Development and validation of a clinical prediction model for concurrent pulmonary infection in convalescent patients with intracerebral hemorrhage

Abstract

AbstractBackground Recovery from intracerebral hemorrhage is an important but underappreciated part of the prognosis of patients with intracerebral hemorrhage. Pulmonary infection (PI) is the most common complication that greatly affects the recovery process of patients with intracerebral hemorrhage. Dynamic nomograms to predict concurrent pulmonary infections in patients recovering from cerebral hemorrhage have not been reported. The aim of this study aims to identify the risk factors for pulmonary infection in convalescent patients with intracerebral hemorrhage, and to build and validate a clinical prediction model. Methods A total of 761 convalescent patients with intracerebral hemorrhage were included in this study. Pulmonary infection was determined based on the clinical manifestations and chest X-ray, and the patients were then divided into the PI group and non-PI group. Baseline and clinical data of the patients were retrospectively analyzed. First, univariate logistic regression was performed to initially screen out predictors. Then, the predictors were optimized using least absolute shrinkage and selection operation (LASSO) regression. Finally, multivariate logistic regression analysis was carried out on the optimized predictors to identify independent risk factors and construct a nomogram prediction model. The area under the curve (AUC), calibration curves, and decision curve analysis (DCA) were used to evaluate the discrimination, calibration, and clinical utility of the model. Results Age, antibiotic use, disturbance of consciousness, tracheotomy, dysphagia, length of bed rest, nasal feeding, and procalcitonin were associated with pulmonary infection in convalescent patients with intracerebral hemorrhage. The consistency index (C-index) of the prediction model was 0.901 (95%CI: 0.878 ~ 0.924). Repeated sampling by Bootstrap for 1000 times yielded a C-index of 0.900 (95%CI: 0.877 ~ 0.923), indicating that the model has excellent discrimination. Moreover, the Hosmer-Lemeshow test revealed a good goodness of fit of the model (P = 0.982). The DCA decision curve showed that the nomogram in PI group has a good net clinical benefit. Conclusion This study constructed a nomogram prediction model based on the demographic and clinical characteristics of convalescent patients with intracerebral hemorrhage. Further studies showed that this model is of great value in the prediction of pulmonary infection in convalescent patients with intracerebral hemorrhage.