Skip to main navigation Skip to main content

CEEM : Clinical and Experimental Emergency Medicine

OPEN ACCESS
ABOUT
BROWSE ARTICLES
FOR CONTRIBUTORS

Articles

Original Article
Airway

Machine learning for the prediction of preclinical airway management in injured patients: a registry-based trial

Clinical and Experimental Emergency Medicine 2022;9(4):304-313.
Published online: November 23, 2022

1Department of Anesthesiology, Intensive Care and Emergency Medicine, Ludwigshafen Municipal Hospital, Ludwigshafen, Germany

2Center for Quality Management in Emergency Medical Services Baden-Wuerttemberg (SQR-BW), Stuttgart, Germany

3Department of Anesthesiology and Intensive Care Medicine, University Medical Center Mannheim, Mannheim, Germany

4Clinic for Anesthesia, Intensive Care and Pain Therapy, BG Trauma Center Tuebingen, Tuebingen, Germany

Correspondence to: André Luckscheiter Department of Anesthesiology, Intensive Care and Emergency Medicine, Ludwigshafen Municipal Hospital, Bremserstrasse 79, Ludwigshafen 67063, Germany E-mail: uckscha@klilu.de
• Received: June 16, 2022   • Revised: September 11, 2022   • Accepted: October 16, 2022

Copyright © 2022 The Korean Society of Emergency Medicine

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/).

  • 8,267 Views
  • 210 Download
  • 12 Web of Science
  • 11 Crossref
  • 14 Scopus
prev next

Citations

Citations to this article as recorded by  Crossref logo
  • Artificial intelligence-driven predictive analytics for postoperative management and recovery in trauma patients
    Olivier Duranteau, David Leon
    Current Opinion in Anaesthesiology.2026; 39(2): 154.     CrossRef
  • Artificial Intelligence in Trauma Care: A Systematic Review of Resuscitation, Diagnosis, Risk Prediction, and Management
    Sukriti Prashar, Youssef Nasef, Alexander Brown, Cameron Nishida, Logan Samuel Rogers, Ian Bundschu, Ruth Zagales, Alexandra Kata, Adel Elkbuli
    Journal of Trauma Nursing.2026; 33(3): 160.     CrossRef
  • From decision support to clinical integration: A scoping review of artificial intelligence in prehospital airway management
    Bai Fangfang, Qiu Wenjuan, Zhu Xiaoting, Feng Yanghui
    International Journal of Medical Informatics.2026; 216: 106480.     CrossRef
  • Automatic Detection of Physiological Attributes from Verbal Communication During Time-Critical Medical Events
    Chenyang Gao, Wenjin Zhang, Aaron H. Mun, Aleksandra Sarcevic, Mary S. Kim, Randall S. Burd, Ivan Marsic
    ACM Transactions on Computing for Healthcare.2026;[Epub]     CrossRef
  • Development and Validation of a Bayesian Network Predicting Intubation Following Hospital Arrival Among Injured Children
    Travis M. Sullivan, Mary S. Kim, Genevieve J. Sippel, Waverly V. Gestrich-Thompson, Caroline G. Melhado, Kristine L. Griffin, Suzanne M. Moody, Rajan K. Thakkar, Meera Kotagal, Aaron R. Jensen, Randall S. Burd
    Journal of Pediatric Surgery.2025; 60(2): 161888.     CrossRef
  • Human intention recognition for trauma resuscitation: An interpretable deep learning approach for medical process data
    Keyi Li, Mary S. Kim, Wenjin Zhang, Sen Yang, Genevieve J. Sippel, Aleksandra Sarcevic, Randall S. Burd, Ivan Marsic
    Journal of Biomedical Informatics.2025; 161: 104767.     CrossRef
  • Utilization of non-invasive ventilation before prehospital emergency anesthesia in trauma – a cohort analysis with machine learning
    André Luckscheiter, Manfred Thiel, Wolfgang Zink, Johanna Eisenberger, Tim Viergutz, Verena Schneider-Lindner
    Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine.2025;[Epub]     CrossRef
  • From algorithms to airways: Applying artificial intelligence to enhance airway assessment, management, and training
    Mingzhu Guo, Yongheng Hou, Yan Liu, Bo Yang, Chuhan Qiao, Jian Li
    Trends in Anaesthesia and Critical Care.2025; 61: 101548.     CrossRef
  • Deep Learning during burn prehospital care: An evolving perspective
    Mohammad Vakili Ojarood, Ramyar Farzan, Seyed Mostafa Mohsenizadeh, Hossein Torabi, Tahereh Yaghoubi
    Burns.2024; 50(5): 1349.     CrossRef
  • Use of artificial intelligence to support prehospital traumatic injury care: A scoping review
    Jake Toy, Jonathan Warren, Kelsey Wilhelm, Brant Putnam, Denise Whitfield, Marianne Gausche‐Hill, Nichole Bosson, Ross Donaldson, Shira Schlesinger, Tabitha Cheng, Craig Goolsby
    JACEP Open.2024; 5(5): e13251.     CrossRef
  • Unravelling intubation challenges: a machine learning approach incorporating multiple predictive parameters
    Parisa Sezari, Zeinab Kohzadi, Ali Dabbagh, Alireza Jafari, Saba Khoshtinatan, Kamran Mottaghi, Zahra Kohzadi, Shahabedin Rahmatizadeh
    BMC Anesthesiology.2024;[Epub]     CrossRef

