Key Points
Question
How does Medicare Advantage (MA) disenrollment change following the development of new complex medical conditions, and does this vary by medical complexity, Medigap protections, and plan type?
Findings
In this cohort study of more than 1 million MA beneficiaries, development of new complex conditions was associated with increased disenrollment, particularly to traditional Medicare. Disenrollment increased with the number of new conditions and was higher in states with Medigap guaranteed issue and community rating protections.
Meaning
In this study, as beneficiaries’ medical complexity increased, they were more likely to leave MA, particularly where supplemental coverage protections made switching to traditional Medicare more feasible.
Abstract
Importance
Prior research has demonstrated higher rates of disenrollment from Medicare Advantage (MA) among people with greater health care needs. Yet, less is known about how disenrollment responds to changes in medical complexity, whether disenrollment varies by the degree of complexity, and how these patterns differ by state Medigap protections and MA plan characteristics.
Objective
To examine changes in MA plan disenrollment following the development of new complex medical conditions and assess heterogeneity by number of conditions, state Medigap policy, and plan characteristics.
Design, Setting, and Participants
This cohort study used Medicare enrollment and claims data from 2016 to 2021. The study population included beneficiaries continuously enrolled in MA in 2016 without complex medical conditions from 2016 through 2018. A treatment group developed a new complex condition in 2019; a comparison group did not develop a complex condition through 2021. Data were analyzed from December 1, 2025, to March 1, 2026.
Exposure
Development of 1 or more new complex medical conditions in 2019, including acute myocardial infarction, Alzheimer disease and related dementia, atrial fibrillation, chronic kidney disease, chronic obstructive pulmonary disease, depression, congestive heart failure, or stroke.
Main Outcomes and Measures
Annual disenrollment from the prior year MA plan, which included disenrollment to traditional Medicare (TM) or switching to another plan within MA. Difference-in-differences models with beneficiary and year fixed effects were estimated and converted to differences in percentage points.
Results
The analytic sample included 1 054 926 beneficiaries, including 219 942 (mean [SD] age, 75.3 [6.3] years; 54.6% female) who developed new complex conditions in 2019 and 834 984 (mean [SD] age, 73.8 [5.6] years; 56.7% female) who did not develop a new complex condition in 2019. Developing any new complex condition was associated with an increase of 3.3 (95% CI, 3.1-3.5) percentage points (pp) in MA disenrollment, including an increase of 2.8 (95% CI, 2.7-2.9) pp in disenrollment to TM. The number of new complex conditions was associated with increased disenrollment, from 1.4 (95% CI, 1.3-1.6) pp for 1 condition to 12.8 (95% CI, 11.6-13.9) pp for 4 or more conditions. Disenrollment to TM was 1.5 (95% CI, 0.9-2.1) pp higher among beneficiaries residing in states with Medigap guaranteed issue and community rating protections.
Conclusions and Relevance
In this study, development of new complex medical conditions was associated with higher MA disenrollment, particularly to TM and in states with Medigap protections, and the magnitude increased with medical complexity. These findings highlight how changes in health status and state policy may shape Medicare coverage decisions.
This cohort study examines changes in Medicare Advantage plan disenrollment following the development of new complex medical conditions and assesses heterogeneity by number of conditions, state Medigap policy, and plan characteristics.
Introduction
Medicare Advantage (MA), the privately managed component of Medicare, accounts for a greater share of total Medicare enrollment each year, comprising 54% of total Medicare enrollment in 2025.1 MA offers several benefits relative to traditional Medicare (TM). These benefits include out-of-pocket maximums, which TM does not guarantee, integrated prescription drug coverage, and supplementary benefits that can include dental and vision coverage, gym memberships, and even pet food.2,3 However, MA plans impose use restrictions, including prior authorization and limited networks, which can introduce administrative friction or limit access to needed care.
Beneficiaries with higher health care needs are more likely to encounter these administrative problems and have been found to disenroll from their MA plans at an elevated rate.4,5,6,7,8,9 Further complicating disenrollment decisions is that most states do not have community rating and guaranteed issue Medigap, the supplemental private insurance that covers out-of-pocket expenditures not covered by Medicare,10 such that enrollees leaving MA may be unable to afford or may be denied issue of supplemental insurance. In theory, beneficiaries may understand these trade-offs when they first enroll in an MA plan. However, if their health needs change, their preferences are likely to change as well.
