Helping Veterans understand their providers' notes with Chatbot-NoteAid

Helping Veterans understand their providers’ notes with Chatbot-NoteAid Background: Allowing patient access to their providers’ notes (OpenNote) helps enhance disease understanding, patient-provider communication, medication safety, self-managed care, and health outcomes. However, many patients do not understand their electronic health record (EHR) notes. Significance: Diabetes affects 25 percent of the VA patient population. The disease is also the leading cause of visual impairment. While physicians face burnout and have limited time for patients, few computational interventions exist to help improve patient education. Innovation & Impact: We will therefore build Chatbot-NoteAid to help mitigate the challenges. Open-domain systems (e.g., GPTs) may be used for education; however, they are not HIPAA compliant and have concerns in hallucination. Chatbot-NoteAid will overcome these limitations. Communication is the central process of education. Chatbot-NoteAid will help patients understand clinical notes with learning as conversation technology, where patients do not read their notes but will gain information and knowledge through conversation with Chatbot-NoteAid. In addition to helping Veterans with low vision, compared to the traditional learning by reading, learning as conversation has the advantages that conversation helps patients stay engaged and that information is provided piece by piece, which helps strengthen learning. Specific Aims: Aim 1: We will develop Chatbot-NoteAid using innovative NLP for empathy and patient engagement, and user-centered design and usability testing. We will implement Chatbot- NoteAid using FHIR, the standard for health information exchange. Aim 2: We will build a novel instrument for note-level comprehension so that the instrument can be used to quantitatively evaluate whether Chatbot-NoteAid improves patients’ note comprehension in Aim 3. Aim 3: We will evaluate whether Chatbot-NoteAid improves EHR note comprehension in patients with diabetes in a pilot study. Methodology: We will develop Chatbot-NoteAid technologies using methods to engage patients in conversation, user-centered design, and usability testing. Chatbot-NoteAid will generate a comprehensive list of questions specific to a patient’s note and will check the accuracy of the patient’s answer. We will develop Chatbot-NoteAid with user-centered design and usability testing. A novel literacy instrument will be developed to quantitatively evaluate whether Chatbot-NoteAid improves patients’ note comprehension. We will conduct a pilot evaluation to assess the feasibility, acceptability, and impact of Chatbot-NoteAid with 32 patients. Next Steps/Implementation: Results will be used to design a future implementation trial to assess comprehension, and behavioral and clinical outcome changes in the Veteran population.