Is Managed Care Superior to Traditional Fee-For-Service among HIV-Infected Beneficiaries of Medicaid? David Zingmond, MD, PhD UCLA Division of General Internal Medicine and Health Services Research June 8, 2004 1 Background (1) Medicaid is the largest payer of healthcare for HIV/AIDS – Annual budget > $4.1B for HIV/AIDS High costs of treating HIV/AIDS (and other diseases) have led to the adoption of managed care (HMO) in place of traditional fee-for-service (FFS) by Medicaid Concerns that HMO enrollment might worsen care & outcomes of HIV/AIDS patients 2 Background (2) In California, Medicaid HMOs and enrollment policy are implemented on a county-by-county basis Depending upon the county, Medicaid managed care is mandatory, voluntary, or not offered. 3 Hypotheses HMO enrollment is associated with lower hospitalization rates. Medi-Cal HMO enrollment is associated with lower antiretroviral medication usage. HMO enrollment reduces survival. 4 Conceptual Model DISEASE STAGE COMORBID DISEASE DEMOGRAPHICS MEDI-CAL HMO ENROLLMENT • ANTIRETROVIRAL THERAPY • HOSPITALIZATION • DISEASE PROGRESSION • MORTALITY COUNTY POLICY FOR MEDI-CAL HMO ENROLLMENT OF HIV/AIDS PATIENTS 5 Methods: Data Sources Data Source Medi-Cal Eligibility File Data Measures Demographics & Enrollment Medi-Cal Claims Antiretroviral Medication Usage OSHPD Discharge File Hospitalizations, SCAH Death Stat’l Master File Time to Death AIDS Registry & HIV Reporting System Exposure Risk, CD4, Time since AIDS diagnosis SCAH - Severity Classification of AIDS Hospitalizations 6 Methods: Cohort Definition Identified all adult HIV/AIDS patients enrolled in Medi-Cal in January 1999 (in counties with mandatory or optional HMO enrollment) who were continuously enrolled until 12/2001 or death. In sensitivity analyses, we relaxed restrictions regarding county of residence and of continuous enrollment. 7 Methods: Dependent Variables Mortality by follow-up Disease progression by follow-up Hospitalization (or death) by follow-up Use of HAART (at study baseline) HAART - Highly Active Antiretroviral Therapy 8 Methods: Independent Variables Baseline HMO enrollment (& home county) Covariates: Demographics - Age, gender, & race Comorbidity - non-HIV hospitalizations Disease severity - HIV hospitalizations, CD4*, & SCAH* Health Habits - Exposure risk category* Treatment - Baseline HAART or ARV * Only AIDS patient analyses 9 Methods: Regression Analyses (1) Bivariate comparison of dependent and independent variables by HMO enrollment We employed standard multivariate probit regression model predicting: Dependent Variable = Function (HMO Enrollment, Demographics, Disease Severity, Comorbidity, Treatment) However, this approach may result in biased estimates if unmeasured severity is correlated with enrollment and outcomes. 10 Methods: Regression Analyses (2) Solution - Treatment Selection Model (bivariate probit): HMO Enrollment = Function (County Plan Type, Demographics, Disease Severity/Stage, Comorbidity, Treatment) + Dependent Variable = Function (HMO Enrollment, Demographics, Disease Severity/Stage, Comorbidity, Treatment) + The error terms of the two equations, and , are modeled as being correlated. 11 Results 12 Results: Demographics N Male (%) Race (%) White Black Latino Age (%) 20-29 30-49 50+ AIDS (%) ** P < 0.01, * P< 0.05 HMO 2,838 51 37 30 23 13 69 16 45 FFS 15,357 72 ** ** 42 30 17 ** 6 68 26 52 13 Results: Unadjusted Outcomes by Disease Stage - HMO vs FFS N Baseline Treatment (%) Any ARV HAART Death (%) Hospitalization (%) ** P < 0.01, * P< 0.05 AIDS HMO FFS 1,299 7,922 64 36 16 60 69 46 18 57 HIV+, No AIDS HMO FFS 1,539 7,435 ** ** 21 10 6 52 25 13 9 56 ** ** ** ** 14 Results: Impact of HMO Enrollment on AIDS Patients Probit RR 95% CI Bivariate probit RR 95% CI P* 1.01 0.88 1.14 1.07 0.88 1.28 0.49 Death Death or hospital’n 1.04 0.99 1.09 0.98 0.90 1.07 0.05 HAART at baseline 0.80 0.74 0.86 0.90 0.80 1.01 0.01 Covariates: age, race, gender, baseline HAART, baseline other ARV, prior hiv- hospitalization, prior non-hiv- hospitalization, lowest CD4, exposure category, SCAH. P* - Chi-square test of rho coefficient different from 0 RR - Relative Risk with 95% CI calculated by bootstrapping with 1000 repetitions. 15 Results: Impact of HMO Enrollment on HIV+ Patients Probit RR Death Disease Progression Death or hospital’n HAART at baseline* 95% CI Bivariate probit RR 95% CI P* 1.01 0.89 1.12 1.21 0.78 1.75 0.17 1.13 0.98 1.31 1.22 0.95 1.54 0.49 0.98 0.93 1.03 0.99 0.89 1.08 0.92 0.79 0.65 0.94 0.87 0.64 1.19 0.51 Covariates: age, race, gender, baseline HAART, baseline other ARV, prior hiv- hospitalization, prior non-hiv-hospitalization P* - Chi-square test of rho coefficient different from 0 RR - Relative Risk with 95% CI calculated by bootstrapping with 1000 repetitions. 16 Discussion (1) HMO enrollment in California appears to have negligible impact on hospitalization and death. Despite concerns that HMOs might provide less necessary medications for AIDS patients, analysis results show no difference. Important treatment guarantees may mediate the effects of plan type on outcomes – Include guaranteed access to medications and specialist providers 17 Discussion (2) Differences in treatment appear to exist among the HIV+, non-AIDS patients. – Treatment criteria are less stringent for non-AIDS patients. – Disease severity is more varied but less well measured as that for AIDS patients. Overall, the bivariate probit approach gives greater confidence to standard regression results. 18 Limitations Single state Limited follow-up No ambulatory care data HIV+ without AIDS patients had fewer casemix measures HMO implementation is heterogeneous and distributed geographically 19 Conclusions and Policy Implications Medicaid HMOs for patients with HIV/AIDS have similar outcomes as standard FFS Medicaid. – Expansion of Medicaid HMOs may be justified if cost beneficial Similar approaches may be used to examine benefits of managed care models for other medically needy Medicaid populations. 20