A machine learning algorithm shows promise for identifying, via data in their electronic health records, patients at high risk of developing HIV, according to research presented in New Orleans last week.

The work focuses on identifying patients who might benefit from pre-exposure prophylaxis (PrEP), a treatment to reduce the risk of developing HIV that’s been found highly effective, reports MedPage Today.

Failure to have sexual health and risk evaluations performed as part of routine care poses the greatest barrier to getting more people on PrEP, according to Douglas Krakower, M.D., of Beth Israel Deaconess Medical Center. Though many physicians lack training or are uncomfortable talking about sexual health, EHR data can provide insight, he contends. A record of sexually transmitted infection testing or treatment suggests someone has had unprotected sex, while a prescription for buprenorphine or methadone indicates a substance use disorder, he said at the IDWeek annual meeting.

Researchers began by looking at more than 100 variables including demographics, diagnoses, drug prescriptions, laboratory tests, and procedures for patients at a large medical practice in Boston. They then compared characteristics and risk factors between newly infected HIV patients with control subjects who remained HIV-negative, according to MedPage Today.

Next, they used logistic regression modeling and machine learning to predict HIV infections, then compared the models to find the best at predicting HIV.

While this subpopulation was only 1.1 percent of the clinic’s patients, it represented more than 8,000 patients who might be good candidates for PrEP.

The researchers plan to further fine-tune the algorithm at a community clinic that specializes in care for sexual and gender minorities where PrEP use is more common.