Developing Medicare and Medicaid Substance Abuse Treatment Spending Estimates

This is a supplemental report to the final report of a study jointly funded by ONDCP and ASPE (Medicaid Substance Abuse Treatment Spending: Findings Report). This technical report provides more detail on the methods used to make estimates, and it also describes how MPR would make similar estimates of Medicare funding for substance abuse treatment, if requested to do so. [29 PDF pages]

Introduction

As federal and state substance abuse (SA) agencies work to establish priorities and coordinate their efforts, policymakers need reliable national and state estimates of Medicaid and Medicare SA treatment spending and accurate methods for projecting these estimates forward. Spending estimates and projections are essential for both aligning funding with policy objectives and developing realistic budgets to support treatment and prevention. Given these needs the Office of the Assistant Secretary for Planning and Evaluation in the U.S. Department of Health and Human Services (HHS) and the Office of National Drug Control Policy (ONDCP) contracted with Mathematica Policy Research to conduct this of Medicare and Medicaid SA treatment spending. This study included both methodological and programmatic issues. This technical report addresses two methodological issues:

• What methods should be used to develop estimates of Medicare SA treatment spending?

• How should estimates of Medicaid SA treatment spending be projected forward when estimates derived from observed data are not available to policymakers?

A separate findings report produced under this project presents the programmatic findings of this study and addresses methodological issues related to estimating Medicaid SA treatment expenditures in Medicaid Analytic eXtract (MAX) data.

This report contains two sections. The first section presents an approach for developing estimates of SA treatment spending in the Medicare program. The second section presents methods for trending estimates of Medicaid SA treatment.

I. Proposed Methodology for Producing Medicare Substance Abuse Treatment Expenditures Estimates

In this section, we propose a methodology for developing estimates of Medicare SA treatment spending. The methods parallel the approach used to develop Medicaid spending estimates in this project. In Section A, we provide an overview of Medicare coverage of SA treatment services and the payment approaches Medicare uses to reimburse providers for these services. In Section B, we discuss Part A and B expenditures. In Section C, we propose methods for imputing SA treatment expenditures for Medicare Advantage (MA) managed care beneficiaries, based on fee-for-service (FFS) spending on SA treatment services and encounter data (when those data become available for 2012 and later years). Finally, in Section D, we discuss Part D, prescribed drug expenditures.

A. Overview of Medicare Coverage of SA Treatment

In this section, we first describe the SA treatment services covered under the Medicare program. Then, we describe the approaches Medicare uses to reimburse providers for these services.

1. Service Coverage

Medicare covers a full spectrum of SA treatment services:

• Inpatient hospital care--Medicare covers medically necessary inpatient hospital care for SA with the same coinsurance levels (for example, $1,156 deductible for 2012) and length of stay restrictions (up to 90 days in a benefit period, with a one-time 60-day reserve) as other types of hospitals stays. Medicare covers only a total of 190 days spent in a psychiatric hospital for an entire lifetime. Medicare may cover further inpatient mental health care in a general hospital, but not a psychiatric hospital.

• Outpatient treatment--Medicare Part B covers outpatient SA treatment. Visits to a doctor or other health professional to diagnose a SA disorder are covered with a 20 percent coinsurance amount. Outpatient treatment of SA is covered with a 40 percent coinsurance in 2012. This coinsurance amount will decrease until 2014, when it will be 20 percent. Partial hospitalization is covered by Medicare if inpatient treatment would be necessary otherwise.

• Preventive treatment--In October 2011, Medicare Part B initiated a new preventive benefit related to SA. This comprises an annual screening for alcohol misuse and up to four counseling visits to reduce alcohol misuse. Medicare fully covers these services from providers who accept assignment.

• Prescribed drugs--Medicare Part D covers prescribed drugs for SA treatment at the same coverage levels as other prescribed drugs.

It should be noted that Medicare coverage of SA treatment is increasing from 2010 through 2014 including the reduction in coinsurance for outpatient services to 20 percent by 2014. This increase in coverage results from the implementation of mental health parity and the addition of preventive alcohol use screening in 2011. These enhanced benefits are likely to result in a higher use rate and greater expenditures for SA treatment over time.

