Patients with stroke or transient ischemic attack (TIA) are at high early risk of mortality and morbidity. Current risk prediction tools focus on patients after hospital discharge but not on those surviving to outpatient follow-up. We examined whether demographic and medical history data could predict 1-year stroke recurrence and mortality, among those discharged alive and event-free for 90 days after stroke and 1 day after TIA.

Data were obtained from the Ontario Stroke Registry (13,848 stroke and 13,059 TIA patients) and linked to administrative databases. Two-thirds of each cohort were used for model derivation and one-third for validation. Multivariable regression models were used to predict stroke recurrence and all-cause mortality.

There were 238 (2.71%) recurrent strokes in the ischemic stroke and 298 (3.44%) in the TIA cohorts at one year. Increasing age and previous stroke/TIA were associated with an increased risk of recurrent stroke in both cohorts. A higher modified Rankin Scale and diabetes were associated with an increased risk of recurrent stroke in the stroke cohort and heart failure, smoking and discharge location in the TIA cohort. Time-dependent areas under the curve were modest, 0.59 (0.54–0.64) and 0.59 (0.55–0.64) for the stroke and TIA validation cohorts, respectively. C-statistics from derivation and validation cohorts for mortality ranged from 0.74–0.78.

The predictive accuracy of the models was quite low after accounting for several risk factors. Additional risk factors associated with stroke recurrence for people seen in outpatient stroke clinics, and innovative approaches to individualized secondary prevention are needed.