Estimating asthma control questionnaire (ACQ) scores from claims data

Background: Asthma control is the main focus of treatment guidelines. Valid instruments such as the Asthma Control Questionnaire (ACQ) require prospective survey. These surveys may be challenging for large population health applications. Objective: To develop an algorithm for estimating ACQ-5 scores from commonly available claims data. Methods: Data was derived from four prospective surveys including the ACQ-5 combined with retrospective claims of Kaiser Permanente of Colorado (KPCO) patients. The statistical approach consisted of derivation and validation of a prediction algorithm including medical and pharmacy claims data using stepwise regression elimination. Validation was conducted by estimating mean squared error (MSE) and mean absolute error (MAE) in one hundred split-sample iterations. Ordinary least squares (OLS), Tobit and Median regression were used. Results: There were 2,657 individuals with valid ACQ-5 scores, claims and eligibility at baseline. The following had statistically significant associations with ACQ-5 scores: gender, use of oral corticosteroids and short-acting beta agonists, the number of asthma drug classes, and emergency and outpatient visits. Average MSE and MAE were similar for the estimation and validation samples. Conclusion: This research provides preliminary results of the feasibility of predicting ACQ-5 scores using commonly available medical and pharmacy claims data. The resulting algorithm may facilitate public health and population level analyses of asthma control. Future studies in different populations will be important to validate the algorithm.