| This research examined a method to use the Medical Expenditure Panel Survey (MEPS) and National Ambulatory Medical Care Survey (NAMCS) databases simultaneously to answer health-policy related questions with sufficient reliability and validity for meaningful inference. Specifically, this research used these databases to determine the rate of initial compliance to two different medications, azithromycin and methylphenidate.;The point estimate of the number of medications prescribed (P) from 1998 to 2005 was estimated from NAMCS and similarly, the number of medications filled (F) was estimated from MEPS. The two estimates were combined to form a single ratio of filled medication versus prescribed medication (F/P), thus estimating the rate of initial compliance (C) for each agent for each of the eight years. The associated variances from each point estimate were calculated and then combined to form the variance for the rate of initial compliance. The feasibility of forming F/P ratio was tested by examining the slopes of the individual lines for each drug, the reliability was tested by examining the consistency of the measures over the eight years and the validity was tested comparing the results to expected values as noted in the literature.;The results showed azithromycin, but not methylphenidate, passed the feasibility test. The chronic nature of methylphenidate use resulted in multiple prescriptions offered during a single visit and confounded the analysis. Azithromycin apparently passed the reliability test, but failed to meet the validity criteria developed through the literature review. Upon further examination, it seems that small sample sizes, leading to unreliable data, may have been the cause for a lack of validity.;Future researchers could improve the reliability of the data by using classes of medications instead of individual medications, and new features in the data collection may allow for cleaner data to be obtained. Together, these two refinements may improve the reliability of the data and allow for better validity testing. Should these refinements prove fruitful, using these two datasets concurrently for other health policy questions can be examined. |
