Effect of an AI-powered clinical workflow tool on CDI queries

Medical facilities and payers seek accurate documentation of patients’ diagnoses and treatments. Clinical documentation integrity (CDI) programs exist to ensure documentation completeness and precision. CDI queries, however, are often interruptive and annoying, increasing physician dissatisfaction and burnout. We analyzed CDI queries 5 months before and 5 months after implementation of a commercially available AI-powered clinical workflow tool to determine its effect on query rate reduction. The site was an urban, 726-bed US hospital for 11 months ending in June 2023. Twenty-five hospitalists were trained and provided with the AI-powered clinical workflow tool leveraging natural language processing (NLP) and rule-based algorithms that analyzed patients’ EHR data. The tool presents a draft assessment and plan to clinicians and suggests improved specificity. We analyzed 43,597 patient encounters and 3,790 CDI queries. CDI query rates decreased from 9.2% before training to 7.8% after training (absolute reduction – 1.4%, relative reduction – 14.8%, p = 0.02). Active tool users’ CDI query rates decreased from 10.1% to 8.1% (absolute reduction – 2.1%, relative reduction – 20.4%, p = 0.04). Extrapolating to the nation’s hospitalists, this represents annual savings of ∼500,000 physician hours – demonstrating the ability for AI-powered tools to reduce CDI queries and improve clinician satisfaction.