Impact of Medicare Managed Care Market Withdrawal on Beneficiaries

Abstract

The 2001 Survey of Involuntary Disenrollees was conducted to investigate the impact of Medicare+Choice (M+C) plan withdrawals on Medicare beneficiaries. Eighty-four percent of a total of 4,732 beneficiaries whose Medicare managed care (MMC) plan stopped serving them at the end of 2000 responded to the survey Their responses indicated that the withdrawal of plans from Medicare affected beneficiaries in terms of concerns about getting and paying for care, increased payments for premiums and out-of-pocket costs, and changes in health care arrangements. Of particular concern were the impacts on those in vulnerable subgroups such as the disabled, less educated, and minorities.

Introduction

The Balanced Budget Act (BBA) of 1997 expanded the health care options potentially available to Medicare beneficiaries through the establishment of M+C, allowing beneficiaries to enroll in a variety of private health care options beyond the original Medicare fee-for-service (FFS) program. However, although BBA 1997 increased the health care provider options available to seniors, the payment rates to plans were lower than expected. These rates, along with new administrative requirements, may have initiated the withdrawal of many managed care organizations from the Medicare market (U.S. General Accounting Office, 2001).

The effect of withdrawals by MMC plans, at the individual beneficiary level, can be disruptive, particularly for any beneficiaries who have to change providers. Statistics compiled from CMS' 1998 reports indicate that Medicare health care plan withdrawals resulted in 407,000 beneficiaries (6.5 percent of M+C enrollees) making a plan change in January 1999. In 2000, 327,000 beneficiaries (5 percent of M+C enrollees) were affected by plans' withdrawals or reduction in service areas. Some plans withdraw completely from MMC while others continue participation in Medicare but reduce their service areas by no longer serving beneficiaries in some counties or ZIP codes. In 2001, 934,000 Medicare beneficiaries (15 percent of total enrollment in M+C) were forced to make new choices about their health plan coverage when their Medicare health care plan withdrew from the program or reduced their service areas. Most recently, 536,000 (10 percent of M+C enrollees) were affected in 2002 (Gold and McCoy, 2002).

Few studies have documented the experiences these involuntary disenrollees face when they are forced to select new coverage upon withdrawal of their plan from the M+C Program (Kaiser Family Foundation, 1999; U.S. General Accounting Office, 1999; Gold and Justh, 2000). For that reason, CMS decided to conduct a survey to assess the ongoing impact of MMC market withdrawal on beneficiary coverage.

Most health maintenance organizations (HMOs) that participate in Medicare offer additional benefits outside the regular Medicare benefit package (Achman and Gold, 2002). Extra benefits commonly include low copayments, prescription drugs, unlimited hospitalization, and preventive services. Many beneficiaries have come to rely on the extra benefits they receive from their HMO, particularly prescription drug coverage. Replacing these benefits through Medigap insurance is usually very expensive, and may be unaffordable for some (Gold and Mittler, 2001). Joining another HMO or going to FFS may also force many beneficiaries to change doctors, creating dissatisfaction and disrupting existing patterns of care. Therefore, there has been concern among policymakers about the impact of the recent HMO withdrawals on the beneficiary population (Barry and Kline, 2002).

There were two previous national efforts specifically designed to assess the impact of the plan withdrawals and service area reductions on beneficiaries.¹ The first consisted of a survey sponsored by the Henry J. Kaiser Family Foundation after the January 1999 withdrawals. A report based on the survey results indicated that although most disenrollees fared relatively well after their HMO withdrew from Medicare, many experienced a reduction in supplemental benefits, an increase in premiums, and/or disruptions in their care arrangements (Kaiser Family Foundation, 1999). Problems were disproportionately experienced by disabled beneficiaries, racial and ethnic minorities, the poor and near poor, and those reporting fair or poor heath. The second effort consisted of a telephone survey of several hundred beneficiaries conducted by the Department of Health and Human Services' Office of the Inspector General. The survey covered enrollee notification; information and assistance in exploring new insurance options; what option beneficiaries selected; changes in benefits and costs; problems encountered; and satisfaction. The Office of the Inspector General survey was conducted twice, following the January 1999 and January 2000 withdrawals. These surveys did not find severe problems, but no analyses were done for vulnerable populations.

A third study, by Gold and Justh (2000), involved a national sample of over 6,000 Medicare beneficiaries of whom 425 were in M+C plans that stopped serving enrollees at the end of 1999. (This survey was conducted as part of the larger monitoring M+C project of Mathematica Policy Research, funded by the Robert Wood Johnson Foundation.) A recent publication from this study summarizes the differences between plans that withdraw from M+C and those that remain (Achman and Gold, 2002). Withdrawing plans tend to have lower enrollments, offer less-generous benefit packages, had less stable benefits from 1999 to 2000, and faced competition problems within their markets.

The purpose of the 2001 Survey of Involuntary Disenrollees (Center for Health Systems Research and Analysis and RTI International, 2001) was to understand how Medicare beneficiaries are affected by Medicare health plan withdrawals and reductions in service areas. This article describes the methods and the results of that survey. Also, we discuss:

• What types of beneficiaries are affected by plan withdrawals.

• What information beneficiaries receive about plan withdrawals.

• The extent to which beneficiaries understand the implications of their plan's withdrawal and their new coverage options.

• The impact on beneficiaries in terms of new coverage, concerns, costs, provider arrangements, and access to care.

Methodology

Design

The survey was conducted by mail with telephone followup of non-respondents. Data were collected between March and June 2001. (A copy of the survey is available on request from the authors.) We designed two versions: the first targeted to sample members who did not have end stage renal disease (ESRD) and the second targeted to those who did. ESRD patients may be adversely impacted by plan withdrawals because they tend to have high health care expenses, and at the time that the survey was designed, they were not permitted to enroll in other MMC plans. Until the Medicare, Medicaid, and State Children's Health Insurance Program (SCHIP) Benefits Improvement and Protection Act of 2000 (BIPA) was enacted in December 2000, ESRD beneficiaries in non-renewing plans could not join a new M+C plan. (The main goal of BIPA was to increase payments to M+C organizations to maintain and expand beneficiary access to M+C plans. However, on implementation of BIPA in March 2001, only seven M+C organizations re-entered counties from which they had previously withdrawn or expanded into new counties [U.S. General Accounting Office, 2001]). Since the implementation of BIPA, ESRD beneficiaries are allowed to enroll in another M+C plan if their plan terminates its contract with CMS. This provision applies to terminations occurring on or after the date of BIPA's enactment and retroactively to terminations on or after December 31, 1998.

