AI-Driven Solution for Automated and Efficient Patient-Provider Message Classification

Stanford researchers have developed an innovative AI-driven solution that leverages the BERT-based AI model to automatically classify patient-provider messages into 12 distinct categories, reducing clinician workload and enhancing workflow efficiency in healthcare settings.

In healthcare communication, there has been a dramatic rise in secure patient-provider messaging, particularly Patient Medical Advice Requests (PMARs), which enables patients to seek medical advice, request medication refills, schedule appointments, and discuss their health concerns directly with their healthcare providers. Currently, there is no effective way of sorting these messages without involving manual examination. Moreover, this surge in PMARs has introduced several challenges like the lack of efficient processing methods and workforce capacity to handle PMARs, leading to clinician burnout and inefficiencies in patient care.

To address these critical challenges, Stanford researchers from Quadrant Health and Stanford's Healthcare Applied Research Team have developed an innovative AI solution using the BERT algorithm to automate categorizing patient-provider messages. BERT is designed to understand the context and semantics of text, making it highly effective at categorizing unstructured and multidimensional messages like PMARs. The model was trained on PMARs from online patient portals to categorize patient-provider messages into 12 distinct classes including medication, scheduling, paperwork, orders, results, data entry, referral, and review level. By automating the classification of PMARs, we eliminate the need for manual review and sorting by clinicians, enhancing workflow efficiency and reducing burnout. In summary, our innovative AI solution addresses the critical problem of managing the surge in PMARs by automating the categorization process, thereby reducing clinician workload and improving overall efficiency in patient-provider communication.

Stage of Development:

Software Prototype. The next steps include EHR integration with EPIC and patient portals.

Poster Presentation of AI-Driven Solution for Automated and Efficient Patient-Provider Message Classification