ORIGINAL RESEARCH Hospitalization Rates for Ambulatory Care Sensitive Conditions in California Medicare HMOs Feng Zeng, PhD, June F. O Leary, PhD, Elizabeth M. Sloss, PhD, Nasreen Dhanani, PhD, and Glenn Melnick, PhD Abstract Objective: To examine annual hospitalization rates for ambulatory care sensitive conditions (ACSCs) among Medicare HMO beneficiaries. Design: Cross-sectional descriptive study. Setting and participants: Medicare beneficiaries aged 65 years and older continuously enrolled in 1 of 15 California Medicare HMOs from 1 January through 31 December 2001. Measurements: Hospitalization rates overall and for each of 15 ACSCs (bacterial pneumonia, cellulitis, dehydration, gastric and duodenal ulcer, hypoglycemia, hypokalemia, severe ear/nose/throat infections, urinary tract infections, asthma/chronic obstructive pulmonary disease, congestive heart failure, diabetes, hypertension, seizure disorder, influenza, and malnutrition). The rate for the 15 ACSCs combined as well as acute, chronic, and preventable indices were also estimated. Results: Of the 1.2 million Medicare beneficiaries enrolled in California HMOs during 2001, 24% were 80 years or older (range among plans, 15% 34%), 5% were African American (range, 1% 25%), and 6% were Medicaid-eligible (range, 3% 16%). Enrollees experienced a total of 315,503 hospitalizations in 2001 (267 per 1000), 22% of which were for an ACSC. ACSC hospitalization rates varied widely by plan and were higher among older enrollees, males, and those eligible for Medicaid. Conclusion: ACSC hospitalization rates are easy to calculate based on administrative data. These rates can be used by individual plans as a method to screen for possible access and quality of care problems. Monitoring performance of Medicare HMOs continues to be a topic of considerable interest to policymakers and health plan administrators. A measure that has been suggested for tracking the performance of these plans is hospitalization rates for ambulatory care sensitive conditions (ACSCs) [1]. This approach is based on the assumption that providing patients with timely and appropriate care in the outpatient setting will lead to reduced hospitalizations for ACSCs [2 4]. High ACSC admission rates may be indicative of problems with availability, access, or appropriateness of primary care, whereas low ACSC admission rates might reflect provision of effective primary care and preventive services [5]. Use of annual ACSC hospitalization rates to monitor access to care in Medicare HMOs builds on existing models developed for monitoring care for Medicare fee-for-service (FFS) beneficiaries. The rates are based on inpatient hospital encounter data that are routinely collected by Medicare HMOs. Tracking ACSC hospitalization rates over time is a practical tool that Medicare HMOs could use to identify clinical areas and subgroups of enrollees in further need of quality assessment and improvement. McCall et al [1] investigated the feasibility of using ACSC hospitalization rates to monitor the performance of Medicare HMOs. They hypothesized that ACSC hospitalization rates could be used for monitoring Medicare HMOs if 3 conditions are met: complete hospital data are available, ACSC hospitalization rates are considered to be valid measures of providing ambulatory care, and ACSC hospitalization rates are statistically reliable [1]. Using hospital inpatient encounter data for July 1997 to June 1998, McCall et al identified 305 Medicare HMOs with up to 255,520 enrollees and up to 54,009 discharges per HMO. They first evaluated the completeness of the inpatient encounter data and found it to be sufficiently complete based on a comparison of hospitalization rates for all conditions versus only ACSCs between Medicare HMO and FFS beneficiaries. Hospitalization rates were then calculated by age and sex and for ACSCs individually and collectively. They found that although Medicare HMO rates were consistently lower than Medicare FFS, the Medicare HMO From the School of Policy, Planning, and Development, University of Southern California, and therand Corporation, Santa Monica, CA, andwashington, DC. www.turner-white.com Vol. 12, No. 11 November 2005 JCOM 559
