Impact of Continued Biased Disenrollment from the Medicare Advantage Program to Fee-for-Service

Abstract

Background

Medicare managed care enrollees who disenroll to fee-for-service (FFS) historically have worse health and higher costs than continuing enrollees and beneficiaries remaining in FFS.

Objective

To examine disenrollment patterns by analyzing Medicare payments following disenrollment from Medicare Advantage (MA) to FFS in 2007. Recent growth in the MA program, introduction of limits on timing of enrollment/disenrollment, and initiation of prescription drug benefits may have substantially changed the dynamics of disenrollment.

Study design

The study was based on MA enrollees who disenrolled to FFS in 2007 (N=248,779) and a sample of “FFS stayers” residing in the same counties as the disenrollees (N=551,616). Actual Medicare Part A and Part B payments (excluding hospice payments) in the six months following disenrollment were compared with predicted payments based on claims experience of local FFS stayers, adjusted for CMS-Hierarchical Condition Category (CMS-HCC) risk scores.

Results

Disenrollees incurred $1,021 per month in Medicare payments, compared with $798 in predicted payments (ratio of actual/predicted=1.28, p < 0.001). Differences between actual and predicted payments were smaller for disenrollees of Preferred Provider Organizations and Private Fee-for-Service plans than of Health Maintenance Organizations. Analysis of 10 individual MA plans revealed variation in the degree of selective disenrollment.

Conclusions

Despite substantial changes in policies and market characteristics of the Medicare managed care program, disenrollment to FFS continues to occur disproportionately among high-cost beneficiaries, raising concerns about care experiences among sicker enrollees and increased costs to Medicare.

Introduction

Since the early 1980s, the Medicare Advantage (MA) program and its predecessors have offered beneficiaries the option of joining Health Maintenance Organizations (HMOs) and other managed care plans. In return for providing covered services, plans receive capitation payments from Medicare that are risk-adjusted to account for differences in expected costs between their enrollees and beneficiaries in the local fee-for-service (FFS) sector. Since its inception, the Medicare managed care program has been characterized by favorable selection among participating plans, due to enrollment of healthier individuals and disenrollment of sicker ones (McGuire, Newhouse, & Sinaiko, 2011). To the extent that favorable selection is not reflected in risk adjustment, Medicare pays more for enrollees in managed care plans than for similar beneficiaries in FFS (Brown, Clement, Hill, Retchin, & Bergeron, 1993; Riley, Tudor, Chiang, & Ingber, 1996). Recent studies have confirmed that MA plans continue to enroll beneficiaries with better health and lower costs than their counterparts in FFS (MedPAC, 2012; Riley, 2012).

Disenrollment of costly beneficiaries from the MA program exacerbates the favorable selection observed among MA plans. Selective disenrollment of high-cost enrollees increases Medicare costs by removing heavy users of services from capitated arrangements and placing them in the FFS sector where Medicare absorbs their full costs. Because the distribution of Medicare costs is heavily skewed, even a small number of high cost disenrollees may substantially increase costs for the Medicare program. Prior to 2006, the departure of chronically ill enrollees from managed care plans was facilitated by the ability of beneficiaries to disenroll from plans on a monthly basis. Consequently, enrollees could leave the managed care sector at the onset of expensive illness to seek an expanded choice of providers or better benefits. There has also been concern that plans may induce chronically ill enrollees to disenroll if they become too expensive.

