Prediction Model Incorporating Infarct Localization for Stress Ulcer Risk Stratification in Acute Ischemic Stroke

While traditional stress ulcer predictors from intensive care settings are well-established, infarct location may represent a unique risk factor in patients with ischemic stroke. This study aimed to develop and validate a predictive model incorporating neuroanatomical localization for stress ulcer risk in acute ischemic stroke.

Using data from a prospective cohort study of patients with acute ischemic stroke, we performed variable selection through univariate and LASSO regression analyses. Infarct locations identified by magnetic resonance imaging (MRI) were specifically included as candidate variables. Significant predictors derived from multivariable Cox proportional hazards modeling were used to construct both a nomogram and a concise scoring model. Internal validation was conducted using 1000 bootstrap iterations. External validation was performed in a separate cohort of patients with ischemic stroke. Model performance was evaluated through discrimination analysis using area under the curve (AUC) metrics.

Three infarct location predictors-insular cortex infarction, cerebellar infarction, and medullary infarction-were included in the predictive models, together with endovascular treatment, admission national institutes of health stroke scale score, and leukocyte count. The nomogram model demonstrated excellent discrimination (AUC 0.88 across 3-, 7-, and 31-day horizons), good calibration, and positive net benefit. The scoring model showed an AUC of 0.85 in the training cohort with an optimal cut-off value of 2.5 points. Bootstrap validation confirmed the robustness of both models (AUC 0.88). In the external validation cohort, the nomogram achieved an AUC of 0.85, while the scoring model reached an AUC of 0.79.

Models incorporating infarct locations-insular cortex, cerebellar, and medullary infarction-demonstrated good discriminative performance for predicting stress ulcers in acute ischemic stroke. External validation in larger and more diverse cohorts is warranted to further establish generalizability.