From Continuous Predictors to Clinical Thresholds: Early Evidence on Performance Trade-offs of Guideline-Based Categorisation for Ischaemic Stroke Outcome Prediction

2026-08-05

AI Summary

Although existing machine learning models effectively predict outcomes in ischemic stroke, their reliance on continuous variables often conflicts with categorical thresholds recommended in clinical guidelines, limiting real-world applicability. This study systematically evaluates, for the first time, a strategy that replaces continuous predictors with guideline-based categorical encodings. Using gradient boosting models applied to a multicenter European stroke cohort, we compared the predictive performance of this categorization approach against standard continuous inputs across three treatment subgroups. Results demonstrate that in two of the three subgroups, the categorized models performed comparably to their continuous counterparts, with no statistically significant differences in predictive accuracy. Moreover, global feature importance rankings remained highly consistent between approaches, supporting the feasibility of using guideline-aligned categorical features to enhance clinical interpretability without compromising predictive performance.

Abstract

Machine learning models achieve strong predictive accuracy for 90-day outcome prediction in acute ischaemic stroke, yet clinical adoption is limited by the misalignment of model explanations with clinicians' reasoning. Motivated by a clinician user study calling for clinical guideline-aligned cut-offs, we ask whether continuous predictors can be replaced by clinically informed categorical encodings without sacrificing performance. On a multi-centre European registry stratified into three treatment cohorts, we compare standard and fully categorised gradient-boosted models, the latter using stroke guideline-aligned, treatment-specific thresholds. The fully categorised models are statistically indistinguishable from their continuous counterparts in two of the treatment cohorts, with a significant drop in predictive accuracy in one cohort. Global feature importance rankings remain consistent, suggesting that discretising continuous predictors into guideline-based categories preserves the core hierarchy of prognostic factors across all treatment groups. Guideline-based categorisation is thus a viable design choice for stroke-outcome models.

Problem

• ischaemic stroke
• clinical thresholds
• predictive modeling
• guideline-based categorisation
• model interpretability

Innovation

• guideline-based categorisation
• ischaemic stroke outcome prediction
• continuous-to-categorical encoding
• clinical interpretability
• gradient-boosted models