Previous research has found annual disenrollment rates to FFS of 4–9 percent (Riley, Ingber, & Tudor, 1997; Laschober, 2005). Researchers have found that disenrollees to FFS have worse health status than continuing enrollees, and worse health status and higher costs than beneficiaries who remain in FFS (MedPAC, 2012; Ng, Kasper, Forrest, & Bierman, 2007; Morgan, Virnig, DeVito, & Persily, 1997; Call et al., 2001; Maciejewski, Dowd, Call, & Feldman, 2001; Physician Payment Review Commission, 1996; Riley, Lubitz, & Rabey, 1991). Atherly, Hebert, & Maciejewski (2005) reported that among managed care enrollees with diabetes, higher pre-enrollment expenditures were associated with earlier disenrollment. Morgan, Virnig, DeVito, & Persily (2000) found high rates of osteoarthritis-related joint replacement among disenrollees to FFS in South Florida. Some chronically ill beneficiaries may disenroll in response to plan restrictions on certain benefits. For example, Rector (2000) found that exhaustion of drug benefits led to disenrollment from Medicare managed care. The tendency of sicker beneficiaries to disenroll may be related to the timing of disease onset. Cancer patients who were diagnosed while enrolled in managed care were less likely to disenroll than cancer-free enrollees (Elkin et al., 2008; Riley, Feuer, & Lubitz, 1996). However, those diagnosed 1–18 months before enrollment were more likely to disenroll (Riley, Feuer, & Lubitz, 1996).

Dissatisfaction with care sometimes leads to disenrollment of higher cost individuals to FFS. Findings from the Medicare Consumer Assessments of Healthcare Providers and Systems (CAHPS) survey have shown that in 2000–2001, beneficiaries in FFS rated the care they received more highly than enrollees in managed care across several domains (Landon, Zaslavsky, Bernard, Cioffi, & Cleary, 2004). Keenan, Elliott, Cleary, Zaslavsky, & Landon (2009) found these differences in ratings between MA and FFS to be larger for sick than healthy enrollees in 2003–2004. Between 2000 and 2005 the CAHPS program fielded surveys of Medicare managed care disenrollees, asking them about their reasons for disenrollment. Plan-specific disenrollment rates and reasons for disenrollment were publicly reported as a tool for beneficiaries to evaluate and compare plans. Analysis of survey results revealed that respondents in fair or poor health were more likely to report problems with care access or meeting specific needs as a reason for disenrollment (Mobley et al., 2005). Problems with costs and benefits were also frequently cited by less healthy disenrollees as reasons for disenrollment (Mobley et al., 2007). Other research has shown that high disenrollment rates are associated with poor performance on CAHPS measures of plan quality among members still enrolled in their plan (Lied, Sheingold, Landon, Shaul, & Cleary, 2003).

The purpose of this study was to measure recent patterns of biased selection in MA disenrollment by analyzing post-disenrollment Medicare payments of beneficiaries who disenrolled from MA to FFS in 2007. These payments were compared to payments incurred by a comparison group of beneficiaries who remained continuously in FFS. Current disenrollment patterns may differ from earlier patterns for several reasons. First, the Medicare managed care program has grown significantly in the 2000's, expanding to include preferred provider organizations (PPOs), private fee-for-service plans (PFFS), and special needs plans (SNPs), which were not available during most earlier studies. Second, beginning in 2006, new rules were put in place limiting the timing of enrollment and disenrollment in MA. Beneficiaries may enroll, disenroll, or switch plans during an annual election period extending from November 15 to December 31, with new enrollments effective January 1. Beneficiaries may also join, switch, or drop plans during an open enrollment period between January 1 and March 31, with some exceptions. Dual eligibles, Part D low income subsidy recipients, nursing home residents, new Medicare beneficiaries, and those who move may still enroll and disenroll outside the normal open enrollment periods. Lastly, enrollment and disenrollment decisions have become more complicated with the introduction of the Part D prescription drug benefit, which is offered through both MA plans and stand-alone prescription drug plans (PDPs). All these developments may have affected selection patterns associated with disenrollment.

Methods

The primary goal of the analysis was to compare Medicare FFS payments of disenrollees, in the 6 months following disenrollment with “predicted payments,” based on payment amounts of “stayers” residing in the same counties as the disenrollees and remaining in FFS. Predicted payments were calculated to reflect differences in CMS-HCC risk scores between the disenrollees and stayers. The difference between actual and predicted payments represents the degree to which disenrollees are more or less costly than similar beneficiaries who stayed in FFS. Disenrollment of beneficiaries with higher-than-expected costs would indicate that the favorable selection experienced by MA plans through enrollment of low-cost beneficiaries is being exacerbated by selective disenrollment of high cost-beneficiaries. The analysis is designed to explore selection at disenrollment from a Medicare payer perspective and the implications for Medicare program costs.