rates varied as expected with regard to geographic region and demographic characteristics. For example, higher ACSC hospitalization rates were observed in the Northeast and among those aged 85 years and older. Many studies have investigated inpatient utilization differences between Medicare FFS and HMO populations, with much of the difference attributed to healthier HMO enrollees (ie, favorable selection) and some to managed care practices of HMOs [6 14]. While differences in enrollee characteristics and utilization among Medicare HMOs are expected to be less than the selection difference between Medicare HMOs and FFS, these differences by plan have not been well documented. Understanding these differences by plan is important because rates of hospitalization for ACSCs have been shown to be sensitive to patient characteristics such as age, race/ethnicity, and income [4,15 19]. In the current study, we examine rates of hospitalization for 15 ACSCs using data on beneficiaries aged 65 years and older enrolled in Medicare HMOs in California during 2001. The specific objectives of the study were to compare annual hospitalization rates in Medicare HMOs for all conditions and for ACSC conditions by beneficiary characteristics and to estimate rates of ACSC hospitalizations at the plan level for Medicare HMOs. Methods Data Sources Data for this study were derived from linking Medicare enrollment data on all beneficiaries in California between January and December 2001 from the Centers for Medicare & Medicaid Services (CMS) denominator files to inpatient discharge data for short-term stays from the California Office of Statewide Health Planning and Development. All nonfederal hospitals in California submit discharge records to the state agency irrespective of payer source. Social Security number served as the starting point for the linkage, and if it was missing from either the Medicare data or discharge data, the record was dropped (< 2% of records from either file). Records were linked using probabilistic matching based on Social Security number, zip code of residence, date of birth, date of death, gender, and race/ethnicity [20 24]. The level of agreement between Social Security numbers in the final linked file was nearly 99%. The linkages were performed by Health Information Solutions with approvals from the Committee for the Protection of Human Subjects, California Health and Human Services Agency, and the University Park Institutional Review Board, University of Southern California (USC), under a CMS data use agreement between all parties with access to the confidential data. The linked data were returned to USC after all potential identifiers were stripped. CMS plan contract numbers from our data were linked to model type from CMS [25]. Selection of Ambulatory Care Sensitive Conditions We identified ACSC hospitalizations based on the principal diagnosis using codes from the ICD-9-CM. We selected the 15 ACSCs that were identified by 2 clinical consultants as particularly relevant to the elderly in a study by McCall et al [1]. The 15 ACSCs were bacterial pneumonia, cellulitis, dehydration, gastric and duodenal ulcer, hypoglycemia, hypokalemia, severe ENT (ear/nose/throat) infections, urinary tract infections (UTIs), asthma/chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), diabetes, hypertension, seizure disorder, influenza, and malnutrition. Individual condition rates as well as a total rate that combined all 15 conditions were estimated. In addition, we aggregated conditions into 3 indices as defined by McCall et al [1]: acute (bacterial pneumonia, cellulitis, dehydration, hypoglycemia, hypokalemia, gastric and duodenal ulcer, severe ENT infections, and UTIs), chronic (asthma/copd, CHF, diabetes, hypertension, and seizure disorder), and preventable (influenza and malnutrition). The ICD-9-CM codes used to identify ACSC hospitalizations are provided in Table 1. Sample Selection The study period extends from 1 January through 31 December 2001. Medicare HMO enrollees were eligible for the study if they met the following criteria: (1) age 65 by 1 January 2001, (2) did not have end-stage renal disease, (3) continuously eligible for both Medicare Part A and Part B, and (4) enrolled in the same California Medicare HMO for the entire study period or until death, whichever came first. Any beneficiaries who disenrolled to the traditional FFS Medicare plan, switched between HMOs, or moved during the study