S2902 Practical Redesign and Implementation of a Computerized Clinical Decision Support Tool for MASLD Screening

Introduction: Metabolic dysfunction-associated steatotic liver disease (MASLD) is highly prevalent, affecting 25-30% of the general population and poses significant risk for progression to advanced fibrosis. Fibrosis screening is underutilized in the primary care setting despite risk scores such as the Fibrosis-4 (FIB-4) leveraging routinely collected blood tests, leading to underdiagnosis. To improve fibrosis screening in our primary care network, our team implemented a computerized clinical decision support (cCDS) tool in 3 primary care clinics to recommend confirmatory testing based on FIB-4. We found the cCDS was disregarded in over 90% of instances; therefore, we aimed to identify areas of improvement in the cCDS tool to redesign to enhance provider engagement. Methods: CDS tool response data recorded in the electronic health record (EHR), including each action taken by providers, was retrospectively evaluated. A multidisciplinary team of informaticists, statisticians, EHR specialists, and physicians interpreted the initial responses to guide redesign. Patterns of negative responses were identified and grouped, then linked to CDS tool design aspects that are shown to contribute to high rates of rejection in the literature. Results: During the first implementation of the CDS tool at pilot site clinics, 73% of physicians (19 out of 26) and 75% of advanced practice providers (APPs) (6 out of 8) placed orders for confirmatory testing using the tool. However, 68% of providers (17 out of 25) canceled the CDS recommendation on its initial firing before placing orders on the second firing. Immediate rejection was linked with decision burden and poorly timed presentation. Orders for confirmatory testing were placed in 1.2% of encounters involving patients with a high-risk FIB-4 score, compared to 2.4% of encounters with patients with indeterminate FIB-4 (P = 0.190). This pattern was linked to there being unnecessary information presented. Conclusion: The response data indicated that the initial OPA was often dismissed due to decision burden, untimely appearance, and unnecessary information, among other causes. Thus, the OPA was redesigned to feature minimized order options, targeted firing times, minimized text, and narrowed firing criteria to include only the highest-risk patients. Given the learnings from the index feasibility implementation, the study team plans to implement the redesigned tool in a study including 14 clinics for a 6-month period.