Both questionnaires contained the same questions, with the ESRD version containing an additional three questions specifically related to dialysis treatment. The topics included:

• Questions about the sample member's former health insurance.

• Choosing new health insurance.

• Questions about the sample member's current health insurance.

• Getting needed care since the sample member left their former plan.

• Impact on dialysis treatment (for ESRD sample members only).

• Respondent health status and demographic characteristics.

Sample Selection and Weighting

For the non-ESRD population, a sampling frame was constructed which included all enrollees, as of October 1, 2000, in plans that terminated or reduced their service areas effective January 1, 2001. The reason for using a 3-month window was to capture people who stayed in the plan until the end of the year, as well as those who may have left earlier, in the event that there were differences between these types of enrollees. All beneficiaries who lived outside the United States, deceased, and institutionalized sample members were excluded from the sampling frame. Once the frame for the non-ESRD population was constructed, beneficiaries were assigned to one of two strata—those who lived in areas where another MMC plan was available after December 31, 2000, and those who lived in areas in which no other Medicare plan was available. A separate sample consisting of all Medicare beneficiaries who had ESRD who were affected by plans' withdrawals and reduction in service areas was selected. No stratum for these sample members was defined since, at the time that the study was designed, ESRD patients would not be able to enroll in another MMC plan even if other Medicare health plans were available in their area on January 1, 2001.

The sample sizes for the choice and no choice strata were based on a goal of obtaining 3,000 completed interviews with non-ESRD Medicare beneficiaries, and 385 completed interviews with Medicare beneficiaries with ESRD. The choice strata included 2,772 Medicare beneficiaries who lived in areas in which another Medicare health plan was available after December 31, 2000. The no choice strata included 1,422 Medicare beneficiaries who lived in areas in which no other Medicare health plan was available after December 31, 2000. The ESRD strata included 538 beneficiaries. Data collection activities resulted in an overall response rate of 83.7 percent. (The response rate was calculated using the following formula: Numerator—the number of completed interviews. Denominator—all sample members in the sample minus those who were institutionalized or deceased, and those who reported that they were still enrolled in the sample plan or left the plan because they moved out of the plan's service area. Questionnaires were considered complete if beneficiaries responded to at least one of the items concerning the impact of plan withdrawal. [Additional information about the completeness criteria are available on request from the authors.])

To adjust for the potential of differential non-response bias, we used logistic regression to model the functional relationship between a set of predictors and a dichotomous response outcome and then used that model to construct response propensity weights. The potential set of predictors available for both respondents and non-respondents included age, race, and geographic area. Black persons and other minority races had less than one-half the odds of a response than did white persons. This was the most significant effect in the model. Age was also a significant factor. The odds of a response steadily decreased with the age of the sample member. No change in odds was noted in the under age 65 population. Address fields that contained rural routes or post office boxes (along with variant abbreviations) had a lower odds of response. Address fields that were abnormally short or long had a higher odds of response. Abnormally long addresses often suggest that an individual is in the care of someone else which, in this case, may have increased the likelihood of response. Responses varied only slightly among the U.S. census divisions.

Data Analysis

Descriptive statistics and chi-square tests of independence were used to assess statistical associations between a number of potential outcomes of the plan withdrawals and beneficiary characteristics. These outcomes include new coverage arrangements and the financial, psychological, and care-related impacts of the plan withdrawals from the Medicare Program. The results presented in this article are based on weighted data. A weighting model was developed incorporating both sample design and response propensity. Consequently, the data reported for the total sample reflects the total population of involuntary disenrollees. (Additional details are available within the final report available on request from the authors [Booske et al., 2002]). Analyses were conducted using SUDAAN^® software that appropriately accounts for the sample weighting approach in calculating standard errors. Findings of significance at the 99 percent probability level and differences of at least 10 percentage points are reported. In addition, where appropriate, results of multivariate analysis (using logistic regression) are reported to further examine the relationships between beneficiary characteristics and the impact of plan withdrawals. Results are reported for significant logit models with a minimum Cox & Snell R-square of at least 0.10 or where the model increases the likelihood of prediction from the logistic model by at least 10 percent (versus a model that simply assigns all responses to the most frequent response category). Additional variables used in the logistic regression analyses, such as the MMC penetration rate groups and the payment rates that M+C organizations receive per enrollee per month, were derived from CMS files, Internet site: http://www.cms.hhs.gov/healthplans/reportfilesdata/

Results

Who are Affected by Plan Withdrawals?

Table 1 shows the characteristics of the three sample groups of beneficiaries who responded to the survey: those who lived in a county with a choice of another Medicare HMO, those in a county without another Medicare HMO alternative, and those with ESRD. The total column represents weighted data from all three strata. Thus, the total columns in selected tables reflect the weighted mix of those in counties with and without a Medicare HMO (as of January 1, 2001) and of those with ESRD, i.e., in proportion to the composition of the entire population of involuntary disenrollees. Overall, 92 percent of the beneficiaries responding to the survey reside in a metropolitan county compared with 76 percent of Medicare beneficiaries nationally (Achman and Gold, 2002). However, there was a significant difference between the geographic location of the choice and no-choice strata: only 3 percent of those with a choice of another HMO lived in a non-metropolitan area while 34 percent of the beneficiaries without another HMO available live in non-metropolitan counties.

Table 1. Survey of Involuntary Disenrollee Sample Strata, by Beneficiary Characteristics: 2001.

Characteristic	Total (n=3,780)	Medicare HMO Available (n=2,215)	No Medicare HMO Available (n=1,195)	ESRD (n=370)
	Percent
Age
Under 65 Years	7	7	7	16
65-74 Years	53	52	55	46
75-84 Years	33	33	31	34
85 Years or Over	7	7	7	5
Sex
Female	57	58	54	44
Male	43	42	46	56
Race/Ethnicity
White Non-Hispanic	84	84	84	61
Black	9	1	7	25
Hispanic	5	4	8	11
Other	2	2	1	2
Education
Less than 9th Grade	13	12	18	19
Some High School	18	18	17	27
High School Graduate	36	36	37	29
Beyond High School	33	34	28	25
Self-Reported Health Status
Excellent	6	6	5	2
Very Good	22	22	19	6
Good	37	37	38	21
Fair	28	28	29	41
Poor	8	7	1	3
Recent Hospitalizations
At Least 1 in Past 12 Months	21	21	21	69
None in Past 12 Months	79	79	79	31
Location
Metropolitan County	92	97	66	95
Non-Metropolitan County	8	3	34	5
Dual Eligibility Status
Not Medicaid Eligible	97	97	96	9
Medicaid Eligible	3	3	4	1

NOTES: HMO is health maintenance organization. ESRD is end stage renal disease. Percentages are based on weighted data. The weighting incorporates both the sample design and response propensity. Consequently, the percentage in the total column cannot be calculated based on the weighted average of the three component columns. Numbers may not sum to 100 percent due to rounding. Metropolitan/non-metropolitan county designation based on the 1993 Office of Management and Budget definition.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Compared with CMS data on voluntary disenrollees (those who leave Medicare HMOs of their own accord), enrollees (those who stay in an HMO), and involuntary disenrollees (those in plans that leave the Medicare Program) are similar in age, sex, and education, but more likely to report their health to be fair or poor and less likely to be Hispanic (Figure 1).

