Assessing Electronic Health Record-Derived Outcomes in a Pragmatic Weight Gain Prevention Trial: Effects, Variability, and Implications for Future Trials

BACKGROUND: Electronic health record (EHR) outcome ascertainment may introduce challenges if measurements are inconsistently or infrequently captured. OBJECTIVE: To evaluate the effects of the "Balance" intervention on EHR-derived secondary/exploratory outcomes and assess outcome frequency, timing, and variability. METHODS: Balance was a 2-arm, 12-month pragmatic randomized trial of a digital weight gain prevention intervention (N = 443). Cardiometabolic outcomes (eg, blood pressure, hemoglobin A1c, triglycerides, and total cholesterol) were extracted from the EHR and analyzed using linear mixed models. RESULTS: The intervention arm (n = 223) had lower systolic (-3.4 mm Hg, 95% confidence interval [CI]: -6.7 to -0.2) and diastolic blood pressure (-2.4 mm Hg, 95% CI: -4.3 to -0.6) at 6 months. Effects attenuated by 12 and 24 months. Other outcomes showed no significant differences. Blood pressure was measured 11.9 times per participant, 4.2 months apart. Other outcomes were measured 12 to 17.4 months apart. Variability averaged 1.0 standardized units. CONCLUSIONS: Trials using EHR data should account for data variability and sparsity in trial planning, implementation, and analysis.