period were excluded. In 2001, there were 15 Medicare HMOs operating under risk contracts in California [25]. Enrollees of HMOs outside California as well as enrollees of plans with nonrisk contracts (eg, HMOs with cost contracts, demonstrations, health care prepayment plans) were excluded. Calculation of Hospitalization Rates Rates of hospitalization for ACSCs were calculated as the number of ACSC hospitalizations per 1000 Medicare HMO person-years. Person-years were used to adjust for partial year enrollment due to death. For example, 2 enrollees who both died on 1 July 2001 each contribute 6 months of enrollment or together 12 months of data, which yields 1 personyear. All hospitalization rates are unadjusted. Statistical significance of differences between groups was tested using analysis of variance, performed using SAS version 8.0 (SAS Institute, Inc, Cary, NC) [26]. Results More than 1.2 million Medicare beneficiaries were enrolled 560 JCOM November 2005 Vol. 12, No. 11 www.turner-white.com
ORIGINAL RESEARCH Table 1. ICD-9-CM Diagnosis Codes for 15 Ambulatory Care Sensitive Conditions Condition Diagnosis Codes Asthma/chronic obstruc- 491xx, 492xx, 493xx, 494xx, 496xx tive pulmonary disease Bacterial pneumonia 481, 482xx, 483xx, 485, 486 Congestive heart failure 40201, 40211, 40291, 428xx, 5184 Cellulitis 681xx, 682xx, 683, 686xx Dehydration 2765 Diabetes mellitus 2500x, 2501x, 2502x, 2503x, 2508x, 2509x Gastric and duodenal 531xx, 532xx, 533xx ulcer Hypertension 401x, 40200, 40210, 40290, 403xx, 40400, 40402, 40410, 40412, 40490, 40492, 405xx, 4372 Hypoglycemia 2510, 2511, 2512 Hypokalemia 2768 Influenza 487x Malnutrition 260, 261, 262, 263x, 264x, 265x, 266x, 267, 268x, 269x Seizure disorder 3450x, 3451x, 3452x, 3453x, 3454x, 3455x, 3457x, 3458x, 3459x, 7803x Severe ear, nose, and 382xx, 462, 463, 465x, 4721 throat infection Urinary tract infection 590xx, 5990, 5950, 5970, 5978, 6010, 6012, 6013 during 2001 in the 15 California HMOs and included in the analysis (Table 2). Of these, more than half were in an IPA (independent practice association) and about a quarter were aged 80 years or older. About 5% were African American while 11% were classified in another or unknown racial category. Almost 6% were eligible for Medicaid (ie, dual eligible). These statistics mask vastly different characteristics across plans (Table 2), whose enrollments ranged from less than 2000 to over 450,000. Among the 15 plans, the youngest beneficiary group (65 69 years) made up 18% to 43% of total enrollment, whereas the oldest beneficiary group ( 85 years) accounted for 6% to 16%. Some plans enrolled relatively few African Americans whereas others had enrollments up to 25%. The percentage of Medicare enrollees also eligible for Medicaid varied from 3% to 16% among the 15 plans. The Medicare beneficiaries enrolled in California HMOs during 2001 experienced 315,503 hospitalizations (Table 3). Of these, 1 of every 5 hospitalizations was for an ACSC. With respect to beneficiary characteristics, the pattern of ACSC hospitalization rates closely followed the hospitalization rates for all conditions. ACSC hospitalization rates were higher among enrollees of IPAs (versus group model HMOs), males, and those eligible for Medicaid (P < 0.01). Table 2. Medicare HMO Enrollee Characteristics for All Plans Combined and Minimum and Maximum by Plan, California 2001 All Plans Combined By Plan Minimum Maximum Characteristic Number Percent Number Number Total beneficiaries 1,208,566 100.0 1753 451,097 HMO model type IPA 703,882 58.2 1953 451,097 Group 504,684 41.8 1753 247,187 Minimum Maximum Age, yr Percent Percent 65 69 314,469 26.0 17.8 42.9 70 74 327,016 27.1 23.6 29.6 75 79 273,051 22.6 15.2 25.1 80 84 169,166 14.0 08.9 18.0 85+ 124,864 10.3 05.6 15.6 Gender Female 693,127 57.4 50.4 62.5 Male 515,439 42.6 37.5 49.6 Race African 65,043 05.4 00.5 24.5 American White 1,006,868 83.3 01.2 96.6 Other/unknown 136,655 11.3 02.9 98.1 Medicaid* Eligible 65,309 05.4 02.6 16.4 Not eligible 1,143,257 94.6 83.6 97.4 IPA = independent practice association. *Includes those eligible for Medicaid for at least 1 month during 2001. ACSC hospitalization rates increased markedly with age (more than 4 times as high among those 85 years compared with those 65 69 years; P < 0.01). Those classified in the other or unknown