Figure 1. Involuntary and Voluntary Disenrollees and Medicare Beneficiaries that Remain with their Managed Care Plan: 2001.

Information about Plan Withdrawals

Sixty-six percent of beneficiaries first found out that their plan was going to stop covering them from the plan itself. The next most common source of information about the plan withdrawal came from the media, 18 percent of beneficiaries first found out that their plan was leaving the Medicare Program from newspapers, radio, or television.² Ninety-six percent of beneficiaries recalled receiving a letter at some point from the plan about its impending withdrawal.³

Only about 6 out of 10 beneficiaries indicated they had enough information about their options when their plan stopped covering them. Disabled beneficiaries (under age 65) and the oldest-old beneficiaries were less likely to indicate that they received enough information about their coverage options when they heard that their plan would stop covering them. Compared with other racial/ethnic groups, black beneficiaries were less likely to indicate that they had received enough information. Those in poor health were also less likely than healthier beneficiaries to indicate that they received enough information about their coverage options (Table 2).

Table 2. Beneficiary Reports of Adequacy of Information and Availability of Supplemental Insurance and their Understanding of Implications of Plan Withdrawal, by Beneficiary Characteristics: 2001.

Characteristic	Reporting Having Enough Information when their Plan Withdrew	Reporting that Supplemental Insurance Option Available	Understood What Would Happen when their Plan Withdrew1
	Percent
All Beneficiaries2	63	68	53
Age3
Under 65 Years	46	51	55
65-74 Years	64	71	55
75-84 Years	65	70	50
85 Years or Over	59	61	46
Race/Ethnicity3
White Non-Hispanic	65	72	56
Black	48	44	39
Hispanic	56	53	34
Other	60	65	35
Education3
Less than 9th Grade	56	55	39
Some High School	60	62	4
High School Graduate	65	71	55
Beyond High School	65	75	63
Self-Reported Health Status3
Excellent	63	62	45
Very Good	67	73	57
Good	66	72	54
Fair	59	63	51
Poor	49	61	47
End Stage Renal Disease3
No	63	68	53
Yes	55	67	47
Information About Plan Withdrawal3
Received Enough information	—	77	58
Did not Receive Enough Information	—	54	45

¹Percentage of beneficiaries who thought they would be covered by the original Medicare plan, covered through their current or former employer, would be able to select a new plan, or would have to purchase supplemental insurance.

²n=3,780.

³Chi-square for each column significant at 0.01 level.

NOTES: Percentages are based on weighted data. The weighting incorporates both the sample design and response propensity.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Knowledge of the availability of supplemental insurance was somewhat common: 68 percent of beneficiaries were aware of the availability. Responses to the question on availability of supplemental options may reflect beneficiary experiences with health screening, i.e., some may have been turned down, and so the insurance is unavailable although it is likely that some of these responses reflect misunderstandings on the part of respondents about what options are available to them. Beneficiaries who were disabled (under age 65), older, in the all other racial group (including Hispanic), and less educated beneficiaries were the least likely to report availability of supplemental insurance options (Table 2).

Beneficiaries were asked what they thought would happen to them when their plan stopped covering them. Those who thought they would be covered by the original Medicare plan, covered through their current or former employer, would be able to select a new plan, or would have to purchase supplemental insurance were considered to understand what would happen. (Those who thought they would have to purchase supplemental insurance may believe that prudence or financial necessity demanded the purchase of supplemental insurance rather than it being a legal or program requirement. Conse-quently, these individuals were also considered to have understood what would happen.) Less than one-half of the beneficiaries thought that the original Medicare plan would cover them. Some beneficiaries thought they would be automatically enrolled in another HMO while only 2 percent thought they would be able to select a new plan. One in 10 beneficiaries either indicated that they did not know what would happen or did not respond to the question. Conversely, those who thought that they would end up with no health insurance or that they would be automatically enrolled in another HMO apparently did not understand the implications of their plan's decision to stop covering them.⁴ Over one-quarter thought they would end up with no health insurance.

Just over one-half of beneficiaries appeared to understand exactly what would happen when their plan left the Medicare Program (Table 2). There was less understanding of what would happen among beneficiaries who were less educated, in the all other racial group, or Hispanic. However, those who reported that they had received enough information about the plan withdrawals were more likely to indicate they understood what would happen than those who did not indicate that they had enough information.

New Coverage Arrangements

Beneficiaries were asked about their new coverage arrangements following their plan's withdrawal from Medicare (Table 3). Respondents could indicate coverage under more than one arrangement so a hierarchical approach was used to assign them to the types of coverage. If respondents reported enrollment in a Medicare HMO, they were assigned to this category. For the remaining respondents (those who did not report enrollment in an HMO), if they reported that Medicaid covered them, they were assigned to this category. This process was repeated for each category so that the final category represented all respondents who did not report that they were enrolled in a Medicare HMO, were covered by Medicaid, were not covered through a current or former employer, and had no supplemental health insurance. Thus, we designated these respondents as covered by original Medicare only.

Table 3. Sample Strata, by Beneficiary Reports of New Coverage Arrangements after Plan Withdrawal: 2001.

