Classification of videolaryngoscopy: a systematic review

Structured classification and consistent documentation of videolaryngoscopy findings are essential for future airway planning. Although numerous classification tools for communication have been proposed, their accuracy and effectiveness remain uncertain. We aimed to synthesise the evidence on the accuracy, performance and inter-rater reliability of existing tools used to classify videolaryngoscopic tracheal intubation.

Databases were searched for original articles framed by the Participants - Index test - Target conditions criteria. We extracted accuracy, performance, concordance and reliability to evaluate the generalisability and clinical utility of each classification. We applied the QUADAS-2 tool for quality assessment.

We identified 13 eligible studies evaluating characteristics of seven distinct classification tools in adults and children, including the Cormack-Lehane classification; percentage of glottic opening score; intubation difficulty scale; Fremantle score; video classification of intubation score; videolaryngoscopic intubation and difficult airway classification (VIDIAC) score; and paediatric difficult airway classification (PeDiAC) score. Eight publications reported clinical studies in patients. Since many studies reported only inter-rater reliability or concordance, the accuracy, performance, generalisability and clinical utility of the respective classification tools remained uncertain. Accuracy or performance metrics were reported for three classification tools - the Cormack-Lehane classification, VIDIAC and PeDiAC scores - with the first showing only limited performance. Empirically-derived thresholds were not reported for most classification tools. Overall risk of bias was high among the reviewed studies, particularly regarding the reference standard, flow and timing, accompanied by substantial applicability concerns.

Despite widespread dissemination of various proposed classification tools, current evidence remains limited. Accuracy, performance and thresholds have been evaluated for only three classification tools, while remaining largely unknown for the other four tools. Further high-quality research is needed.