The study sample was selected from the Medicare Enrollment Data Base (EDB), which contains beneficiary-level data on entitlement, demographics, and a complete history of managed care plan enrollment and disenrollment, including plan identifiers. An initial sample of EDB records was selected consisting of all non-end-stage renal disease (ESRD) beneficiaries enrolled in MA plans as of January 1, 2007 who disenrolled to FFS in the subsequent 12 months. The sample did not include disenrollees from non-capitated plans, demonstration plans, or program of all-inclusive care for the elderly (PACE) plans, because these organizations tend to differ from standard MA plans in various ways and have different rules regarding enrollment, disenrollment, and payment. A disenrollment was defined to be to FFS if the month following disenrollment was spent in FFS; if an individual had more than one disenrollment to FFS in 2007, the first disenrollment was selected. Because voluntary disenrollments were the focus of the analysis, the sample excluded disenrollments due to death, a change of residence in the month of disenrollment or in the month preceding or following disenrollment, or withdrawal of the plan from the enrollee's area of residence. The disenrollee sample was limited to residents of counties with 20 or more MA disenrollees (94 percent of all disenrollees) to limit the size of the comparison group (described below).

A comparison sample of “FFS stayers” was selected consisting of a 2 percent sample of non-ESRD beneficiaries who resided in the same counties as the disenrollees, were entitled to Part A and Part B and in FFS on January 1, 2007, and who remained in FFS through June 30, 2008 (or until death if they did not survive through June 2008). Each FFS stayer was randomly assigned a “pseudo-disenrollment date” in 2007; these dates were assigned in such a way that the distribution of pseudo-disenrollment dates matched the distribution of true disenrollment dates among the disenrollee sample. If a FFS stayer died before his/her assigned pseudo-disenrollment date, that observation was eliminated from the analysis.

For individuals in both samples, Medicare Part A and B FFS claims data were obtained for the 6 months following the date of disenrollment or pseudo-disenrollment. Claims for all services except hospice were obtained; hospice claims were excluded, because hospice payments are not incorporated in the CMS-Hierarchical Condition Category (CMS-HCC) risk adjuster, which was the basis for calculating predicted costs (see below). Part D prescription drug records were not included, because a substantial portion of the samples was not enrolled in Part D. Sample records were linked to denominator file records for 2007 to obtain demographic information and to ascertain Medicaid eligibility at the time of disenrollment or pseudo-disenrollment. CMS-HCC risk scores were also obtained for all sample members for 2007 and 2008. CMS-HCC risk scores, which are computed for all Medicare beneficiaries, are used to establish capitated payment rates under the MA program and are derived from diagnostic and demographic factors predictive of future Medicare costs (Pope et al., 2004). By construction, the average risk score for all Medicare beneficiaries is approximately 1.0, with higher risk scores indicating higher expected costs. For each sample member, a weighted average of the 2007 and 2008 CMS-HCC risk scores was calculated according to the proportion of post-disenrollment or post-pseudo-disenrollment months that was spent in each of the two calendar years. After eliminating 0.4 percent of the combined sample because of missing risk scores, there were 248,779 disenrollees and 551,616 comparisons in the final sample.

To obtain predicted monthly payments for disenrollees, monthly payment data for the FFS stayer sample were regressed on CMS-HCC risk scores using ordinary least squares (OLS) regression. OLS regression is appropriate for analyzing health care costs in large samples (Lumley, Diehr, Emerson, & Chen, 2002) and was used in the development of CMS-HCC risk scores (Pope et el., 2011). The stayer observations were weighted in such a way to produce identical distributions across counties of disenrollees and stayers in order to control for market factors that may influence beneficiary payments (e.g., geographic variation in prices, availability of supplemental insurance, physician practice patterns). Observations were also weighted to reflect the number of months each stayer was alive during the 6 months following pseudo-disenrollment. Model coefficients were then applied to the disenrollee sample to generate predicted monthly payments.