race category experienced fewer ACSC hospitalizations than those in the white (27% more; P < 0.01) or African American categories (61% more; P < 0.01). Medicare beneficiaries were hospitalized infrequently for several of the ACSCs during 2001, resulting in a small number of hospitalizations for individual plans (Table 4). For 4 of the ACSC conditions (hypoglycemia, severe ENT infections, influenza, and malnutrition), 5 or more of the 15 plans had no hospitalizations during the study year. For another 4 conditions (hypokalemia, diabetes mellitus, hypertension, and seizure disorder), at least 1 plan did not have a hospitalization during 2001. For each of these 8 conditions, the median annual hospitalization rate for the 15 plans was less than 2 per 1000 person-years. For the remaining 7 more frequent www.turner-white.com Vol. 12, No. 11 November 2005 JCOM 561
Table 3. Medicare HMO Hospitalization Rates for All Conditions and 15 ACSCs, California 2001 All Conditions 15 ACSCs Characteristic No. of Hospitalizations Person-years No. of Hospitalizations Person-years Total person-years 315,503 266.9 67,953 57.5 Total beneficiaries 315,503 261.1 67,953 56.2 HMO model type IPA 204,346 290.3 44,306 62.9 Group 111,157 220.3 23,647 46.9 Age, yr 65 69 52,319 166.4,08774 27.9 70 74 71,655 219.1 13,634 41.7 75 79 76,556 280.4 16,060 58.8 80 84 58,948 348.5 13,816 81.7 85+ 56,025 448.7 15,669 125.50 Gender Female 167,082 241.1 36,218 52.3 Male 147,082 285.4 31,735 61.6 Race African American 18,280 281.0 4683 72.0 White 268,2840 266.5 57,178,0 56.8 Other/unknown 28,939 211.8 6092 44.6 Medicaid* Eligible 28,538 522.1 8081 147.80 Not eligible 286,9650 320.7 59,872,0 66.9 Note: All comparisons are statistically significant at the 0.01 level based on analysis of variance. ACSC = ambulatory care sensitive conditions; IPA = independent practice association. *Includes those eligible for Medicaid for at least 1 month during 2001. ACSCs, large differences between the minimum and maximum rates illustrate the wide variation in ACSC hospitalization rates among plans. As shown in Table 5, correlations among the more common ACSC hospitalization rates and indices ranged from 0.53 (pneumonia and chronic ACSCs) to 0.99 (CHF and chronic ACSCs). The correlation between the acute and chronic ACSC indices was 0.82, while the correlation among the single most common conditions within the acute and chronic indices (bacterial pneumonia and CHF, respectively) was 0.56. Discussion This study of beneficiaries in Medicare HMOs revealed large differences in patient sociodemographic characteristics across 15 plans in California. We also found that ACSC hospitalization rates varied substantially by patient characteristic and by plan. This wide variation in rates of ACSC hospitalization across plans is likely due, at least in part, to the differences in enrollee characteristics. Our results also indicate that estimating statistically reliable rates for the less frequently occurring 15 individual ACSCs may not be possible for plans with fewer enrollees (6 of 15 plans had less than 10,000 enrollees; data not shown). For these plans, estimating the rate for the 15 ACSCs combined or the acute and chronic ACSC indices may be necessary. Correlations between the individual ACSC rates and their corresponding indices also indicate that the indices are dominated by a few ACSCs (as was also demonstrated by the rates provided in Table 4). In addition, the high correlation between the acute and chronic indices (0.82) suggests a consistency in style of care across plans. The data also show that except for the very largest HMOs (ie, over 200,000 enrollees), there are few hospitalizations for the 2 preventable conditions (influenza and malnutrition). Seven and 5 of the 15 HMOs had zero hospitalizations during the study period for influenza and malnutrition, respectively. The median number of hospitalizations for these 2 conditions for all plans was 1. This suggests that for these 2 conditions, monitoring access and quality of care in plans may not be feasible. The pattern of ACSC hospitalization rates by patient characteristics in our results is consistent with previous studies [4,15 19]. In a study of Medicare FFS beneficiaries, Blustein et (continued on page 566) 562 JCOM November 2005 Vol. 12, No. 11 www.turner-white.com
