OBJECTIVES: Analytic datasets based on electronic Medical Records (eMRs) offer long and detailed patient follow-up, large sample sizes, and a rich set of variables.  However, the existence of a particular variable within an eMR is no assurance of actual data availability for all patients.  Patients may receive lab tests irregularly or miss visits.  Reporting standards can vary within an eMR system, creating large numbers of patients with missing data. We present details from a three step imputation process to improve eMR data availability from an observational study of hypertension outcomes.  METHODS: We used an analytic dataset comprised of 227,257 patients from 10 participating medical centers.  One important element of the study was the computation of a cardiovascular risk score based on 11 demographic and clinical variables available from eMRs.  However, only 43,676 patients (19%) had complete data for all 11 variables.  To increase the sample size of patients with complete data for all 11 variables, we used three imputation techniques: we first allowed for lab and blood pressure values to be carried forward to replace missing values; we then augmented a missing diabetes status with pharmacy data by using diabetes-specific medications to indicate a positive diabetes history; and, finally, we applied a multiple imputation (MI) technique, based on a Markov Chain Monte Carlo (MCMC) approach, implemented using SAS® PROC MI.  RESULTS: After application of the three data imputation techniques, the sample size of patients with complete or imputed data for all 11 variables increased to 75,209 (33%).  This amounts to a substantial increase (72%) in the number of patients with all critical variables available to be used for further analysis.  CONCLUSIONS: Proper application of imputation techniques, including MI, can yield a substantial increase in the number of patients available for analysis.