2. Payment Method

Medicare provides Part A and B services to beneficiaries through the traditional Medicare FFS or MA managed care plans. When beneficiaries elect to enroll in MA, Medicare pays health plans chosen by beneficiaries a monthly premium to manage their care. For beneficiaries covered under the traditional Medicare FFS program, Medicare maintains administrative data on the services received and the associated Medicare payment amounts. However, until 2012, for beneficiaries enrolled in MA, Medicare maintained administrative data only on premium payment amounts but did not require the MA plans to submit data on the services they provided to their enrollees. Thus, through 2011 for individuals enrolled in FFS Medicare, Medicare administrative data can be used to calculate the number of beneficiaries using SA treatment services and expenditures related to these services. However, for individuals covered under MA, Medicare administrative data cannot be used to analyze service use.

Beginning in 2012, MA plans are required to submit encounter data enumerating the services provided to MA enrollees. It is expected that initial submissions may be incomplete or that other data quality issues may exist (for example, non-uniform coding across plans). The quality of the MA data likely will improve over time, but the analysis of the MA encounter data likely will be more resource intensive in the initial years, as it will require assessment of reporting quality and adjustments to allow for incomplete, inconsistent, or inaccurate reporting. Analyses based on the initial encounter data are also likely to be less accurate. Thus, it likely will be necessary to allow time for the Centers for Medicare and Medicaid Services (CMS) to assess encounter data reporting and work with plans to improve such data before undertaking research using them. Thus, in the near future, we recommend that estimates of Medicare SA spending not rely on MA encounter data reporting and instead use FFS data to impute SA spending among MA enrollees.

Medicare prescription drug coverage, known as Medicare Part D, is provided through private companies approved by Medicare. Medicare pays these plans a monthly premium amount for each Medicare beneficiary covered. Utilization of Part D services is recorded in the CMS administrative data. These utilization data can be used to estimate SA related expenditures for prescribed drugs under Part D.

Table 1 displays the number of MA enrollees in each state and the District of Columbia for the last five years. MA enrollment has been increasing at a rate of 9.4 percent annually, with the rate of increase varying across states. Table 2 displays MA enrollment as a share of overall Medicare enrollment by state for 2011. Overall, almost one-quarter of Medicare enrollees are in an MA plan. Since MA enrollees represent a significant share of overall Medicare enrollment, the expenditures estimated for this population will represent a significant share of the overall estimate of SA treatment spending in Medicare.

B. Part a & B FFS Expenditures

Medicare's Chronic Condition Warehouse (CCW) can provide claims data to estimate total Medicare SA treatment spending for traditional Medicare FFS beneficiaries. Data files with 100 percent of FFS beneficiaries are available, as well as sample files. However, use of SA treatment is rare in the population ages 65 and older. Thus, there may be an insufficient sample of users to develop precise estimates in smaller states if a sample file is used. However, the 5 percent sample should provide sufficient precision for national estimates of SA treatment expenditures.

We recommend using the CCW Institutional and Non-Institutional claims files to estimate SA treatment spending in institutional and non-institutional settings, respectively. In parallel to the analysis of Medicaid SA treatment spending, we recommend that SA expenditures be divided into the following six types:

• Core--Services included in the Substance Abuse and Mental Health Services Administration's (SAMHSA) definition of SA treatment.

• Fetal exposure--Medical services primarily resulting from fetal exposure to alcohol or drugs. We expect that few of these services will be identified in the Medicare population, however, they are included as a separate category to parallel the Medicaid estimates.

• Poisoning--Medical services primarily resulting from poisoning by alcohol or drugs.

• Supplemental--Medical services primarily related to medical conditions fully attributable to alcohol or drug use.

• Mental health claim with secondary SA diagnosis--Individuals with an SA disorder often have a co-morbid mental health condition. Thus, it will be desirable to identify claims with a primary mental health diagnosis and a secondary SA diagnosis.

