Trends in Cumulative Disenrollment in the Medicare Advantage Program, 2011-2020

Abstract

Importance: The Medicare Advantage (MA) program is rapidly growing. While previous work has found that beneficiaries with substantial health needs disenroll from plans at higher rates, the long-term frequency of disenrollment is not well understood.

Objective: To compare cumulative disenrollment trends in the MA program by beneficiary and plan characteristics.

Design, setting, and participants: This retrospective, serial cross-sectional study included beneficiaries with any MA enrollment from January 1, 2011, to December 31, 2020. Data analysis took place from September 2022 to March 2023.

Exposures: Beneficiary characteristics, including race and ethnicity, length of Medicare enrollment, dual eligibility, and comorbidity burden, and contract characteristics, including vertical integration status, premium, and MA star rating.

Main outcomes and measures: The main outcome was disenrollment from an MA contract within 5 years. Rates of cumulative disenrollment by beneficiary and contract characteristics were compared. Pearson correlation coefficients were calculated to assess the correlation between a contract's 1-year disenrollment and the contract's disenrollment over a longer period.

Results: The sample included 82 377 917 beneficiaries (524 442 225 beneficiary-year observations; 56.7% female; mean [SD] age, 71.9 [10.3] years). After 1 year, 13.2% of nondually enrolled and 15.9% of dually enrolled beneficiaries had left their contract, increasing to 48.3% and 53.4%, respectively, after 5 years. Black enrollees disenrolled at the highest rates among race and ethnicity categories, with 14.8% disenrolling after 1 year and 52.6% disenrolling after 5 years. Contracts had a median disenrollment rate of 9.8% (IQR, 4.5%-19.0%) after 1 year and 56.1% (IQR, 23.1%-79.0%) after 5 years. Contracts rated 5 stars had substantially lower 5-year disenrollment rates (23.0% after 5 years compared with 41.2% for 4- to 4.5-star contracts and 67.2% for 3- to 3.5-star contracts). Disenrollment from a contract after 1 year was not well correlated with disenrollment after 5 years (r, 0.46).

Conclusions and relevance: This cross-sectional study found substantial cumulative rates of disenrollment from MA plans within 5 years between 2011 and 2020, with wide variation in 5-year disenrollment by contract. The findings suggest that evaluating long-term disenrollment rates in MA performance measures may capture different outcomes than single-year disenrollment alone.

Figure 1.. Percentage of Beneficiaries Who Disenrolled or Switched to Traditional Medicare Over Time by Dual Eligibility With Medicaid

Disenrollment was defined as a beneficiary voluntarily leaving their contract for either traditional Medicare or another Medicare Advantage contract. Switching was defined as leaving a Medicare Advantage contract and enrolling in traditional Medicare. Error bars represent 95% CIs calculated from bootstrapping.

Figure 2.. Percentage of Beneficiaries Who Disenrolled From Their Contract by Beneficiary Characteristics

Disenrollment was defined as a beneficiary voluntarily leaving their contract for either traditional Medicare or another Medicare Advantage contract. A, Race and ethnicity were included from the Medicare Master Beneficiary Summary File. B, The comorbidity risk score was calculated using Johns Hopkins pharmaceutical-based ambulatory care groups for beneficiaries with Medicare Part D coverage. Error bars represent 95% CIs calculated from bootstrapping. ^aIncludes beneficiaries for whom the Centers for Medicare & Medicaid Services did not have race and ethnicity information and beneficiaries who may identify as having more than 1 race and ethnicity.

Figure 3.. Contract-Level Disenrollment Rates by Overall Medicare Advantage (MA) Star Rating

Disenrollment was defined as a beneficiary voluntarily leaving their contract for either traditional Medicare or another MA contract and then aggregated to the contract level. Lines are stratified by publicly reported overall MA star rating. Error bars represent 95% CIs calculated via bootstrapping.