In this context, recent studies11,12 have focused more specifically on enrollees’ disenrollment decisions following the development of a new condition that they did not have when they first enrolled in MA. One study12 focused on new cancer diagnoses and the other study11on a broader definition of serious illness. The authors focused on understanding the extent to which Medigap protections affected disenrollment to TM. However, there remain several gaps, which we aim to address. First, these recent studies11,12 have focused on a binary characterization of new conditions. There may be substantial heterogeneity in the relative need and complexity of individuals with these conditions, which is likely to influence individuals’ preferences and the frequency with which they encounter administrative challenges within their MA plans. Second, prior research11,12 has primarily focused on only disenrollment to TM and not plan switching within MA, which accounts for most plan disenrollment. In particular, switching within MA may be most pronounced for individuals with lower levels of medical complexity, whereas disenrollment to TM may be most common among those with the highest levels of complexity. Finally, this disenrollment and switching is likely to vary not only by the Medigap policies where the beneficiary lives and the characteristics of their MA plan but also by the level of medical complexity itself. As shown in prior studies,11,12 disenrollment to TM is likely to be highest among beneficiaries with new conditions when those beneficiaries live in states with Medigap community rating and guaranteed issue policies, while beneficiaries in states without those protections may instead switch within MA to plans with fewer administrative challenges. Whether Medigap protections matter more for beneficiaries as a function of medical complexity remains an open question. Characterizing these interactions is important because they identify which beneficiaries are most affected by state Medigap policy and under what conditions, informing the case for and against expanding such protections.
In this cohort study, we used a difference-in-differences design comparing disenrollment decisions over time in MA enrollees who developed a new complex medical condition and in MA enrollees who did not. In all analyses, we explored heterogeneity across the number of new complex conditions, a proxy for relative levels of medical complexity. We further examined heterogeneity in disenrollment to TM and within MA across varying state Medigap protection policies and MA plan characteristics, considering relative levels of medical complexity. We contribute to the literature by (1) quantifying how disenrollment varies with the number of new complex conditions, (2) separating disenrollment to traditional Medicare from switching within Medicare Advantage, and (3) testing how these responses vary by Medigap policy and plan type.
Methods
Study Data and Population
We built a cohort comparing beneficiaries who developed a new complex condition in 2019 to beneficiaries who never developed a complex condition. This study was approved by the Johns Hopkins Institutional Review Board with a waiver of informed consent because the study used enrollment and claims data. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Data were analyzed from December 1, 2025, to March 1, 2026. We used the Medicare Master Beneficiary Summary File and Medicare claims data from 2016 through 2021. We first identified a cohort of beneficiaries who were enrolled in MA for the full year in 2016 and did not have a complex medical condition for 2016 through 2018. Beneficiaries were not required to remain in MA after 2016. Beneficiaries were required to be alive for the full year from 2016 through 2018. From 2019 onward, decedents were retained in the sample in all years in which they were alive to begin the year. We did not require survival through 2021. Complex medical conditions included acute myocardial infarction, Alzheimer disease and related disorders, atrial fibrillation, chronic kidney disease, chronic obstructive pulmonary disease and asthma, depression, congestive heart failure, or stroke. We selected this list of complex conditions as they have been identified as conditions that segment high-cost Medicare beneficiaries into clinically coherent cohorts with elevated care needs, and because they have been used in prior studies6,13 examining MA disenrollment. To identify these conditions, we applied the Chronic Conditions Data Warehouse14 algorithm to each year’s claims, and for MA, we excluded diagnoses in medical record reviews. Within our study population, we then identified a treatment group that developed a new complex condition in 2019 as well as a comparison group that continued to show no evidence of a complex condition through 2021. In all years, beneficiaries were required to reside in the 50 US states or Washington, DC. Additional plan information, including the plan type (ie, health maintenance organization [HMO] or other plan type), plan average enrollee risk score, and contract star rating were obtained from Centers for Medicare & Medicaid Services public use files.
Outcome and Exposure Variables
Our primary outcome measure was whether an individual disenrolled from their prior year MA plan, which included either disenrollment to TM or switching to another plan within MA. We measured the outcome for each year from 2017 through 2021. For each year, we measured disenrollment conditional on enrollment in an MA plan in the prior year. Beneficiaries who disenrolled to TM in a prior year were therefore no longer at risk of disenrollment in subsequent years; our outcomes captured annual, rather than cumulative, disenrollment rates. We further deconstructed this outcome into its component parts of disenrollment to TM and switching within MA.
