CMS-10316 - Supporting Statement part A [rev 7-8-2013]

SUPPORTING STATEMENT PART A

Implementation of the Medicare Prescription Drug Plan (PDP) and Medicare

Advantage (MA) Plan Disenrollment Reasons Survey

Background

The Centers for Medicare & Medicaid Services (CMS) requests clearance from the Office of Management and Budget (OMB), under the Paperwork Reduction Act of 1995, for the Medicare Prescription Drug Plan (PDP) and Medicare Advantage (MA) Plan Disenrollment Reasons Survey.

Under contract to CMS between 2009 and 2012, the RAND Corporation developed and tested a Medicare Part D Disenrollment Survey to capture the reasons why beneficiaries disenroll from their MA-PD and PDP contracts. This was the first time that a disenrollment survey was conducted on the Medicare Part D population. The survey focused only on beneficiaries who voluntarily disenrolled from their PDP or MA-PD between November 2010 and July 2011. The survey excluded beneficiaries who were involuntarily disenrolled from contracts because of eligibility reasons, movement out of the Part D contract’s service area, or death. This initial survey effort served as a large- scale field test of methods and to shed light on some of the most important reasons for beneficiary disenrollments. The sampling approach used in the first fielding of the

survey was not designed to produce contract-level estimates for every PDP and MA-PD contract; however, RAND did produce contract-level estimates for the contracts in its sample to assess the number of beneficiaries that would need to be sampled in future survey iterations to generate reliable estimates for all Part D contracts. Through this work, RAND identified several improvements for future iterations of the survey,

_including

• refinements to survey wording regarding contract name recognition,

• efficiencies in the administration of the survey by eliminating the phone interview survey mode (which yielded little in terms of response rates), and

• efficiencies in the total sample needed to generate reliable contract-level estimates of reasons for disenrollment for both MA-PD and PDP contracts (i.e., plans).

Under the follow-on contract (2012 through 2014), RAND will refine and implement this Disenrollment Survey to capture the reasons for beneficiary disenrollments from Medicare Advantage (MA and MA-PD) plans and stand-alone Prescription Drug Plans (PDPs). Implementation of the survey for 2012 through 2014 will differ from the earlier large-scale field test, which excluded the small number of MA only plans. RAND will sample and survey beneficiaries who disenroll from MA plans that do and do not offer prescription drug coverage, as well as beneficiaries who disenroll from stand-alone PDP plans. As with the first survey, this follow-on survey will exclude beneficiaries who were involuntarily disenrolled due to eligibility reasons, moved out of the service area, or died. In addition, in the 2012-2014 survey work, the RAND team will produce individual health plan contract reports, which will assist contracts in their quality improvement efforts. RAND will also develop data analysis procedures and formats for possible future public reporting by CMS of the disenrollment reasons information.

A. Justification

A1. Need and Legal Basis

The Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA) provides a requirement to collect and report performance data for Part D prescription drug plans. Specifically, the MMA under Sec. 1860D-4 (Beneficiary Protections for Qualified Prescription Drug Coverage) requires CMS to conduct consumer satisfaction surveys regarding PDPs and MA-PDs – pursuant to section

1860D-4(d).

A2. Information Users

This data collection complements the satisfaction data collected through the Medicare Consumer Assessment of Healthcare Providers and Systems (MCAHPS) survey by providing dissatisfaction data in the form of reasons for disenrollment from a Prescription Drug Plan.

The data collected in this survey can be used to improve the operation of Medicare Advantage (both MA and MA-PD) contracts and standalone Part D Plans (PDP) through the identification of beneficiary disenrollment reasons. Plans can use the information to guide quality improvement efforts. The data can also be used by beneficiaries who need to choose among the different MA and PDP options. To the extent that these data identify areas for improvement at the contract level they can be used to inform CMS contract oversight.

A3. Use of Information Technology

The survey vendor will collect the data via a mail data collection strategy that involves two rounds of mailed surveys. The mailed survey will be formatted for data scanning, and data from all returned surveys will be scanned into an electronic data file.

A4. Duplication of Efforts

A survey for individuals disenrolling from a Medicare managed care plan was last fielded in 2011. No standardized survey for disenrollees from Medicare Part C or Part D plans is currently in use.

A5. Small Businesses

Survey respondents are disenrollees from Medicare Prescription Drug Plans (PDPs), and Medicare Advantage plans (both MA-only plans and MA-PD plans). The survey should not impact small businesses or other small entities.

