Epidemiological characteristics of nosocomial infection in hospitalized children with burns and the establishment and verification of a risk prediction model
10.3760/cma.j.cn501225-20230812-00046
- VernacularTitle:住院烧伤患儿发生医院感染的流行病学特征及其风险预测模型的建立与验证
- Author:
Chao HAN
1
;
Peng JI
;
Yage SHANG
;
Jin LI
;
Kejia WANG
;
Tao CAO
;
Dahai HU
;
Ke TAO
Author Information
1. 空军军医大学第一附属医院全军烧伤中心，烧伤与皮肤外科，西安 710032
- Keywords:
Burns;
Child;
Nomograms;
Inflammation;
Cross infection;
Risk factors
- From:
Chinese Journal of Burns
2023;39(11):1006-1013
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze the epidemiological characteristics of hospitalized children with burns who developed nosocomial infection, and screen their independent risk factors, based on which, a risk prediction model was established and evaluated.Methods:A retrospective cohort study was conducted. From May 2010 to April 2023, 417 children with burns who met the inclusion criteria were admitted to the First Affiliated Hospital of the Air Force Medical University, including 248 males and 169 females, aged ≤14 years. Statistics on the composition and source distribution of pathogenic bacteria in children were detected. According to the occurrence of nosocomial infection, the children were divided into infected group (216 cases) and uninfected group (201 cases), and the children gender, age, total area of burns, presence of full-thickness burns, cause of the injury, and season of the injury of the children in the 2 groups were collected, as well as presence of an abnormal serum albumin level, delayed resuscitation, combination of inhalation injury at admission, and early shock, tracheotomy, admission to the intensive care unit, and deep venous catheterization after post-hospitalization, and more or less times (>2 times being more and ≤2 times being less) of surgeries, indwelling catheter days, and length of hospitalization stay on post-hospitalization. The burned children were divided into modeling group (291 cases) and validation group (126 cases) according to the ratio of 7∶3, and the data of the 2 groups were recorded as before. Data were statistically analyzed with Mann-Whitney U test, chi-square test, and Fisher's exact probability test. The least absolute value selection and shrinkage operator (LASSO) regression analysis was used to reduce the risk factors of nosocomial infection in the children in modeling group. Multivariate logistic regression analysis was used to further screen the above screened risk factors, and the nomogram prediction model was drawn based on the further screened independent risk factors. The Bootstrap method was used for internal validation of the aforementioned predictive models, and the receiver operator characteristic (ROC) curves, calibration curves, and clinical decision curves of the predictive models were plotted in modeling group and validation group in order to assess its discriminative power, calibration, and clinical utility, respectively. Results:A total of 245 strains of pathogenic bacteria were detected, with Staphylococcus aureus (101 strains, accounting for 41%), Pseudomonas aeruginosa (54 strains, accounting for 22%), and Acinetobacter baumannii (33 strains, accounting for 13%) dominating, and the wound secretions were the most frequent source of pathogenic bacteria (211 strains, accounting for 86%), followed by blood (10 strains, accounting for 4%), and sputum (5 strains, accounting for 2%). There were statistically significant differences between infected group and non-infected group in the total burn area, indwelling catheter days, length of hospitalization stay, presence of full-thickness burns, combined with inhalation injury, and deep vein catheterization, and more or less times of surgeries (with Z values of -2.32, -3.29, and -3.85, respectively, with χ2 values of 26.36, 7.03, 10.13, and 10.53, respectively, P<0.05); there was statistically significant difference in cause of the injury between the two groups ( P<0.05). All clinical characteristics of children with burns in the modeling and validation groups were similar ( P>0.05). The six risk factors obtained from the LASSO regression analysis were full-thickness burns, deep vein catheterization, abnormal serum albumin level, multiple surgeries, indwelling catheter days, and length of hospitalization stay; the multivariate logistic regression analysis showed that full-thickness burns, abnormal serum albumin level, deep vein catheterization, and multiple surgeries were the independent risk factors for the occurrence of nosocomial infection in burned children (with odds ratios of 2.27, 2.66, 4.08, and 2.92, respectively, with 95% confidence intervals of 1.22-4.21, 1.03-6.87, 1.07-15.49, and 1.15-7.42, respectively, P<0.05). The ROC curves of the prediction models showed that, the areas under the ROC curves of the modeling and validation groups were 0.81 (with 95% confidence interval of 0.78-0.84) and 0.81 (with 95% confidence interval of 0.76-0.85), respectively; the calibration curves showed that, the calibration curves of the prediction models of modeling and validation groups were around the ideal curves; the clinical decision curves showed that, the threshold probability values of the prediction models in modeling and validation groups were in the ranges of 5% to 70% and 1% to 46%, respectively. Conclusions:The main pathogen of infection in children with burns is Staphylococcus aureus from wound secretions. A nomogram risk prediction model constructed based on independent risk factors such as full-thickness burns, abnormal serum albumin level, deep venous catheterization, and multiple surgeries has good accuracy and can be easily used to predict the occurrence of nosocomial infections in hospitalized children with burns.