The key exposure variable was a binary variable for whether a beneficiary had a complex medical condition in a given year. For all beneficiaries, this value was equal to 0 for 2016 through 2018. From 2019 and on, the value was 1 for the treatment group and 0 for the comparison group. In addition to the binary measure, we also constructed a measure that was the total number of complex medical conditions an individual developed in 2019 (0, 1, 2, 3, or ≥4 new complex conditions).
In addition to the key exposure variable, we also assessed heterogeneity across state Medigap protection policies, plan characteristics, and beneficiary characteristics. State Medigap protection policies were dichotomized into states with guaranteed issue and community rating policies (Connecticut, Maine, Massachusetts, and New York) or states without either policy based on the beneficiary’s state. States with only community rating policies were excluded from these analyses (Arkansas, Minnesota, Vermont, Washington). Plan characteristics, whether a plan was an HMO, a low-risk enrollee plan (mean risk score below 1.0), and contract star rating (2.0-2.5, 3.0-3.5, 4.0-4.5, 5, or no star rating) were based on enrollee’s 2018 plan so that they would reflect the plan the beneficiary was enrolled in immediately prior to the onset of a new complex condition in 2019. Beneficiaries not enrolled in MA in 2018 were excluded from heterogeneity analyses by plan characteristics. Beneficiary characteristics included dual eligibility, race and ethnicity (as classified in Medicare enrollment files), sex, and zip code level median income and were based on an enrollee’s 2018 status.
Statistical Analysis
We first compared unadjusted trends from 2016 through 2021 in disenrollment from their MA plan among our treatment group of beneficiaries, who developed a new complex condition in 2019, and our comparison group with no new complex condition, stratified by the number of new complex conditions. We then estimated changes in disenrollment from MA following the development of a new complex condition using a difference-in-differences approach in which we compared changes in disenrollment before and after 2019 among beneficiaries with and without new complex conditions. All regressions included beneficiary and year fixed effects. We calculated effect estimates of the development of any new complex condition and separately calculated effect estimates of developing 1, 2, 3, or 4 or more new complex conditions in 2019. We then estimated heterogeneity in the effect estimate of the new complex condition by interacting this key independent variable with binary variables for whether the beneficiary’s state had Medigap community rating and guaranteed issue policies and whether the beneficiary was enrolled in an HMO plan in 2018. In our difference-in-differences analysis, beneficiary fixed effects accounted for baseline differences in beneficiary characteristics and disenrollment rates. However, we also evaluated whether enrollees differed in disenrollment rates prior to 2019, when new complex conditions first appeared, using an event study analysis. Further, we tested the sensitivity of our difference-in-differences estimates to adjustment for time-varying changes in beneficiaries’ count of the chronic conditions in the Chronic Conditions Warehouse’s algorithm not designated as complex conditions.14 Data reported in the main article as differences in percentage points (pp) appear in the tables and figures as unconverted ordinary least squares coefficients with SE. Statistical significance was set at 2-sided P < .05. Data were analyzed using Stata version 17.0 (StataCorp LLC).
Results
Trends in MA Disenrollment in Enrollees With New Complex Conditions
Our final analytic sample included 1 054 926 beneficiaries (6 200 678 beneficiary-year observations), including 219 942 (mean [SD] age, 75.3 [6.3] years; 54.6% female) who developed new complex conditions in 2019 and 834 984 beneficiaries (mean [SD] age, 73.8 [5.6] years; 56.7% female) who did not develop a new complex condition in 2019. There were 179 056 beneficiaries with exactly 1 new complex condition, 31 009 with 2 new complex conditions, 8867 with 3 new complex conditions, and 1010 with 4 or more new complex conditions. Sample size and unadjusted disenrollment rates by the number of new complex conditions in 2019 are reported in eTable 1 in Supplement 1. In 2018, prior to the development of new complex conditions, beneficiary characteristics were similar between the treatment and comparison groups (eTable 2 in Supplement 1).
