van Diepen 2014.

Study characteristics
General information	Objective  Prediction model compared Journal American Heart Journal Country Canada Study design Retrospective cohort study
Participants	Number of included patients 32160 Surgical specialty Noncardiac surgery Age Not reported Male sex Not reported High‐risk surgery Not reported   Insulin‐dependent diabetes mellitus Not reported History of ischaemic heart disease Not reported History of congestive heart failure Not reported History of cerebrovascular events Not reported Elevated creatinine Not reported 0 to 1 RCRI factors 73.4% 1 RCRI factor Not reported 2 RCRI factors 16.8% 3 or more RCRI factors 9.8%
Predictors	Predictor 1: CHADS2  Objective: prediction model compared Category: prediction model Scale: not applicable Threshold: not applicable Assay/device: not applicable   Predictor 2: CHADS2‐Vasc  Objective: prediction model compared Category: prediction model Scale: not applicable Threshold: not applicable Assay/device: not applicable   Predictor 3: R2CHADS2  Objective: prediction model compared Category: prediction model Scale: not applicable Threshold: not applicable Assay/device: not applicable
Outcome	Outcome category Other; all‐cause mortality Full outcome definition Composite outcome of all‐cause mortality, stroke, TIA or systemic embolism Prediction horizon 30‐day events
Analysis	Number of outcomes 1363 Handling missing data Complete case analysis Discrimination reported? Yes Calibration reported? No Reclassification reported? Yes
PROBAST: Applicability	Domain 1: Participant selection High Justification: only patients with nonvalvular atrium fibrillation were included Domain 2: Predictors High Justification: some definitions of the RCRI did not match the definitions used for this article Domain 3: Outcome High Justification: outcome is composite of mortality, stroke, TIA and systemic embolism and not MACE Overall judgement High Justification: only a selected group of patients was included which are not generalisable to the RCRI development cohort. No/unclear information on predictor definitions for some items and other predictors of the original RCRI were not included or had a different definition. Outcome definition was different compared to the RCRI development study.
Notes	—
Item	Authors' judgement	Support for judgement
Domain 1: Participant selection	Yes	Appropriate participant selection in which patients were selected in whom the RCRI model can be applied.
Domain 2: Predictors	No	Some definitions of the RCRI did not match the definitions used for this article.
Domain 3: Outcome	Yes	some definitions of the RCRI did not match the definitions used for this article.
Domain 4:  Analysis	Yes	Clear methodology and appropriate number of outcomes.
Overall judgement	No	Patient selection was appropriate, outcome definitions were clearly defined and comparable to the definitions used in the development study. Methodology used was appropriate including the number of outcomes. However, some predictor definitions were defined differently compared to the definitions used in the RCRI development study.