CMS-10166.Final Supporting Statement-8-23-05

Supporting Statement for the Payment Error Rate Measurements in Medicaid and the State Children’s Health Insurance (SCHIP) Program

A. Background

The Improper Payments Information Act (IPIA) of 2002 requires CMS to produce national error rates for Medicaid and State Children’s Health Insurance Program (SCHIP). To comply with the IPIA, CMS will engage a Federal contractor to produce the error rates in Medicaid and SCHIP. Initially, based on states’ annual medical expenditures from the previous year, the Federal contractor will group all states into three equal strata of small, medium and large medical expenditures. A statistically valid random sample of states will be drawn from each strata. However, CMS may revise its sampling methodology and may use a methodology to select states that will ensure that each state is selected at least every three years but that no state is sampled more than once every three years. The error rates produced by this selection methodology will provide the state with a state-specific error rate estimated to be within 3% precision at the 95% confidence level.

The states selected for review will submit the previous year’s claims data and expenditures on which the contractor will determine each State’s sample size and the sample size for each strata (i.e., claims are stratified by service). These states also will submit quarterly claims data to the contractor who will pull a statistically valid random sample, each quarter, by strata, so that medical and data processing reviews can be performed. State-specific error rates will be based on these review results.

CMS needs to collect the claims data, annual expenditures, medical policies, and other information from states as well as medical records from providers in order for the contractor to sample and review adjudicated claims in those states selected for review. Based on the reviews, State-specific error rates will be calculated which will serve as the basis for calculating national Medicaid and SCHIP error rates.

B. Justification

1. Need and Legal Basis

The collection of information is necessary for CMS to produce national error rates for Medicaid and SCHIP as required by Public Law 107-300, the IPIA of 2002. A copy of the Act is attached.

2. Information Users

The information collected from the states selected for review will be used to conduct claims reviews on which state-specific error rates will be calculated. The annual claims data expenditures for the previous fiscal year submitted by the states will be used by the Federal contractor to determine each state’s sample size and for each strata (i.e., claims are stratified by service). The current fiscal year’s quarterly claims data will be used to sample claims for reviews. The medical policies will be used by the contractor to guide the medical review of the claims. Providers within the selected states whose claims were sampled for review will submit medical records on which the medical reviews also will be based. The review findings will be used to calculate state-specific error rates on which national error rates for Medicaid and SCHIP will be calculated.

3. Use of Information Technology

This information collection involves the use of electronic submission of information to the extent that states have the technological capability. CMS will not require states or providers to provide information electronically if they do not have secure systems in place to do so. While most states have claims information electronically, some states will likely submit information regarding claims in a hard copy format such as a tape. The percentage expected to be received electronically is less than 1%. The collection of information does not require a signature from the respondents.

4. Duplication of Efforts

This information collection does not duplicate any other effort and the information cannot be obtained from any other source.

5. Small Businesses

The collection of information does not impact small businesses or other small entities.

6. Less Frequent Collection

Failure to acquire this certification form will prevent CMS from effectively measuring state-specific payment error rates on which to base national error rates for Medicaid and SCHIP. Consequently, CMS will not be able to produce error rates in a timely manner and would cause CMS to be out of compliance with IPIA.

7. Special Circumstances

CMS does not anticipate that states would be required to submit information more often than quarterly. States will provide annual expenditures once a year and quarterly claims data at the end of each quarter. States will also be required to submit medical policies at the beginning of being selected and updates on a quarterly basis at the end of each quarter.

8. Federal Register/Outside Consultation

The first 30-day Federal Register notice published on July 22, 2005.

9. Payments/Gifts to Respondents

There is no provision for any payment or gift to respondents associated with this reporting requirement.

10. Confidentiality

Confidentiality has been assured in accordance with Section 1902(a) (7) of the Social Security Act.

11. Sensitive Questions

No questions of a sensitive nature are asked.

12. Burden Estimate (Total Hours & Wages)

The number of respondents is estimated to be up to 36 states (18 Medicaid and 18 SCHIP states). The annualized number of hours that may be required to respond to requests for information equals 58,680 hours ( 1630 hours per state, per program) at a GS-12, step one rate of pay for a total cost of $ 1,524,506.

18 states will be selected at random to participate in the Medicaid error rate measurement project and 18 states in the SCHIP error rate measurement project. Even if the same 18 states were randomly selected for both Medicaid and SCHIP, requests for information would require two responses from each of the 18 states, one response for Medicaid and one for SCHIP. Since each program will be sampled separately, it is possible that 36 unique states could be sampled. Therefore, estimates were calculated for 36 responses to each request for information.

It is estimated that each state will spend up to 1630 hours of time annually, per program, to support this collection of information. The states will provide reports on a quarterly basis for the following requests, per program:

1. FY Quarter 1, Expenditure data for the previous fiscal year (up to 10 hours per request), adjudicated and stratified claims data for the current quarter (up to 210 hours), policy in effect at the time of state selection (up to 10 hours) and policy updates at the end of the current quarter (up to 10 hours). This would be four (4) total responses per state, per program for an estimated time of 240 hours per state, per program.

2. FY Quarters 2, 3, and 4, Stratified claims data and policies for the current quarter. This would be 2 total responses each quarter, per program, for a total estimated time of 220 hours per state, per program, per quarter.

