Evaluating the Efficacy of Different Electronic Medical Record Alerts to Increase Pediatric Lipid Screening Across a Large Integrated Health System

Brief Summary

Detailed Description

Patients who are eligible for this study will be randomized into one of four groups via an Epic electronic medical record (EMR) randomization algorithm run automatically at the time of the visit: 1. Control group (6-month delay before their providers will receive an alert) 2. Health maintenance topic (HMT) 3. Best practice alert (BPA) 4. Best practice alert and health maintenance topic (BPA+HMT) Geisinger Health System will introduce Epic's Storyboard panel (a novel way of summarizing patient information in the EMR) approximately one month into this study. The analysis plan will therefore test for the potential impact of this change. The providers will be prompted to discuss and order screening lipid study that is non fasting at the time of the visit with the patient, based on the alerts above. Some families will have an alert in their MyGeisinger portal stating that a health maintenance test is due and to discuss with their provider. Outcomes will be reviewed and classified as followed, Outcomes will include lipid screening orders by providers (yes/no) and screening completions by patients (yes/no). The following descriptive results will also be provided: 1. Lipid screening ordered 2. Lipid screening ordered and completed 3. Lipid screening ordered but not completed 4. Lipid screening declined with reason why 5. Alert not acted on at all Analysis will account for the nesting of patients within providers; this will include provider as a random effects variable in a series of multilevel binomial logistic regression models, to account for potential correlation with patients. If the intraclass correlation coefficient is low, only the patient-level logistic regression models will be conducted. In the first model, the passive control will serve as the reference group, to test whether each of the active alert conditions have a significant impact on the outcomes. In the second model, the BPA-only condition will serve as the reference group, to test whether HMT and BPA+HMT offer significant improvements in performance. Finally, the third model will use the HMT-only condition as the reference, to test whether BPA+HMT has a significantly greater impact on the outcomes. Storyboard X Condition interactions will be tested within the models, and if any are significant, the series of models will be conducted separately on patients prior to, and after, implementation of Storyboard in Epic, to test whether and how results replicate in the different contexts.

Primary Outcomes