AUTHOR=He Yaqiong , Liu Peng , Xie Leyun , Zeng Saizhen , Lin Huashan , Zhang Bing , Liu Jianbin TITLE=Construction and Verification of a Predictive Model for Risk Factors in Children With Severe Adenoviral Pneumonia JOURNAL=Frontiers in Pediatrics VOLUME=10 YEAR=2022 URL=https://www.frontiersin.org/journals/pediatrics/articles/10.3389/fped.2022.874822 DOI=10.3389/fped.2022.874822 ISSN=2296-2360 ABSTRACT=Objective

To construct and validate a predictive model for risk factors in children with severe adenoviral pneumonia based on chest low-dose CT imaging and clinical features.

Methods

A total of 177 patients with adenoviral pneumonia who underwent low-dose CT examination were collected between January 2019 and August 2019. The assessment criteria for severe pneumonia were divided into mild group (N = 125) and severe group (N = 52). All cases divided into training cohort (N = 125) and validation cohort (N = 52). We constructed a prediction model by drawing a nomogram and verified the predictive efficacy of the model through the ROC curve, calibration curve and decision curve analysis.

Results

The difference was statistically significant (P < 0.05) between the mild adenovirus pneumonia group and the severe adenovirus pneumonia group in gender, age, weight, body temperature, L/N ratio, LDH, ALT, AST, CK-MB, ADV DNA, bronchial inflation sign, emphysema, ground glass sign, bronchial wall thickening, bronchiectasis, pleural effusion, consolidation score, and lobular inflammation score. Multivariate logistic regression analysis showed that gender, LDH value, emphysema, consolidation score, and lobular inflammation score were severe independent risk factors for adenovirus pneumonia in children. Logistic regression was employed to construct clinical model, imaging semantic feature model, and combined model. The AUC values of the training sets of the three models were 0.85 (0.77–0.94), 0.83 (0.75–0.91), and 0.91 (0.85–0.97). The AUC of the validation set was 0.77 (0.64–0.91), 0.83 (0.71–0.94), and 0.85 (0.73–0.96), respectively. The calibration curve fit good of the three models. The clinical decision curve analysis demonstrates the clinical application value of the nomogram prediction model.

Conclusion

The prediction model based on chest low-dose CT image characteristics and clinical characteristics has relatively clear predictive value in distinguishing mild adenovirus pneumonia from severe adenovirus pneumonia in children and might provide a new method for early clinical prediction of the outcome of adenovirus pneumonia in children.