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

Include:

Machine learning for the prediction of preclinical airway management in injured patients: a registry-based trial
Clin Exp Emerg Med. 2022;9(4):304-313.   Published online November 23, 2022
Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:
Include:
Machine learning for the prediction of preclinical airway management in injured patients: a registry-based trial
Clin Exp Emerg Med. 2022;9(4):304-313.   Published online November 23, 2022
Close

Figure

  • 0
  • 1
  • 2
Machine learning for the prediction of preclinical airway management in injured patients: a registry-based trial
Image Image Image
Fig. 1. Flowchart for patient selection, dataset creation, and analysis. SQR-BW, Center for Quality Management in Emergency Medical Services Baden-Wuerttemberg; MIND, minimal emergency dataset; SMOTE, synthetic minority oversampling method. a)A total of 24 attributes included: >550 attributes filtered by causality or potential correlation, then selected by principal component analysis (Wrapper).
Fig. 2. Averaged (A) receiver operator curves for the overall performance and (B) precision-recall curves for the prediction of airway management by the Naive Bayes and random forest algorithms. AUC, area under the curve; CI, confidence interval.
Fig. 3. Attribute weighting in the random forest model, given as means with standard deviation error bars.
Machine learning for the prediction of preclinical airway management in injured patients: a registry-based trial
Attribute Airway management
P-value
Yes (n=1,451) No (n=24,105)
Auscultation 1 < 0.01a)
 Obstruction/gasping/apnea 15.0 0.3
 Bronchial spasm 18.0 0.3
 Rhonchi 2.0 0.2
 Other 31.0 13.0
Auscultation 2
 Dyspnea ± cyanosis 37.0 3.0 < 0.01a)
Head injury
 None 36.0 65.0 < 0.01a)
 Mild 5.0 20.0 < 0.01a)
 Moderate 15.0 13.0 0.05a)
 Severe 44.0 2.0 < 0.01a)
Face injury < 0.01a)
 None 77.0 83.0
 Mild 5.0 10.0
 Moderate 11.0 7.0
 Severe 7.0 0.6
Cervical spine injury
 None 90.0 88.0 0.07
 Mild 2.0 6.0 < 0.01a)
 Moderate 4.0 5.0 0.40
 Severe 4.0 0.7 < 0.01
Thoracic/lumbar spine injury < 0.01a)
 None 90.0 85.0
 Mild 1.0 5.5
 Moderate 4.0 8.0
 Severe 4.0 1.0
Thoracic injury
 None 68.0 77.0 < 0.01a)
 Mild 3.0 9.0 < 0.01a)
 Moderate 10.0 12.0 0.10
 Severe 19.0 2.0 < 0.01a)
Abdominal injury
 None 85.0 92.0 < 0.01a)
 Mild 1.0 2.0 < 0.01a)
 Moderate 4.0 4.0 0.50
 Severe 10.0 1.0 < 0.01a)