• Other claim with secondary SA diagnosis--Expenditures on these claims are related primarily to a medical condition other than SA; however, the SA co-morbidity increases the cost of this care to the Medicare program.

SA treatment claims within each of these types will be identified based on diagnosis code. Appendix A, Table A.1 and Table A.2, list diagnosis codes indicating SA treatment. For the first four SA treatment types only the first listed diagnosis on the Medicare claim should be examined. The final column of the Table A.1 and Table A.2 identifies which of these four SA treatment types is associated with each diagnosis code (Core, Fetal, Poisoning, or Supplemental). Claims will be assigned to the fifth group if they have a first listed mental health diagnosis, including any of the codes listed in Table A.3, and a secondary or later diagnosis listed in Table A.1 and Table A.2. All other Medicare claims with a secondary or later diagnosis in Table A.1 and Table A.2 will be classified into the sixth category.

Once claims with an SA diagnosis are identified, the total Medicare payment amount will be summed across the claims to determine the Medicare expenditures for these services.

C. Part C: Medicare Advantage

Because encounter data reporting is not mature for MA enrolled beneficiaries, we propose initially estimating SA treatment expenditures for the MA population based on the average level of expenditures for non-MA enrolled Medicare beneficiaries with similar characteristics. As complete encounter data become available for MA enrollees, the second approach, described below, entailing an estimation of the value of the utilization represented in the encounter data, could be used.

1. Imputation Based on FFS Experience

The following steps can be used to impute SA treatment expenditures for the MA enrolled population:

• Step 1: Develop homogenous tiers--The Beneficiary Summary File can be used to divide Medicare enrollees into two groups based on whether they were enrolled in an MA plan at any point in the year. Within each of these two groups beneficiaries then should be divided into tiers based on their personal characteristics, including age, gender, state of resident, Medicare status code (aged, disabled, End Stage Renal Disease), and dual-eligible status.

• Step 2: Calculate mean FFS expenditures per enrolled month--Calculate mean SA treatment expenditure per enrolled month among Medicare beneficiaries never enrolled in MA during the year for each tier created in Step 1 for each of the six SA treatment service types listed in Section B above.

• Step 3: Calculate total MA enrolled months--For each tier in Step 1, calculate the total number of MA enrolled months.

• Step 4: Calculate total imputed MA SA expenditures--For each SA service type, multiply the estimated mean FFS expenditure per enrolled month from Step 2 times the total number of MA enrolled months in Step 3 for each tier and sum across the tiers to obtain the total imputed MA SA expenditures for the service type.

This approach adjusts for differences in cost related to beneficiary characteristics observable in the administrative data. However, unobservable characteristics of MA enrollees may influence their treatment use. For example, wellness benefits offered by MA plans may appeal to healthier Medicare beneficiaries. Meanwhile, sicker beneficiaries may be less likely to make the effort to sign up for MA and may desire the broader choice of providers obtainable under traditional Medicare. Adjustment differences in these two groups of beneficiaries not linked to characteristics observable in the administrative data could be assessed by using multiple years of Medicare data and assessing whether MA enrollees had a lower SA treatment use rate prior to enrollment in MA relative to their counterparts who chose to remain in traditional FFS Medicare.

2. Imputation Based on Encounter Data

Once encounter data reporting is mature, encounter claims for SA treatment can be identified based on diagnosis codes in a manner similar to the identification of FFS SA treatment claims, as described in Section B. The encounter claims may not include accurate information on payment amount. If accurate information on the payment amount is not available from the encounter claims, we recommend estimating the price per unit of service based on mean expenditures per unit of service among traditional Medicare beneficiaries receiving the same service types. We also recommend that inpatient and other institutional care be priced per treatment day. Table A.5 can be used to classify outpatient treatment visits into homogeneous service types. The units of service observed in the encounter data then would be multiplied by the unit prices from traditional FFS Medicare and the total expenditures summed across the services types to determine total expenditures.

