Abstract
Identifying effective indicators and developing predictive models for the early detection of severe adenoviral pneumonia (SAP) is critical to safeguarding patients' lives. This study examined differences between 428 patients with SAP and those with non-severe adenoviral pneumonia (NSAP) from March 2022 to January 2023, focusing on variables such as age, sex, type of coinfection, and a range of clinical laboratory indicators. SAP was significantly more common in children aged 3-6 years (20/54 of all SAP cases, p = 0.0258) and among those with polymicrobial coinfections (p < 1.20 × 10(-11)). Patients with SAP exhibited significantly higher prealbumin (PA) level, while C-reactive protein (CRP) level was significantly lower. Composite indicator, such as CRP -to- prealbumin ratio (CPAR), was also significantly elevated (p < 0.05). The random forest model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.699, with an accuracy of 84.5% and a precision of 91.5%. Analysis of the data revealed key predictive parameters for early-stage SAP. Indicators such as CPAR, PA, and CRP are valuable for assessing SAP risk. Moreover, commonly available clinical indicators can effectively construct a random forest-based predictive model for SAP.