Explaining Prescription Drug Use and Expenditures Using the Adjusted Clinical Groups Case-Mix System in the Population of British Columbia, Canada

Abstract

Given that prescription drugs have become a major financial component of health care, there is an increased need to explain variations in the use of and expenditure on medicines. Case-mix systems built from existing administrative datasets may prove very useful for such prediction. We estimated the concurrent and prospective predictive validity of the adjusted clinical groups (ACG) system in pharmaceutical research and compared the ACG system with the Charlson index of comorbidity. We ran a generalized linear models to examine the predictive validity of the ACG system and the Charlson index and report the correlation between the predicted and observed expenditures. We reported mean predictive ratios across medical condition and cost-defined groups. When predicting use of medicines, we used C-statistics to summarize the area under the receiver operating characteristic curve. The 3,908,533 British Columbia residents who were registered for the universal health care plan for 275+ days in the calendar years 2004 and 2005. Outcomes were total pharmaceutical expenditures, use of any medicines, and use of medicines from 4+ different therapeutic categories. The ACG case mix system predicted drug expenditures better than the Charlson index. The mean predictive ratios for the ACG system models were all within 4% of the actual costs when examining medical condition group and the C-stats for the 2 dichotomous outcomes were between 0.82 and 0.89. ACG case-mix adjusters are a valuable predictor of pharmaceutical use and expenditures with much higher predictive power than age, sex, and the Charlson index of comorbidity.