Standards for Reproducible, Accurate Measurement of Clinical Decision Support (STREAM-CDS)

Healthcare providers nationwide have spent billions of dollars adopting electronic health records (EHRs) that incorporate clinical decision support (CDS) to improve patient safety, reduce costs, and comply with federal regulations, but evaluations of CDS implementations have not consistently reported improved patient outcomes or lower costs. Alerts that prompt clinicians about drug interactions or provide dosing guidance are the most common form of CDS, but high rates of alert overrides prevent healthcare settings from realizing the expected improvement in clinical processes and patient care. EHR adoption is nearly universal, 95% of EHRs in use are commercially-developed systems, and CDS is becoming interoperable across institutions and EHRs. Despite an extensive history of CDS alert evaluation through locally-developed approaches and commercial software within or external to EHRs, as well as increasing generalizability of CDS across systems, there is no standard set of measures that can capture completely the workflow of CDS alerts and resulting effects, positive and negative, on healthcare quality and safety. Thus, there is a critical need to develop, evaluate, and adopt reproducible, accurate measures for evaluating CDS alerts. Our long-term goal is to prevent adverse events and improve patient safety through effective design and implementation of CDS across EHRs and healthcare settings. Based on needs identified in prior research, our overall objective in this proposal is to develop a standard set of measures that accurately capture the CDS alert workflow and can be reproduced in diverse clinical and informatics settings (STREAM-CDS: Standards for Reproducible, Accurate Measurement of Clinical Decision Support). The rationale for the proposed work is that a standard approach to evaluating and reporting CDS alerts across all phases – triggers, inputs, interventions, and actions – can better overcome existing barriers to identifying best practices that will result in improved outcomes. We plan to achieve our overall objective by pursuing the following three specific aims: 1) Identify measures of clinical decision support alerts, 2) Implement measures across multiple organizations and develop benchmarks for clinical decision support alerts, and 3) Disseminate measures of clinical decision support alerts to healthcare organizations, policymakers, and vendors. Creating comprehensive, standards-based measures that are generalizable across EHRs and healthcare organizations will allow us to continue working toward our long-term goal by pursuing future research that will lead to more effective CDS with improved provider performance and satisfaction and patient outcomes. This work has significant implications, given the persistently high rates of CDS overrides and provider dissatisfaction despite widespread use of commercially-developed EHRs and shareable CDS. The ability to consistently and completely evaluate the entire CDS workflow to identify and implement best practices can enable the delivery of safer and higher quality healthcare overall.