Figure 1 shows trends in disenrollment from 2016 through 2021 across these groups. After being enrolled in MA for 12 months in 2016, approximately 20% had either disenrolled into TM or switched plans within MA in 2017 across all groups (Figure 1, A). Disenrollment rates remained similar across all groups in 2018 before diverging in 2019, the year in which the treatment group first showed evidence of a new complex condition, and diverging again in 2020 and 2021. There was a clear dose-response relationship between the number of new complex conditions and disenrollment. In 2020, 37.3% of people with 4 or more new complex conditions who had been enrolled in MA in 2019 had disenrolled from their plan compared with 30.9% among people with 3 new complex conditions, 25.6% among people with 2 new complex conditions, 21.0% among people with 1 new complex condition, and 17.6% among people with 0 new complex conditions. Disenrollment rates decreased in 2021; however, this is calculated among those who stayed enrolled in MA in 2020.
Figure 1. Line Graphs of Plan Switching Among Beneficiaries With and Without New Complex Conditions in 2019, 2017-2021.
Complex conditions include acute myocardial infarction, Alzheimer disease, atrial fibrillation, chronic kidney disease, chronic obstructive pulmonary disease, major depression, congestive heart failure, or stroke, identified using the Chronic Conditions Data Warehouse algorithm applied to Medicare claims; diagnoses from MA medical records were excluded. Enrollment outcomes were measured annually from 2017 through 2021, conditional on MA enrollment in the prior year. Data are from the Medicare Master Beneficiary Summary File and Medicare claims data from 2016 through 2021. MA indicates Medicare Advantage; TM, traditional Medicare.
Differences in disenrollment were primarily based on disenrollment to TM, especially among those with 2, 3, or 4 or more new complex conditions (Figure 1, B). While close to 0% of enrollees across all groups had disenrolled to TM in 2017 or 2018, 7.0% disenrolled to TM in 2019, 23.4% disenrolled to TM in 2020, and 15.9% disenrolled to TM in 2021 among those that developed 4 or more new complex conditions and were still enrolled in MA in the prior year. While rates of plan switching within MA remained fairly similar between all the groups (Figure 1, C), beneficiaries who developed 1 new complex condition were most likely to switch plans within MA, whereas those with 4 or more new complex conditions were less likely than those with no new complex conditions to switch within MA.
Difference-in-Differences Estimates of Disenrollment Following New Complex Conditions
Figure 2 shows the estimated changes in disenrollment, either to TM or switching plans within MA, following the development of any, 1, 2, 3, or 4 or more new complex conditions. Developing any new complex condition was associated with an increase of 3.3 (95% CI, 3.1-3.5) pp in MA plan disenrollment, including an increase of 2.8 (95% CI, 2.7-2.9) pp in disenrollment to TM and an increase of 0.5 (95% CI, 0.3-0.7) pp in switching plans within MA. The number of new complex conditions was associated with increased disenrollment, from 1.4 (95% CI, 1.3-1.6) pp in beneficiaries with 1 new complex condition to 5.1 (95% CI, 4.7-5.4) pp in beneficiaries with 2 new complex conditions, 8.7 (95% CI, 8.0-9.5) pp in beneficiaries with 3 new complex conditions, and 12.8 (95% CI, 11.6-13.9) pp in beneficiaries with 4 or more new complex conditions. Increased rates of switching were based solely on disenrollment to TM, except for those with exactly 1 new complex condition, for which developing a new complex condition was also associated with increased switching within MA among beneficiaries (0.4 [95% CI, 0.3-0.6] pp). We did not find that our estimates were sensitive to controlling for changes in the number of other chronic conditions (eTable 3 in Supplement 1). Event study estimates do not show evidence of differential pretreatment trends in disenrollment between beneficiaries who developed new complex conditions and those who did not (eFigure 1 in Supplement 1).
Figure 2. Dot and Whisker Plot of Changes in Plan Disenrollment Following New Complex Conditions, by Number of Conditions.
Difference-in-differences estimates between developing a new complex medical condition in 2019 and MA plan disenrollment, stratified by the number of new complex conditions (any, 1, 2, 3, or ≥4). Estimates compare changes in outcomes before and after 2019 among beneficiaries who developed new complex conditions and beneficiaries who did not. All regressions include beneficiary and year fixed effects. Beneficiaries were required to be enrolled in MA in the prior year and had no traditional Medicare enrollment in the prior year. Complex conditions include acute myocardial infarction, Alzheimer disease, atrial fibrillation, chronic kidney disease, chronic obstructive pulmonary disease, major depression, congestive heart failure, or stroke, identified using the Chronic Conditions Data Warehouse algorithm applied to Medicare claims; diagnoses from MA medical record review records were excluded. Error bars represent 95% CI. Data are from the Medicare Master Beneficiary Summary File and Medicare claims data from 2016 through 2021. MA indicates Medicare Advantage; TM, traditional Medicare.
