Author(s)

Achala Vagal^1*, Heidi Sucharewv², Christopher Lindsell³, Dawn Kleindorfer⁴, Kathleen Alwell⁴, Charles J. Moomaw⁴, Daniel Woo⁴, Matthew Flaherty⁴, Pooja Khatri⁴, Opeolu Adeoye³, Simona Ferioli⁴, Jason Mackey⁵, Sharyl Martini⁶, Felipe De Los Rios La Rosa F.⁶, Brett Kissela³

¹Departments of Radiology, University of Cincinnati Medical Center, Cincinnati, OH, USA.
²Department of Biostatistics and Epidemiology, Cincinnati Children’s Hospital, Cincinnati, OH, USA.
³Department of Emergency Medicine, University of Cincinnati Medical Center; Department of Neurology, Cincinnati, OH, USA.
⁴University of Cincinnati Medical Center, Cincinnati, OH, USA.
⁵Department of Neurology, Indiana University School of Medicine, Indianapolis, Indiana, USA.
⁶Baylor College of Medicine, Baptist Health Neuroscience Center, Miami, Florida, USA.

Background: We previously developed predictive models for 3-month mortality and modified Rankin Score (mRS) after ischemic stroke. Aim: The aim was to test model validity for 3-month mortality and mRS after ischemic stroke in two independent data sets. Methods: Our derivation models used data from 451 subjects with ischemic stroke in 1999 enrolled in the Greater Cincinnati/Northern Kentucky Stroke Study (GCKNSS). We utilized two separate cohorts of ischemic strokes through GCKNSS (460 in 2005 and 504 in 2010) to assess external validity by utilizing measures of agreement between predicted and observed values, calibration, and discrimination using Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis. Results: The 3-month mortality model performed well in the validation datasets with an average prediction error (Brier score) of 0.045 for 2005 and 0.053 for 2010 and excellent discrimination with an area under the curve of 0.86 (95% CI: 0.79, 0.93) for 2005 and 0.84 (0.76, 0.92) for 2010. Predicted 3-month mRS also performed well in the validation datasets with R2 of 0.57 for 2005 and 0.50 for 2010 and a root mean square error of 0.85 for 2005 and 1.05 for 2010. Predicted mRS tended to be higher than actual in both validation datasets. Re-estimation of the model parameters for age and severe white matter hyperintensity in both 2005 and 2010, and for diabetes in 2005, improved predictive accuracy. Conclusions: Our previously developed stroke models performed well in two study periods, suggesting validity of the model predictions.

Ischaemic Stroke, Epidemiology, Stroke, Leukoaraiosis, African American, Stroke Prevalence

Vagal, A. , Sucharewv, H. , Lindsell, C. , Kleindorfer, D. , Alwell, K. , Moomaw, C. , Woo, D. , Flaherty, M. , Khatri, P. , Adeoye, O. , Ferioli, S. , Mackey, J. , Martini, S. , De Los Rios La Rosa F., F. and Kissela, B. (2018) Predicting Mortality and Functional Outcomes after Ischemic Stroke: External Validation of a Prognostic Model. Journal of Behavioral and Brain Science, 8, 587-597. doi: 10.4236/jbbs.2018.810036.