MedicalResearch.com Interview with:
Douglas Krakower, MD
Infectious Disease Division
Beth Israel Deaconess Medical Center
MedicalResearch.com: What is the background for this study? What are the main findings?
Response: There are 45,000 new HIV infections in the US annually, so effective HIV prevention strategies are needed. HIV pre-exposure prophylaxis (PrEP), whereby a person who is HIV-uninfected uses an HIV treatment medication on a daily basis to protect themselves from becoming infected with HIV, is over 90% effective when taken with high adherence. The Centers for Disease Control and Prevention estimates that there are 1.2 million Americans who are likely to benefit from using PrEP. However, only 80,000 persons have been prescribed PrEP. One of the barriers to implementing PrEP is that clinicians face challenges with identifying persons who are most likely to benefit from PrEP, given infrequent sexual health history assessments during routine clinical care. We thus sought to develop an automated algorithm that uses structured data from electronic health records (EHRs) to identify patients who are most likely to benefit from using PrEP. Our methods included extracting potentially relevant EHR data for patients with incident HIV and without HIV from nearly a decade of EHR data from a large ambulatory practice in Massachusetts. We then used machine learning algorithms to predict HIV infection in those with incident HIV and those without HIV. We found that some algorithms could offer clinically useful predictive power to identify persons who were more likely to become infected with HIV as compared to controls. When we applied these algorithms to the general population and identified a subset of about 1% of the population with risk scores above an inflection point in the total distribution of risk scores; these persons may be appropriate for HIV testing and/or discussions about PrEP.