Project Context
To ensure and improve the quality care to the healthcare beneficiaries, this prominent federal healthcare agency needs to validate hospitals’ self-reported healthcare quality data submitted to the Agency by abstracting data from the medical records.
Challenge
Medical records can be very large and received from 400 different hospitals in various layouts. The medical records range between 150-2000 pages with an average of 450 pages. The paper records need to be digitized through OCR.
Once digitized, human abstractors need to track timelines and record multiple clinical parameters against quality measures. This is a highly specific and repetitive task and resource-consuming.
Solution
Pulsebyte developed and trained a Deep Learning language model that is capable of labeling clinical entities of interest from any text. The LLM summarizes these clinical parameters for timeline tracking. The AI solution significantly improves the efficiency and accuracy of the medical records validation. This solution can be scaled and customized to identify patterns in new workflows.