Researchers analyzed data in patients’ electronic health records to find details associated with a higher likelihood of test positivity.
Despite high volumes of SARS-CoV-2 tests and the deployment of COVID-19 vaccines, and increased use of electronic health records, little is known about whether the data in those records can predict SARS-CoV-2 test positivity.
A new study examines sociodemographic and clinical features that might limit the spread of COVID-19. The findings were published this month in PLOS ONE.
Research led by the University of Washington analyzed almost 64,000 patients who had at least one clinical visit in the UW Medicine healthcare system between Jan. 1 and Aug. 7, 2020, prior to receiving an initial reverse transcription polymerase chain reaction (RT-PCR) test for SARS-CoV-2.
The investigators studied patient sociodemographic and clinical features for predictive modeling, accounting for the type of healthcare visit and the information available to indicate patient risks at the time of testing.
Patients’ geographic, socioeconomic and demographic factors were relatively stronger predictors of SARS-CoV-2 positivity than individual clinical characteristics.
“Among these patients, we found the two-week test positivity rate in patient ZIP code was the most informative feature toward test positivity across visit types,” said lead investigator Jimmy Phuong, an IT research specialist with UW Medicine.
The findings from this study were consistent across different visit types, he said. They may offer context to COVID-19 risk factors and how routinely collected health information can inform the deployment of testing, outreach, and population-level prevention efforts.
“This new study can help provide an efficient method of identifying patients with a high probability of newly testing positive, and represents a potential approach to characterizing vulnerable populations,” Phuong said.
Other authors in this work include, Stephanie L. Hyland, Microsoft Research Cambridge, Stephen J. Mooney, UW School of Public Health, Department of Epidemiology and Harborview Injury Prevention & Research Center, Dustin R. Long, UW School of Medicine, Department of Anesthesiology and Pain Medicine, Kenji Takeda, Microsoft Research Cambridge, Monica S. Vavilala, Director of Harborview Injury Prevention & Research Center and UW School of Medicine, Department(s) of Anesthesiology and Pain Medicine Pediatrics and Kenton O’Hara, Microsoft Research Cambridge.
Funding for this research was grant supported in part by Microsoft Research, the University of Washington Population Health Initiative, UW Department of Anesthesiology and Pain Medicine, and NIH T32 funding.