Among the individual new complex conditions, Alzheimer disease and related disorders was associated with the largest increase in disenrollment (8.6 [95% CI, 8.1-9.1] pp), while the other complex conditions ranged between 2.9 (95% CI, 2.0-3.8) pp for acute myocardial infarction and 5.0 (95% CI, 4.6-5.5) pp for chronic obstructive pulmonary disease. All estimates were statistically significant with P < .001. eFigure 2 in Supplement 1 depicts disenrollment across different complex conditions.
Heterogeneity in Changes in Disenrollment Following New Complex Condition
Table 1 examines heterogeneity in changes in disenrollment following the development of new complex conditions by state Medigap policy overall and by the number of conditions. Table 2 examines heterogeneity in changes in disenrollment following the development of new complex conditions by plan HMO status overall and by the number of conditions. Enrollees in states with Medigap community rating and guaranteed issue policies that developed new complex conditions were 1.5 (95% CI, 0.9-2.1) pp more likely to disenroll to TM than enrollees in states without those policies that developed new conditions (Table 1). Consistent with prior results,11,12 state Medigap policy accounted for the largest differential effect estimates in TM disenrollment among those with the greatest number of complex conditions (Table 1). However, the finding of developing a new complex condition and switching plans within MA was not modified by state Medigap policy.
Table 1. Difference-in-Differences Estimates of Plan Disenrollment Following New Complex Conditions, by State Medigap Protection Policy.
| Condition statusa | Difference-in-differences estimate, (SE)b | ||
|---|---|---|---|
| Any switch | Disenroll to TM | Switch within MA | |
| Any complex conditiona | |||
| New complex condition | 0.033 (0.001)c | 0.027 (0.000)c | 0.006 (0.001)c |
| Complex condition × guaranteed issue stated | 0.015 (0.003)c | 0.015 (0.002)c | −0.000 (0.003) |
| Meane (SD) | 0.153 (0.360) | 0.0152 (0.122) | 0.138 (0.344) |
| No. of complex conditionsa | |||
| 1 | 0.024 (0.001)c | 0.017 (0.001)c | 0.008 (0.001)c |
| 2 | 0.053 (0.002)c | 0.052 (0.001)c | 0.000 (0.002) |
| 3 | 0.094 (0.005)c | 0.096 (0.003)c | −0.003 (0.004) |
| ≥4 | 0.128 (0.008)c | 0.144 (0.006)c | −0.016 (0.007)f |
| 1 Complex condition × guaranteed issue state | 0.011 (0.003)c | 0.011 (0.002)c | 0.000 (0.003) |
| 2 Complex condition × guaranteed issue state | 0.029 (0.008)c | 0.026 (0.005)c | 0.003 (0.007) |
| 3 Complex condition × guaranteed issue state | 0.014 (0.015) | 0.030 (0.010)c | −0.015 (0.013) |
| ≥4 Complex condition × guaranteed issue state | 0.054 (0.028)g | 0.056 (0.020)c | −0.002 (0.021) |
| Meane (SD) | 0.153 (0.360) | 0.0152 (0.122) | 0.138 (0.344) |
Abbreviations: MA, Medicare Advantage; TM, traditional Medicare.
Complex condition information is listed in the caption to Figure 1.
To convert difference-in-differences estimates to difference in percentage points as reported in Results, multiply by 100.
Statistical significance indicated by P < .01.
Guaranteed issue states are identified in Outcome and Exposure Variables in Methods.
Mean denotes the mean of the dependent variable in the preperiod for the estimation sample.
Statistical significance indicated by P < .05.
Statistical significance indicated by P < .10.
