. 2020 Jul 21;76(4):442–453. doi: 10.1016/j.annemergmed.2020.07.022

Table 3.

Performance characteristics for the COVID-19 Severity Index, quick COVID-19 Severity Index, and comparison models on independent validation.

	AU-ROC	Accuracy	Sensitivity	Specificity	AU-PRC	Brier Score	F1 Score	Average Precision
CSI	0.76 (0.65–0.86)^∗	0.79 (0.72–0.86)	0.73 (0.56–0.88)	0.81 (0.72–0.89)	0.38 (0.23–0.54)	0.25 (0.25–0.25)	0.47 (0.34–0.61)	0.40 (0.25–0.56)
CURB-65	0.50 (0.40–0.60)	0.64 (0.42–0.89)	0.57 (0.03–0.97)	0.52 (0.18–1.00)	0.18 (0.09–0.30)	0.12 (0.09–0.15)	0.13 (0.00–0.27)	0.16 (0.10–0.24)
Elixhauser	0.61 (0.51–0.70)	0.49 (0.26–0.74)	0.82 (0.45–1.00)	0.42 (0.15–0.78)	0.19 (0.11–0.29)	0.12 (0.09–0.15)	0.28 (0.20–0.37)	0.20 (0.13–0.30)
qCSI	0.81 (0.73–0.89)	0.82 (0.77–0.88)	0.79 (0.63–0.93)	0.79 (0.71–0.87)	0.47 (0.30–0.64)	0.10 (0.07–0.13)	0.49 (0.36–0.62)	0.44 (0.29–0.60)
qSOFA	0.59 (0.50–0.68)	0.83 (0.79–0.88)	0.47 (0.06–0.66)	0.72 (0.64–1.00)	0.22 (0.11–0.35)	0.12 (0.09–0.15)	0.08 (0.00–0.23)	0.20 (0.12–0.29)

AU-ROC, Area under the ROC; AU-PRC, area under the precision-recall curve.

Point estimates for model performance are provided at Youden’s index.

^∗

The COVID-19 Severity Index area under the ROC was statistically greater than qSOFA and Elixhauser after testing with Welch’s t test.²⁹^,⁴⁶