A6. Less Frequent Collection

The consequence of not collecting data as soon as possible after a beneficiary disenrolls from a health or prescription drug plan is that the beneficiary will be less able to recall their specific reasons for disenrolling from a health or Medicare Part D prescription drug plan and their experiences under their previous plan, information that is critical for Medicare health plan and Part D program improvement. PDP and MA plans (both MA only and MA-PD) frequently make changes to the types of medications covered, to beneficiary costs, and to other plan features that impact beneficiaries. As such, it is important to conduct annual assessments of the performance of the plans from the perspective of the beneficiary.

A7. Special Circumstances

None of the special circumstances described on Form OMB 83-I are applicable to this survey.

A8. Federal Register/Outside Consultation

The 60-day Federal Register notice published on April 19, 2013 (78 FR 23566). Public comments were received and our response has been added to this package.

A9. Payment/Gifts to Respondents

This data collection will not include respondent incentive payments or gifts.

A10. Confidentiality

Individuals contacted as part of this data collection will be assured of the confidentiality of their replies under 42 U.S.C. 1306, 20 CFR 401 and 422, 5 U.S.C. 552 (Freedom of Information Act), 5 U.S.C. 552a (Privacy Act of 1974), and OMB Circular A-130.

A11. Sensitive Questions

The survey does not include any questions of a sensitive nature.

A12. Burden Estimate (Hours & Wages)

Exhibit 1 shows the estimated annualized burden for the respondents' time to participate in this data collection. The Medicare Prescription Drug Plan (PDP) and Medicare Advantage (MA) Plan Disenrollment Reasons Survey will be administered to 176,985 beneficiaries in 2013 and 2014 using three survey versions: 1) Stand Alone Prescription Drug Plan or PDP version (sample size approximately 39,582); 2) Medicare Advantage with Prescription Drug Plan Coverage or MA-PD version (sample size approximately 123,662); and 3) Medicare Advantage Only or MA-only version (sample size approximately 13,740) (see attachment 3, 4, and 5). We anticipate a response rate of approximately 50% based on the large-scale field test; therefore, we estimate that the PDP version will be completed by about 19,791 persons; the MA-PD version will be completed by about 61,831 persons; and the MA-only version will be completed by 6,870 persons. The estimated response time of 0.24 hours or 14 minutes for the PDP version of the survey is based on the length of that survey version, a pace of 4.5 items per minute, and CMS’ experience with surveys of similar length that were fielded with Medicare beneficiaries. Similarly, the estimated response time of 0.27 hours or 16 minutes for the

MA-PD version of the survey is based on the length that survey version, a pace of 4.5 items per minute, and CMS’ experience with surveys of similar length that were fielded with Medicare beneficiaries. Finally, the estimated response time of 0.21 hours or 12 minutes for the MA-only version of the survey is based on the length that survey version, a pace of 4.5 items per minute, and CMS’ experience with surveys of similar length that were fielded with Medicare beneficiaries. As indicated below, the total burden hours are estimated to be 22,887 hours.

_{Exhibit 1. Estimated annualized burden hours}

Survey Version		Number of Respondents	Number of responses per respondent	Hours per response	Total Burden hours
Medicare Disenrollee Survey, Stand Alone Prescription Drug Plan (PDP) Version		19,791	1	0.24	4,750
Medicare Disenrollee Survey, Medicare Advantage with Prescription Drug Plan Coverage (MA-PD) Version		61,831	1	0.27	16,694
Medicare Disenrollee Survey, Medicare Advantage without Prescription Drug Plan Coverage (MA-only) Version		6,870	1	0.21	1,443
	Total	88,492	1	-	22,887

Exhibit 2 shows the survey participants’ cost burden associated with their time to complete a survey. The total cost burden is estimated to be $447,670.

_{Exhibit 2. Estimated annualized cost burden}

Survey Version		Number of Respondents	Total Burden hours	Average Hourly Wage Rate*	Total Cost Burden
Medicare Disenrollee Survey, Stand Alone Prescription Drug Plan Version		19,791	4,750	$19.56	$92,910
Medicare Disenrollee Survey, Medicare Advantage with Prescription Drug Plan Coverage Version		61,831	16,694	$19.56	$326,535
Medicare Disenrollee Survey, Medicare Advantage without Prescription Drug Plan Coverage Version		6,870	1,443	$19.56	$28,225
	Total	88,492	22,887	-	$447,670

*Based upon the average wages, “National Compensation Survey: Occupational Wages in the United

States, May 2007,” U.S. Department of Labor, Bureau of Labor Statistics.