Pelvic injury
 None 83.0 87.0 < 0.01a)
 Mild 2.0 5.0 < 0.01a)
 Moderate 4.0 6.0 0.02a)
 Severe 1.0 2.0 < 0.01a)
Lower limb injury < 0.01
 None 76.0 72.0
 Mild 4.0 12.0
 Moderate 7.0 12.0
 Severe 13.0 3.0
Oxygen therapy 57.0 35.0 < 0.01a)
Noninvasive ventilation 32.0 0.2 < 0.01a)
Thoracic drainage 14.0 0.2 < 0.01a)
Colloid infusion 7.0 0.2 < 0.01a)
Tranexamic acid 40.0 4.0 < 0.01a)
Pelvic sling 27.0 4.0 < 0.01a)
Catecholamine 43.0 1.0 < 0.01a)
 Systolic blood pressure (mmHg) 137 ± 29 138 ± 28 0.29
 Oxygen saturation (%) 94 ± 7 95 ± 6 < 0.01a)
 Heart rate (beats/min) 90 ± 20 89 ± 19 < 0.01a)
 Respiratory rate (breaths/min) 16 ± 5 16 ± 5 0.24
Pain level (0–10)b) 5 (0–10) 5 (0–10) < 0.01a)
Shock index 0.7 ± 0.3 0.6 ± 0.6 0.03a)
Preemergency status (1–4)c) 2 (1–3) 2 (1–3) < 0.01a)
Glasgow Coma Scale (3–15) 15 (14–15) 15 (15–15) < 0.01a)
Age (yr)d) 54.88 ± 21.44 55.80 ± 22.28 0.13
Male sexd) 72.0 60.0 < 0.01a)
Variable Random forest Naive Bayes P-value
Overall correctness (%) 97.80 ± 0.37 (97.57–98.03) 93.55 ± 0.71 (93.11–93.99) < 0.01a)
Kappa 0.78 ± 0.04 (0.75–0.80) 0.54 ± 0.03 (0.52–0.56) < 0.01a)
AUC-ROC 0.96 ± 0.01 (0.96–0.97) 0.93 ± 0 (0.92–0.93) < 0.01a)
MCC 0.78 ± 0.04 (0.76–0.80) 0.56 ± 0.02 (0.54–0.57) < 0.01a)
Sensitivity 0.73 ± 0.05 (0.71–0.76) 0.75 ± 0.02 (0.73–0.76) 0.38
 Positive predictive value 0.85 ± 0.03 (0.84–0.87) 0.46 ± 0.03 (0.44–0.49) < 0.01a)
 PRC areab) 0.83 ± 0.04 (0.80–0.85) 0.66 ± 0.09 (0.61–0.72) < 0.01a)
Specificity 0.993 ± 0.002 (0.992–0.994) 0.947 ± 0.008 (0.942–0.952) < 0.01a)
 Negative predictive value 0.984 ± 0.006 (0.980–0.987) 0.984 ± 0.001 (0.983–0.985) 0.85
 PRC areab) 0.996 ± 0.001 (0.996–0.997) 0.992 ± 0.001 (0.992–0.993) < 0.01a)
Table 1. Clinical findings and medical treatments for both classes with the attributes selected through the principal component analysis

Values are presented as percentage, mean±standard deviation, or median (interquartile range).

Statistically significant value (P<0.05).

No pain, 0.

Healthy, 1; moribund, 4.

Baseline characteristics not used in the algorithm.

Table 2. Model performance and evaluation of random forest versus Naive Bayes

Values are presented as standard deviation (95% confidence interval).

AUC-ROC, area under the receiver operator curve; MCC, Matthews correlation coefficient; PRC, precision-recall.

Statistically significant value (P<0.05).

Given for the prediction and exclusion of airway management.