Using AI to Summarize Patient Medical Records
Retrieving and synthesizing information from the EHR can be difficult due to the abundance and complexity of patient data. This competition seeks solutions that retrieve and synthesize medical records from the EHR for primary care health professionals (e.g., MDs, PAs, NPs).
Background
AMIA set an ambitious goal for documentation burden reduction for U.S. health professionals to 25% of current state in five years. To that end, we seek to optimize electronic health record (EHR) and related vendor solutions.
Documentation burden goes beyond data entry into the electronic medical record (EMR). Information retrieval and synthesis also contribute to burden. Simplifying retrieval and synthesis of patient information will reduce documentation burden and is essential for optimizing healthcare delivery, improving patient outcomes, and creating a more efficient and effective healthcare system.
The 25x5 Technology Requirement Workstream issued a challenge focused on patient information synthesis and retrieval from EMRs. This challenge supports the overarching goals of the 25x5 Task Force to reduce documentation burden by addressing the challenges associated with the increasing volume and complexity of health information while delivering high-quality, patient-centered care.
Timeline
| Date | Milestone |
|---|---|
| April 29 | Challenge Launch |
| May 24 | Round One Submission Deadline (written proposals) |
| June 17 | Round One Winners Announced |
| September 13 | Round Two Submission Deadline (20 patient summaries) |
| October 18 | Winners announced via press release |
Prize
The winner and top finalists will receive:
- AMIA press release recognizing the finalist for finalist’s own promotional use
- A forum to address structural barriers to scaling and adoption of the innovation including:
- A consortium of vendors to collaborate with on interoperability needs
- An advocacy group actively focused on driving policy change
- A collective of progressive health systems leaders willing to promote and experiment with new innovations.
Competition Information
Who Can Participate?
This challenge is open to everyone. AMIA membership is not required and there is no cost to submit.
Competition Breakdown
Round One
Participants will submit a 300-word description of their proposed tool that can retrieve and synthesize electronic medical records for primary care health professionals and a 200-word description of how their proposed tool aligns with risk mitigation principles to reduce bias and advance health equity.
Up to 10 entries will be invited to participate in Round Two.
Round Two
Participants will be given access to 20 synthesized deidentified patient medical records (e.g., text files, FHIR, JSON. Note-handwritten scans will not be included) to develop a tool that produces a patient summary for primary care health professionals seeing the patient for the first time. Participants will submit the resulting AI-generated 20 patient summaries for evaluation. In addition, participants must sign a public commitment that the produced summaries submitted were completely generated by a computer system and were not produced or edited by a human.
Please note: If selected to go on to round two, AMIA may use participant name, logo, and trademarks solely to identify participant as competing in the "25x5 Task Force Technology Challenge: Using AI to Summarize Patient Medical Records" on AMIA's website and other marketing materials in accordance with participant's trademark usage guidelines.
Participant may likewise use AMIA's name, logo and trademark solely to identify themselves as a participant in the challenge on their own website and other marketing materials in accordance with AMIA's trademark usage guidelines. AMIA reserves the right to make participant's round one proposal answers and performance metrics during round two publicly available on its website and other marketing materials.
Judging Information
The challenge judges are a diverse and knowledgeable pool of healthcare professionals, ensuring a representative group with expertise spanning various specialties and facets of the healthcare industry.
Review Criteria
Round One
Round One will be evaluated for innovation, functionality, feasibility, integration with existing EHRs, health equity, and overall impression.
Round Two
Round Two submissions will be evaluated using the following criteria:
Readability
The summary is well-written using professional language as a trained health professional. The organization and structure of the summary promotes readability.
Completeness
The key information important to a PCP is included in the summary. A very low score signifies that the information missed would create a potential patient-safety risk.
Curation
The summary excludes irrelevant information.
Correctness
The summary does not introduce hallucinations, faulty logic, or bias. A very low score signifies the incorrect information introduced in the summary would create a potential patient safety risk.
Overall Utility
Rate the summary from your own subjective context of what you view as a well-written summary.