Modeling the Effects of Expanded Viral Testing on Diagnosis and Antibiotic Use Among Patients With Upper Respiratory Illness

Rapid viral polymerase chain reaction (PCR) testing has evolved since COVID-19 emerged. While targeted respiratory virus panels (1-3 pathogens), were used extensively early in the pandemic, comprehensive panels can test many more pathogens. However, the clinical benefits of broader testing remain unproven. We evaluated whether testing for more viruses correlates with a higher viral detection rate and lower antibiotic use.

We conducted a retrospective cohort study of all discharged patients with a chief complaint suggestive of a viral respiratory illness (e.g., “cough” or “congestion”) who underwent PCR testing were at a large integrated health system in the Northeast with 9 EDs (academic, urban, and rural), and 22 urgent care centers between 2015-2025. Generalized logistic regression models predicted viral detection and antibiotic prescription likelihood, with covariates including the number of viruses tested, age (years), sex, self-identified race/ethnicity, history of diabetes, heart failure, chronic lung disease, and active cancer. Model selection used Akaike information criterion, with all interaction terms tested, and a significance level of p<0.001 to reflect correction for multiple comparisons.

During the study period, 137,895 patients (5.3% of all visits) met criteria. Baseline odds ratio of viral detection was 0.150 (95% CI 0.145-0.154, p<0.001), with an adjusted odds ratio (aOR) of 1.271 (95% CI 1.263-1.279, p<0.001) for each additional virus tested. Baseline odds of antibiotic prescription were 1.352 (95% CI 1.320-1.385, p<0.001), with an aOR of 0.781 (95% CI 0.777-0.785, p<0.001) for each additional virus tested. No other factors or interaction terms were significant.

Testing for more viruses was associated with significantly increased viral detection rates and reduced antibiotic prescribing. These benefits should be weighed against test costs and may vary by setting and patient population.