Table 2. Difference-in-Differences Estimates of Plan Disenrollment Following New Complex Conditions, by Plan HMO Statusa.
| Condition status | Difference-in-differences estimate, (SE)b | ||
|---|---|---|---|
| Any switch | Disenroll to TM | Switch within MA | |
| Any complex conditionc | |||
| New complex condition | 0.006 (0.001)d | 0.034 (0.001)d | −0.028 (0.001)d |
| Complex condition × HMO | 0.044 (0.002)d | −0.010 (0.001)d | 0.054 (0.002)d |
| Meane (SD) | 0.153 (0.360) | 0.0152 (0.122) | 0.138 (0.344) |
| No. of complex conditionsc | |||
| 1 | −0.003 (0.002)f | 0.023 (0.001)d | −0.026 (0.001)d |
| 2 | 0.027 (0.004)d | 0.060 (0.002)d | −0.032 (0.003)d |
| 3 | 0.070 (0.007)d | 0.102 (0.004)d | −0.032 (0.006)d |
| ≥4 | 0.091 (0.012)d | 0.138 (0.008)d | −0.047 (0.010)d |
| 1 Complex condition × HMO | 0.045 (0.002)d | −0.010 (0.001)d | 0.055 (0.002)d |
| 2 Complex condition × HMO | 0.045 (0.008)d | −0.009 (0.003)d | 0.054 (0.004)d |
| 3 Complex condition × HMO | 0.037 (0.009)d | −0.005 (0.006) | 0.042 (0.008)d |
| ≥4 Complex condition × HMO | 0.065 (0.015)d | 0.017 (0.011) | 0.048 (0.013)d |
| Meanf (SD) | 0.153 (0.360) | 0.0152 (0.122) | 0.138 (0.344) |
Abbreviations: HMO indicates health maintenance organization; MA, Medicare Advantage; TM, traditional Medicare.
Plan type is based on the beneficiary’s 2018 MA plan; beneficiaries not enrolled in MA in 2018 were excluded from these analyses. All regressions include beneficiary and year fixed effects. Beneficiaries were required to be enrolled in MA in the prior year and had no TM enrollment in the prior year.
To convert difference-in-differences estimates to difference in percentage points as reported in Results, multiply by 100.
Complex condition information is listed in the caption to Figure 1.
Statistical significance indicated by P < .01.
Mean denotes the mean of the dependent variable in the preperiod for the estimation sample.
Statistical significance indicated by P < .10.
In contrast, enrollees enrolled in HMO plans in 2018 who developed new complex conditions were less likely to disenroll to TM than enrollees not enrolled in HMOs who developed new conditions (1.0 pp; 95% CI, 0.8-1.2 pp) and were more likely to switch plans within MA (5.4 pp; 95% CI, 5.0-5.8 pp) (Table 2). However, this finding did not vary considerably as a function of the number of new complex conditions a beneficiary developed.
We further examined heterogeneity in disenrollment rates across beneficiary and plan characteristics (eFigure 3 and eTable 4 in Supplement 1). We found that dual-eligible enrollees with new complex conditions had elevated disenrollment rates to TM and lower rates of switching plans within MA. There was limited heterogeneity across other beneficiary and plan characteristics. For example, we did not find that enrollees in 4-star or 5-star plans who developed new complex conditions differed significantly in disenrollment from those enrolled in plans rated 3 stars.
Discussion
This cohort study found that developing a new complex condition was associated with substantially higher rates of disenrollment from MA. This finding was concentrated in beneficiaries with multiple new conditions and was primarily based on disenrollment to TM rather than switching within MA. Disenrollment to TM was particularly elevated in enrollees living in states with Medigap community rating and guaranteed issue protections, while disenrollment within MA was higher among beneficiaries enrolled in HMOs.
Our study adds methodologic rigor to the literature demonstrating that beneficiaries with higher health care needs are more likely to disenroll from MA plans. Relative to the most closely related prior studies, our approach allows us to examine differences in disenrollment levels and trends prior to the development of new complex conditions. Nonetheless, our findings are consistent with prior work. Our estimates are comparable with estimates from cross-sectional studies between having a health condition and disenrollment, despite differences in methods.4,5,6,7,8,9
Our findings highlight that estimates that focus only on a binary measure of medical condition may mask heterogeneity in how the complexity of a beneficiary’s medical needs may shape their MA disenrollment decisions. For example, we found that disenrollment rates increased substantially with a greater number of complex medical conditions. Beneficiaries with 4 or more new complex medical conditions were more than 5-fold more likely to disenroll to TM than enrollees with just a single new complex condition. While enrollees with just 1 new complex condition were slightly more likely to switch plans within MA, enrollees with 4 or more were actually less likely to switch plans within MA.
