Electronic medical records (EMR) have been incorporated into many healthcare settings to assist physicians in sharing patient information effectively across health care providers. Most EMRs were designed to optimize sharing of clinical information and not to accommodate the needs of health services researchers. Building a comprehensive database that includes information from multiple databases and EMR sources can be challenging. OBJECTIVE: To develop a comprehensive diabetes dataset of clinical resource utilization information and costs using a community clinics and hospital based EMR and charge data. This allows a direct comparison of resource utilization and economics associated with various anti-diabetic treatments. METHODS: Prescription order data for patients on anti-diabetic medications or with a diagnosis of diabetes in 2001-2003 were obtained from the community clinics EMR system. Other healthcare information was collected from the hospital outpatient clinics and hospital EMR systems. Pertinent healthcare information included site of care, procedures performed, laboratory tests results, and diagnosis. Missing information was retrieved manually from the EMR chart from physician notes which are not transferred to the EMR database or imputed from retrieved data sources. Resource use information was matched to financial data based on patient visit numbers. Professional charges were matched to each visit based on patient identifier and approximate visitation date, with Institutional Review Board approval. RESULTS: The final dataset includes pertinent clinical, costs resource utilization information for patients suffering from diabetes across 810 patients receiving insulin. This dataset was used to determine the differences in resources and cost differences between different insulin regimens. CONCLUSION: Integrated data systems across outpatient and inpatient settings can be very useful in outcomes research however pulling together information from various datasets can be challenging.