The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #62575

Howard M Townsend, Delmarva Foundation for Medical Care, 9240 Centreville Road, Easton, MD 21601, 410-763-6252, townsendh@dfmc.org

Administrative data hold a lot of promise for assessing quality of care because they are "are readily available, are inexpensive to acquire, are computer readable, and typically encompass large populations.” In general, this makes administrative data sources cost effective and statistically friendly; however, "questions about the accuracy and completeness of administrative data abound." Multiple sources of administrative data can be combined such that the strengths of one source compensate for shortcomings in others. We adapted capture-recapture techniques to develop a technique for determining the optimal combination medical administrative data sources to assess the quality of prenatal care services received by Medicaid-eligible women. We identified women from four different sources using unique identifiers and listed whether a woman was absent or present in each source. Data were arranged in a 24 incomplete multiway contingency table with one missing cell corresponding to women who were not listed in any source. We developed a priori hypotheses about which source main effects and interactions effects would be important to in determining the population size and establish a candidate set of models (GLMs with log link function using Poisson distribution) based on the hypotheses. We used Akaike Information Criterion (AIC) for model selection. Additionally, maximum likelihood estimates were used to estimate the missing cell count. The model with the lowest AIC value of the candidate set included only the main effects of birth registry data and contained an estimated 95% of the Medicaid-eligible women who were entitled to receive prenatal care services.

Learning Objectives:

At the end of this presentation, participants will have an understanding of the techniques used to:
2. Assess the optimal combination of administrative data sources,
4. Test for interrelatedness of administrative data sources, and
6. Estimate the number of persons eligible for a service (i.e., the denominator of a quality of care indicator).

Keywords: Data/Surveillance, Quality of Care

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.