Coverage	Total (n=3,780)	Medicare HMO Available (n=2,215)	No Medicare HMO Available (n=1,195)	ESRD (n=370)
	Percent
New Arrangements1
Enrolled in Medicare HMO	52	56	34	37
Covered by Medicaid	3	3	4	11
Covered Through Current or Former Employer	8	8	9	7
Have Supplemental Insurance	22	19	35	34
Covered by Original Medicare Only	15	15	18	11
Satisfaction with New Health Insurance
Less Satisfied Now	37	37	4	36
About the Same Now	38	39	3	34
More Satisfied Now	17	17	19	24
Don't Know or Missing	8	8	11	6
Payments for Monthly Premiums
Pay More Now	56	54	63	57
Pay Same Amount Now	13	13	10	13
Pay Less Now	8	8	6	9
Don't Pay Premiums2	13	14	9	9
Don't Know or Missing	11	11	11	13
Former Plan Paid Cost of Medicines
Yes	74	76	65	75
No	16	14	23	14
Don't Know or Missing	10	10	12	11
Health Insurance Now Pays Cost of Medicine
Yes	53	55	41	42
No	38	36	49	49
Don't Know or Missing	10	9	11	10
Paying for Prescription Medicines
Pay More Now	51	52	49	55
Pay Same Amount Now	25	25	25	18
Pay Less Now	10	10	10	17
Don't Know or Missing	14	14	16	10

¹Respondents could indicate coverage under more than one arrangement so a hierarchical approach was used to assign new coverage arrangements. First, if respondents reported enrollment in a Medicare HMO, they were assigned to this category. Next, if applicable, they were assigned to Medicaid. This process was repeated for each category. The final category includes all respondents who did not report enrollment in a Medicare HMO, Medicaid, coverage through an employer, or supplemental insurance.

²Beneficiaries who paid no premiums both before and after plan withdrawal.

NOTES: HMO is health maintenance organization. ESRD is end stage renal disease. Percentages are based on weighted data. The weighting incorporates both the sample design and response propensity. Consequently, the percentage in the total column cannot be calculated based on the weighted average of the three component columns. Numbers may not sum to 100 percent due to rounding.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Just over one-half of the involuntary disenrollees reported enrollment in another HMO after their plan withdrew (Table 3). Fifteen percent of the involuntary disenrollees reported being covered by original Medicare only. The results of a logistic regression model to predict which beneficiaries would end up with only original Medicare are shown in Table 4. As might be expected, a market characteristic (the market penetration of MMC in an area) was a predictor of whether a beneficiary would end up with original Medicare only. Beneficiaries in areas of high managed care penetration were less likely to end up with only original Medicare only after their plan withdrew. However, the model also suggests that those in vulnerable subgroups, such as the disabled, and those who report their health to be fair or poor, were also more likely than other beneficiaries to only have original Medicare coverage. In addition, those who reported that they did not have enough information about plan withdrawals were also more likely to end up with Medicare only. However, caution is advised in interpretation of these results since the multivariate model using logistic regression to predict whether a beneficiary reported having traditional Medicare coverage did not have as much explanatory power as other models derived from this study.

Table 4. Results of Logistic Regressions of New Coverage for All Involuntary Disenrollees: 2001.

Independent Variable	Having Original Medicare Only After Plan Withdrawa1		Having Supplemental Insurance After Plan Withdrawal2
	Odds Ratio	95 Percent Confidence Interval	Odds Ratio	95 Percent Confidence Interval
Intercept	0.15	0.10-0.22	3.33	2.42-4.59
Age
Under 65 Years (Disabled)	*1.89	1.29-2.77	*0.41	0.27-0.60
65 Years or Over	1.00	—	1.00	—
Sex
Female	0.81	0.63-1.02	1.07	0.88-1.31
Male	1.00	—	1.00	—
Race/Ethnicity
All Other Racial Groups and Hispanic	1.45	0.07-1.98	*0.54	0.41-0.72
White Non-Hispanic	1.00	—	1.00	—
Education
Less than 9th Grade	1.36	0.98-1.89	*0.67	0.50-0.91
9th Grade or More	1.00	—	1.00	—
Self-Reported Health
Fair or Poor	*1.58	1.22-2.05	0.83	0.67-1.03
Good or Excellent	1.00	—	1.00	—
Hospitalized in Past 12 Months
Yes	*0.65	0.48-0.89	1.15	0.89-1.47
No	1.00	—	1.00	—
CMS Region
Region 63	*1.82	1.38-2.41	*0.58	0.45-0.74
Other Regions	1.00	—	1.00	—
County
Non-Metropolitan	1.42	0.96-2.11	0.80	0.57-1.14
Metropolitan	1.00	—	1.00	—
Reported HMO Enrollment
Yes	NA	NA	*0.70	0.57-0.85
No	—	—	1.00	—
Reported Medicaid Enrollment
Yes	NA	NA	1.01	0.67-1.51
No	—	—	1.00	—
Medicare Managed Care Market Penetration
High (35-45 Percent)	*0.44	0.22-0.86	*0.38	0.24-0.58
Moderate (15-34 Percent)	*0.46	0.32-0.67	*0.62	0.47-0.82
Limited (6-14 Percent)	0.98	0.68-1.40	0.74	0.54-1.01
Minimal (1-5 Percent)	0.78	0.55-1.09	*1.39	1.02-1.89
None (No HMOs)	1.00	—	1.00	—
Medicare Monthly Payment Rate to Medicare+Choice Organization
Less than $525	*0.47	0.32-0.69	*2.90	2.08-4.04
$525	0.91	0.66-1.25	*1.52	1.19-1.93
More than $525	1.00	—	1.00	—
Information about Plan Withdrawal
Not Enough Information	*1.61	1.26-2.06	0.84	0.69-1.03
Enough Information	1.00	—	1.00	—

^*p<0.05.

¹Model is significant. Cox & Snell R-square for dependent variable having Medicare only = 0.07.

²Model is significant. Cox & Snell R-square for dependent variable having supplemental insurance = 0.12.

³Centers for Medicare & Medicaid Services (CMS) Region 6 consists of the following States: Arizona, Louisiana, New Mexico, Oklahoma, and Texas.

NOTES: Overall sample size for these logistic regressions was 3,780. HMO is health maintenance organization. NA is not available. Metropolitan/non-metropolitan county designation based on the 1993 Office of Management and Budget definition. Medicare managed care market penetration groups based on characterization of Medicare markets by the Center for Studying Health System Change.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Those living in areas without a Medicare HMO and those with ESRD were far more likely to report having supplemental insurance than beneficiaries in areas with a Medicare HMO (Table 5). Beneficiaries who indicated that they had enough information about the plan withdrawals and those with more education were also more likely to report that they have supplemental coverage. Beneficiaries who were disabled, under age 65, in the all other racial group, Hispanic, and living in CMS region 6⁵ were all less likely to report having supplemental insurance (Table 4).

Table 5. Results of Logistic Regressions of Enrollment in Another HMO for Beneficiaries Living in Counties with Choice of Another Medicare HMO: 2001.

