Sensitivity of Lung Point-of-Care Ultrasound (POCUS) to Predict Oxygen Requirements in Emerging Viral Infections

The prognostic characteristics of lung point-of-care ultrasound (L-POCUS) to predict respiratory decompensation in patients with emerging infections remains unstudied. Our objective was to examine whether scored lung ultrasounds predict hypoxia among a nonhypoxic, ambulatory population of patients with COVID-19.

This was a diagnostic case-control study. Three academic emergency departments across the United States collected a convenience sample of nonhypoxic subjects with COVID-19, scored subjects' hemithorax at 7 locations using lung ultrasound, and followed outcomes for 40 days. We defined cases as hypoxia (≤91% by pulse oxygenation) from 2 hours after index presentation to day 40. Follow-up was by telephone plus home pulse oximeter and by chart review. We conducted a logistic regression to test the association between L-POCUS scores and hypoxia. To evaluate lung ultrasound score prediction of a hypoxic event, we calculated sensitivity and specificity at optimal cut off scores and report receiver operating characteristic curve and area under the curve.

We enrolled 163 subjects but excluded 15 (3 duplicate entries; 12 lost to follow up). Median age was 41 years (interquartile range [IQR] 31-56); 83 (56%) were female, and median body mass index was 29 (IQR 25-35). We classified 47 of 148 as hypoxic cases (32%, 95% confidence interval [CI]: 25-40), leaving 101 controls. L-POCUS scores associated with hypoxia by logistic regression (odds ratio = 1.05, 95% CI: 1.02-1.08), with a 5% increase in odds of hypoxia for each 1-unit increase in L-POCUS score. The optimal cut-off score was 15 (sensitivity, 0.60; specificity, 0.73) and the area under the curve was 0·66 (95% CI 0·58-0·75). The correctly classified proportion was 69% (95% CI: 61-76).

Among nonhypoxic COVID-19 patients, higher L-POCUS rubric scores were associated with hypoxia but no scoring threshold strongly predicts hypoxia at 40 days.