Augmented intelligence decision tool for stroke prediction combines factors from CHA(2)DS(2)-VASc and the intermountain risk score for patients with atrial fibrillation

摘要

Introduction CHA(2)DS(2)-VASc and CHADS(2) are computationally simple risk prediction tools used to guide anticoagulation decisions for stroke prophylaxis, but they have modest risk discrimination ability and use static dichotomous variables. The Intermountain Mortality Risk Scores (IMRS) are dynamic decision tools using standard clinical laboratory tests. This study derived new stroke prediction scores using variables from both CHA(2)DS(2)-VASc and IMRS. Methods and Results In outpatients with first atrial fibrillation (AF) diagnosis at the Intermountain Healthcare (females, n = 26 063 males, n = 29 807), sex-specific "IMRS-VASc" scores were derived using variables from CHA(2)DS(2)-VASc, warfarin use, the complete blood count, and the comprehensive metabolic profile. Validation was performed in an independent Intermountain outpatient AF cohort (females, n = 11 021; males, n = 12 641). Stroke occurred among 3.1% and 3.1% of females and 2.3% and 2.5% of males in derivation and validation groups, respectively. IMRS-VASc stratified stroke with similar ability in derivation (c-statistics, females: c = 0.703, males: c = 0.697) and validation groups (females: c = 0.681, males: c = 0.685). CHA(2)DS(2)-VASc (females: c = 0.581 and c = 0.605; males: c = 0.616 and c = 0.613 in derivation and validation, respectively) and CHADS(2) (females: c = 0.581 and c = 0.608; males: c = 0.620 and c = 0.621 in derivation and validation, respectively) were substantially weaker stroke predictors. IMRS was the strongest mortality predictor (females: c = 0.783 and c = 0.782; males: c = 0.796 and c = 0.794 in derivation and validation, respectively) and all scores were poor at predicting bleeding risk. Conclusions A temporally dynamic risk score, IMRS-VASc was derived and validated as a predictor of stroke in outpatients with AF. IMRS-VASc requires further validation and the evaluation of its use in guiding care and treatment decisions for patients with AF.