Diese Seite ist für Menschen und Suchmaschinen gestaltet. Zur Version für KI-Systeme (LLM) →

· Science Digest (engl.) · 6 Min. · Von Jan-Patrik Novoa Lill [Qualitätssicherung: L. Mehulic]

Using AI in emergency room triage

How can analytics and AI transform human-made triage algorithms?

For most of the general public, the term “triage” had entered their vocabulary during the height of the COVID-19 pandemic, when it was mentioned on the News on, what felt like, a daily schedule. However, for most health professionals, the assessment and prioritization of patients based on the urgency of their need for treatment represents business as usual. It being such an integral aspect of daily emergency operations, it is worth asking if artificial intelligence (AI) technologies could increase accuracy and efficiency gains in the triage process. With its ability to quickly compute large and complex data inputs, AI can be a powerful and effective resource and decision support tool, both on a daily basis and particularly in high-stress scenarios.

AI in the emergency department

Bartenschlager et al. (2023) compared a standard triage algorithm recommended by the German Society for Interdisciplinary Emergency and Acute Medicine (DGINA) [1] an analytics-based extended triage algorithm [2], three AI machine learning (ML) triage algorithms (Multi-Layer Perceptron, Random Forest, and Extreme Gradient Boosting) and an integrated human-AI triage algorithm. Based on data from the Lean European Open Survey on SARS-CoV-2 infected patients (LEOSS), their study encompassed over 4,000 patients, significantly surpassing prior studies in scale [2].

The results show that AI and integrated human-AI algorithms generally exceed both standard and extended triage algorithms in most performance metrics (accuracy, sensitivity, precision, specificity). Accuracy ranged from 27% for the standard triage algorithm, 51% for the extended triage algorithm, 73% for the integrated human-AI algorithm, and up to 78% for the AI ML algorithms. Results for sensitivity, precision, and specificity overall followed this trend, if with some variation [2]. However, it should be noted that the ML algorithms where operating within a “black box”, making their decisions non-traceable and -transparent and thereby creating ethical concerns.

Limitations of the Study

Some limitations exist within the LEOSS dataset. There is a bias toward inpatients, incomplete values for some patients (addressed through approximations during imputation), and its representation of a predominantly European, particularly German, sample. Therefore, these findings mainly represent emergency room situations in developed countries. Additional diverse data is necessary, to extend these conclusions to ambulatory care settings and nations with differing healthcare standards [2].

Challenges for the implementation of AI in triage decision-making processes

It is shown that AI-driven emergency room triage surpasses humans in all measured aspects. However, ethically, AI lacks transparency and traceability in results, potentially perpetuating health disparities among different demographics. A lack of shared ethical standards in AI complicates addressing these issues [3]. Nonetheless, steps are taken to approach these challenges. For example, Hopster and Maas (2023) propose a triadic model integrating values, technology, and regulations to guide moral and legal responses to such disruptive technologies [4]. With such concerns in mind, the study at hand suggests the use of the integrated human-AI triage algorithm, as it performs similarly to the AI ML triage algorithms, however, with less transparency and traceability problems.

Once ethical concerns are addressed, implementation remains challenging. Data protection and ownership questions remain, especially in collaborations between tech giants and health professionals [5]. Furthermore, integrating these technologies with hospital systems presents significant challenges. Nonetheless, advances in healthcare digitalization in many countries, for example, Germany’s Hospital Future Act, may facilitate the adoption of digital decision support tools [2].

Conclusion

The study highlights AI’s potential in emergency room triage, aiding physicians’ decision-making. With technological and regulatory advancements, implementing these tools will likely be feasible in the near future. Addressing ongoing ethical concerns will likely involve integrated human-AI solutions [2, 6]. These approaches will help physicians in the emergency rooms expedite decision-making and enhance accuracy by swiftly processing and preparing data while allowing for human intervention if needed.

Further readings in German: Maschinelle Lernalgorithmen als neuer Screening-Ansatz für Patientinnen mit Endometriose

References

  1. Pin C, Künstler C, Jersualem K. Behandlung COVID-19 Verdachtsfälle in der Notaufnahme, modified version according to Klaus Weber, Klinikum Kassel; COVID-19 Abklärungsalgorithmus Erwachsene (nach UCSF COVID-19 ID Clinical Working Group) and Zhang et al.: Therapeutic and triage strategies for 2019 novel coronavirus disease in fever clinics. Lanc Resp Med 8(3), e11-e12. https://notfall-campus.de/wp-content/uploads/2020/03/covid-19-prozess-zina_update_02.pdf (Accessed 14.11.2023)
  2. Bartenschlager C, Grieger M, Erber J, Neidel T, Borgmann S, Vehreschild J, Steinbrecher M, Rieg S, Stecher M, Dhillon C, Ruethrich M, Jakob C, Hower M, Heller A, Vehreschild M, Wyen C, Messmann H, Piepel C, Brunner J, Hanses F, Römmele C. Covid‑19 triage in the emergency department 2.0: how analytics and AI transform a human‑made algorithm for the prediction of clinical pathways. Health Care Management Science. 2023, 26(3), 412-429. doi: 10.1007/s10729-023-09647-2
  3. Tang L, Li J, Fantus S. Medical artificial intelligence ethics: A systematic review of empirical studies. Digital Health. 2023, 9, 1-22. doi: 10.1177/20552076231186064
  4. Hopster J, Maas M. The technology triad: disruptive AI, regulatory gaps and value change. AI and Ethics. 2023. doi: 10.1007/s43681-023-00305-5
  5. Martin C, DeStefano K, Haran H, Zink S, Dai J, Ahmed D, Razzak A, Lin K, Kogler A, Waller J, Kazmi K, Umair M. The ethical considerations including inclusion and biases, data protection, and proper implementation among AI in radiology and potential implications. Intelligence-Based Medicine. 2022, 6. doi: 10.1016/j.ibmed.2022.100073
  6. Weisberg E, Chu L, Fishman E. The first use of artificial intelligence (AI) in the ER: triage not diagnosis. Emergency Radiology. 2020, 27(4), 361-366. doi: 10.1007/s10140-020-01773-6

Verwandte Artikel

Verwandte Artikel