Independent Variable	Report Enrolling in Another HMO after Plan Withdrawal1		Enrolled in Medicare+Choice Plan after Plan Withdrawal, per CMS Administrative Records2
	Odds Ratio	95 Percent Confidence Interval	Odds Ratio	95 Percent Confidence Interval
Intercept	0.63	0.47-0.84	0.17	0.12-0.26
Age
Under 65 Years (Disabled)	0.92	0.61-1.40	1.50	0.95-2.39
65 Years or Over	1.00	—	1.00	—
Sex
Female	1.16	0.94-1.42	*1.29	1.02-1.62
Male	1.00	—	1.00	—
Race/Ethnicity
All Other Racial Groups and Hispanic	1.03	0.76-1.39	0.94	0.68-1.30
White Non-Hispanic	1.00	—	1.00	—
Education
Less than 9th Grade	1.04	0.75-1.45	0.97	0.68-1.39
9th Grade or More	1.00	—	1.00	—
Self-Reported Health
Poor or Fair	0.85	0.67-1.07	0.82	0.64-1.05
Good or Excellent	1.00	—	1.00	—
Hospitalized in Past 12 Months
Yes	1.07	0.82-1.38	0.95	0.72-1.25
No	1.00	—	1.00	—
CMS Region
Region 63	*0.72	0.55-0.94	0.88	0.65-1.17
Other Regions	1.00	—	1.00	—
County
Non-Metropolitan	1.59	0.83-3.07	0.69	0.23-2.03
Metropolitan	1.00	—	1.00	—
Medicare Managed Care Market Penetration
High (35-45 Percent)	*6.58	4.15-10.43	*22.89	13.59-38.54
Moderate (15-34 Percent)	*3.91	2.99-5.12	*9.21	6.35-13.37
Limited (6-14 Percent)	*1.74	1.27-2.38	*3.25	2.12-4.99
Minimal (1-5 Percent)	1.00	—	1.00	—
Monthly Medicare Payment Rate to Medicare+Choice Organization
Less than $525	0.93	0.53-1.64	*0.35	0.14-0.83
$525	*0.70	0.54-0.90	0.87	0.66-1.14
More than $525	1.00	—	1.00	—
Information about Plan Withdrawal
Not Enough Information	0.81	0.65-1.01	*0.67	0.52-0.85
Enough Information	1.00	—	1.00	—

^*p<0.05.

¹Model is significant. Cox & Snell R-square for dependent variable having Medicare only = 0.10.

²Model is significant. Cox & Snell R-square for dependent variable HMO enrollment according to administrative records = 0.20

³Centers for Medicare & Medicaid Services Region 6 consists of the following States: Arizona, Louisiana, New Mexico, Oklahoma, and Texas.

NOTES: Initial sample size for these logistic regressions was 2,215. HMO is health maintenance organization. Metropolitan/non-metropolitan county designation based on the 1993 Office of Management and Budget definition. Medicare managed care market penetration groups based on characterization of Medicare markets by the Center for Studying Health System Change.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

These frequencies are based on beneficiaries' reporting of their current health insurance coverage—and often do not correspond with administrative data. Consequently, caution is advised in interpreting these numbers, e.g., the 34 percent of beneficiaries who live in an area without another Medicare HMO, but reported belonging to an HMO. In general, there was over 90 percent agreement between CMS enrollment records and beneficiaries' reports of not being in an HMO. However, when beneficiaries reported that they do belong to an HMO, CMS records only confirmed HMO membership in about 50 percent of these cases.

Beneficiaries in counties without another Medicare HMO available who reported enrollment in an HMO may have been correct at the time of their survey response. The definition of whether a county offered a choice reflects whether there was a choice of another M+C HMO available for enrollment as of January 1, 2001. Subsequently, a few plans did expand their service into some counties (e.g., Texas, New York, and New Mexico as of March 1, 2001). Also, some Medicare HMOs imposed capacity limits. There are two types of capacity limits: (1) plans can request capacity limits be established prospectively to be applied when their enrollment reaches a certain level, or (2) plans can request that their enrollment level be limited to the number of beneficiaries currently enrolled. Consequently, at any point in time a plan with capacity limits may or may not actually be accepting new enrollees. Other HMOs only cover parts of counties (particularly group and staff models).

Also, the definition of choice does not address the availability of cost contract or private FFS plans or of HMO coverage that is offered to beneficiaries by a current or former employer or via participation in Medicaid. In addition, some beneficiaries may not understand that if they continue to see a provider that was affiliated with their former HMO, their services may now be covered under the original Medicare. Of the beneficiaries who live in an area without another Medicare HMO, CMS administrative files showed no record of M+C enrollment for 80 percent of the beneficiaries who reported HMO membership. Of the remaining 80 beneficiaries, administrative records showed that as of January 1, 2001, 27 were enrolled in cost plans and 24 enrolled in a private FFS which were not included in the definition of the choice/no-choice counties.

While over one-half of all beneficiaries reported membership in an HMO, this proportion was not constant across the three sample strata (Table 3). Beneficiaries in counties without a Medicare HMO and beneficiaries with ESRD were far less likely to report membership in an HMO. In fact, since only 37 percent of ESRD beneficiaries reported membership in an HMO (similar to the level for those in counties without Medicare HMOs), this suggests that few ESRD beneficiaries have benefited so far from the legislative changes in BIPA) of 2000. (CMS records actually only confirmed reports of HMO enrollment for 11 percent of the 38 percent of ESRD beneficiaries who reported enrollment in a M+C plan.) By far the most significant predictor of whether a beneficiary reported enrollment in another HMO in multivariate modeling was the market penetration of Medicare managed care in their county (Table 6). None of the beneficiary demographic characteristics were associated with a significant increase or decrease in the likelihood of reporting enrollment in another HMO. The odds of CMS Region 6 beneficiaries reporting enrollment in another HMO were 28 percent lower than those for beneficiaries living elsewhere in the United States. A logistic regression model of administrative records of HMO enrollment produced similar results with respect to market penetration; however, Region 6 was not significant in this model (Table 5). Beneficiaries who indicated that they had received enough information about the plan withdrawals were more likely to enroll in an HMO, according to administrative records, than those who said they did not get enough information.

Table 6. Results of Logistic Regressions of Concerns about Getting Care and Having to Pay More for Premiums after Plan Withdrawal: 2001.