D. Part D: Prescribed Drug Expenditures

Expenditures for prescribed drugs can be estimated based on the CCW Part D Event file. Prescribed drug expenditures are categorized as core SA treatment services. The Part D Event file includes event records for both traditional Medicare FFS and MA enrollees, as well as the variable product service identification number, which provides the National Drug Code (NDC) of the prescribed drug. Table A.4 lists the NDC codes of pharmaceuticals used to treat SA. Only the first 8 digits of the 11-digit NDC code are used to identify the SA treatment pharmaceuticals. Event records listing these codes should be obtained. The values for the covered D plan paid amount and the non-covered plan paid amount are included in the Part D Event file. However, these values typically cannot be used for research. If they are available to this study then they should be summed to estimate payments under Part D. If they are not available national average payment amounts can be used to value the utilization observed in the Part D Event file.

II. Proposed Methods and Resources Required for Updating Medicaid Estimates with Future Years of Data

ONDCP develops the annual National Drug Control Strategy Budget Summary and so needs an approach for updating the estimates of Medicaid SA treatment spending annually. We recommend direct derivation of the estimates of SA treatment spending from MAX files every 5-7 years or after major policy changes take place, such as implementation of the Affordable Care Act. For the intervening years, we recommend projecting base year estimates forward using the CMS-64 reporting and Medicaid enrollment data. The CMS-64 report summarizes annual Medicaid expenditures for each state. Information from the forms is currently available through FY 2011 for each state by service category. The benefit of using the information from the CMS-64 reports is that the available trends are state and service specific, and are available with only limited lag. One drawback to the CMS-64 data is that they are not broken out by eligibility group. Another limitation is that they are based on date of payment, therefore lump sum adjustments or lags in payment processing can impact trends artificially. Since SA treatment utilization varies by age, gender, and eligibility, we propose using Medicaid enrollment information in combination with the CMS-64 data.

We propose using the following steps to develop projections for each state:

• Step 1: Obtain data. The CMS-64 reports, containing cost and service data, can be downloaded from the CMS website.¹ Medicaid monthly enrollment data for December and June can be obtained from the Kaiser Commission on Medicaid and the Uninsured website.² Data currently are compiled through June 2011. Projections of SA treatment spending by SA care group and type of SA Spending for FY 2011 are provided in Appendix B. These estimates would be trended forward until updated estimates are directly derived from MAX.

• Step 2: Map CMS-64 service categories to the five SA care groups. Map the service categories available in CMS-64 data to the five SA care groups included in the FY 2011 projections (Table B.1) developed from MAX data for the base year of the data (FY 2011 for the first round) and the desired year of the estimates. Map services in two steps. (a) First, group together individual categories from the CMS-64 into broader service groups shown in Table 3. Include relevant “C” categories, which represent expenses reported for Medicaid expansion Children's Health Insurance Program (CHIP) populations, and “T” categories, which account for expenses in those states that qualify to expend 20 percent of their CHIP allotment on the Medicaid program and still receive the enhanced CHIP match rate. Table 3 shows the groups by care setting developed based on the 2011 CMS-64 categories. (b) Next, crosswalk these groups to the five SA care groups included in the FY 2011 projections. Table 4 shows this crosswalk.³

• Step 3: Estimate overall Medicaid expenditures. Estimate the overall level of Medicaid spending represented in the CMS-64s for each state in each SA care group in the base year and the most recent year of CMS-64 data available. Also, estimate the overall level of Medicaid spending for the state in the base year and the most recent year of data available.

• Step 4: Calculate spending per enrolled month. Divide the total Medicaid spending for each state in each SA care group and overall by 12 times the number of Medicaid enrolled months in the state in June of the given year. The June months are multiplied by 12 to represent a full year of enrollment.

• Step 5: Estimate the trend in spending per enrolled month. Estimate the overall Medicaid expenditure trend for each state for each SA care group and overall from the base period through the most recent year available.⁴ In rare cases, where trends for a particular service category indicates more than a 35 percent increase or decrease, replace the service category-specific trend with the overall trend in state Medicaid spending.

