Construction and validation of a nomogram prediction model for death risk in patients with chronic obstructive pulmonary disease complicated by hypercapnic respiratory failure in the intensive care unit.

BACKGROUND

A nomogram prediction model was developed to estimate the death risk in patients with chronic obstructive pulmonary disease (COPD) complicated by hypercapnic respiratory failure (HRF). The prediction performance and clinical applicability were validated.

METHODS

The clinical data of 2454 COPD patients with HRF from the MIMIC-IV (Medical Information Mart for Intensive Care IV, Version 3.0) database were included and randomized into a training set (n = 1717) and a validation set (n = 737). A nomogram prediction model for the death risk was constructed using two methods: the least absolute shrinkage and selection operator (LASSO) regression analysis and the multifactorial logistic regression. The model was evaluated and validated using several analytical methods, including receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and Kaplan-Meier (KM) curves.

RESULTS

The findings indicated that age, red cell distribution width, white blood cell count, acute Physiology Score III, partial pressure of oxygen, lung cancer, vasopressor use, and lack of mechanical ventilation were independent predictors for death in COPD patients with HRF (P < 0.05). The nomogram prognosis model demonstrated that the area under the ROC curve (AUC) for predicting the death risk within 30, 60, and 90 days was 0.767 (0.738-0.796), 0.750 (0.721-0.779), and 0.737 (0.708-0.767), respectively. Calibration plots and DCA curves demonstrated strong consistency and favorable clinical applicability.

CONCLUSION

A nomogram incorporating 8 variables was developed to predict the death risk in COPD patients with HRF. It is a simple, convenient, and relatively accurate tool that can be used to guide clinical decision-making and enhance patients' outcomes.