Average monthly payments for disenrollees were compared with average payments for FFS stayers and with average predicted payments using t-tests. Actual and predicted payments were weighted to reflect the number of months each disenrollee was alive and in FFS during the 6 months following disenrollment. Comparisons of actual and predicted monthly payments were made for several beneficiary subgroups and for disenrollees from different types of MA plans. Such comparisons were also made at the individual plan level for plans with more than 5,000 disenrollees. For all subgroup analyses, predicted payments were based on separate regression models in which stayer observations were weighted to produce the same distribution across counties as the relevant disenrollee group. Predicted payments for dually eligible disenrollees (i.e., those eligible for Medicare and Medicaid) were based on payments for dually eligible FFS stayers, and similarly for non-dually eligible disenrollees.

Results

In January 2007 there were 7.1 million enrollees in Medicare Advantage, excluding non-capitated, demonstration, and PACE plans (Centers for Medicare & Medicaid Services, 2007). The fact that 248,779 individuals disenrolled to FFS suggests that the voluntary disenrollment rate in that year was approximately 3.5 percent (data not in exhibits). This is consistent with the lower range of findings from prior studies. Disenrollees tended to be somewhat younger and were much more likely to be of Black race (19.0 percent) or dually eligible (31.1 percent) than FFS stayers (10.6 percent Black and 21.0 percent dually eligible) (Exhibit 1). About 90 percent of disenrollees had Part D coverage at the time of disenrollment, compared to slightly over half of FFS stayers.

Exhibit 1. Characteristics of disenrollees from Medicare Advantage plans and fee-for-service stayers, 2007–2008.

Characteristic	Disenrollees	Fee-for-Service Stayers
N	248,779	551,616
	Percent
Age	100.0	100.0
< 65	17.2	15.5
65–74	42.1	38.5
75–84	29.4	31.9
85+	11.3	14.2
Gender
Male	41.4	39.7
Female	58.6	60.3
Race/ethnicity
Non-Hispanic white	64.6	77.3
Non-Hispanic black	19.0	10.6
Hispanic	11.9	8.2
Other/unknown	4.4	4.0
Dual eligible (Medicaid and Medicare eligible)
No	69.0	79.0
Yes	31.1	21.0
Institutionalized
No	97.2	97.7
Yes	2.8	2.3
Enrolled in Part D drug plan at the time of disenrollment or pseudo-disenrollment
No	10.1	49.4
Yes	89.9	50.6

NOTES. p < 0.001 for differences between disenrollees and fee-for-service (FFS) stayers with respect to all characteristics. Disenrollees went from Medicare Advantage (MA) to fee-for-service. Switches from one MA plan to another were not counted as MA disenrollments. Data for FFS stayers were weighted to match the distribution of disenrollees across states and counties of residence. All data were weighted by the number of months alive and in FFS during the six months following disenrollment or pseudo-disenrollment.

SOURCE: Medicare administrative records.

Average CMS-HCC risk scores were higher among disenrollees (1.26) than among FFS stayers (1.12), indicating higher predicted Medicare payments and presumably worse health status (Exhibit 2). Average monthly Medicare FFS payments were also higher among disenrollees ($1,021 compared to $710). Disenrollees incurred significantly higher average payments for all categories of service, with the largest relative difference occurring for skilled nursing facility and home health care services.

Exhibit 2. Average risk scores and monthly Medicare fee-for-service payments for Medicare Advantage disenrollees and fee-for-service stayers, by type of service, 2007–2008.

Measure	Disenrollees	Fee-for-service stayers
Average risk score (CMS-HCC)	1.26	1.12
Average monthly Medicare payments
Total	$1,021	$710
Inpatient hospital	$425	$294
Physician/supplier	$268	$228
Outpatient hospital	$92	$72
Skilled nursing facility	$141	$62
Home health care	$96	$54

NOTES. p < 0.001 for differences between disenrollees and fee-for-service stayers with respect to average risk scores and all types of service. Disenrollees went from Medicare Advantage (MA) to fee-for-service (FFS). Switches from one MA plan to another were not counted as MA disenrollments. Medicare payments exclude payments for hospice care and were measured over the six months following dates of disenrollment or pseudo-disenrollment. Data for FFS stayers were weighted to match the distribution of disenrollees across states and counties of residence. All data were weighted by the number of months alive and in FFS during the six months following disenrollment or pseudo-disenrollment.

