Objective:

To develop and validate a predictive model for tracheostomy in mechanically ventilated stroke patients and investigate the impact of early versus late tracheostomy on in-hospital outcomes.

Approach:

• Study Design: Retrospective study of 508 mechanically ventilated stroke patients, divided into tracheostomy and non-tracheostomy groups.
• Data Analysis: Used LASSO regression and Random Forest feature importance to identify predictors; multivariable logistic regression for independent predictors; nomogram construction; model performance evaluated with ROC curves, calibration curves, and DCA.
• IPTW Analysis: Applied inverse probability of treatment weighting to assess early versus late tracheostomy effects.

Key Findings:

• Independent predictors for tracheostomy included midline shift, hypoalbuminemia, admission GCS score, CRP, and PNI (p < 0.05).
• The area under the curve (AUC) was 0.855 in the training set and 0.849 in the validation set, indicating good discriminative ability.
• Early tracheostomy significantly reduced ICU length of stay (p < 0.05) but did not significantly affect post-tracheostomy ventilation duration, antibiotic use, total hospital stay, or hospitalization costs.

Interpretation:

Limitations:

• Retrospective design may introduce bias.
• Single-center study limits generalizability.

Conclusion: