Development and external validation of a risk stratification model for stroke-associated pneumonia

Stroke-associated pneumonia (SAP) is a common complication after acute stroke and worsens clinical outcomes. We developed and externally validated SENTRY, an early prediction model using routine admission variables and ultra-early chest computed tomography (CT) findings from eight stroke centers. The final model included age, dysphagia, NIHSS (National Institutes of Health Stroke Scale) score, neutrophil-to-lymphocyte ratio, and ultra-early chest CT abnormalities. Compared with a base model, adding chest CT findings improved discrimination (area under the curve [AUC]: 0.811 vs. 0.886) and lowered the Brier score from 0.146 to 0.122. Across validation cohorts, SENTRY showed higher discrimination than established scores (A²DS² and ISAN), and decision curve analysis showed greater net benefit across clinically relevant thresholds. SENTRY stratified patients into low-, medium-, and high-risk groups, with stepwise increases in pneumonia incidence. We applied target trial emulation to further illustrate its potential use for risk-enriched trial. A web calculator implements both base and chest CT-inclusive versions of SENTRY to support early risk estimation and risk-stratified workflows.