Classification of patient-ventilator asynchronies, development of an updated conceptual framework: Scoping review

Patient-ventilator asynchrony (PVA), a mismatch between a patient's respiratory effort and the ventilator's support, affects up to 85% of patients receiving invasive mechanical ventilation (IMV). It is associated with prolonged ventilation, increased complications, and higher mortality. Existing classification systems for PVA are inconsistent, difficult to apply in clinical settings, and not validated for integration with advanced technologies. This scoping review aimed to evaluate current PVA classification systems and propose an operational, flexible framework suitable for intensive care unit (ICU) settings and future integration with artificial intelligence (AI).

A comprehensive literature search was conducted across major databases and gray literature sources, including studies on adult ICU patients undergoing IMV. The review followed Joanna Briggs Institute methodology. Studies were included if they proposed, evaluated, or validated PVA classifications. Two reviewers independently selected and analyzed the studies; a third reviewer resolved disagreements.

Fifteen studies met inclusion criteria. Most demonstrated methodological heterogeneity and limited external validity. Approaches included waveform analysis, respiratory drive assessment, and AI-based models such as convolutional neural networks. However, inconsistencies in terminology, thresholds, and detection methods limited their clinical implementation. Based on the identified gaps, we developed a novel five-level hierarchical classification framework that integrates clinical context, waveform analysis, functional categorization, operational subtypes, and severity assessment.

Current PVA classification systems lack standardization and practical applicability. This review supports the development of a novel five-level classification framework designed to be operational, ventilator-mode agnostic, reproducible, and adaptable for future artificial intelligence-based detection systems, thereby facilitating both clinical application and research standardization. Such a system can enhance bedside decision-making, promote consistency across studies, and enable future AI-based automation in PVA detection and management.