Table 2

Model discrimination and calibration in retrospective and prospective datasets

ModelOutcomeRetrospective dataset (n=1948)Prospective cohort (n=315)
AUROC (95% CI)Brier scoreAUROC (95% CI)CITL (95% CI)Calibration slope (95% CI)Brier score
CAPEInpatient mortality0.88 (0.86 to 0.90)0.0690.81 (0.73 to 0.88)0.14 (−0.26 to 0.53)0.67 (0.39 to 0.95)0.092
30-day mortality0.79 (0.73 to 0.85)0.84 (0.46 to 1.22)0.58 (0.39 to 0.76)0.137
Recalibrated CAPE30-day mortality0.79 (0.73 to 0.85)0.00 (−0.31 to 0.31)1.00 (0.67 to 1.33)0.130
PSI30-day mortality0.80 (0.74 to 0.86)0.00 (−0.32 to 0.32)1.00 (0.72 to 1.28)0.121
CURB-6530-day mortality0.76 (0.70 to 0.81)0.00 (−0.30 to 0.30)1.00 (0.66 to 1.34)0.131
CAPE+PSI30-day mortality0.84 (0.79 to 0.89)0.00 (−0.33 to 0.33)1.00 (0.73 to 1.27)0.116
CAPE+CURB-6530-day mortality0.83 (0.77 to 0.88)0.00 (−0.32 to 032)1.00 (0.71 to 1.29)0.121
  • AUROC, area under the receiver operating characteristic curve; CAPE, community-acquired pneumonia artificial intelligence predictive engine; CITL, calibration-in-the-large; CURB-65, Confusion of new onset, blood Urea nitrogen, Respiratory rate, Blood pressure, 65 years old; PSI, Pneumonia Severity Index.