In addition, we found that Medigap community rating and guaranteed issue policies were also associated with a 1.5-pp increase in disenrollment, broadly consistent with 2 recent studies that examined new conditions, specifically, but focused on differences between beneficiaries across states with different Medigap policies.11,12 These findings underscore the difficulty of making an initial enrollment decision in Medicare. Beneficiaries cannot foresee all of their future health needs when they first enroll in MA, but the consequences of that decision may depend on health events that occur years later. Outside of states with guaranteed issue and community rated Medigap policies, enrollees may effectively be unable to switch even after their preferences change. In addition, we also observed elevated rates of disenrollment to TM among dual-eligible enrollees, whose eligibility for Medicaid reduces the need for Medigap.
Finally, switching plans within MA was elevated only among beneficiaries enrolled in HMO plans, consistent with more restrictive plans driving switching. Contrary to prior cross-sectional work,6 contract-level star ratings were not associated with disenrollment among beneficiaries with new complex conditions. Beneficiaries enrolled in plans rated 4 stars or 5 stars who develop a new complex condition are just as likely to disenroll from their plans as those in plans rated 3 stars.
There are several specific takeaways from our findings. State Medigap community rating and guaranteed issue policies, and Medicaid eligibility, likely ease the decision of enrollees with complex medical conditions to disenroll to TM, but combined with selection into the market may increase overall Medigap premiums.11,15 Further, MA star ratings may not reflect how well plans serve individuals with elevated medical complexity.
Limitations
This study has several limitations. First, we identify new complex conditions as they appeared in medical claims, which likely appear after the actual onset of the condition. Our estimates should not be interpreted as the causal effect of a diagnosis in isolation. The emergence of a complex condition in claims data likely coincides with increased health care use. Enrollees may also experience symptoms of their conditions prior to the formal diagnosis, which may bias estimates toward 0. Second, we explore heterogeneity in the degree of medical complexity using the count of complex conditions as a proxy. True medical complexity is a qualitative measure that does not vary linearly in this way. Third, the subgroup of beneficiaries who developed 4 or more new complex conditions in a single year is small and estimates for this group should be interpreted with corresponding caution. The dose-response pattern across the 1, 2, and 3 condition groups, which are substantially larger, supports the broader inference that disenrollment rises with medical complexity.
Conclusions
In this cohort study, developing a new complex medical condition was associated with higher disenrollment from an MA plan, particularly to TM, and the difference-in-differences estimates increased with the number of complex conditions. Disenrollment to TM was particularly elevated in enrollees living in states with Medigap community rating and guaranteed issue protections, while disenrollment within MA was higher among beneficiaries enrolled in HMOs. State Medigap protections likely make switching to TM easier for those with the greatest levels of medical complexity enrolled in MA.
eTable 1. Sample Size and Unadjusted MA Plan Disenrollment Rates, Before and After 2019, by Number of New Complex Conditions Developed in 2019
eTable 2. Beneficiary Characteristics in Patients with and without New Complex Conditions in 2019, 2018
eTable 3. Sensitivity of Difference-in-Differences Estimates, Controlling for Changes in Other Chronic Conditions
eFigure 1. Event Study Regression of Any Switch
eFigure 2. Disenrollment Following New Complex Conditions, by Condition
eFigure 3. Plan Disenrollment Following New Complex Conditions, by Beneficiary and Plan Characteristics
eTable 4. Beneficiary Characteristics in Patients with New Complex Conditions in 2019 by Plan Switching Status, 2020
Data Sharing Statement
References
- 1.Ochieng N, Freed M, Fuglesten Biniek J, Damico A, Neuman T. Medicare Advantage in 2025: enrollment update and key trends. 2025. Accessed Aug 21, 2025. https://www.kff.org/medicare/medicare-advantage-enrollment-update-and-key-trends/