(continued from page 562) Table 4. Distribution of Number and Unadjusted Rate of ACSC Hospitalizations for 15 Medicare HMOs, California 2001 No. of Plans with Zero No. of ACSC Hospitalizations by Plan Rate per 1000 Person-years by Plan ACSC Hospitalizations Median Range Mean SD Median Range All 15 ACSCs 0 10900 82 28,534 59.930 14.320 56.890 38.11 94.12 Acute ACSC 0 503 43 14,917 30.880 6.05 30.940 19.70 41.86 Bacterial pneumonia 0 258 20 7529 15.910 3.11 16.360 10.32 21.33 Cellulitis 0 041 1 1139 2.23 0.94 2.41 0.52 3.99 Dehydration 0 062 3 1906 3.86 1.18 4.13 1.55 5.98 Gastric and duodenal ulcer 0 041 4 1397 3.35 1.60 2.96 2.20 8.77 Hypoglycemia 7 001 0 68 0.06 0.08 0.03 0 0.21 Hypokalemia 1 004 0 135 0.25 0.15 0.27 0 0.59 Severe ENT infections 6 001 0 59 0.07 0.06 0.09 0 0.16 UTI 0 095 2 2684 5.15 2.12 5.45 1.03 9.54 Chronic ACSC 0 586 38 13,545 28.920 8.93 26.980 18.21 51.97 Asthma/COPD 0 167 13 3862 8.55 1.78 8.09 6.22 11.40 CHF 0 330 19 7451 15.950 5.42 15.890 9.53 30.62 Diabetes mellitus 1 038 0 1016 2.09 1.20 1.99 0 5.41 Hypertension 2 033 0 708 1.42 0.87 1.40 0 3.13 Seizure disorder 1 019 0 508 0.90 0.39 0.96 0 1.42 Preventable ACSC 3 002 0 72 0.14 0.14 0.10 0 0.52 Influenza 7 001 0 15 0.07 0.14 0.01 0 0.52 Malnutrition 5 001 0 57 0.07 0.07 0.05 0 0.21 ACSC= ambulatory care sensitive condition; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; ENT = ear, nose, and throat; UTI = urinary tract infection. Table 5. Correlation Matrix of Unadjusted ACSC Hospitalization Rates for 15 Medicare HMOs, California 2001 Acute Chronic Asthma/ Bacterial All 15 ACSCs ACSCs ACSCs COPD CHF Pneumonia UTI All 15 ACSCs 1.00 0.93 0.97 0.86 0.97 0.70 0.88 Acute ACSCs* 1.00 0.82 0.73 0.83 0.89 0.88 Chronic ACSCs* 1.00 0.88 0.99 0.53 0.82 Asthma/COPD 1.00 0.83 0.54 0.66 CHF 1.00 0.56 0.83 Bacterial pneumonia 1.00 0.66 UTI 1.00 ACSC = ambulatory care sensitive condition; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; UTI = urinary tract infection. *Acute ACSCs include bacterial pneumonia, cellulitis, dehydration, gastric and duodenal ulcer, hypoglycemia, hypokalemia, severe ear/ nose/throat infections and urinary tract infections. Chronic ACSCs include asthma/chronic obstructive pulmonary disease, congestive heart failure, diabetes mellitus, hypertension, and seizure disorder. al [16] demonstrated that sequentially adjusting ACSC hospitalization rates for several demographic and utilization variables can have a large effect on differences between rates for different socioeconomic subgroups (ie, race, education, and income). Based on a model using survey and Medicare claims data, Culler et al [17] identified age over 75; being black; in fair or poor health; having coronary heart disease, myocardial infarction, or diabetes; being limited in 2 or more activities of daily living; and living in a core standard metropolitan statistical area or rural county as associated with increased ACSC hospitalization. Using a similar model, Laditka [18] found ACSC hospitalization rates were higher among elderly 566 JCOM November 2005 Vol. 12, No. 11 www.turner-white.com
ORIGINAL RESEARCH African American and Hispanic women than among non- Hispanic white women. Arecently published study [19] based on data collected as part of the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality indicates that rates of hospitalization for ACSCs have been changing over time and vary by geographic region, age, gender, income, and urban versus rural residence. Previous research has also shown large risk selection differences between the traditional Medicare FFS population and the Medicare HMO population [6 13]. Our study showed large variation in patient characteristics among the individual plans in California, indicating that risk differences exist not only between FFS and HMOs but also between individual plans in the HMO population. ACSC hospitalization rates are an attractive option for screening for potential access and quality of care problems in HMOs for several reasons. First, they can be generated with relative ease using readily available inpatient encounter data. Second, they are simple measures that are easy to understand and easy to explain. Third, hospitalization rates in general are familiar to clinicians and administrators and therefore will be more likely to be accepted and used. Finally, due to