Abstract
OBJECTIVE: To analyze microbial infection patterns and develop a predictive model for hospital-acquired infection (HAI) in rehabilitation inpatients. METHODS: A retrospective cohort study included 635 patients admitted between August 2018 and February 2025; 4,523 clinical specimens were analyzed. After exclusions, 361 patients were classified into HAI (n=213) and non-HAI (n=148) groups. Significant variables from univariate analysis were incorporated into LASSO and logistic regression to build a prediction model, which was visualized as a nomogram. A simplified scoring tool and a web application were developed. External validation was performed using 332 patients from three hospitals. RESULTS: Among 4,523 specimens from 635 rehabilitation inpatients, the overall positivity rate was 61.2%. Sputum cultures were most frequent, while urine cultures increased over time. Key pathogens like Klebsiella pneumoniae and Pseudomonas aeruginosa showed distinct temporal trends. High antimicrobial resistance was prevalent, especially among multidrug-resistant organisms, with carbapenem-resistant Enterobacteriaceae being the most common MDRO type. Regression analyses identified age, prothrombin time, D-dimer, and C-reactive protein as key risk factors of HAI, while albumin and high-density lipoprotein cholesterol were protective.The nomogram demonstrated good discriminatory ability and calibration internally (AUC = 0.741) and maintained robust, generalizable performance in external validation, with the simplified risk score achieving an AUC of 0.799, while also showing stable performance before and during the COVID-19 pandemic. The tool is publicly accessible at: https://wjing-enzemed.shinyapps.io/hospital-infection-risk-en/. CONCLUSION: Our findings elucidate key microbiological patterns and predictive factors for HAI in rehabilitation inpatients. The developed model, utilizing readily available clinical parameters, shows robust and generalizable performance in stratifying infection risk, which can aid early intervention and optimize resource allocation in rehabilitation care.