Risk adjusted performance measures may not be an accurate measure of health plan performance

There may be substantial residual confounding in risk-adjustment models used to evaluate health plan performance due to differences in patient characteristics between plans. This means that they may not be able to accurately or fairly identify differences between plans and should caution policymakers against assuming that risk adjustment is sufficient to isolate real differences in plan performance. These findings are published in Annals of Internal Medicine.

Nearly 70% of the Medicaid-eligible population is enrolled in a Medicaid managed care plan. Managed care plans are private healthcare plans that receive prospective per-enrollee per-month capitation payments from states and are then responsible for managing and paying for enrollees’ health care. Capitation payments to plans are “risk-adjusted”, meaning that they differ to reflect differences in health care needs across patient populations. However, our results suggest that inadequate adjustment for patient risk penalizes plans (and providers) with unobservably higher-risk patients, incentivizes plans and providers to engage in risk-selection strategies that are wasteful and can undermine quality of care, and leads public-reporting initiatives to potentially misinform patients.

Researchers from Yale School of Public Health analyzed Louisiana Medicaid data to assess the degree to which risk-adjusted measures of health plan performance reflect differences in performance across plans versus differences in patient characteristics (residual confounding). The authors examined data from 2013 and 2014, the period in which Louisiana Medicaid transitioned to Medicaid managed care. The analyses focused on 137,933 eligible residents in the first region to transition to Medicaid managed care. Of those, 94,972 did not select a plan and were randomly assigned to one of 5 plans, creating a natural experiment. The remaining 42,961 chose among the same 5 plans. The authors compared each of the 5 plans’ risk-adjusted performance between the patients who selected a plan and “gold standard” estimates of plan performance based on patients who were randomly assigned. The authors found that risk-adjusted measures of plan performance based on enrollees that chose plans differed substantially from estimates based on randomly assigned enrollees, with residual confounding only modestly reduced by risk adjustment . The authors suggest that the results should serve as a warning to policymakers who assume current risk adjustment is sufficient to measure the performance of plans (or providers) and the study discusses several implications of the findings for how payers and providers assess performance and deploy risk-adjustment in public insurance programs.