the cost and inherent risks associated with hospitalization, minimizing potentially avoidable hospitalizations is an important public health goal. Comparisons of unadjusted rates should not be used as a means to infer access or quality of care differences among plans. Instead, the monitoring of rates for an individual plan over time is the preferred method. For example, in the first year of estimation, a plan might compare their plan s ACSC hospitalization rates with published estimates. We found, similar to the results of others [1,17,19], that bacterial pneumonia and CHF tend to yield the highest hospitalization rates when compared with the other selected ACSCs. A change in a plan s ranking from 1 year to the next might indicate that a particular subgroup of enrollees needs further assessment (eg, patients with diabetes). Changes in rates of hospitalization for a single ACSC requires further investigation by plan administrators and providers to understand why the change occurred. It might reflect the effectiveness of a quality improvement program, the removal of certain services, or a change in enrollee sociodemographics. Several limitations of the study are worth mentioning. First, our results may not represent the experience of HMOs in other states. The sample is derived from 15 HMOs in California with large numbers of Medicare enrollees. These HMOs may have more experience in the delivery of coordinated, integrated ambulatory care and provision of preventive care than HMOs in states with lower Medicare HMO penetration. Second, the ACSC hospitalization rates have not been adjusted for demographic characteristics (ie, age, gender, or race), or socioeconomic characteristics (eg, education, income), all of which influence ACSC hospitalization rates. Finally, our 2001 data represents a period of time when Medicare managed care enrollment was starting to experience plan withdrawals and disenrollment rather than expansion in California [27] and the rest of the country [28]. Although this trend has continued since 2001 [28], our results may not be generalizable to Medicare HMOs today for this and other reasons. Further research related to adjusting ACSC hospitalization rates is needed to better understand how much of the difference in unadjusted rates among plans is due to patient characteristics and how much reflects an underlying difference in access and quality of care. In addition, studying how changes in care or enrollee characteristics affect ACSC hospitalization rates within the same plan over time will improve our understanding of how ACSC hospitalization rates relate to access and quality of care. Acknowledgments: The authors are grateful to the Agency for Healthcare Research and Quality; Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services; and The Robert Wood Johnson Foundation for support of this work. They thank Robert Reddick for the computer programming of the original data files and Beate Danielson of Health Information Solutions for linking the data files. They are also grateful to the Centers for Medicare and Medicaid Services and the California Office of Statewide Health Planning and Development for providing the data that made this study possible. Corresponding author: Feng Zeng, PhD, School of Policy, Planning and Development, University of Southern California, Los Angeles, CA 90089. Funding/support: Grant #R01HS10256 from the Agency for Healthcare Research and Quality; Office of the Assistant Secretary for Planning and Evaluation, U.S. Dept. of Health and Human Services; and grant #41289 from the Robert Wood Johnson Foundation, Princeton, NJ. Financial disclosures: None. Author contributions: conception and design, FZ, JFO, EMS, ND, GM; analysis and interpretation of data, FZ, JFO, EMS; drafting of the article, FZ, JFO, EMS; statistical expertise, FZ; obtaining of funding, ND, GM; administrative support, JFO. References 1. McCall N, Harlow J, Dayhoff D. Rates of hospitalization for ambulatory care sensitive conditions in the Medicare+Choice population. Health Care Financ Rev 2001;22:127 45. 2. Millman M, editor. Access to health care in America. Committee on Monitoring Access to Personal Health Care Services, Institute of Medicine. Washington (DC): National Academy Press; 1993. 3. Billings J, Zeitel L, Lukomnik J, et al. Impact of socioeconomic www.turner-white.com Vol. 12, No. 11 November 2005 JCOM 567
