A Dynamic Model for Predicting Survival up to 1 Year After Ischemic Stroke

Area of Science:

• Neurology
• Clinical Epidemiology
• Biostatistics

Background:

• Ischaemic stroke poses a significant challenge to patient survival.
• Accurate prognostic models are crucial for effective clinical management and patient counseling.
• Existing models may not fully capture the dynamic nature of survival post-stroke.

Purpose of the Study:

• To develop and validate a dynamic prediction model for survival.
• To assess the model's performance in predicting 1-year survival after ischaemic stroke.
• To identify key predictors of survival in ischaemic stroke patients.

Main Methods:

• Model derivation used data from 425 ischaemic stroke patients (2002-2004) from the South London Stroke Register (SLSR).
• An extended Cox model with time-dependent covariates was employed.
• Temporal validation was performed on two independent cohorts (n=1735, 1995-2002; n=2155, 2004-2016).

Main Results:

• The model incorporated six strong predictors: age, sex, stroke subtype, stroke severity, and pre-stroke/post-stroke disabilities.
• The model achieved a c-statistic of 0.822 at 1 year in the derivation cohort.
• Prognostic accuracies ranged from 77%-83% in validation cohort 1 and 70%-75% in validation cohort 2, with good calibration observed.

Conclusions:

• The developed dynamic prediction model accurately forecasts survival up to 1 year post-ischaemic stroke.
• The model demonstrates good discrimination and calibration, making it a valuable tool for clinical use.
• This model enhances prognostic capabilities for ischaemic stroke patients.