Independent Variable	Being Very Concerned about Getting Needed Health Care after Plan Withdrawal1		Having to Pay More for Premiums after Plan Withdrawal2
	Odds Ratio	95 Percent Confidence Interval	Odds Ratio	95 Percent Confidence Interval
Intercept	0.75	0.52-1.09	11.73	7.32-18.78
Age
Under 65 Years (Disabled)	*2.30	1.53-3.47	*0.52	0.31-0.88
65 Years or Over	1.00	—	1.00	—
Sex
Female	*1.37	1.14-1.64	1.19	0.93-1.52
Male	1.00	—	1.00	—
Race/Ethnicity
All Other Racial Groups and Hispanic	*1.66	1.26-2.18	0.77	0.52-1.13
White Non-Hispanic	1.00	—	1.00	—
Education
Less than 9th Grade	*1.55	1.17-2.04	1.27	0.84-1.90
9th Grade or More	1.00	—	1.00	—
Self-Reported Health
Poor or Fair	*1.56	1.29-1.89	1.07	0.81-1.41
Good or Excellent	1.00	—	1.00	—
Hospitalized in Past 12 Months
Yes	*1.38	1.10-1.72	1.30	0.95-1.76
No	1.00	—	1.00	—
CMS Region
Region 63	1.05	0.83-1.33	2.33	1.63-3.33
Other Regions	1.00	—	1.00	—
County
Non-Metropolitan	*1.42	1.01-2.00	*0.79	0.47-1.33
Metropolitan	1.00	—	1.00	—
New Coverage Arrangement
Medicare HMO	*0.72	0.53-0.96	1.00	—
Medicaid	0.75	0.44-1.28	NA	0.70-1.54
Employer-Provided	*0.31	0.21-0.47	1.04	3.72-7.37
Supplemental	*0.43	0.31-0.59	*5.24	—
Original Medicare only	1.00	—	NA	—
Medicare Managed Care Market Penetration
High (35-45 Percent)	0.69	0.46-1.05	*0.44	0.26-0.73
Moderate (15-34 Percent)	*0.74	0.57-0.96	*0.55	0.39-0.77
Limited (6-14 Percent)	1.05	0.78-1.41	*0.64	0.43-0.96
Minimal (1-5 Percent)	0.98	0.74-1.30	1.30	0.85-1.99
None (No HMO)	1.00	—	1.00	—
Monthly Medicare Payment Rate to Medicare+Choice Organization
Less than $525	*0.67	0.47-0.95	1.36	0.84-2.18
$525	0.89	0.72-1.12	0.88	0.65-1.19
Greater than $525	1.00	—	1.00	—
Information about Plan Withdrawal
Not Enough Information	*2.55	2.11-3.07	NA	—
Enough Information	1.00	—	NA	—

^*p<0.05.

¹Initial sample size for this logistic regression was 3,780. Model is significant. Cox & Snell R-square for dependent variable having Medicare only = 0.40.

²Initial sample size for this logistic regression was 2,604 (excludes those with Medicare only, those with Medicaid coverage, and those who paid no premiums before and after plan withdrawal and those who did not know whether they paid higher or lower premiums). Model is significant. Cox & Snell R-square for dependent variable having to pay = 0.13

³Centers for Medicare & Medicaid Services Region 6 consists of the following States: Arizona, Louisiana, New Mexico, Oklahoma, and Texas.

NOTES: Metropolitan/non-metropolitan county designation based on the 1993 Office of Management and Budget definition. Medicare managed care market penetration groups based on characterization of Medicare markets by the Center for Studying Health System Change. HMO is health maintenance organization. NA is not available.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Impact of Plan Withdrawals on Beneficiaries

This section of results summarizes the impact of plan withdrawals on beneficiaries' care and provider arrangements. When asked about their level of satisfaction with their new coverage following the withdrawal of their former plan from Medicare, 37 percent of beneficiaries indicated that they were less satisfied with their insurance coverage now (Table 3). However, this may be due in part to lack of experience with their new coverage.

Another set of questions addressed the concerns that beneficiaries faced when they found out that their plan was withdrawing from Medicare. Fifty-one percent of the beneficiaries reported being very concerned about getting care that they needed. Respondents in counties with or without another Medicare HMO appeared not to differ with respect to concerns regarding ability to get care. However, there were significant differences between various subgroups with the more vulnerable expressing more concerns about their ability to get care after their plan withdrew: the beneficiaries who were disabled, less educated in the all other racial group or Hispanic, and reported fair or poor health were all more likely to report more concerns than other beneficiaries (Table 6). In particular, when all other conditions are held constant, the odds of those who are disabled being very concerned about getting care after their plan withdrew from Medicare are over 100 percent higher than those for aged beneficiaries. Information clearly reduced the likelihood of concerns: beneficiaries who did not have adequate information about the plan withdrawals were far more likely to have concerns about getting care than those with enough information (Table 6). Beneficiaries who live in nonmetropolitan areas or areas with low MMC penetration were also more likely to be concerned.

Fifty-six percent of the beneficiaries who responded to the survey reported that they had to pay more in premiums after their plan withdrew (Table 3). However, many HMOs that have remained in Medicare have increased premiums and reduced benefits in the last few years. In their continuing analysis of trends in benefits and premiums funded by the Commonwealth Fund, Achman and Gold (2002) noted that despite the congressional action to increase the payment rates that M+C organizations receive, mean premium and cost-sharing levels in M+C plans continued to increase in 2001. For example, average monthly premiums went from $14.43 in 2000 to $22.94 in 2001. (These average monthly premiums are for all M+C plans including those with zero premiums. In 2001, 46 percent of M+C plans offered zero premium packages.)

The odds of the disabled having to pay more for premiums were lower than those of aged beneficiaries. Fewer disabled beneficiaries may have experienced increases in premiums because many are dually eligible for Medicare and Medicaid and pay no premium for Medicaid. However, the logistic regression excluded those who are dually eligible and those who report having Medicare only and controlled for other characteristics (Table 6). The odds of beneficiaries with supplemental insurance reporting that they had to pay more for premiums after their plan withdrew were far greater when compared with beneficiaries reporting enrollment in another Medicare HMO (Table 6). Compared with beneficiaries in other regions, the odds of beneficiaries from CMS Region 6 having to pay more for premiums were also higher.

Seventy-four percent of beneficiaries reported that their former plan paid all or some of the cost of their prescription medicines whereas, after the withdrawing plans stopped covering them, this percentage dropped to 53 percent (Table 3). This compares with a decrease in the proportion of all M+C enrollees with prescription drug coverage that went from 78 percent in 2000 to 70 percent in 2001 (Achman and Gold, 2002).

