Validation of Lung Ultrasound for Coronavirus Disease 2019 Prognostication in an International Multicenter Cohort Study

Abstract

Background Despite many studies evaluating lung ultrasound (LUS) for COVID-19 prognostication, the generalizability and utility across clinical settings is uncertain. Methods Adults (≥18 years of age) with COVID-19 were enrolled at two military hospitals, an emergency department, home visits, and a homeless shelter in the United States, and in a referral hospital in Uganda. Participants had a 12-zone LUS scan performed at time of enrollment and clips were read off-site. The primary outcome was progression to higher level of care after the ultrasound scan. We calculated the cross-validated area under the curve for the validation cohort for individual LUS features. Results We enrolled 191 participants with COVID-19 were enrolled (57.9% female, median age 45.0 years, interquartile range [IQR]: 31.5, 58.0). Nine participants clinically deteriorated. The top predictors of worsening disease in the validation cohort measured by cross-validated area under the curve (cvAUC) were B-lines (0.88, 95% confidence interval [CI]: 0.87, 0.90), discrete B-lines (0.87, 95% CI: 0.85, 0.88), oxygen saturation (0.82, 95%: CI:0.81, 0.84), and A-lines (0.80, 95% CI: 0.78, 0.81). Conclusions In an international multisite POCUS cohort, LUS parameters had high discriminative accuracy. Ultrasound can be applied towards triage across a wide breadth of care settings during a pandemic.