Real-Time Predictive Analysis of ICU Ventilator Weaning Failure: A Prospective Validation Study

To construct and validate a nomogram model for predicting the risk of ICU ventilator weaning failure (reintubation or death within 48 h after extubation) based on multidimensional clinical indicators.

A total of 485 patients in the ICU who needed ventilator weaning were selected and divided into a training set (n = 340) and a validation set (n = 145) at a ratio of 7:3. Baseline data and weaning-related indicators of the patients were collected. Weaning failure (reintubation or death within 48 h after weaning) was regarded as the outcome event. Independent risk factors were screened through univariate and multivariate logistic regression. A nomogram model was constructed, and the model's performance was evaluated using the C-index, AUC, calibration curve, Hosmer-Lemeshow test, and decision curve.

The weaning failure rate in the training set was 28.53% (97/340), and that in the validation set was 29.66% (43/145). Multivariate regression showed that age, APACHE II score, duration of mechanical ventilation, spontaneous breathing frequency, Glasgow Coma Scale score, and the use of sedatives were independent influencing factors (p < 0.05). The C-index of the nomogram model in the training set and the validation set was 0.829 and 0.826, respectively. The AUC was 0.828 (95% CI: 0.762-0.893) and 0.823 (95% CI: 0.732-0.915), respectively. The sensitivity and specificity were 0.811, 0.751 and 0.662, 0.702, respectively. The calibration curve and Hosmer-Lemeshow test (p = 0.109, 0.402) showed that the model had a good fit.

The nomogram model constructed based on the above indicators can effectively predict the risk of ICU ventilator weaning failure and provide a basis for formulating individualized weaning strategies.