With nearly 500 clinical studies on cancer being published every day, it is impossible for oncologists to keep up with the developments in the field. This information overload does not even take into account the new cancer therapies coming to market and what fellow doctors are doing right now in similar cases.

To address this information avalanche, Dr. Andrew Pecora, chief innovation officer at Hackensack Meridian Health, created a tool that combines all of these data points (including the specific characteristics of the patient) and comes up with treatment suggestions for the doctor right at the moment when it is most needed.

Tom Castles explains how the data aggregation point-of-care tool works in the medical setting in this post from Healthcare Analytics News:

Pecora set out to create such an artificial intelligence-driven point-of-care data aggregation tool with the help of IBM’s Watson for Oncology. The tool would simultaneously pull data from clinical trials and real-world results, present unique patient characteristics from the EHR, include each patient’s social determinants of health, and benchmark against patients with similar characteristics who had been previously treated. Then, that data would be combined to arrive at an AI-generated treatment suggestion.

“We thought we were geniuses,” Pecora said, describing the moment his idea became reality. “But the physicians said we had to show them data that the experts actually agreed with.” Pecora tapped three leading breast cancer experts to suggest treatment approaches on 88 cases. Of the 223 responses, 78.5% agreed with the Watson for Oncology recommendation for the best possible treatment approach, 9.4% of recommendations were listed as clinically acceptable, and 12.1% of recommendations were listed as not recommended.

The next hurdle was to demonstrate that presentation of the data would result in changed clinical decision making at the point of care, Pecora said. He polled four solid tumor doctors and 6 hematologic malignancy doctors who didn’t typically take care of breast cancer on 339 breast cancer cases. 62% of the time, their choice agreed with Pecora’s results, 13% were clinically acceptable, and 24% of the time their choice was deemed by the system to be inappropriate care, he said, adding that their care pattern match was only 4%.

“So, if you show doctors at the point of care what experts think is appropriate and what the literature says might be best, the actually change their behavior,” Pecora said, adding that he believes he now has evidence to support moving forward with this cognitive computing point of care decision support system, supplemented by real-world data to improve treatment selection of complex disease such as breast cancer, especially when delivered in non-expert care settings.