NLP EHR Integration Identifies Acute COVID-19 Cases with 94% Sensitivity

EHR integration of natural language processing (NLP) models can effectively triage patients reporting positive at-home COVID-19 test results through the patient portal, according to a study published in JAMA Network Open.

Researchers developed an NLP model called eCOV to triage self-reported positive COVID-19 cases in real time. The model identified acute COVID-19 cases from patient portal messages with 94 percent sensitivity.

Additionally, the researchers found that when responses to patient messages occurred faster, patients were more likely to receive antiviral medical prescriptions within a five-day treatment window.

Health IT tools such as the NLP model can help triage clinically urgent messages for large healthcare systems that receive thousands of patient portal messages every day, the authors emphasized.  

“Whether a physician acts on a message the day it is sent or multiple days later might determine whether oral antiviral treatment can be appropriately administered and whether benefits of reduced risk of hospitalization or mortality can be realized,” the researchers wrote.

The model can also help improve timely care by allowing patients with positive test results for COVID-19 to send an EHR message reporting positive test results. Primary care physicians (PCPs) can then treat some patients with COVID-19 remotely, thereby reducing hospitalization rates, infection risk for healthcare workers, and burden on the healthcare system.

The authors also noted that the NLP model allows for more accurate case reporting, as only 2.2 percent of COVID-19 cases identified in the cohort were otherwise documented in the structured EHR data elements.

“By classifying patient messages accurately and improving the speed of treatment access, NLP, when integrated into the EHR, has the potential to improve clinical outcomes while simultaneously reducing health care system burden,” the study authors wrote. “Additional analyses on outcomes following clinical integration are needed to quantify true clinical impact.”

The authors pointed out that their study is limited due to the absence of visual validation of test results and the inability to verify treatment adherence or prescriptions from other facilities. However, underreporting rates likely substantially outweigh false-positive reporting, they suggested.

“Additionally, symptom onset could not be systematically evaluated, so a subset of patients identified as antiviral candidates may have been out of the treatment window by the time a message was sent,” the researchers said.