Epidemiological characteristics and the establishment and evaluation of a risk prediction model for nosocomial infection in burn patients

Abstract

Objective: To find the epidemiological characteristics of nosocomial infection in burn patients, to establish a risk prediction model for nosocomial infection in burn patients based on the screened independent risk factors of the infection, and to analyze its predictive value. Methods: A retrospective case series study was conducted. From May 2016 to December 2019, 3 475 burn patients who were admitted to the Department of Burns of Affiliated Hospital of Jiangnan University met the inclusion criteria, including 2 290 males and 1 185 females, aged from 1 to 94 years. The incidence of nosocomial infection, the detection site and specific composition of pathogenic bacteria were counted. The patients were randomly divided into training group (2 434 cases) and verification group (1 041 cases) in R 4.1.3 statistic software with a ratio of about 7∶3. Factors including gender, age, total burn area, combination of full-thickness burn/inhalation injury/shock/diabetes on admission, admission to intensive care unit (ICU), status of central venous catheterization/endotracheal intubation/urethral catheter indwelling/surgery, nosocomial infection status, days of antibiotic use, and days of hospital stay of patients were compared between the two groups. According to the occurrence of nosocomial infection, the patients were divided into nosocomial infection group (102 cases) and non-nosocomial infection group (3 373 cases), and in addition to the aforementioned data, non-nosocomial infection related data, the season of admission and types of antibiotics used were compared between the two groups. The above-mentioned data were statistically analyzed with one-way analysis of independent sample t test, chi-square test, and Mann-Whitney U test, and the indicators with statistically significant differences between nosocomial infection group and non-nosocomial infection group were included as variables in multivariate logistic regression analysis to screen independent risk factors for the development of nosocomial infection in 3 475 burn patients. On the basis of independent risk factors and important clinical characteristics, a nomogram prediction model was constructed for the risk of developing nosocomial infection of burn patients in training group. In both training group and verification group, receiver operating characteristic (ROC) curves for prediction of nosocomial infection by the prediction model were plotted, and the area under the ROC curve was calculated; calibration curves were plotted to evaluate the conformity between the predicted results of the prediction model and the actual situation; clinical decision curves were plotted to evaluate the clinical utility of the prediction model. Results: The incidence of nosocomial infection of patients included in this study was 2.94% (102/3 475); pathogens were detected from 212 specimens, mainly wound (78 cases, accounting for 36.79%) and blood (64 cases, accounting for 30.19%) specimens; 250 strains of pathogenic bacteria were detected, mainly gram-negative bacteria (153 strains, accounting for 61.20%). All clinical characteristics of patients between training group and verification group were similar (P>0.05). There were statistically significant differences between patients in nosocomial infection group and non-nosocomial infection group in the aspects of age, total burn area, days of antibiotic use, antibiotic use type, days of hospital stay, combination of full-thickness burn, combination of inhalation injury, combination of shock, ICU admission status, central venous catheterization status, endotracheal intubation status, urethral catheter indwelling status, surgery status (with Z values of 4.41, 14.95, 15.70, 650.32, and 13.73, χ2 values of 151.09, 508.30, 771.20, 955.79, 522.67, 967.40, 732.11, and 225.35, respectively, P<0.01). ICU admission, endotracheal intubation, urethral catheter indwelling, and days of hospital stay were independent risk factors for developing nosocomial infection by 3 475 burn patients (with odds ratios of 5.99, 3.39, 9.32, and 6.21, 95% confidence intervals of 2.25-15.99, 1.56-7.39, 2.77-31.31, and 2.48-15.92, respectively, P<0.01). In training group and verification group, the area under ROC curves of the nosocomial infection prediction model based on independent risk factors, total burn area, and central vein catheterization were both 0.97 (with both 95% confidence intervals being 0.95-0.99); the calibration curve analysis showed that the prediction results of the prediction model were in good agreement with the actual situation; the clinical decision curve analysis showed that the prediction model had good clinical utility. Conclusions: The nosocomial infection in burn patients is mainly caused by gram-negative bacteria, with wound as the main infection site, and the independent risk factors including ICU admission, endotracheal intubation, urethral catheter indwelling, and days of hospital stay. Based on independent risk factors and important clinical features, the risk prediction model for nosocomial infection has a good ability to predict nosocomial infection in burn patients.