The loss of prescription drug coverage meant that 51 percent of involuntary disenrollees found themselves paying more for prescription medicines after their former plan withdrew from Medicare (Table 3). However, about 1 in 10 beneficiaries did not know or did not answer questions about their prescription drug coverage before or after the plan withdrawals.

About one in five beneficiaries in areas with another Medicare HMO indicated that they had to change their personal doctor or nurse after their former plan stopped covering them compared with 12 percent in areas without another HMO. Twenty-four percent of the beneficiaries in areas with another HMO who had to change their personal provider indicated that it was a big problem to find a new provider. Beneficiaries in areas with a choice of another Medicare HMO reported a higher incidence of changing providers compared with those in areas without a choice of another HMO (Table 7).

Table 7. Beneficiary Reports of Impact on Provider Arrangements: 2001.

Report	Medicare HMO Available (n=2,215)	No Medicare HMO Available (n=1,195)
	Percent
Had to Change Personal Doctor or Nurse	22 (n=478)	12 (n=144)
A Big Problem to Get a Personal Doctor or Nurse	20	13
A Small Problem to Get a Personal Doctor or Nurse	26	17
Not a Problem to Get a Personal Doctor or Nurse	45	58
Don't Know or Missing	5	1
Have Not Found a New Doctor Yet	6	11
Seeing a Specialist in Former Plan	41 (n=895)	38 (n=448)
Had to Stop Seeing Specialist	24	16
Did Not Have to Stop Seeing Specialist	66	73
Don't Know or Missing	8	10
Did Not Need to See a Specialist	2	2

NOTES: HMO is health maintenance organization. Percentages are based on weighted data. The weighting incorporates both the sample design and response propensity. For example, 22 percent of beneficiaries in areas with another Medicare HMO available had to change providers after plan withdrawal. Twenty percent of this 22 percent, or 20 percent of 478, reported a big problem getting a different personal provider.

SOURCE: Booske, B.C., University of Wisconsin at Madison, Lynch, J., RTI International, and Riley, G., Centers for Medicare & Medicaid Services, 2001.

Four out of 10 beneficiaries reported that they had been seeing a specialist when their plan stopped covering them (Table 7). One in four of these beneficiaries (9 percent) reported that they had to stop seeing their specialist.

Twelve percent of the involuntary disenrollees indicated that they had trouble getting care they wanted or needed since their plan stopped covering them (Figure 2). This compares with 4 percent of all Medicare beneficiaries in HMOs (Centers for Medicare & Medicaid Services, 2002). Twenty-four percent, however, indicated that they had delayed seeking medical care because they were worried about the cost, compared with 3 percent of Medicare HMO beneficiaries. Fifteen percent reported that there were medicines prescribed for them that they did not get. Cost and the lack of insurance coverage were the most common reasons for not getting prescribed medicines.

Figure 2. Beneficiary Reports of Impact on Access to Care, by Beneficiary Characteristics: 2001.

There were clear and significant differences in problems with access to care (Figure 2) between the beneficiaries who were disabled, age 65 or over, white non-Hispanic and those in the all other racial group or Hispanic, and between those who reported good or excellent health and those who reported themselves to be in fair or poor health.

Discussion and Implications

The characteristics of involuntary disenrollees living in areas with and without a choice of another HMO were quite similar with one major exception: only 3 percent of beneficiaries affected by the 2001 plan withdrawals with a choice of another HMO lived in non-metropolitan counties, while 34 percent of beneficiaries without another HMO option lived outside of metropolitan areas. This difference highlights the continued disparity noted by Achman and Gold (2002) in choices available to metropolitan and non-metropolitan beneficiaries. In contrast to the similarities between beneficiaries in the other two sample strata, beneficiaries in the ESRD sample were quite different from those without ESRD: beneficiaries with ESRD were more likely to be female, black, in fair or poor health, and had to have been hospitalized during the past year.

The findings show that there are some clear information and understanding gaps among beneficiaries, particularly those in more vulnerable subgroups, regarding the options available to them and the implications of plans withdrawing from the Medicare Program. Letters from the non-renewing plans were by far the most frequent first source of information about the plan withdrawal for the majority of disenrollees. The media (television, radio, or newspaper) was the next most frequent source of information except for those in the more vulnerable subgroups such as the less educated and all racial groups other than white non-Hispanic. Disabled beneficiaries who were in the all other racial group and in fair to poor health were less likely to indicate they received enough information about the plan withdrawals. These same groups were less likely to be aware of the availability of supplemental health insurance. Other groups who were less aware of supplemental insurance options included those beneficiaries less-educated who were in the all other racial group (i.e., other than white) and less-educated beneficiaries were also less likely than other beneficiaries to understand what would happen to them with respect to health care coverage when their plan left the Medicare Program. About 4 in 10 beneficiaries in the all other racial group or beneficiaries with less than a high school education understood what would happen compared to one-half of all beneficiaries. Beneficiaries' reports of having enough information were clearly associated with their understanding of what would happen when their plan withdrew and awareness of supplemental health insurance options.

Compared with results from the Kaiser Family Foundation (1999) study of beneficiaries affected by plan withdrawals in January 1999, those affected by the January 2001 withdrawals were less likely to report enrollment in another HMO (52 percent in 2001 compared with 77 percent in 1999) even though similar proportions of beneficiaries (four out of five) still had an HMO option available to them. As was true in 1999, reports of enrollment in another HMO were strongly related to the number of Medicare plans and their market penetration in an area. Furthermore, it should be noted that when compared with CMS enrollment records, beneficiaries apparently overestimate membership in HMOs. Consequently, the true percentage that switches to another M+C is less than that derived from survey responses. While there are legitimate reasons why many beneficiaries believe that they are enrolled in a Medicare HMO even though this is not confirmed by CMS data, the high level of discordance with CMS enrollment records may cause some to wonder about the validity of other responses that beneficiaries gave. Although such discrepancies highlight the lack of understanding that many beneficiaries have about their coverage, beneficiaries' reports of coverage arrangements do reflect their perceptions of what coverage they do or do not have. Since CMS records do not include information on supplemental or employer coverage, asking beneficiaries about the coverage does provide some incremental information about levels of these types of coverage.

After MMC's role in a local market, having enough information was the next most likely predictor of beneficiaries choosing to enroll in another HMO. As opposed to reverting to original Medicare coverage, joining another HMO requires a conscious action on behalf of the beneficiary. Those who felt they did not have enough information may not have known that there was another HMO that they could join.