SOURCE: Medicare administrative records.

The difference in average Medicare monthly payments between disenrollees and FFS stayers largely reflects the influence of the highest cost disenrollees (Exhibit 3). In the bottom half of the distribution, monthly payments were similar for disenrollees and stayers. In the upper half, payments for disenrollees and stayers diverged, with disenrollees experiencing much higher payment levels. For example, the median monthly Medicare payment for disenrollees was $143, compared to $122 for FFS stayers; the 90^th percentile of payments was $2,888 for disenrollees and $1,738 for stayers.

Exhibit 3. Distribution of monthly Medicare fee-for-service payments for Medicare Advantage disenrollees and fee-for-service stayers, 2007–2008.

Percentile	Disenrollees	Fee-for-Service stayers
	Average monthly Medicare payments
10	$0	$0
25	$25	$27
50	$143	$122
75	$672	$434
90	$2,888	$1,738
95	$5,461	$3,696
99	$12,046	$9,182

NOTES. Disenrollees went from Medicare Advantage (MA) to fee-for-service (FFS). Switches from one MA plan to another were not counted as MA disenrollments. Medicare payments exclude payments for hospice care and were measured over the six months following dates of disenrollment or pseudo-disenrollment. Data for FFS stayers were weighted to match the distribution of disenrollees across states and counties of residence. All data were weighted by the number of months alive and in FFS during the six months following disenrollment or pseudo-disenrollment.

SOURCE: Medicare administrative records.

Average monthly FFS payments for disenrollees were 28 percent higher than predicted (Exhibit 4). The ratio of actual to predicted FFS payments was similar for dually eligible and non-dually eligible disenrollees, although the absolute difference between actual and predicted payments was much higher for dual eligibles ($331 compared to $159). Average monthly payments for rapid disenrollees (those who had been enrolled only 1–3 months before disenrolling) were closer to predicted payments (actual/predicted=1.15). Disenrollees who re-enrolled in another MA plan within six months incurred lower-than predicted average monthly FFS payments (actual/predicted=0.90). The ratio of actual to predicted payments was larger for disenrollees from HMOs and similar managed care plans (1.36) than for disenrollees from preferred provider organizations (1.18) or private fee-for-service (PFFS) plans (1.14).

Exhibit 4. Actual and predicted monthly fee-for-service payments for Medicare Advantage disenrollees, by selected plan and disenrollee characteristics, 2007–2008.

Plan and disenrollee characteristics	N	Actual post-disenrollment payments	Predicted post-disenrollment payments	Actual/predicted payments
Total	248,779	$1,021	$798	1.28
Disenrollee characteristics
Dual (Medicare and Medicaid) eligible	78,494	$1,464	$1,133	1.29
Non-dual eligible	170,285	$822	$663	1.24
Rapid disenrollee (enrolled 1–3 months)	44,524	$942	$816	1.15
Re-enrolled in Medicare Advantage within six months	33,720	$732	$809	0.90
Type of Medicare Advantage plan
Health Maintenance Organization/Managed Care Plans	146,151	$1,165	$855	1.36
Preferred Provider Organizations	22,545	$882	$748	1.18
Private Fee-for-Service Plans	80,083	$810	$711	1.14

NOTES. Disenrollees went from Medicare Advantage (MA) to fee-for-service (FFS).

Switches from one MA plan to another were not counted as MA disenrollments.

Medicare payments exclude payments for hospice care and were measured over the six months following dates of disenrollment or pseudo-disenrollment. Predicted payments for MA disenrollees are based on payments for FFS stayers adjusted for differences in CMS-HCC risk scores between MA disenrollee and local FFS stayer populations.

Data for FFS stayers were weighted to match the distribution of disenrollees across states and counties of residence. All data were weighted by the number of months alive and in FFS during the six months following disenrollment or pseudo-disenrollment. p < 0.001 for ratio of actual/predicted payments = 1.0 for each disenrollee and plan characteristic.

