User login
Health care providers who do not have the time or tools to screen patients for HIV risk also may not be prescribing preexposure prophylaxis (PrEP). But help is on the way: NIH-funded researchers have come up with novel computerized methods to identify the patients PrEP could benefit.
In 2 separate studies, the researchers developed and validated algorithms that analyze electronic health records (EHR). In the first study, Harvard researchers used machine learning to create an HIV prediction algorithm using 2007 to 2015 data from > 1 million patients in Massachusetts. The model included variables such as diagnosis codes for HIV counseling or sexually transmitted infections (STIs), laboratory tests for HIV or STIs, and prescriptions for medications related to treating STIs.
The model was validated using data from nearly 600,000 other patients treated between 2011 and 2016. The prediction algorithm successfully distinguished with high precision between patients who did or did not acquire HIV and between those who did or did not receive a PrEP prescription.
The researchers found hundreds of potential missed opportunities. They point to > 9,500 people in the 2016 dataset with particularly high-risk scores who were not prescribed PrEP. A “striking outcome,” the researchers say, is that their analysis suggests that nearly 40% of new HIV cases might have been averted had clinicians received alerts to discuss and offer PrEP to patients with the highest 2% of risk scores.
In the second study, researchers used the EHRs of > 3.7 million patients who entered the Kaiser Permanente System Northern California between 2007 and 2014 to develop a model to predict HIV incidence, then validated the model with data from between 2015 and 2017. Of the original patient group, 784 developed HIV within 3 years of baseline. The study found that the model identified nearly half of the incident HIV cases among males by flagging only 2% of the general patient population.
Embedding the algorithm into the EHR, the lead investigator says, “could prompt providers to discuss PrEP with patients who are most likely to benefit.”
Health care providers who do not have the time or tools to screen patients for HIV risk also may not be prescribing preexposure prophylaxis (PrEP). But help is on the way: NIH-funded researchers have come up with novel computerized methods to identify the patients PrEP could benefit.
In 2 separate studies, the researchers developed and validated algorithms that analyze electronic health records (EHR). In the first study, Harvard researchers used machine learning to create an HIV prediction algorithm using 2007 to 2015 data from > 1 million patients in Massachusetts. The model included variables such as diagnosis codes for HIV counseling or sexually transmitted infections (STIs), laboratory tests for HIV or STIs, and prescriptions for medications related to treating STIs.
The model was validated using data from nearly 600,000 other patients treated between 2011 and 2016. The prediction algorithm successfully distinguished with high precision between patients who did or did not acquire HIV and between those who did or did not receive a PrEP prescription.
The researchers found hundreds of potential missed opportunities. They point to > 9,500 people in the 2016 dataset with particularly high-risk scores who were not prescribed PrEP. A “striking outcome,” the researchers say, is that their analysis suggests that nearly 40% of new HIV cases might have been averted had clinicians received alerts to discuss and offer PrEP to patients with the highest 2% of risk scores.
In the second study, researchers used the EHRs of > 3.7 million patients who entered the Kaiser Permanente System Northern California between 2007 and 2014 to develop a model to predict HIV incidence, then validated the model with data from between 2015 and 2017. Of the original patient group, 784 developed HIV within 3 years of baseline. The study found that the model identified nearly half of the incident HIV cases among males by flagging only 2% of the general patient population.
Embedding the algorithm into the EHR, the lead investigator says, “could prompt providers to discuss PrEP with patients who are most likely to benefit.”
Health care providers who do not have the time or tools to screen patients for HIV risk also may not be prescribing preexposure prophylaxis (PrEP). But help is on the way: NIH-funded researchers have come up with novel computerized methods to identify the patients PrEP could benefit.
In 2 separate studies, the researchers developed and validated algorithms that analyze electronic health records (EHR). In the first study, Harvard researchers used machine learning to create an HIV prediction algorithm using 2007 to 2015 data from > 1 million patients in Massachusetts. The model included variables such as diagnosis codes for HIV counseling or sexually transmitted infections (STIs), laboratory tests for HIV or STIs, and prescriptions for medications related to treating STIs.
The model was validated using data from nearly 600,000 other patients treated between 2011 and 2016. The prediction algorithm successfully distinguished with high precision between patients who did or did not acquire HIV and between those who did or did not receive a PrEP prescription.
The researchers found hundreds of potential missed opportunities. They point to > 9,500 people in the 2016 dataset with particularly high-risk scores who were not prescribed PrEP. A “striking outcome,” the researchers say, is that their analysis suggests that nearly 40% of new HIV cases might have been averted had clinicians received alerts to discuss and offer PrEP to patients with the highest 2% of risk scores.
In the second study, researchers used the EHRs of > 3.7 million patients who entered the Kaiser Permanente System Northern California between 2007 and 2014 to develop a model to predict HIV incidence, then validated the model with data from between 2015 and 2017. Of the original patient group, 784 developed HIV within 3 years of baseline. The study found that the model identified nearly half of the incident HIV cases among males by flagging only 2% of the general patient population.
Embedding the algorithm into the EHR, the lead investigator says, “could prompt providers to discuss PrEP with patients who are most likely to benefit.”