Medicaid and CHIP retention among children in 12 states

Administrative databases

Administrative healthcare databases are created from records of service utilization and payments for payer or hospital billing purposes. Such databases contain large amounts of information on the diagnoses, procedures, resource utilization, and costs or charges of thousands or even millions of patients. However, the patient populations included and data elements available vary widely by database, resulting in different strengths, limitations, and potential uses for each. Therefore, selection of an administrative database should be guided by the research question of interest. This review discusses important considerations when selecting an administrative healthcare database to use to answer a research question and describes key characteristics of the administrative databases most frequently used for pediatric surgical research.

Using Appendicitis to Improve Estimates of Childhood Medicaid Participation Rates

OBJECTIVE

Administrative data are often used to estimate state Medicaid/Children's Health Insurance Program duration of enrollment and insurance continuity, but they are generally not used to estimate participation (the fraction of eligible children enrolled) because administrative data do not include reasons for disenrollment and cannot observe eligible never-enrolled children, causing estimates of eligible unenrolled to be inaccurate. Analysts are therefore forced to either utilize survey information that is not generally linkable to administrative claims or rely on duration and continuity measures derived from administrative data and forgo estimating claims-based participation. We introduce appendectomy-based participation (ABP) to estimate statewide participation rates using claims by taking advantage of a natural experiment around statewide appendicitis admissions to improve the accuracy of participation rate estimates.

METHODS

We used Medicaid Analytic eXtract (MAX) for 2008-2010; and the American Community Survey for 2008-2010 from 43 states to calculate ABP, continuity ratio, duration, and participation based on the American Community Survey (ACS).

RESULTS

In the validation study, median participation rate using ABP was 86% versus 87% for ACS-based participation estimates using logical edits and 84% without logical edits. Correlations between ABP and ACS with or without logical edits was 0.86 (P < .0001). Using regression analysis, ABP alone was a significant predictor of ACS (P < .0001) with or without logical edits, and adding duration and/or the continuity ratio did not significantly improve the model.

CONCLUSION

Using the ABP rate derived from administrative claims (MAX) is a valid method to estimate statewide public insurance participation rates in children.