Martin BC, McMillan JA.
AHSR FHSR Annu Meet Abstr Book. 1994; 11: 45.
College of Pharmacy, University of Georgia, Athens 30605.
PROBLEM AND OBJECTIVES: A notion that Medicaid recipients rarely use personal financial resources to purchase medical therapies not reimbursed by Medicaid systems has led many researchers to analyze claims data exclusively. Given the current reliance on Medicaid claims data, this study determined if and to what extent omitted data, namely out-of-pocket expenditures, bias estimates of interrupted time series analysis routinely used in Medicaid policy studies. DATA AND METHODS: In order to determine if there is bias associated with omitted data, two data sets were created for comparison. The first data set termed "Medicaid based", consisted exclusively of prescription records generated from Medicaid claims tapes. The second data set termed "Comprehensive", includes the Medicaid based data set match merged with data gathered directly from pharmacy providers to ascertain prescription purchases outside the Medicaid system. Identical interrupted time series analyses were performed on the two data sets to ascertain the effect of the omitted data on the series estimates modeling a decrease in an existing prescription limit. RESULTS AND CONCLUSIONS: For a cohort of 743 high prescription utilizers, the mean monthly decline in prescription procurement was 318 (6.6%) prescriptions per month for the comprehensive data base and 397 (9.9%) prescriptions per month using the Medicaid data. The Medicaid coefficient modeling the change in intercept is over twice the absolute value of the same estimate utilizing the comprehensive data set (-526.45 vs -256.24; t=2.23, P less than 0.025). Coefficients estimated from the Medical based data indicate that in addition to those drug classes exhibiting decline in the comprehensive data base, decreases were also estimated for central nervous system and hormone/antidiabetic therapeutic categories (alpha = 0.05). IMPLICATIONS FOR AUDIENCE: Analysis based solely on Medicaid claims data may be biassed when monitoring prescription utilization, incorrectly attributing decreases in prescription use to some policy portraying a more dehabilitating effect of the policy on prescription use.
Publication Types:
Keywords:
- Bias (Epidemiology)
- Health Expenditures
- Medicaid
- Pharmaceutical Preparations
- Pharmacies
- Prescriptions, Drug
- economics
- hsrmtgs
Other ID:
UI: 102212001
From Meeting Abstracts