OBJECTIVES: Electronic Medical Records often lack accurate problem lists. The objectives of this study were to examine if alerts requesting indication information during  order entry of medications used to treat a common disease, hypertension, can improve problem lists.   METHODS: Inpatient orders generated by previously implemented clinical decision support (CDS) in an  academic hospital using CPOE were reviewed. CDS was developed for classes of antihypertensive and 3 additional diagnoses:  CHF, benign prostatic hypertrophy and nephropathy. Five logical medication groups were developed. A random sample of 250 charts were collected for each medication group and reviewed by physicians to determine  alert yield  and added problem accuracy. The accuracy and yield (problems added per order) were analyzed across the medication groups, clinician types, and clinical venues. RESULTS: Preliminary results obtained from  the group of medications that only treat HTN and no other diagnoses (HTN group) showed an accuracy of additions to the problem list of 92% and the yield, was 76%. For the group of medications which treated HTN and CHF (CHF group), the accuracy and yield were lower, 81% and 62% respectively. Looking at the HTN group alone, the accuracy was 100% for attending physicians and 55% for residents physicians. Similarly the accuracy was higher in the ambulatory setting than for inpatient, 89% vs. 62%. The yield for the HTN group was also higher with attendings than residents, 100% vs. 67%, and the yield was higher in ambulatory versus inpatient settings, 87% vs. 64%. CONCLUSIONS: CDS based on medication orders during CPOE may improve problem lists. The accuracy and yield of diagnostic information is sensitive to medication type, clinical venue and possibly clinician type. Mandatory indication based prescribing may not produce very accurate indication data. There is a substantial need for further research in this area.