A13. Capitol Cost

We have no capital costs

A14. Cost to Federal Government

The total cost for design, data collection, analysis, and contract-level report production is $2,462,540.

A15. Changes to Burden

Implementation of the survey for 2012 through 2014 will differ from the earlier large-scale field test, which excluded the small number of MA only plans. The MA only plan survey has been added to this ICR.

In addition to adding the MA only survey, the number of responses for the PDP survey has decreased from 60,000 to 19,791 and the hours have decreased from 16,800 to 4,750 hr.

While the MA-PD survey has a slight increase in the number of responses from 60,000 to 61,831, the response time decreased from .3 hr to .27 hr and a number of hours has decreased from 18,000 to 16,694 hr.

Overall, this ICR reduces the number of respondents from 120,000 to 88,492 and reduces the burden hours from 34,800 to 22,887 hr.

A16. Publication/Tabulation Dates

We anticipate that the analysis plan will include (1) psychometric evaluation of the survey items, (2) development and evaluation of case-mix adjustment models and nonresponse weights, (3) development of adjusted MA-PD, MA-only, and PDP contract-

level results, (4) development of national, regional, and subgroup estimates, (5) analyses of the relationship between disenrollment rates at the contract level and evaluations of the prescription drug plan, (6) contract-level linkages to Medicare CAHPS Results, (7) developing consumer reporting displays, and (8) develop and implement data analysis procedures for public reporting. All aspects of these analyses will be described in a final project report to CMS.

(1) Psychometric Revaluation. RAND will update and verify these psychometric analyses in the new data where appropriate and conduct more complete analyses of any new items developed. Analyses will include evaluation of item missing data, item distribution (including ceiling and floor effects), and assessment of contract-level reliability of items. We will compute these statistics overall and separately by mode of administration, language, and MA-only, MA-PD, and PDP.

(2) Case-mix adjustment and nonresponse. RAND will begin by verifying the applicability of the MCAHPS model according to the criteria of exogeneity (only control for factors that are not a consequence of care), reliability (only adjust for factors that are precisely measured), and parsimony (only adjust for factors that meaningfully impact scores – i.e., scores that are both predictive of individual responses within contracts and which vary between contracts). Among other factors, the current model considers age, education, self-rated health status, and low income subsidy status.

(3) National, Regional, and Subgroup Estimates. RAND will use adjustments, as appropriate, to produce national and regional estimates of reasons for disenrollment. These models will update and build on analyses performed in the previous project and will test for evidence of trends. Hierarchical variance- component models will assess the extent to which variation in each measure reflects contracts within sponsors, sponsors, geography, and interactions between geography and contracts and sponsors. Those receiving the low income subsidy differ from other beneficiaries with respect to Part D in a number of important ways that include but are not limited to different prior experiences with prescription drug coverage, auto enrollment into Part D coverage, and different disenrollment timing options. In previous work, the RAND team has identified

racial/ethnic differences in experience with Part D coverage and variations in experiences for vulnerable subgroups of beneficiaries that include those receiving the low income subsidy, beneficiaries with no high school degree, disabled beneficiaries, those 85 and older. We propose to assess differences in the prevalence of reasons for disenrollment nationally across these subgroups and to use mixed random and fixed effect models to assess the extent to which any differences in experiences are consistent across contracts. Such analyses may identify subgroups and contracts which might especially benefit from CMS outreach and intervention. Additional analyses may consider the role of enrollment history as a predictor of disenrollment and may distinguish beneficiaries on the basis of whether they were changing between PDP and MA- PD system or within those systems. Particular attention will be devoted to distinguish the roles of Part D and non-Part D aspects of Medicare experiences in these decisions.

(4) Contract-Level Linkages to Medicare CAHPS Results. Substantial insight may be gained by linking disenrollment survey data to corresponding data from the Medicare CAHPS Health Plan survey at the contract level. The extent to which contract-level disenrollment rates for the applicable reasons and the reasons themselves do (or do not) correlate with contract-level CAHPS beneficiary assessments has important implications for the extent to which current CAHPS scores reflect the experiences of all beneficiary with a given contract and may influence CMS’ future interest in reporting disenrollment information. Analyses that compare a limited set of CAHPS Part D items from the disenrollment surveys to corresponding items as answered in the Medicare CAHPS survey for the same contracts could estimate disenrollment behavior as a function of beneficiary experiences in a manner as a complementary source of information to disenrollees’ directly reported reasons for disenrollment. Analyses that break this information down by beneficiary subgroup might determine whether different groups (those eligible for low income subsidy, racial/ethnic subgroups) use different criteria or thresholds for disenrollment decisions.

