Kehl says the aim is to apply the model to EHRs so that researchers can connect data on patient tumors to data on patient outcomes to learn what characteristics of the tumors predict benefits from specific treatments. This research will enable providers to develop more personalized treatment strategies.

In addition, the system, he adds, could potentially enable hospitals to run near-instantaneous reports to identify all of its patients who have worsening cancer at any given time and provide treating oncologists information on relevant clinical trials or symptom relief strategies. Further testing, though, is required.


"We are currently developing a proposal with researchers at a sister institution to see how well our models perform on EHR data collected at their site," said Kehl. "We also currently have curators reviewing records for patients with other common types of cancer, including colorectal, kidney, prostate, and bladder cancer, and we plan to validate and refine our models for those groups as well."

The findings were published in the journal, JAMA Oncology.

Back to HCB News

Artificial Intelligence Homepage