Replication Data for: Effect of an AI-powered Clinical Workflow Tool on CDI Queries

This dataset includes information for 18,284 CDI queries between June 2022 to June 2023 at a single urban hospital that is part of an academic medical center in the eastern United States. The sheet "Providers" includes the training date and use categories for 25 hospitalists who were trained on the tool utilized in the study. Twenty-five hospitalists were trained on the use of a tool utilizing natural language processing (NLP) and rule-based algorithms to analyze patient EMR data. The tool then presents information to the clinician structured as a draft assessment and plan to integrate into their clinical documentation. The tool also identifies opportunities for additional appropriate specificity in clinical documentation to reduce CDI queries and claim denials. This training was mandatory; however, continued use of the tool post-training was optional. Training consisted of two, at-the-elbow, one-on-one 1-hour sessions, with the option of a third session if the user needed extra assistance. The user was shown a 10 minute overview presentation and then practiced creating 2-3 clinical documentation notes on their patients with coaching from the trainer. Refresher sessions were provided semi-annually as necessary with additional refresher sessions available by user request, however none were requested. The tool was then made available for clinician use in regular medical practice. Participants’ use of the tool was categorized as: “full user,” “reference user,” or “non-user.” “Full user” indicates the user loads the clinical workflow tool and uses the tool to finalize their clinical documentation. “Reference user” indicates the user loads the tool and reviews the output but does not use the tool to edit their clinical documentation. “Non-user” indicates the hospitalist underwent training but did not load and thus did not employ the tool in clinical practice. The sheet "Encounters" includes total number of patient encounters for each provider between January 2022 and June 2023.