FHAIR: AI generated medical records

[pavlushkin]

Problem

• Physicians spend a significant amount of time on documentation. A study found that physicians spend around 35% of their time documenting patient data
• A 2022 report from the CDC Quality of Worklife survey found that 46% of health workers reported feeling burned out often or very often compared to 32% in 2018

Solution

• Automatically generating medical records.
• Using HIPAA compliant GPT-4 provided by Microsoft Azure we can generate FHIR QuestionnaireResponses and other FHIR resources from a brief description provided by a medical worker.
• GPT assistant can interact with medical workers by asking additional questions to fulfil all required parts of questionnaire.
• Another GPT assistant verifies that a generated record meets quality criteria and if not initiates further activity.
• Generated QuestionnaireResponse is considered as a draft and must be reviewed by the medical worker.

Expected effects

• Reduce time to document medical activity by tens of percents
• Reduced burnout of medical workers
• The quality of generated medical records is acceptable
• Positive feedback from medical workers

Competitors

• Automatically document care with Dragon Ambient eXperience https://www.nuance.com/healthcare/ambient-clinical-intelligence.html

Scope of the project at the hackathon

• Use FHIR-EMR project as a base
• Use regular (not HIPAA compliant) GPT models since we don't use it for real patients
• Generate progress notes as a QuestionnaireResponse based on a concise free-text description
• Respect and refer other patient's medical records to produce the output

Open AI docs: https://platform.openai.com/docs/overview

Appendix 1 - Surgeon's Medical Document List In a typical US hospital, a surgeon would be responsible for filling out and managing various medical documents as part of their duties. Here's a list of common documents that surgeons often deal with:

Operative Report: This is a detailed report created by the surgeon following a surgical procedure. It includes information about the surgery, such as the type of surgery, the technique used, findings during the surgery, and any complications or unexpected events.

Consent Forms: Surgeons are required to obtain informed consent from patients or their legal guardians before performing any surgical procedure. This form details the risks, benefits, and alternatives to the surgery.

Pre-Operative Note: Before surgery, a surgeon fills out a pre-operative note which includes the patient's history, physical examination findings, and the plan for the surgery.

Post-Operative Note: Immediately after the surgery, the surgeon writes a post-operative note summarizing the procedure and any immediate post-operative care instructions.

Discharge Summaries: When a patient is discharged from the hospital post-surgery, the surgeon prepares a discharge summary that includes information about the surgery, the patient's condition at discharge, and follow-up care instructions.

Progress Notes: Throughout a patient’s hospital stay, the surgeon will write progress notes to document the patient’s ongoing status, any changes in their condition, and adjustments to their treatment plan.

Orders: Surgeons write various orders for patient care, including medication orders, orders for tests or procedures, and orders related to the patient's diet or activity level.

Referral Letters: If a patient needs to be referred to another specialist or for further treatment, the surgeon will prepare a referral letter detailing the patient’s condition and the reason for the referral.

Death Certificate: In the event of a patient's death, the attending surgeon might be responsible for filling out and signing the death certificate, stating the cause of death.

Anesthesia Record: While typically filled out by the anesthesiologist, the surgeon also reviews and contributes to this document, which records details about the anesthesia given during surgery.

[pavlushkin]

Webinar: The next evolution of AI: Fully automated clinical documentation https://www.digitalhealth.net/events/webinar-the-next-evolution-of-ai-fully-automated-clinical-documentation/