status on hospital use in New York City. Health Aff (Millwood) 1993;12:162 73. 4. Bindman AB, Grumbach K, Osmond D, et al. Preventable hospitalizations and access to health care. JAMA 1995;274:305 11. 5. Garg A, Probst JC, Sease T, Samuels ME. Potentially preventable care: ambulatory care-sensitive pediatric hospitalizations in South Carolina in 1998. South Med J 2003;96:850 8. 6. Riley G, Lubitz J, Rabey E. Enrollee health status under Medicare risk contracts: an analysis of mortality rates. Health Serv Res 1991;26:137 63. 7. Miller RH, Luft HS. Managed care plan performance since 1980. A literature analysis. JAMA 1994;271:1512 9. 8. Riley G, Tudor C, Chiang YP, Ingber M. Health status of Medicare enrollees in HMOs and fee-for-service in 1994. Health Care Financ Rev 1996;17:65 76. 9. Cox DF, Hogan C. Biased selection and Medicare HMOs: analysis of the 1989 1994 experience. Med Care Res Rev 1997;54:259 85. 10. Morgan RO, Virnig BA, DeVito CA, Persily NA. The Medicare- HMO revolving door the healthy go in and the sick go out. N Engl J Med 1997;337:169 75. 11. Hamilton BH. HMO selection and Medicare costs: Bayesian MCMC estimation of a robust panel data tobit model with survival. Health Econ 1999;8:403 14. 12. Call KT, Dowd B, Feldman R, Maciejewski M. Selection experiences in Medicare HMOs: pre-enrollment expenditures. Health Care Financ Rev 1999;20:197 209. 13. Dhanani N, O Leary JF, Keeler E, et al. The effect of HMOs on the inpatient utilization of medicare beneficiaries. Health Serv Res 2004;39:1607 27. 14. Revere L, Large J, Langland-Orban B. A comparison of inpatient severity, average length of stay, and cost for traditional fee-for-service medicare and medicare HMOs in Florida. Health Care Manage Rev 2004;29:320 8. 15. Billings J, Anderson GM, Newman LS. Recent findings on preventable hospitalizations. Health Aff (Millwood) 1996;15: 239 49. 16. Blustein J, Hanson K, Shea S. Preventable hospitalizations and socioeconomic status. Health Aff (Millwood) 1998;17:177 89. 17. Culler SD, Parchman ML, Przybylski M. Factors related to potentially preventable hospitalizations among the elderly. Med Care 1998;36:804 17. 18. Laditka JN. Hazards of hospitalization for ambulatory care sensitive conditions among older women: evidence of greater risks for African Americans and Hispanics. Med Care Res Rev 2003;60:468 508. 19. Kruzikas DT, Jiang HJ, Remus D, et al. Preventable hospitalizations: window into primary and preventive care, 2000: HCUP Fact Book No. 5. Rockville (MD): Department of Health and Human Services, Agency for Healthcare Research and Quality; 2004. AHRQ Publication No. 04 0056. 20. Kaiser Family Foundation. State health facts online. Undated. Menlo Park, California. Henry J. Kaiser Family Foundation. Available at www.statehealthfacts.kff.org. Accessed 13 Sep 2004. 21. Fellegi IP. Section II: overview of applications and introduction to theory. Tutorial on the Fellegi-Sunter model for record linkage. In: Kliss B, Alvey W, editors. Record linkage techniques. 1985: Proceedings of the Workshop on Exact Matching Methodologies; 1985 May 9 10; Arlington, VA. Washington (DC): Dept. of the Treasury, Internal Revenue Service; 1985: 127 38. 22. Fellegi IP, Sunter A. A theory for record linkage. JASA 1969; 64:1183 210. 23. Newcombe HB. Handbook of record linkage: methods for health and statistical studies, administration, and business. New York: Oxford University Press; 1988. 24. Jaro MA. Advances in record-linkage methodology as applied to matching the 1985 Census of Tampa, Florida. JASA 1989; 84:414 20. 25. Centers for Medicare and Medicaid Services. Monthly report Medicare coordinated care health plans. Available at http://cms.hhs.gov/healthplans/statistics/monthly/. Accessed 26 Oct 2005. 26. SAS Institute. SAS procedure guide, version 8.0. Cary (NC): SAS Institute; 1999. 27. Gold M, Lake T. Medicare+Choice in California: lessons and insights. Kaiser Family Foundation. Available at www.kff.org/ medicare/6047-index.cfm. Accessed 26 Oct 2005. 28. Kaiser Family Foundation. Medicare advantage fact sheet. Available at www.kff.org/medicare/2052-07.cfm. Accessed 26 Oct 2005. Copyright 2005 by Turner White Communications Inc., Wayne, PA. All rights reserved. 568 JCOM November 2005 Vol. 12, No. 11 www.turner-white.com