Abstract
BACKGROUND: Stroke-associated pneumonia (SAP) is a common complication in stroke patients, and the Barthel Index (BI) is a well-established metric for assessing activities of daily living (ADL). However, the association between BI and SAP in acute ischemic stroke (AIS) patients remains unclear. This study aims to investigate the relationship between BI at admission and SAP, and explore the factors in AIS elderly patients. METHOD: Retrospective data were collected from ischemic stroke patients hospitalized at the Second Affiliated Hospital of Nanchang University between January 2018 and July 2021, including their basic demographic and laboratory test results. Restricted cubic spline regression, multivariate logistic regression analysis, and receiver operating characteristic (ROC) curve analysis were employed to investigate the relationship between BI and SAP. Additionally, the Shapley Additive exPlanations (SHAP) method was used to identify the factors influencing SAP. RESULTS: The study included 7,548 eligible stroke patients with a mean age of 75.1 ± 7.6 years, among which 41.14% were female. The SAP group demonstrated significantly lower BI compared to the non-SAP group (50.86 ± 35.60 vs. 75.27 ± 26.33, P < 0.001). Additionally, a conspicuous trend of decreasing SAP risk across the Q1-4 groups was observed (P < 0.001). The RCS analysis further confirmed a gradual reduction in SAP risk with increasing BI. Based on the clinical model, both the BI (NRI = 0.014, P = 0.005; IDI = 0.04, P < 0.001) and the NIHSS score (NRI = 0.09, P = 0.03; IDI = 0.025, P < 0.001) demonstrated additional predictive value for SAP. Multivariate logistic regression and SHAP analysis identified WBC, CONUT, TG, UA, and RBC levels, as well as the type of health insurance (urban employee basic medical insurance), as important independent predictors of SAP. CONCLUSION: BI at admission constitutes a risk factor for the onset of SAP in elderly patients with AIS, Compared to the NIHSS and mRS score, BI may be a more reliable and practical predictor of SAP.