Abstract
Background
The problem of delirium caused by fluoroquinolones, especially moxifloxacin, has posed a great challenge to clinical practice. Currently, there is a shortage of predictive models for predicting moxifloxacin-induced delirium. Therefore, this study aims to develop and validate a predictive model for moxifloxacin-induced delirium in community-acquired pneumonia (CAP) patients.
Methods
This retrospective cross-sectional study was conducted in Taizhou, Jiangsu Province, China. A total of 488 CAP patients who had received moxifloxacin treatment were included between June 2023 and March 2024. Least absolute shrinkage and selection operator (LASSO) regression and multiple logistic regression were used to identify predictive factors for moxifloxacin-induced delirium. A nomogram containing four predictive factors was created. Discrimination, calibration, and clinical utility were employed to evaluate the model's performance, with internal validation using the bootstrap method. Feasibility analysis of the model was conducted based on the respective prediction probabilities and nomogram scores.
Results
Among the 488 patients, 9.63% (47/488) exhibited moxifloxacin-induced delirium, while the remaining 90.37% (401/488) did not encounter such adverse effects. Through LASSO and multiple logistic regression analysis, we identified increasing age, weight loss, elevated bilirubin levels, and a history of ischemic heart disease as significant predictive factors. These four predictors were utilized to construct a predictive nomogram. The area under the receiver operating characteristic curve (AUC) was determined to be 0.889 (95% CI 0.841–0.937), which was further validated through bootstrap sampling analysis with an AUC of 0.897 (95% CI 0.846–0.949). The Hosmer-Lemeshow test yielded a p-value of 0.257, and the calibration curve also indicated that the model exhibits good calibration ability. The decision curve analysis (DCA) demonstrated a positive net benefit within a risk range from 0.5–77%. The clinical impact curve demonstrated a strong alignment between the model's predictions and actual occurrences when the risk threshold exceeded 0.3. The feasibility analysis not only demonstrated the model's advantages over internal variables but also revealed significantly elevated nomogram scores in delirium patients.
Conclusions
This study has developed a predictive model for identifying moxifloxacin-induced delirium in CAP patients, exhibiting excellent performance and providing valuable assistance to clinicians in identifying high-risk individuals.