Speech recognition application for surgical safety checklist verification: A systematic review

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Surgical checklists (SCs) are designed to reduce medical errors and improve staff coordination and accountability. However, the effectiveness of SCs depends on strict adherence to the rules for their use. Voice assistants (VAs) based on artificial intelligence may improve patient safety during surgical verification procedures, but data are fragmentary. The aim of the review was to examine VAs designed to record speech in the operating room, check SCs items, and interact with the surgical team. From 102 references identified, 44 publications were selected for full-text analysis. Of these, four studies met the selection criteria. One was a clinical trial comparing VA with manual SC completion, while the others were descriptions of VA pilot testing. The risk of bias for non-randomized studies, according to the ROBINS-I criteria, was considered high for the comparative study and moderate for the descriptive study. All studies used different automated speech recognition (ASR) models; one additionally used speaker verification (SV) and natural language processing technologies. VA accuracy reported in three studies ranged from 90.7% to 97.3% for ASR; 94.1% for SV; 92.8% to 94.8% for formalized response /word verification. VA usability was rated positively in one study (System Usability Scale score 76.04). In a comparative study, VA usability was rated higher than manual SC completion (p0.05). In the same study, VA increased the SC full completion by 16.4% (from 81.8% to 92.5%); the odds ratio for SC full completion was 2.74 (95% CI: 1.55–4.82; p=0.00032). No studies described recording adverse events with VA. Thus, analyzed studies confirmed the feasibility of voice-based SC completion with sufficient accuracy and intelligent support for surgical verification. However, a lack of high-quality studies confirming the effectiveness of the GA in ensuring patient safety during surgical interventions, identifying and preventing adverse events, and deviations from the surgical process