Results: Problem: Data safety reporting is required for most diabetes care intervention trials, but study protocols often include more frequent contact between research staff and study subjects in intervention groups compared to control groups. This may lead to apparently higher rates of adverse event reporting in the interventions group due to more frequent contact and reporting. Here we describe our experience with a system to passively identify adverse events in all study subjects using data from Electronic Medical Records (EMR).
Approach: We applied EMR-based passive safety surveillance in a group-randomized trial including 17 clinics, 61 consented primary care physicians, and about 2,600 adults with diabetes mellitus (R01DK68314) receiving care at a single large medical group with an EMR. Safety was monitored at the patient level by following these steps: (a) Identify potential safety concerns related to the interventions including but not limited to drug safety monitoring and side effects, as well as clinical states (hypoglycemia, hyperglycemia) related to more intensive diabetes care. (b) Define a set of diagnostic and procedural codes related to the potential safety concerns and a set of discrete time periods for comparison across intervention groups that included periods prior to, during, and after intervention. (c) Quantify all procedures and diagnostic codes that occurred in study subjects assigned to each of the 3 study arms, and compare event rates within each study arms and time period, to identify potentially significant safety issues that were further assessed using more sophisticated analytic techniques.
Conclusions: Using EMR-derived automated data for passive patient safety surveillance may provide an efficient and unbiased method of obtaining safety data in some settings. This approach reduces biased reporting of adverse events due to unequal research team-study subject exposure in both randomized and group-randomized study designs.