(5) Design and Produce Individual Plan Reports. We will design individual contract reports that will display the results of the survey from each contract’s own enrollees with comparisons to state, region and national estimates. We will provide both summary measures and drill-down item information to maximize information for quality improvement purposes. Graphical representations will be used to improve the ability to easily interpret the results. In presenting state benchmarks, we will likely need to combine data across states when data from a single state are based entirely on one plan, making the statewide average uninformative. We may also need to combine states when the sample size for the disenrollment reasons data is too small.

(6) Develop and Test Consumer Reporting Displays. The analyses will take the following general plan. We will first recode some variables. The CAHPS Macro, developed and maintained at Harvard, will be used for most comparative analyses, including case-mix adjustment (CMA), calculation of composite scores,

and statistical tests of above- or below-average performance. We will employ the analysis-macro feature that “smoothes” contract estimates of sampling variances toward a pooled variance estimate, to avoid implausibly small estimated contract standard errors (even 0) when by chance almost all of a contract’s respondents give the same answer. Within this process we may also further adjust PDP contracts’ scores for state effects using code written in the R programming language, which is also used for some output management. We also will calculate the reliability of each contract’s score, R=1−V/(V+T²), where T² is the between- contract model variance of the means and V is the variance of the estimate of the measure for that contract. Finally, we will compile for transmission to CMS the adjusted mean scores, and flags for “low-reliability” scores (those with both reliability < 0.75 and in the lowest 12% of contracts ordered by reliability) and “nonreportable” scores (those based on <10 respondents). We also prepare reports on distributions of sample sizes and reliabilities as an aid to assessing the quality of the data. In developing methods to publicly report the disenrollment reasons information, it is essential to assess how consumers understand the proposed groupings of reasons and the labels used to describe them. It is also important to assess their understanding and interpretation of the numerical information and/or graphic displays used to summarize the data. Comparison across plans requires first, comparing disenrollment rates across plans, and second, comparing the distribution of reasons for disenrollment across plans. Displays presenting both of these distinctive items of information can easily confuse consumers with low or even average numeracy skills, while displays that present only the distribution of reasons without the underlying rates are incomplete or misleading. We will approach this task by first conducting an environmental scan to seek examples of similar cases in which displays of “nested” information have been designed for a broad consumer audience.

Based on results of this scan, we will generate alternative approaches to presenting these data that can be tested on individual consumers. RAND will also draw upon our experience developing consumer reports in both English and Spanish as part of the CAHPS project. RAND will conduct two rounds of cognitive testing, with 9

English-speaking consumer participants and 9 Spanish-speaking consumer participants in each round (total n = 36), selected to be diverse in terms of age, education, and race/ethnicity. We propose to conduct the interviews in Los Angeles and in Baltimore so that CMS may observe the cognitive interviews. In order to ensure comparability, RAND proposes to test and then review and refine the Spanish version of displays at the same time as the English so that any changes required in the Spanish version can inform any other changes that may be required in the English version.

Exhibit 3 details the timeline for data collection, analysis and delivery of the analytic report.

_{Exhibit 3. Timeline}

Task	Planned Start Date	Planned End Date
Sample selection and file preparation	OMB approval	290 days after OMB approval
Data collection	30 days after OMB approval	384 days after OMB approval
Data analysis	230 days after OMB approval	420 days after OMB approval
Prepare and submit data analysis report	300 days after OMB approval	450 days after OMB approval

Publication of Results: CMS may confidentially share sponsor or contract-level disenrollment estimates with individual plan sponsors for quality improvement purposes. However, sponsor or contract-level disenrollment data from this survey will not be made publicly available to Medicare beneficiaries or the general public. CMS may present more general disenrollment data and patterns in a publicly available report format.

RAND will work with CMS to develop peer-reviewed publications to extend the impact of this work

A17. Expiration Date

The expiration date for OMB approval of this information collection will be displayed on the survey.

A18. Certification Statement

There are no exceptions to the certification statement identified in item 19 of OMB Form 83-I associated with this data collection effort.

1