Formal and informal prediction of recurrent stroke and myocardial infarction after stroke: a systematic review and evaluation of clinical prediction models in a new cohort

Table 2 Performance of models in the Edinburgh Stroke Study

	Recurrent stroke (50/671)^a			Any vascular event (80/671)^a			Original development outcome^b
Model	n/N	AUROCC	95% CI	n/N	AUROCC	95% CI	n/N	AUROCC	95% CI
Clinical gestalt	40/575	0.53	0.44 to 0.63	63/574	0.56	0.48 to 0.64	-	-	-
ESRS	50/664	0.56	0.48 to 0.64	80/664	0.57	0.50 to 0.63	101/1,224	0.54	0.49 to 0.60
SPI-II	50/669	0.58	0.49 to 0.66	80/669	0.59	0.52 to 0.66	274/1,253	0.63	0.59 to 0.67
RRE-90^c	50/671	0.61	0.52 to 0.69	80/671	0.59	0.53 to 0.66	52/1,254	0.59	0.51 to 0.67
Putaala	50/669	0.48	0.39 to 0.57	80/669	0.56	0.49 to 0.63	269/1,247	0.65	0.61 to 0.68
Dhamoon	50/668	0.60	0.52 to 0.68	80/668	0.61	0.54 to 0.67	205/1,253	0.73	0.69 to 0.76

Cell entries are AUROCCs for a recurrent stroke by one year, all vascular events by one year, and the outcome as defined in development. Few patients were deleted due to missingness for prediction models (671 outpatients with 50 strokes in follow-up and 80 vascular events; and for all patients 1,257, 102 and 274 respectively). ^aOutpatients only; ^bone year follow-up was available for all patients in the ESS; ^cModel A was the clinical based model. AUROCC, area under the receiver operating characteristic curve; CI, confidence interval; ESRS, ESSEN Stroke Risk Score; RRE-90, Recurrence Risk Estimator at 90 days; SPI-II, Stroke Prognosis Instrument II.

ISSN: 1741-7015