OBJECTIVES: Emergence of retrospective data sources without diagnosis has necessitated development of methods of assigning diagnoses using medications as a proxy. The objective of this study is to define a method for this allocation of multi-indication products using several data sources. METHODS: The study was conducted using three IMS HEALTH databases: National Disease and Therapeutic Index (NDTI), a compilation of the treatment of disease by office-based physicians; LifeLink database, an employer claims database; and LRx, a longitudinal retail prescription database. Using NDTI, medications commonly prescribed for treatment of specific diseases were identified. Patients with these drugs and diagnoses separately or in combination were selected from LifeLink(tm) to examine convergence of drug and diagnosis information. Based on the findings, a clinical algorithm to allocate prescription use by indication was designed. RESULTS: Medications used for treatment of asthma or allergic rhinitis (AR) were identified using NDTI. When applying this to LifeLink data, findings indicated diagnosis alone was not an adequate means of identifying indication: 49.7% of patients filling only asthma medications had only an asthma diagnosis; 15.0% of those receiving AR medications had a corresponding AR diagnosis; and 31.2% had neither diagnosis, yet filled prescriptions for both types of medications. Using the information gained from this analysis, algorithms were built assigning the most likely diagnosis based on the patient's drug history. Using this algorithm in LRx, only 8% of asthma/allergic rhinitis prescriptions could not be categorized. Similar patterns were seen in other therapeutic areas. CONCLUSIONS: Relying on a single source for defining medications as a proxy for diagnosis can result in under-identification of patients with a condition. Use of multiple data sources allows for increased accuracy in identifying drugs or drug combinations used as a proxy for diagnosis in a prescription database.