Pediatric Cardiologist and Intensivist Congenital Heart Center at Mott Children's Hospital/University of Michigan Medical School Ann Arbor, Michigan, United States
Abstract:
Introduction: Electronic Health Record (EHR) solutions have minimal data visualization capabilities and poor workflow integration limiting their utility as clinical decision support systems (CDSS). In complex environments like critical care medicine (CCM), these limitations do not support decision-making or team communication. More effective CDSS may decrease cognitive load and improve the accuracy, timeliness, and reproducibility of care, but characteristics of an optimal CDSS are not known. Our multidisciplinary team (clinicians and computer scientists) sought to define optimal CDSS characteristics for CCM.
Methods: We identified a common, data-rich use case associated with harm when inaccurate: daily decision-making in fluid, electrolyte, and nutrition (FEN) domains. Task analysis was used to measure current EHR workflows identifying “pain points” to target with better CDSS. Participatory design, agile scrum, and focus groups identified major challenges in current workflows. Recommendations for optimal CDSS that emerged were developed into a prototype CDSS called Insight that was tested by 48 clinicians (residents, fellows, staff) at 3 institutions (Boston Children’s Hospital, SickKids, Rambam Medical Center) to ascertain whether it functioned more effectively as a CDSS than the current EHR.
Results: Current workflow required 14 steps, 29 different screens, 43 clicks, and ~ 7 minutes to answer common, predefined questions for a given patient. Recommendations for optimal CDSS (figure 1) included that they should: 1) Facilitate problem-based discovery (organize data around specific clinical problems rather than requiring clinicians to ingest bulk data to discover potential problems) 2) Reinforce team communication (elicit a shared mental model among all care team members rather than just a subset) and 3) Improve situational awareness (clearly and continuously indicate when problems change rather than requiring individual determinations during discrete time intervals of data review). Participants rated our recommendation-based prototype Insight as more ideal in workflow, efficiency, user-friendliness, and flexibility than the EHR while accurately making inferences about patient problems (figure 2).
Conclusions: CDSS for CCM should facilitate problem-based discovery, reinforce team communication, and improve situational awareness as key objectives of their design. Novel CDSS should be rigorously evaluated against current workflows to ascertain how successful they are in facilitating clinician workflows and improving care processes and outcomes.
