Pharmaceutically-based severity stratification of an asthmatic population

Abstract

Algorithms designed to precisely identify disease severity for a given patient within a managed care population are helpful in organizing targeted interventions. These algorithms are also attracting considerable attention within the medical research community. Several health risk screening instruments have been developed; however, these involve survey methodologies and have several shortcomings. We present a valid and efficient method for predicting healthcare resource utilization among asthmatics in an Health Maintenance Organization (HMO) population. First, various diagnosis, procedure and pharmacy billing codes were used to identify the asthmatics within the database. The screening algorithm awards points each time one of these codes is identified for an HMO member. By varying the number of points necessary to consider a patient asthmatic, the sensitivity, specificity, positive and negative predictive values of the algorithm can be adjusted. Once identified as asthmatic, subjects were then stratified into severity levels based on pharmacy data. Severity stratification was validated directly by measuring asthma-related bed days utilized during the 12 months following the date of stratification. Our identification algorithm estimated an asthma prevalence of 3·84% within the studied population, with age-specific prevalence estimates that closely mirrored previously published survey data. There was a monotonic relationship between pharmacy severity levels and inpatient resource utilization. For example, asthmatics in severity level 1 used only 92 hospital days per 1000 asthmatics in the year following characterization, while those in levels 2–5 used 133, 156,277 and 1168 hospital days ( P < 0·001), respectively. Results from this model can be used as adjusters in other predictive models or stand alone to represent a patient's severity of illness.