Fig. 2

Three validation measures. Discrimination (panel a + b), calibration (panel c + d) and net benefit analysis (panel e + f) are shown for the outcomes mRS score 0–3 (left column), and mRS score 0–5 (=survival, right column). Panels c + d: the dotted indicates the ideal ratio where expected and observed outcomes are identical. The red line indicates the actually observed ratios. Panels e + f: the red line indicated the net benefit when using the Essen ICH score on the full range of threshold probabilities, the curved dotted line indicates the net benefit of treating all patients, and the straight dotted line on the x-axis indicated the net benefit of treating no one. Panel e shows an overall benefit (mRS 0–3) while panel f (mRS 0–5) does not indicate a clear benefit (red line mostly lower than the curved dotted line). Abbreviations: mRS modified Rankin scale, E:O expected/observed ratio, AUC area under the receiver-operating characteristic curve