Extract Medical Conditions from EHR Doctor's Notes Using AI

Developed by Prof. Lu Xiong, Ph.D., ASA, MAAA

Why This Tool Is Needed:

A significant challenge in EHR data integration and analysis is the presence of unstructured information, such as doctors' notes. Unlike structured data, these free-text notes are not easily analyzable and often require manual effort to extract insights. This app addresses that challenge by leveraging advanced AI, including large language models (LLMs), to convert unstructured notes into structured, analyzable variables. This capability allows life insurers, underwriters, and insurance carriers to seamlessly incorporate EHR data into their analytic workflows, unlocking its full potential for risk assessment, underwriting, and predictive modeling.

Instructions:

Feature 1: Use the text box below to paste a single doctor's note and click Extract Information. The app will extract the medical conditions, medications, family history, and social history.

If you don't have doctor's notes, here is a specific example you can copy-paste:

Patient is a 17 yo male presenting to student health clinic with permission of parent complaining of episodic heart pounding. States that they started 2-3 months ago, come on suddenly, and there have been 5-6 episodes in total lasting 3-4 minutes each with no increase in the frequency. However he does state that the most recent episode 2 days ago was the worst yet. Feels well between episodes. Episodes of heart pounding, shortness of breath, chest pressure, and lightheadedness. Episodes apparently precipitated by taking Adderal occasionally from his roommate, who has it as a prescription, before tests to help with studying.
ROS: no nausea, no changes in bowel or bladder habits, no tingling, no numbness, no changes in vision, no fevers.
PMH: none
Allergies: nkda
Medications: takes Adderal prescribed to roommate
SH: college freshman, acknowledges stress of adjusting to college life, no smoking, social alcohol use, cannabis 1x

For more doctor's notes to test, you can use this sample.

Feature 2: Scroll to the bottom to upload a CSV file for bulk processing of multiple doctor's notes. Each doctor's note should be in the last column of the CSV file. The app will process all rows and return a CSV with the extracted information.

Notes:

Upload a CSV File:

You can upload a CSV file containing doctor's notes in the last column. The app will extract conditions, medications, family history, and social history for each row and provide a downloadable CSV file.

Download a sample CSV file here.

Note: For now, we suggest limiting the CSV file size to fewer than 100 rows.