Table 3

Accuracy of clinician probability estimates using different double thresholds (including indeterminate values)

 Estimate cut-off Prediction* Outcome (within 72 hours) Outcomes Proportion of indeterminate cases Died Survived Total N PPV NPV >50% Die 787 478 1265 62% <50% Survive 163 456 619 74% 50%–50% Indeterminate 40 56 96 5% ≥60% Die 766 446 1212 63% ≤40% Survive 144 422 566 75% 41%–59% Indeterminate 80 122 202 10% ≥70% Die 685 356 1041 66% ≤30% Survive 87 338 425 80% 31%–69% Indeterminate 218 296 514 26% ≥80% Die 577 247 824 70% ≤20% Survive 39 256 295 87% 21%–79% Indeterminate 374 487 861 43% ≥90% Die 440 149 589 75% ≤10% Survive 19 141 160 88% 11%–89% Indeterminate 531 700 1231 62%
• *Each estimate was categorised in to a group if the estimate was above or below the cut-off. Scores that were indeterminate were omitted from the analysis. PPV (positive predictive value; the proportion of patients who died when the clinician predicted dying, eg, 787/1265 for the 50% cut-off), NPV (negative predictive value; the proportion of patients who survived when the clinician predicted survival, eg, 456/619 for the 50% cut-off).