- 2.Roberts ET, Burke R, Haddad K. Medicare Advantage supplemental benefits: origins, evolution, and issues for policy making. Health Aff Forefront. Published online September 19, 2024. doi: 10.1377/forefront.20240913.836055 [DOI]
- 3.Ochieng N, Freed M, Fuglesten Biniek J, Damico A, Neuman T. Medicare Advantage in 2025: premiums, out-of-pocket limits, supplemental benefits, and prior authorization. 2025. Accessed Aug 21, 2025. https://www.kff.org/medicare/medicare-advantage-premiums-out-of-pocket-limits-supplemental-benefits-and-prior-authorization/
- 4.Meyers DJ, Rahman M, Rivera-Hernandez M, Trivedi AN, Mor V. Plan switching among Medicare Advantage beneficiaries with Alzheimer’s disease and other dementias. Alzheimers Dement (N Y). 2021;7(1):e12150. doi: 10.1002/trc2.12150 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rahman M, Keohane L, Trivedi AN, Mor V. High-cost patients had substantial rates of leaving Medicare Advantage and joining traditional Medicare. Health Aff (Millwood). 2015;34(10):1675-1681. doi: 10.1377/hlthaff.2015.0272 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Meyers DJ, Belanger E, Joyce N, McHugh J, Rahman M, Mor V. Analysis of drivers of disenrollment and plan switching among Medicare Advantage beneficiaries. JAMA Intern Med. 2019;179(4):524-532. doi: 10.1001/jamainternmed.2018.7639 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Liu A, Pittman D, Anderson G, Xu J. Medigap-guaranteed issue associated with Medicare Advantage disenrollment for beneficiaries administered a part B drug. Health Aff Sch. 2024;2(11):qxae136. doi: 10.1093/haschl/qxae136 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lee B, Pittman D, Cooper S, Gupta R, Anderson GF, Liu A. State Medigap policies and disenrollment from Medicare Advantage among beneficiaries with cancer. JCO Oncol Pract. 2025;OP2500069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Parsons HM, Greenwald SJ, Jarosek S, et al. Switching between Medicare Advantage and Traditional Medicare for individuals newly diagnosed with cancer 2015-2019. J Natl Cancer Inst. 2025;117(6):1218-1227. doi: 10.1093/jnci/djaf036 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Freed M, Ochieng N, Cubanski J, Neuman T. Key Facts About Medigap Enrollment and Premiums for Medicare Beneficiaries. KFF. Published October 18, 2024. Accessed Aug 25, 2025. https://www.kff.org/medicare/key-facts-about-medigap-enrollment-and-premiums-for-medicare-beneficiaries/
- 11.Sun J, Chernew ME, McWilliams JM, Vabson B, Layton TJ. Medigap regulations provide protections for beneficiaries, especially after health shocks, but may raise premiums. Health Aff (Millwood). 2025;44(7):855-861. doi: 10.1377/hlthaff.2024.01361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kwon Y, Jazowski SA, Hu X, et al. Medigap Protection and Plan Switching Among Medicare Advantage Enrollees With Cancer. JAMA Health Forum. 2025;6(6):e252018. doi: 10.1001/jamahealthforum.2025.2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Joynt KE, Figueroa JF, Beaulieu N, Wild RC, Orav EJ, Jha AK. Segmenting high-cost Medicare patients into potentially actionable cohorts. Healthc (Amst). 2017;5(1-2):62-67. doi: 10.1016/j.hjdsi.2016.11.002 [DOI] [PubMed] [Google Scholar]
- 14.Chronic Conditions Data Warehouse Web site . Condition categories. Accessed February 10, 2021. https://www2.ccwdata.org/web/guest/condition-categories
- 15.Ryan AM, Warrier A, Bejarano G, Whaley CM, Meyers DJ, Dixit MN. Trends in plan offerings, enrollment, and premiums in Medicare Advantage and Medigap. Health Serv Res. 2025;60(3):e14456. doi: 10.1111/1475-6773.14456 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. Sample Size and Unadjusted MA Plan Disenrollment Rates, Before and After 2019, by Number of New Complex Conditions Developed in 2019
eTable 2. Beneficiary Characteristics in Patients with and without New Complex Conditions in 2019, 2018
eTable 3. Sensitivity of Difference-in-Differences Estimates, Controlling for Changes in Other Chronic Conditions
eFigure 1. Event Study Regression of Any Switch
eFigure 2. Disenrollment Following New Complex Conditions, by Condition
eFigure 3. Plan Disenrollment Following New Complex Conditions, by Beneficiary and Plan Characteristics
eTable 4. Beneficiary Characteristics in Patients with New Complex Conditions in 2019 by Plan Switching Status, 2020
Data Sharing Statement
