Drug therapy problems and hospitalization risk in an elderly Medicaid population.

Abstract

The primary objective of this investigation was to evaluate a hypothesized causal relationship between drug therapy problems and hospitalization. A secondary objective was to compare two sources of morbidity information in insurance claims datasets: ICD9CM codes and drug claims. To assess these objectives, an observational study using a retrospective cohort design was conducted on a sample of 8,755 persons aged 60 and over who were enrolled in the Michigan Medicaid program. Two time periods of exposure to drug therapy problems were studied, January through April and April through June 1989. The outcome measure was hospitalization in the three month period following the exposure period. Multivariable logistic regression analyses were performed. Confounding information tested included demographic, prior utilization, and chronic disease variables. Both ICD9CM codes and drug data were used to identify morbidity such as cardiac disease, diabetes, osteoarthritis. Results indicate that persons taking drugs associated with falls and/or persons with drug-disease interaction problems are at higher risk for hospitalization. A logistic regression model for the first exposure period included drug-disease interactions (OR = 1.22, CI =1.02,1.44), age, a coronary artery disease indicator, a long-term care indicator, and the number of different ICD9CM codes. A validation model examined exposures in the same population over another time period and included drugs associated with falls (OR = 1.14, CI =1.01,1.3), a congestive heart failure indicator, a long-term care indicator, number of different physicians, and number of different ICD9CM codes. A statistical relationship appears to exist between drug therapy problems and hospitalization. Results of the comparison between ICD9CM codes and drug claims revealed that while both sources can provide chronic morbidity information, both sources have significant limitations. ICD9CM codes in Medicaid data are gathered primarily from hospitalization claims and may not reflect current health status. Drug claims seem to provide good proxy measures of treated, chronic conditions, however, they may not accurately classify some cardiac conditions.