Abstract
Background: Accurately assessing SARS-CoV-2 infection severity is essential for understanding the health impact of the infection and evaluating the effectiveness of interventions. This study investigated whether SARS-CoV-2–associated hospitalizations can reliably measure true COVID-19 severity. Methods: The diagnostic accuracy of SARS-CoV-2–associated acute care and ICU hospitalizations as indicators of infection severity was assessed in Qatar from 6 September 2021 to 13 May 2024. WHO criteria for severe, critical, and fatal COVID-19 served as the reference standard. Two indicators were assessed: (1) any SARS-CoV-2–associated hospitalization in acute care or ICU beds and (2) ICU-only hospitalizations. Results: A total of 644 176 SARS-CoV-2 infections were analyzed. The percent agreement between any SARS-CoV-2–associated hospitalization (acute care or ICU) and WHO criteria was 98.7% (95% confidence interval (CI), 98.6–98.7); however, Cohen's kappa was only 0.17 (95% CI, 0.16–0.18), indicating poor agreement. Sensitivity, specificity, PPV, and negative predictive value were 100% (95% CI, 99.6–100), 98.7% (95% CI, 98.6–98.7), 9.7% (95% CI, 9.1–10.3), and 100% (95% CI, 100–100), respectively. For SARS-CoV-2–associated ICU-only hospitalizations, the percent agreement was 99.8% (95% CI, 99.8–99.9), with a kappa of 0.47 (95% CI, 0.44–0.50), indicating fair-to-good agreement. Sensitivity, specificity, PPV, and negative predictive value were 46.6% (95% CI, 43.4–49.9), 99.9% (95% CI, 99.9–99.9), 47.9% (95% CI, 44.6–51.2), and 99.9% (95% CI, 99.9–99.9), respectively. Conclusions: Generic hospital admissions are unreliable indicators of COVID-19 severity, whereas ICU admissions are somewhat more accurate. The findings demonstrate the importance of applying specific, robust criteria—such as WHO criteria—to reduce bias in epidemiological and vaccine effectiveness studies.

This publication has 40 references indexed in Scilit: