graduate medical education. JAMA.1998;279:767-771.Google Scholar 22. Kronick R, Dreyfus T. The Challenge of Risk Adjustment for People With Disabilities: Health-Based Payment for Medicaid Programs. Princeton, NJ: The Robert Wood Johnson Foundation's Medicaid Managed Care Program for the Center for Health Care Strategies Inc; 1997:1-64. 23. Robinson JC, Gardner LB. Adverse selection among multiple competing health maintenance organizations. Med Care.1995;33:1161-1175.Google Scholar 24. Smith NS, Weiner JP. Applying population-based case mix adjustment in managed care: the Johns Hopkins ambulatory care group system. Managed Care Q.1994;2:21-34.Google Scholar 25. Ellis RP, Pope GC, Iezzoni LI. et al. Diagnosis-based risk adjustment for Medicare capitation payment. Health Care Financ Rev.1996;17:101-128.Google Scholar 26. Clark DO, Von Korff M, Saunders K, Baluch WM, Simon GE. A chronic disease score with empirically derived weights. Med Care.1995;33:783-795.Google Scholar 27. Rogal DL, Gauthier AK. Are health-based payments a feasible tool for addressing risk segmentation? Inquiry.1998;35:115-121.Google Scholar 28. Newhouse JP, Buntin MB, Chapman JD. Risk adjustment and Medicare: taking a closer look. Health Aff (Millwood).1997;15:27-43.Google Scholar 29. Hillman AL, Goldfarb N, Eisenberg JM, Kelley MA. An academic medical center's experience with mandatory managed care for Medicaid recipients. Acad Med.1991;3:134-138.Google Scholar

Journal

JAMA – American Medical Association

Published: Sep 15, 1999

Keywords: acquired immunodeficiency syndrome,blood coagulation disorders,cystic fibrosis,chronic disease,managed care programs,medicaid,tennessee,transplantation,premature birth,managed care organizations,infant, premature,organ transplantation,pregnancy,academic medical centers,observational studies

There are no references for this article.