SOURCE: Medicare administrative records.

Among the 10 plans with the most disenrollees, 5 were HMO-type plans and 5 were PFFS plans (Exhibit 5). The number of disenrollees varied between 5,218 and 23,083 (accounting for 39 percent of all disenrollees in the sample) and the number of corresponding FFS stayers varied between 92,826 and 491,534. For 7 of the 10 plans, disenrollees incurred significantly higher monthly FFS payments than predicted, with the ratio of actual to predicted payments varying between 1.15 and 1.55. Average payments in the other three plans did not differ significantly from predicted payments. There was more variation among the HMO-type plans than among the PFFS plans with respect to the ratio of actual/predicted FFS payments.

Exhibit 5. Ratio of actual to predicted monthly fee-for-service payments for Medicare Advantage disenrollees, for plans with more than 5,000 disenrollees, 2007-2008.

Discussion

Despite substantial changes in policies and market characteristics of the Medicare managed care program, disenrollment to FFS continues to occur disproportionately among high-cost beneficiaries. Disenrollees had higher risk scores and incurred higher risk-adjusted payments than beneficiaries in FFS. Their high risk scores are in contrast to the risk scores of the general MA population, most of which is enrolled in plans with average risk scores similar to or less than local FFS experience (United States Government Accountability Office, 2010). Recent studies have also shown that MA plans continue to experience favorable selection through enrollment of low-cost beneficiaries (MedPAC, 2012; Riley, 2012). These research findings suggest a pattern of selective disenrollment whereby disenrollees are sicker and more expensive than the beneficiaries who remain enrolled in MA plans. This selective disenrollment potentially increases Medicare costs through the return of high-cost beneficiaries to the FFS sector, leaving behind a healthier and lower-cost population in the capitated MA sector.

The Affordable Care Act mandated changes to MA payment methods that will result in significant decreases in payment rates and bring them closer in line with plan costs. This may intensify pressure on plans to encourage selective disenrollment. It is also noteworthy that the MA program has expanded rapidly in recent years, attaining an enrollment of 13.1 million in 2012 (Gold et al., 2012). High MA enrollment levels increase the population of disenrollees, even at relatively low disenrollment rates. Consequently, selective disenrollment of high cost enrollees from MA plans to the FFS sector remains of significant policy concern for the Medicare program.

Disenrollment of high-cost individuals may be related to persistent lower levels of satisfaction among sicker MA enrollees. Elliott, Haviland, Orr, Hambarsoomian, & Cleary (2011) found more problems getting needed care among MA enrollees than beneficiaries in FFS, and lower ratings of physicians, after case-mix adjustment in 2007. They also found larger disparities in care experiences between sick and healthy beneficiaries in the MA sector. MA outperformed FFS on measures related to Part D, however. Beginning in 2012, MA plans will be eligible to receive quality bonus payments based on quality ratings, including measures of enrollee satisfaction derived from CAHPS. Quality bonus payments are designed to provide incentives to plans to improve access to and quality of care for their enrollees, which may in turn affect future patterns of disenrollment.

Among individual MA plans, the magnitude of favorable selection at disenrollment varied considerably. Additional research is needed to determine why some plans exhibit greater selection at disenrollment than others and whether certain plans have a persistent pattern of disenrollment by high cost beneficiaries over time. One concern would be whether high cost enrollees are being induced to leave such plans. Monitoring disenrollment patterns by plan, and tracking health status indicators such as risk scores, post-disenrollment payments, or mortality rates, may help identify potential plan abuses. For example, disenrollment patterns could be used as a screening tool to trigger reviews of plan policies regarding provider recruitment, coverage, and appeals procedures. Such reviews could focus resources more efficiently toward plans where improvement is more likely to be needed.

