A new clinical validation study, published in the Journal of Clinical Oncology CCI, found that Medial EarlySign’s artificial intelligence tool flagged about 8 percent of patients in a large trial group that were later diagnosed with cancer or precancerous lesions.
“The implementation of machine-learning based algorithms to risk stratify populations at risk is commonly feared by many clinicians and population health managers as a hard-to-trust black box,” Dr. Ran Goshen, chief medical officer of Medial EarlySign, said in a statement. “In collaboration with Maccabi, we have shown for the first time that given the correct set-up of clinical leadership and top-notch technology, obstacles and hurdles can be overcome.”
The study was led by Israeli HMO Maccabi Health Services using EarlySign’s ColonFlag, an AI tool. The goal was to increase earlier detection of cancer among reluctant patients through the use of advanced screening with EHR data, according to a statement.
Researchers looked at the EHRs of 79,671 patients that were identified as noncompliant with current screening recommendations. Out of those participants the ColonFlag pinpointed 688 men and women who were at risk for colorectal cancer.
Researchers then notified those individual’s doctors. The physicians were asked to follow up with their patients. A total of 254 patients actually went in for a colonoscopy, which was performed by Maccabi doctors. Physicians found colorectal cancers in 19 of the participants and advanced adenomas in 22 of the participants. There were 15 additional cancers that were primarily identified outside of Maccabi through the code matching, according to the trial. In a general population that undergoes a colonoscopy, about 0.1 to 0.2 percent of cases present cancer. However, in this trial where the patients were selected based on risk identified by ColonFlag, 8.1 percent of patients came back positive for cancer or advanced adenomas.
“The ColonFlag test is a rapid, efficient, and inexpensive test that can be applied to scan electronic medical records to identify individuals at high risk of CRC who would otherwise avoid screening,” authors of the study wrote.
Currently Medial EarlySign is conducting clinical studies in 14 locations worldwide including cities across the US, Canada, UK, Poland, Taiwan and China.
The company was founded in 2009 with the intention of using AI to help detect early warning signals and health risks in medical data. The technology is aimed at helping expose hidden patterns and blind spots, according to its website.
EarlySign has also been used to look at other medical conditions such as diabetes, kidney disease, and gastrointestinal disorders.
"What is particularly encouraging about these results is that they have been achieved in a real-world setting, where physicians were asked to follow through on the flagged patients. They did, and now these individuals can be treated appropriately,” Dr. Varda Shalev, head of the Maccabi Institute for Research & Innovation (Maccabitech), said in a statement. “It is noteworthy that these results were similar to the findings of our simulations and previous studies. We have found this solution to be very effective for us both clinically and financially.”