Creating and Evaluating a Dynamic Study Randomization and Enrollment Tool within a Robust EHRs

Abstract

Randomized trials are difficult to perform in resource-limited settings. We developed a Randomization and Enrollment Tool (RET) within a live EHRs which automated enrollment, randomization, and data-collection in support of robust EHRs-based randomized interventions. We describe an observational assessment of RET which we piloted at three Kenyan HIV clinics for a decision support trial. We manually evaluated RET’s adequacy and accuracy in its core functions. RET enrolled 327/6626 patients, 100% meeting criteria based on EHRs data. Human reviews reveal that only 250 patients (76.5%) should have been enrolled as the EHRs contained inaccurate data for the other 77 (23.4%). 23 eligible patients were also missed through sole reliance on EHRs data. 18 (5.5%) RET-enrolled patients never received the intervention because of missed appointments. An automated randomization tool has potential to reduce human and financial costs of conducting EHRs-based randomized trials, but remains vulnerable to data quality and workflow limitations.