Comprehensive systematic review and meta-analysis: Evaluating artificial intelligence (AI) effectiveness and integration obstacles within anesthesiology
Umfassende systematische Übersichtsarbeit und Metaanalyse: Wirksamkeit von künstlicher Intelligenz (KI) und Integrationshindernisse in der Anästhesiologie
Journal of Emergency Medicine, Trauma & Acute Care · 14.05.2025 · 2025 (2)
DOI: 10.5339/jemtac.2025.22 ↗Auf einen Blick
20 Studien zeigen eine gute Trennschärfe von KI zur Vorhersage von Hypotonie und Hypoxämie. Für PONV war die Trennschärfe geringer; robotisch assistierte Intubation und Nervenblockaden waren meist erfolgreich.
Fragestellung
Wie wirksam ist KI in der Anästhesie für die Vorhersage anästhesiebezogener Ereignisse und die Unterstützung von Prozeduren?
Methode
Systematische Suche englischsprachiger Arbeiten bis Juni 2024 in PubMed, Google Scholar, IEEE Xplore und Web of Science; Metaanalyse mit CMA und STATA 16.0; Qualität mit Newcastle–Ottawa Scale bewertet.
Ergebnisse
20 eingeschlossene Studien. AUROC für Hypotonie 0,81; im Subgruppenvergleich bei Modellen mit Machine Learning 0,93. AUROC für Hypoxämie 0,81, für PONV 0,68. Erfolgsrate robotischer Intubation: 98 % in Mannequins, 92 % in Menschen; robotisch assistierte ultraschallgeführte Blocks: 96 % in Menschen und Mannequins.
Einschränkungen
Hindernisse für die Integration von KI: Kosten, Datenschutz- und Sicherheitsbedenken, Datenqualität, „Black Box“ und ethische Fragen.
Schlussfolgerung der Autor:innen
Die Autor:innen schließen, dass KI für die Vorhersage anästhesiebezogener Ereignisse und für die Automatisierung von Intubation und ultraschallgeführten Nervenblockaden nützlich ist, die Integrationshindernisse aber adressiert werden müssen.
Kernaussagen
- KI sagte Hypotonie und Hypoxämie mit AUROC 0,81 voraus.
- Für PONV war die Trennschärfe geringer (AUROC 0,68).
- Robotische Intubation und Nervenblockaden zeigten hohe Erfolgsraten.
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Systematische Übersicht in der Zeitschrift Journal of Emergency Medicine, Trauma & Acute Care, erschienen am 14.05.2025. Verfasst von Hany A. Zaki und 6 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 96 Literaturangaben.
Abstract (Originalsprache)
Background
Artificial intelligence (AI) is a multidisciplinary field that focuses on developing intelligent computer algorithms to carry out simple to complex tasks traditionally performed using human intelligence. In anesthesia, AI is rapidly becoming a transformative technology. However, its efficacy remains unknown. Therefore, this study aims to analyze the efficacy of AI in anesthesia by studying two main applications of AI: predicting anesthesia-related events and assisting with anesthesia-related procedures.
Methods
A systematic search was performed for English-language records published from inception until June 2024 in PubMed, Google Scholar, IEEE Xplore, and Web of Science databases. Studies were included in this meta-analysis if they examined the role of any AI model in predicting hypotension, hypoxemia, and post-operative nausea and vomiting (PONV). Moreover, studies investigating the role of AI in guiding anesthesia-related procedures, such as tracheal intubation and ultrasound-guided nerve blocks, were included. The CMA software and STATA 16.0 were used for statistical analyses, while the Newcastle–Ottawa Scale was used for quality evaluation.
Results
Twenty studies meeting the eligibility criteria were included in the analysis. The pooled results indicated that AI demonstrated good discrimination ability in predicting hypotension (area under the receiver operating characteristic curve (AUROC): 0.81), with the subgroup analysis showing that models incorporating machine-learning algorithms outperformed other models (AUROC: 0.93). Similarly, the pooled analysis showed that AI models had a good discriminatory capacity for predicting hypoxemia (AUROC: 0.81). However, AI demonstrated poor discriminatory capability in predicting PONV (AUROC: 0.68). Our analysis also showed that robotically assisted intubations were successful in both mannikins and humans (success rate: 98% and 92%). Similarly, robotically assisted ultrasound-guided blocks were successful in mannikins and humans (success rate: 96% for humans and mannikins).
Conclusion
This study suggests that AI is useful for predicting anesthesia-related events and automating procedures such as intubation and ultrasound-guided nerve blocks. However, multiple barriers hindering the integration of AI into anesthesia, such as cost, privacy and security concerns, data quality, “black box”, and ethical issues need to be addressed.
Zitieren
APA
Zaki, H. A., Elmelliti, H., Shaban, E. E., Shaban, A., Shaban, A., Elgassim, M., & Shallik, N. (2025). Comprehensive systematic review and meta-analysis: Evaluating artificial intelligence (AI) effectiveness and integration obstacles within anesthesiology. Journal of Emergency Medicine, Trauma & Acute Care, 2025(2). https://doi.org/10.5339/jemtac.2025.22
Vancouver
Zaki HA, Elmelliti H, Shaban EE, Shaban A, Shaban A, Elgassim M, et al. Comprehensive systematic review and meta-analysis: Evaluating artificial intelligence (AI) effectiveness and integration obstacles within anesthesiology. Journal of Emergency Medicine, Trauma & Acute Care. 2025;2025(2). doi:10.5339/jemtac.2025.22
BibTeX
@article{Zaki2025,
author = {Zaki, Hany A. and Elmelliti, Hussam and Shaban, Eman E. and Shaban, Ahmed and Shaban, Amira and Elgassim, Mohamed and Shallik, Nabil},
title = {Comprehensive systematic review and meta-analysis: Evaluating artificial intelligence (AI) effectiveness and integration obstacles within anesthesiology},
journal = {Journal of Emergency Medicine, Trauma & Acute Care},
year = {2025},
volume = {2025},
number = {2},
doi = {10.5339/jemtac.2025.22}
}
EUVIMED