The absolute difference between actual and predicted payments was greater for dual eligibles than non-dual eligibles and the ratio of actual to predicted payments was somewhat higher. Dual eligibles have historically exhibited high rates of disenrollment (Riley, Ingber, and Tudor, 1997), and comprised a large proportion of the study sample. The disenrollment of high-cost dual eligibles may be related to the difficulty in combining Medicare and Medicaid contracting for managed care services for this population (Grabowski, 2009). MA members who become eligible for Medicaid while enrolled may face problems with access to Medicaid-covered services if their state Medicaid agency has no formal contracting arrangements with their MA plan, or if their plan does not provide certain Medicaid-covered services. Acute care providers in the MA plan network may also be less accessible to enrollees who have qualified for Medicaid coverage after entering a nursing home. Dually eligible MA enrollees at greatest risk for problems with access to and coordination of services are likely those with multiple chronic conditions, who also tend to be the most costly. Such enrollees may choose to disenroll to FFS if they find they experience problems receiving appropriate care. Disenrollment of high-cost dual eligibles may result in part from the challenges of coordinating Medicare and Medicaid program administrative requirements rather than dissatisfaction with quality of care per se. Dual-eligible Special Needs Plans (SNPs) are designed to coordinate Medicare and Medicaid benefits for enrollees, but states have not contracted extensively with SNPs, limiting the integration of Medicaid with Medicare services (Schmitz, Merrill, Schore, Shapiro, & Verdier, 2008).

Disenrollees from PPOs and PFFS plans incurred lower payments post-disenrollment than disenrollees from HMOs and similar types of plans, and their average payments were closer to predicted levels. Possible explanations include a less chronically ill disenrollee population from PPO and PFFS plans, or less unmet demand for services when they transitioned to FFS. Less selective disenrollment from PPO and PFFS plans may be attributable to the more extensive network of providers available under these types of plans. Beneficiaries with chronic illnesses have a greater choice of physicians to manage their conditions and have more opportunities to switch providers if they become dissatisfied with their care. This expanded choice of providers may reduce the incentives for chronically ill enrollees to leave these types of plans.

Unlike other disenrollees, those who re-enrolled in a MA plan within 6 months did not incur higher than predicted Medicare payments. This is not consistent with the findings of Morgan, Virnig, DeVito, and Persily (1997), who observed an elevated use of inpatient hospital services among HMO disenrollees who later re-enrolled. In that study use of hospital care dropped to FFS levels shortly before re-enrollment. The implication of the finding by Morgan, Virnig, DeVito, and Persily was that some beneficiaries temporarily disenrolled from a Medicare HMO to FFS in order to receive selected services that they wished to obtain outside the HMO's provider network, or that were more expensive in their HMO. Given the restrictions that were placed on disenrollment beginning in 2006, most beneficiaries can no longer disenroll for brief periods to receive a particular service in FFS, but must wait for the annual election period.

Several study limitations should be kept in mind. First, the analysis did not attempt to estimate the full financial impact of biased disenrollment. Other features of the MA payment system, such as the methodology for establishing local benchmarks, also determine the financial impact of MA disenrollment patterns. This analysis was limited to measuring biased selection among MA disenrollees and did not compare the FFS costs of disenrollees with actual payments under the MA program. Second, it was not possible to determine the reasons why beneficiaries included in this study disenrolled from their plans. Third, the CMS-HCC scores of disenrollees are based largely on diagnoses reported by MA plans while the individual was enrolled. These scores may be inflated because of “coding creep” related to plan incentives to report as many diagnoses as possible (Medicare Payment Advisory Commission, 2012). If differential reporting of diagnoses did produce artificially high CMS-HCC scores for disenrollees, the true difference between actual and predicted payments may be greater than reported here. Lastly, the study was based on disenrollments occurring in 2007, which may not be representative of underlying steady-state patterns. Major program changes occurred in 2006, such as introduction of MA enrollment lock-in and the Part D prescription drug benefit, whose effects on MA disenrollment patterns may not have fully played out by 2007.

The Affordable Care Act mandated changes in payment policies for MA plans that are expected to reduce program costs in future years. Disenrollments may increase if plans respond by increasing enrollee cost sharing, reducing benefits, or imposing more utilization controls. Chronically ill enrollees may be more inclined to disenroll if access to care deteriorates or if plans cover a smaller portion of the costs of their care. Disenrollment will continue to be of interest to policymakers as the MA program evolves and financial pressures affect both plans and beneficiaries.