Construction and validation of a prognostic prediction model for critically ill lung cancer patients based on...

Critically ill lung cancer (LC) patients are often admitted to the Department of Respiratory and Critical Care Medicine (RCCM), the Intensive Care Unit (ICU), or the Emergency Intensive Care Unit (EICU), where early risk stratification is essential because of their high short-term mortality. Despite close monitoring, their short-term mortality remains high, and general scoring systems fail to account for tumor-specific pathology. A total of 541 patients with pathologically or cytologically confirmed LC hospitalized in the RCCM, ICU, or EICU were enrolled. They were randomly divided into a training set (n=379) and a validation set (n=162) at a 7:3 ratio. The least absolute shrinkage and selection operator (LASSO) Cox regression with 10-fold cross-validation was used for variable selection. Restricted cubic spline analysis examined nonlinearity, and the Schoenfeld residual test verified the proportional hazards assumption. Multivariable Cox regression identified five independent predictors of 28-day mortality: pulse oxygen saturation (SpO₂), baseline ratio of partial pressure of arterial oxygen to fraction of inspired oxygen (PaO₂/FiO₂), C-reactive protein (CRP) level, hemoglobin level, and bone metastasis (the latter possibly due to its association with pneumonia). These were incorporated into a nomogram. In the training set, the area under the curve (AUC) for 14-day and 28-day mortality was 0.871 and 0.947, respectively, with a concordance index (C-index) of 0.874. In the validation set, the AUCs were 0.870 and 0.946, with a C-index of 0.852. Calibration curves and Brier scores demonstrated good agreement between predicted and observed outcomes. Decision curve analysis demonstrated net clinical benefit across a wide range of threshold probabilities. The 28-day survival rates for low-, intermediate-, and high-risk groups were 98.9%, 82.0%, and 13.8%, respectively (P<0.001). DeLong testing indicated that our model outperformed Acute Physiology and Chronic Health Evaluation II (APACHE II) in both datasets (all P<0.05). This nomogram, based on five routine clinical variables, offers good discrimination, calibration, and risk stratification, serving as a practical tool for early risk assessment and clinical decision-making in critically ill LC patients.