The completeness and reliability of each states MAX data may vary from year to year. We examine the completeness and reliability of all key variables in the eligibility and claims files in each state using 2007 MAX data--the latest calendar year of data available at the start of our project. We identified 22 states with MAX data of sufficient completeness and reliability for inclusion in the study; below, we summarize the state selection process.
Given that the completeness and reliability of each states data may vary between years, we recommended limiting the study to populations enrolled in Medicaid in calendar year 2007 and the services they received during that year. The inclusion of other years of data would require additional project resources to investigate the reliability of the eligibility and FFS claims in each state in other years, to determine the completeness of the encounter data in states that used managed care in other years, and to evaluate outside sources to understand any potentially relevant policy or program changes between the years.
A. Populations Excluded from the Study
To maximize the number of states included in the study, we limited our analysis to adults who qualified for Medicaid on the basis of disability, because this population is often excluded from managed care programs into which other beneficiaries are required to enroll. In addition, we limited the analysis to community-dwelling disabled beneficiaries because of differences in the service delivery systems and thus state Medicaid policies affecting the institutionalized population.
The study excluded beneficiaries who qualified for only partial Medicaid benefits and beneficiaries with other sources of health insurance coverage (including those enrolled simultaneously in Medicare or private insurance). With much of such individuals’ service use not Medicaid-funded, the MAX data would present an incomplete picture of their conditions and treatments. As suggested by our expert panel, to ensure an adequate amount of data to evaluate service use, the analysis only included beneficiaries with at least 10 months of continuous enrollment in Medicaid.
Table A.1 summarizes the Medicaid populations to be excluded from the study due to the limitations of the MAX data.
B. State Selection Process
We examined three types of records to determine whether each state’s MAX data were sufficiently complete and reliable to identify beneficiaries with schizophrenia and bipolar disorder and to measure such beneficiaries’ use of Medicaid-funded services. First, we examined the enrollment records in the person summary file to ensure that the coding of key variables used to include or exclude certain populations was reliable. Second, we examined the FFS claims records in the IP, other acute care medical (OT), and RX files to ensure that no data reporting or data quality issues would cause omitted claims or key variables to be incompletely coded. Last, in states that enroll nondual disabled beneficiaries into risk-based managed care plans that cover some of the EBPs measured in this study,13 we examined the IP, OT, and RX files to ensure that each managed care plan reported encounter data and that the diagnosis coding was complete on the encounter records. As a final check, we compared across states the proportion of nondual disabled adults who had a diagnosis of schizophrenia or bipolar disorder to check for outlier values. Table A.2 summarizes the results of the state selection process and identify the states for inclusion in the analysis.