• Step 6: Adjust overall Medicaid general health expenditure trends for the historical difference in growth between SA treatment and general health care spending. Between 1986 and 2005, the estimated trend in Medicaid SA treatment spending based on the SAMHSA Spending Estimates (SSE) was 98 percent of the National Health Expenditure Accounts (NHEA) estimated trend in Medicaid spending. Given that the rate of growth in SA treatment expenditures (as identified in the SSE) historically has fallen below that of general health care expenditures (as identified in the NHEA), multiply the annual trend estimates developed in Step 5 by 0.98.

• Step 7: Estimate the trend in adult/disabled enrollment. Medicaid expenditures on SA are concentrated in the adult/disabled populations. Ideally, expenditure trend estimates specific to SA treatment services and Medicaid adults would be developed, however, since the CMS-64s are not developed by eligibility nor demographic group and do not include categories tailored to SA, the method proposed in Steps 1-6 uses the expenditure trends per enrolled month across all Medicaid enrollees in each SA care group as a proxy for trends in SA expenditures which are primarily for adults. Since enrollment information is available by eligibility group in this step the Medicaid enrollment information obtained from the June Kaiser enrollment report should be limited to the adult and disabled populations. Then, the enrollment trend between the base and the most recent year of data available should be calculated.

• Step 8: Project the base period MAX estimates to the most recent year of data available. Multiply the base year SA spending in each SA care group by the trend in expenditures per enrolled month for the group, adjusted for the historical difference between the general health and SA expenditure trends (the latter is 98 percent of the former). Then multiply this product by the trend in enrollment for the adult and disabled population for the same period. Estimates for the non-core SA treatment expenditures reported in Table B.2 were not developed by type of care, so they cannot be trended by SA care groups. Instead, trend these estimates based on the overall spending per enrolled month trend for Medicaid times 0.98 times the enrollment trend for the adult and disabled population.

• Step 9: Project beyond the observed data. If projections beyond the most recent period of available CMS-64 and Kaiser enrollment data are needed, assess available information on the likely trend in Medicaid expenditures. Although the NHEA projections largely reflect trends in services other than SA treatment, these trends incorporate both anticipated enrollment changes and change in economic conditions. Since the SA treatment expenditure trend is 98 percent of the general health trend in the historical data, 98 percent of the projected increase in Medicaid spending in the NHEA projections could be used to project SA treatment expenditures beyond the period for which observable data is available. Another approach would be to assume that the trend observed in CMS-64 and Kaiser enrollment data for the recent historical period will continue. If there are no economic or policy-related factors that caused shifts in the recent past or are likely to cause a substantial shift in the future trends, then this approach would be reasonable. Under this approach, annualize the trend in expenditures per enrolled month between the base year and most recent year of data available as calculated in Step 5. Multiply the annualized trend by 0.98 reflecting the historical difference between growth in SA treatment spending and overall health care spending. Then annualize the enrollment trend between the base year and the most recent year of data available as calculated in Step 7. Multiply the projection for the most recent period of CMS-64 data available as calculated in Step 8 by the annualized trend in expenditures per enrolled month and the annualized trend in enrollment in the historical data. Repeat this multiplication for each additional year of trend desired.

Notes

1. Available at https://www.cms.gov/Research-Statistics-Data-and-Systems/Computer-Data-and-Systems/MedicaidBudgetExpendSystem/CMS-64-Quarterly-Expense-Report.html. Accessed July 26, 2012.

2. Available at http://www.kff.org/medicaid/enrollmentreports.cfm. Accessed July 26, 2012.

3. Note that we found some differences between the CMS-64 categories for 2008 and 2011. Some new categories were added to the CMS-64 in 2010 and 2011 that were not included in the CMS-64 in prior years. If the CMS categories change again in the future, it will be necessary to map the categories in the CMS-64 in the base and estimate year as closely as possible so that trends can be developed.

4. Estimate the trend as T_t = (S_t/S_b)^(1/(t-b)) . Where T is the trend, S is Medicaid spending per enrolled month from Step 4, t is current year, and b is base year.

Appendices

Appendix A: Diagnosis Codes

Appendix B. FY 2011 SA Treatment Spending