AUTHOR=Lee Jong Weon , Lee Hyun-Joung , Jang Hyeon Ju , Yun Yeseul , Kim Deog Young 

TITLE=Predicting pneumonia algorithm in stroke patients

JOURNAL=Frontiers in Neurology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1690049

DOI=10.3389/fneur.2025.1690049

ISSN=1664-2295

ABSTRACT=BackgroundPneumonia is a serious complication of stroke, particularly in patients with dysphagia during inpatient rehabilitation, as it significantly increases morbidity, prolongs hospital stays, and impairs functional recovery. Early identification of patients at risk for pneumonia is crucial for improving outcomes and reducing post-stroke complications. This study aimed to develop a comprehensive algorithm for predicting post-stroke pneumonia risk by integrating clinical assessments of defense mechanisms against pneumonia.MethodsThis case-control study enrolled stroke patients at a single tertiary hospital and followed them for 4 weeks to assess pneumonia incidence. A total of 812 patients aged 20 years or older with ischemic or hemorrhagic stroke and signs of dysphagia were screened. Of these, 484 were excluded based on the following criteria: inability to maintain a sitting posture with back support, dyspnea requiring oxygen supplementation, concurrent aspiration pneumonia before enrollment, infectious diseases requiring isolation, and refusal to participate. Final cohort of 328 patients was enrolled. All participants underwent evaluations, including a videofluoroscopic swallowing study (VFSS), a modified cough reflex test (mCRT), and assessments of nutritional status (serum albumin) and cognitive function [Mini-Mental State Examination (MMSE)]. Pneumonia was diagnosed using the Mann criteria, and predictive factors were analyzed using univariate logistic regression and classification and regression tree (CART) analysis.ResultsAmong 328 participants, 28 (8.5%) developed pneumonia. Significant predictors included tracheostomy status (OR 9.34), VFSS-confirmed aspiration (OR 8.21) and bilateral stroke lesions (OR 5.91). CART analysis revealed tracheostomy, VFSS-confirmed aspiration, cough frequency, albumin levels, and MMSE scores as key predictors. The algorithm demonstrated a predictive accuracy of 92.7% with an AUC of 0.89 (95% CI: 0.82–0.95).ConclusionThis study developed a highly accurate predictive algorithm for post-stroke pneumonia, emphasizing the role of defense mechanisms against pneumonia. Implementing this algorithm in clinical practice could enable early preventive measures, reduce pneumonia incidence, and improve patient outcomes.