Contribution of Medicaid Managed Care to the Increasing Undercount of Medicaid Beneficiaries in the Current Population Survey Arpita Chattopadhyay, Ph.D. & Andrew B. Bindman, MD Primary Care Research Center & the Division of General Internal Medicine University of California, San Francisco Background • CPS is an important source of data for comparing • • • • beneficiaries across insurance groups CPS routinely underestimates the size of Medicaid beneficiaries. This affects Medicaid counts, AND estimates of one or both uninsured/privately insured groups. The extent of Medicaid undercount has been increasing over time At the same time managed care has increased in Medicaid 2 Study Objective • To determine if Medicaid managed care is associated with the undercount of Medicaid beneficiaries in the CPS. Study Population • • Medicaid beneficiaries in California < 65 years Approximately 16% of Medicaid beneficiaries in the country Study Period • 1995-1999: A period of managed care expansion among Medicaid beneficiaries in California 3 Data • Current Population Survey 1996-2000 – Have you had Medicaid in the past year? – For how many months did you have Medicaid last year? • Medi-Cal Monthly Eligibility Files 19951999 – Monthly enrollment data with information on county and health plan 4 Calculating the Size of the Medicaid Population CPS estimate = (S wi * mcaidi*monthsi)/12 wi = individual weight; mcaidi=1 if the ith respondent ever had Medicaid the previous year; 0 otherwise; monthsi= No. of months the ith respondent had Medicaid the previous year MMEF count = S No. of beneficiary each month/12 5 Method • Person-Year estimates of Medicaid beneficiaries for the years 1995 to 1999 derived from CPS • Compared to annual counts (person-years) of Medicaid beneficiaries from administrative records for the same years • County level estimates from CPS derived for 24 of the 58 California counties and a “super county” consisting of the 34 remaining counties 6 Variable Definition Dependent Variable: Percentage of CPS undercount Yit = 100*(MMEF countit-CPS estimateit)/MMEF countit i=1….25; t=1…5; Where ‘i’ represents a county/super-county & ‘t’ represents year Explanatory variables: Managed care penetration MCPit = (MMEF managed care countit/ MMEFit)*100 Year 1995-1999 7 Model Yit = a0+a1MCPit+a2YEAR +eit; i=1….25; t=1…5; Where a0 is the intercept term; a1 and a2 are coefficients associated with managed care penetration and year; eit is the error term with a variance component correlation structure 8 Medicaid Population Aged < 65 Years Estimated from Current Population Survey (CPS) and Medicaid Monthly Eligibility File (MMEF) Year % undercount % Managed in CPS Care 1995 28.98 20.24 1996 32.07 25.53 1997 35.00 37.88 1998 32.17 48.76 1999 33.22 56.23 Source: Current Population Survey 1996-2000 & Medical Care Statistics Section, Department of Health Services, California 9 Unadjusted Association Between Percent Underestimate in Medicaid Beneficiaries <65 Years from the Current Population Survey (CPS) and Medicaid Monthly Eligibility Files (MMEF) 1995-1999 Percent Underestimate (CPS estimate) 150 100 50 0 -50 -100 -150 Weighted R = 0.23 (p =0.01) -200 0 20 40 60 80 100 Percent in Managed Care (MMEF count) Source: Current Population Survey (CPS) 1996-2000 and California Department of Health Services Medicaid Monthly Eligibility File (MMEF)1995-1999. Percent underestimate corresponds to CPS estimate as compared to MMEF counts as gold-standard 10 Estimated Coefficients for Current Population Survey (CPS) Underestimate of Medicaid Population <65 Years Variable Coefficient1 SE2 P-Value Year -2.71 2.43 0.267 Medi-Cal managed care penetration Constant 0.40 0.14 0.006 23.92 6.78 <0.001 Source: Current Population Survey 1996-2000 & Medical Care Statistics Section, Department of Health Services, California Note: 1 Regressions were weighted by CPS sample size. 2 Robust standard errors 11 Implication • National data show that during 1995-1997 – Medicaid managed care penetration increased by 18.4% – Underestimate of Medicaid beneficiaries increased by 9% • Applying our results to national data implies that – 0.4*18.4= 7.4% increase in the underestimate of Medicaid beneficiaries in the CPS can be attributed to increases in managed care penetration • Thus Medicaid managed care explains more than 80% ( 7.4% of 9.0%) of the increasing underestimates of Medicaid beneficiaries in the CPS 12 Conclusion • The CPS underestimated the Medicaid population by approximately a third during the study period. • At the county level, each percentage point increase in the penetration of managed care was associated with an underestimate in the CPS of 0.4 percentage points. • Medicaid managed care explains more than 80% of the increasing undercount of Medicaid beneficiaries in CPS 13 Relevance for Survey Design • Include question on health plan characteristics to improve estimates of Medicaid population in CPS 14 Limitations • Single state study • Unable to exclude institutional population from administrative records • County level estimates from CPS are not always stable • Finally, confounding by some other county characteristic remains a possibility 15