The other significant predictor of beneficiaries reporting HMO enrollment was living in a region other than CMS Region 6. Even after accounting for the lower availability of plans to beneficiaries in Region 6 (the average number of plans for a Region 6 beneficiary was one versus two for beneficiaries in other parts of the country), beneficiaries in Region 6 were less likely to enroll in another HMO and they were also less likely to have supplemental insurance coverage. A more detailed examination in particular markets within Region 6 might shed some light on the particular environment that beneficiaries in this region face, e.g., case studies by Grad and Hassol (2002) of involuntary disenrollees identified a unique situation in Houston, Texas in 2000 when the only remaining HMO in the area reached its capacity limit in September of that year and no longer accepted any new enrollees. Stuber et al. (2002) also examined the Houston market in their case studies of seven markets from which M+C plans withdrew. They noted the disruption of provider networks that often precedes a plan's decision to withdraw from a M+C market: in Houston, 27 percent of primary care physicians left a plan's network after one year. Eighty-six percent of Region 6's beneficiaries and 22 percent of all beneficiaries who responded to the 2001 survey were from Texas.

Reports of supplemental health insurance were strongly related to low MMC penetration, i.e., beneficiaries with fewer managed care options did turn to supplemental insurance. However, beneficiaries in vulnerable subgroups, such as the disabled, less educated, and those in the all other racial groups or Hispanic beneficiaries, were less likely to report having supplemental insurance than other beneficiaries. Sixty-six percent of those with original Medicare coverage only, i.e., without supplemental insurance, reported that they did not have supplemental insurance because it cost too much. This percentage was even higher (79 percent) for those in areas without a choice of another HMO. However, ESRD beneficiaries without supplemental insurance were almost as likely to cite not applying for it or thinking that they would be turned down as they were to cite its cost as the main reason for not having supplemental insurance. In addition to the financial barriers to acquiring supplemental coverage, lack of information appears to be a barrier to having supplemental insurance. Those who indicated that they had enough information about the plan withdrawals were more likely to have supplemental coverage than those who did not have enough information.

In examining the impact of plan withdrawals on beneficiaries, we looked at beneficiaries' concerns when hearing about the plan withdrawals, the impact on beneficiaries' costs, provider arrangements, and access to care. Approximately three out of every four beneficiaries reported that they were somewhat or very concerned about being able to pay for health care when their plan withdrew from the Medicare Program. A similar, but not entirely overlapping, proportion of beneficiaries were also concerned about getting care while concerns about having to change providers were not quite as widespread, but still considerable. Again, the more vulnerable subgroups, including the disabled, less-educated, people in the all other racial group (other than white non-Hispanic), and in fair or poor health, were disproportionately affected. Beneficiaries in non-metropolitan areas with low MMC penetration were also more concerned by the plan withdrawals. Having enough information reduced, but did not eliminate concerns about the impact of plan withdrawals.

Plan withdrawals affected just over one-half of beneficiaries with higher premiums. However, as previously mentioned, the average monthly premiums increased for all M+C enrollees along with increases in copayments and reductions in benefits. For involuntary disenrollees, higher premiums were more likely to come from acquiring supplemental insurance than from higher premiums due to enrollment in another HMO. Consequently, those in the more vulnerable subgroups were less likely to report paying more since they were less likely to have supplemental coverage. It is therefore likely that in the tradeoff between higher premiums versus lower benefits, the vulnerable were forced by circumstance to go with lower benefits. In an effort to keep our survey to a manageable length for respondents and due to the known unreliability of reports of specific benefit details, the only specific benefit about which beneficiaries reported was prescription drug coverage. About one in five beneficiaries lost prescription drug coverage as a result of their plan withdrawal. The disabled and those in fair or poor health were again impacted more than others by having to pay more for prescription medicines. Those in areas without a choice of another HMO were less likely to report having prescription drug coverage than those in areas with at least one HMO option.

Disruptions in provider arrangements were less widespread than some of the other outcomes of the plan withdrawals: only one in five beneficiaries indicated that they had had to change their personal doctor or nurse after their plan withdrew from Medicare. This was similar to the rate found in the Kaiser Family Foundation (1999) study. One out of 10 total beneficiaries had to stop seeing a specialist (somewhat lower than the rate found in the Kaiser study). Although the rate of disruption was not as high as might have been predicted, there were significant differences within the involuntary disenrollee population. Those in areas with a choice of another HMO were more likely to report having to change their personal doctor or nurse than those without another HMO option. Among those who were seeing a specialist when their plan withdrew, the disabled and those in fair or poor health were more likely to have to stop seeing their specialist. As would be expected, the potential for disruption in provider arrangements was a tradeoff that beneficiaries had to deal with in exchange for HMO coverage and the potential for more comprehensive benefits. Those who did not enroll in a different HMO were less likely to have to change providers.

In terms of access to care, one in 10 beneficiaries who needed care after their plan withdrew indicated that they had trouble getting care. However, one in five beneficiaries delayed seeking care because of being worried about the cost and 15 percent did not get medications that had been prescribed for them. This rate was three times higher than the rate found in the Kaiser study reflecting, perhaps, the increasing concerns about the costs of prescription drugs. Again, it was the beneficiaries who were disabled, in the all other racial group or Hispanic, and in fair or poor health whose access to care was most affected.

Clearly, the withdrawal of plans from the Medicare Program affected large numbers of beneficiaries causing at least short-term anxiety, higher costs, and/or disruptions in health care. However, what is of continuing concern is the effect of these withdrawals on the most vulnerable. While HMOs that have managed to stay in the Medicare market may continue to offer beneficiaries an attractive alternative to original Medicare, if and when the plans leave Medicare, thousands of disabled, minority, less educated, and sick beneficiaries are left uncertain about their options, concerned about costs (particularly of prescription drugs), and less likely to get needed care. One strategy that appears to ameliorate some, but not all of these problems is the provision of information about what will happen when a plan leaves and what, if any, options are available. However, the beneficiaries most likely to report they did not have enough information were those with the least education, the very old and the disabled, i.e., the beneficiaries who are probably least able to process additional information. Consequently, the problem may not have been a lack of information, but beneficiaries may have found it to be too much or too confusing. Providing more information in the future to them without simplifying it may actually be counterproductive. Furthermore, as an anonymous reviewer pointed out, it seems that, no matter how hard agencies such as CMS try to provide simplified material, there is still going to be a population that does not understand their choices or the implications of these choices while others are simply not motivated enough to try to understand. Consequently, rather than simply providing more information, descriptions of the specific implications of plan withdrawals on beneficiaries and their options for coverage should be tailored to meet the specific needs and interests of these vulnerable subgroups.