3. States will also be required to provide ad hoc information to:

   1. Re-price claims determined to be in error for a total of 200 hours per state, per program. Errors are expected in less than 10% of sampled claims or < 100 requests at an estimated time of two hours per request;

   2. Inform the contractor of claims that were included in the sample but the adjudication decision changed due to the provider appealing the determination and the state overturning the claim, for a total of 20 hours per state, per program. (The appeals process is part of the state’s customary and usual course of business.) Adjudication decision changes are expected in less than 2% of sampled claims or < 20 requests and an average of one hour per request; and

   3. Inform the contractor of provider enrollment information to assist in finding a provider associated with a sampled claim so that documentation can be obtained for a total of 10 hours per state, per program. It is expected that the contractor use common resources such as the internet, the phone book, and directory assistance before consulting the state. Erroneous provider demographic information, where resources other than the state are not available, is expected rare and estimated in less than 1% of sampled claims or < 10 requests.

4. States will also be required to prepare and submit corrective action plans after error rates are determined for each program. This will be a single submission in the third year after state selection. Example, Maryland is selected in FY 2006, the error rate can be calculated and reported only after FY 2006 is concluded or in FY 2007. The corrective action plan is completed in FY 2008. The total estimate for preparing a corrective action plan is 500 hours per state, per program beginning in the third year of this project.

It was determined that the request for medical documentation to substantiate claim submission is not a burden to individual providers nor is the request outside the customary and usual business practices of a Medicaid and/or SCHIP provider. It is highly unlikely for a provider to be selected more than once, per program, per year to provide supporting documentation and due to the timeliness of the request for documentation, that information should be readily available and responses should take minimal time. Therefore, this request for information from providers is within the customary and usual business practice of a provider who accepts payment from an insurance provider whether it is a private organization, Medicaid or SCHIP.

The following assumptions were used:

• The estimated number of states needed to produce a national error rate with the confidence and precision to meet the IPIA is up to 36 annually; 18 for Medicaid and 18 for SCHIP.

• The estimated number of claims needed from each state to produce a state specific error rate with the confidence and precision needed to meet IPIA standards is estimated to be 1,000 per program.

• These 1,000 claims are going to be further stratified, based on annual expenditures, in approximately eight (8) strata: 1) inpatient hospital, (2) long term care, (3) practitioners and clinics, (4) pharmacy, (5) home and community-based services, (6) other services and supplies, and (7) fixed payments such as Medicare Parts A and B premiums) and (8) denied claims.

• The 1,000 claims will be sampled over a full fiscal year of adjudicated claims by sampling a weighted number of claims each quarter, with the weight determined by quarterly expenditure data.

12. Capital Cost

There are no capital costs associated with this collection of information.

12. Cost to the Federal Government

The Federal Government is going to engage a national contractor to determine the error rate for both Medicaid and SCHIP. The estimated cost claim per review is $360. The estimated number of states is 18 and number of claims per state is 1,000. In addition to the cost for review, the Federal government also included $260,000 per year for the contractor’s administrative and travel expenses. Thus the total per state is $620,000 per program (Medicaid and SCHIP).

12. Changes to Burden

This is a new requirement and this is not applicable.

12. Publication/Tabulation Dates

The calculated national error rate for both Medicaid and SCHIP will be published annually in the Performance and Accountability Report (PAR).

12. Expiration Date

This collection does not lend itself to the displaying of an expiration date.

12. Certification Statement

There are no statistical aspects of the certification form.

C. Collections of Information Employing Statistical Methods

1. The universe for this project is the 50 states’ and the District of Columbia’s Medicaid and SCHIP programs. The states will be stratified, in each program, according to annual medical expense expenditures, into three equal strata: small, medium, and large expenditures. Six (6) states will be randomly selected from each stratum.

The potential number of respondent universe is 36 unique states (18 states’ Medicaid and 18 states SCHIP). It is estimated that 1,000 claims will then need to be randomly selected, from each state in each program, to achieve a state specific, program specific error rate. These results will be used to calculate a national error rate in compliance with IPIA.

Number of States (51 total universe, per program)	Small Strata	Medium Strata	Large Strata
Total Universe	17	17	17
Random Sample	6	6	6
Expected Response-Medicaid	6	6	6
Expected Response-SCHIP	6	6	6

The anticipated response rate is 100% due to the statutory requirements at section 1902(a)(6) of the Act and section 2107(b)(1) of the Act that require States to provide information necessary for the Secretary to monitor program performance. Section 1902(a)(27) of the Act requires providers to retain records necessary to disclose the extent of services provided to individuals receiving assistance and furnish the Secretary with information regarding any payments claimed by the provider for furnishing the services as the Secretary may request.

2. States will be stratified by program by annual expenditures with three equal strata of 17 states each. Note, since stratification is based on expenditures, the same states may not be in the same stratum for each sample. A random selection of six (6) states will be selected from each strata for each program.

We determined a claim sample size between 800-1200 (averaged to 1,000) of 1,000 per state using an assumed error rate of 6.8%. This error rate was based on the Payment Accuracy Measurement pilot project results from FY 2004. In order for a state to determine vulnerabilities and have a meaningful corrective action plan, the program level error rate for each state will be estimated to achieve a 95% confidence level within 3% precision.

In order to meet the requirements of IPIA, all selected states must participate.

3. We expect to get reliable data because the states are reporting in the aggregate and the national contractor is completing the sampling and stratification of claims and expenditure data. This method was tested in a pilot project in previous fiscal years.

4. Not applicable.

5. The Lewin Group (Paul Hogan 703-269-5545 and Brian Simonsen 703-269-5702) consulted on the statistical methodology of this project.

7