Sunday, May 19, 2024

Revolutionary AI Model Precisely Forecasts Treatment Outcomes Using Vast Patient Data

A groundbreaking artificial intelligence model, leveraging an extensive dataset of patient information, accurately predicts treatment outcomes for heart disease patients at risk of stroke, outperforming seven existing models and aligning with results from randomized clinical trials.

Developed by scientists at The Ohio State University, this innovative AI model, known as CURE (CaUsal tReatment Effect estimation), is based on a massive cache of de-identified patient data gathered from health care claims submitted by various sources, including employers, health plans, and hospitals. By initially training the model on this vast dataset, the researchers were able to fine-tune it to focus specifically on stroke risk and treatment effectiveness for individuals with heart disease.

Published in the journal Patterns, the study reveals that CURE not only surpassed existing models by 7% to 8% but also provided treatment recommendations consistent with four randomized clinical trials. Lead author Ruoqi Liu, a PhD student in computer science and engineering, highlighted the model’s potential to accelerate clinical trials and personalize patient care.

CURE’s strength lies in its ability to pre-train on large-scale datasets without restriction to any specific treatments. This pre-training, conducted on MarketScan Commercial Claims and Encounters data from 2012-2017, equipped the model with 3 million patient cases, 9,435 medical codes, and 9,153 medication codes.

Key to CURE’s success are two model-building techniques introduced by Liu: filling in patient record gaps using biomedical knowledge graphs and pre-training a deep synergized patient data-knowledge foundation model. The incorporation of knowledge graphs significantly improved the model’s performance.

By considering pre-trained data along with specific medical information, CURE predicts patient outcomes corresponding to different treatments. This comprehensive approach, validated against clinical trial results, underscores the model’s efficacy and potential impact on medical decision-making.

Senior author Ping Zhang envisions a future where clinicians can use AI algorithms like CURE, loaded with electronic health record data, to guide treatment decisions, essentially creating a “digital twin” of the patient. With this research funded by the National Institutes of Health, Zhang and his team are pioneering a new era in AI-assisted medical decision-making.

Related Articles

Latest Articles

Most Popular