Abstract:

Objective To systematically evaluate the risk prediction models for mechanical ventilation weaning failure. Methods Computerised searches of PubMed, Embase, Web of Science, Cochrane Library, CBM, CNKI, Wanfang and VIP databases were conducted to collect risk prediction models for weaning failure from mechanical ventilation from the establishment of the database to February 2025. The risk of bias and applicability of risk prediction models were evaluated using the prediction model risk of bias assessment tool (PROBAST) after the literatures were independently screened and data extracted by two researchers. Results Twenty-one literatures representing 44 risk prediction models for weaning failure from mechanical ventilation were included. The number of predictors ranged from 3-21, with the most common predictors being the duration of mechanical ventilation, age, and the acute physiology and chronic health evaluation Ⅱ(APACHE Ⅱ). The area under the receiver operating characteristic curve (AUC) of the included models ranged 0.689-0.926. The models had a good overall predictive performance, but the overall risk of bias was high. Conclusion Currently, the risk prediction models for weaning failure from mechanical ventilation has well overall predictive performance, but the clinical applicability of the model requires further validation because most studies have not been externally validated and there are variations in the definitions of outcome metrics.

Key words: mechanical ventilation, weaning failure, prediction model, systematic review