HealthDay News — Two new research papers, published online July 5 in The Lancet HIV, present algorithms that can help identify patients at risk for HIV and candidates for preexposure prophylaxis (PrEP).
Julia L. Marcus, PhD, from Harvard Medical School in Boston, and colleagues applied least absolute shrinkage and selection operator (LASSO) regression using 81 electronic health record (EHR) variables to predict incident HIV diagnosis within three years (development dataset; 3,143,963 patients) and assessed the 10-fold cross-validated area under the receiver operating characteristic curve (AUC). Models were validated prospectively using data from an independent set of patients (606,701 patients). The researchers found that 44 predictors were retained in the LASSO procedure, with an AUC of 0.86. In the validation dataset, model performance remained high (AUC, 0.84).
Douglas S. Krakower, MD, from the Beth Israel Deaconess Medical Center in Boston, and colleagues used machine learning algorithms to predict incident HIV infections using 180 potential predictors of HIV risk drawn from EHR data. The development cohort included 1,155,966 patients, while the internal and external validation cohorts included 537,257 and 33,404 patients, respectively. The researchers found that the best-performing algorithm was obtained with LASSO and had an AUC of 0.86 (95% confidence interval [CI], 0.82 to 0.90) for identification of incident HIV infections in the development cohort and 0.91 (95% CI, 0.81 to 1.00) and 0.77 (95% CI, 0.74 to 0.79) on prospective and external validation, respectively. The LASSO model also successfully identified patients independently prescribed PrEP by clinicians in both validation cohorts (AUC, 0.93 [95% CI, 0.90 to 0.96] and 0.79 [95% CI, 0.78 to 0.80], respectively).
“Additional studies are needed to further optimize these models, integrate them into EHRs at the point of care, and evaluate their impact on PrEP prescribing and HIV prevention,” Krakower and colleagues write.
Several authors from both studies disclosed financial ties to the pharmaceutical and medical technology industries.