Development and validation of a prediction model for adverse outcomes in viral pneumonia using CT-based quantitative body composition analysis.

BACKGROUND

Given the association between body fat distribution and prognosis in viral pneumonia, computed tomography (CT) -as a routine imaging modality-offers an opportunity to quantify body composition and enhance risk stratification. This study aimed to investigate the association between CT-based quantitative body composition measurements and adverse outcomes in patients with viral pneumonia, and to develop and validate a predictive model for adverse outcomes.

METHODS

This retrospective study included 140 hospitalized patients with viral pneumonia admitted to Fushun Central Hospital between December 2022 and February 2023. Clinical parameters, including lymphocyte percentage, neutrophil percentage, and C-reactive protein (CRP), were collected. Quantitative body composition measurements-specifically, muscle, subcutaneous adipose tissue, and visceral adipose tissue volumes-were obtained using 3D Slicer software at the T4, T8, and T12 vertebral levels. Patients were randomly assigned to a training set (n=98) and a test set (n=42) in a 7:3 ratio. Logistic regression was used to construct predictive models. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA), evaluating discrimination, calibration, and clinical utility, respectively.

RESULTS

Among the 140 patients, 26 (18.6%) experienced adverse outcomes, while 114 (81.4%) had favorable outcomes. Significant differences were observed between the two groups in sex, age, oxygen therapy status, and several laboratory and imaging parameters (all P<0.05). Multivariate logistic regression identified T12 visceral adipose tissue volume (T12VAV) and standardized T4 subcutaneous adipose tissue volume (ST4SAV) as independent imaging predictors of adverse outcomes. Based on these factors, an imaging model (T12VAV + ST4SAV) and a combined model incorporating clinical features were constructed. The combined model demonstrated superior predictive performance, with area under the curve (AUC) values of 0.852 and 0.882 in the training and test sets, respectively. Calibration curves showed good agreement between predicted probabilities and observed outcomes, and DCA confirmed its favorable clinical utility. Sensitivity analyses using Firth's penalized regression and confounding assessment supported the robustness of the model.

CONCLUSIONS

The nomogram model integrating CT-based quantitative body composition measurements with clinical features effectively predicts adverse outcomes in patients with viral pneumonia, offering a non-invasive tool for early risk stratification and clinical decision-making.