Hostname: page-component-78c5997874-8bhkd Total loading time: 0 Render date: 2024-11-19T02:10:58.277Z Has data issue: false hasContentIssue false

The role of ChatGPT in enhancing ENT surgical training – a trainees’ perspective

Published online by Cambridge University Press:  16 August 2023

Laura Brennan*
Affiliation:
Department of ENT, St Michaels Hospital, University Hospitals Bristol and Weston, Bristol, UK
Ramkishan Balakumar
Affiliation:
Department of ENT, St Michaels Hospital, University Hospitals Bristol and Weston, Bristol, UK
Warren Bennett
Affiliation:
Department of ENT, St Michaels Hospital, University Hospitals Bristol and Weston, Bristol, UK
*
Corresponding author: Laura Brennan; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Objective

ChatGPT, developed by Open AI (November 2022) is a powerful artificial intelligence language model, designed to produce human-like text from user-written prompts. Prompts must give context-specific information to produce valuable responses. Otolaryngology is a specialist field that sees limited exposure during undergraduate and postgraduate education. Additionally, otolaryngology trainees have seen a reduction in learning opportunities since the coronavirus disease 2019 pandemic.

Method

This article aims to give guidance on optimising the ChatGPT system in the context of education for otolaryngology by reviewing barriers to otolaryngology education and suggesting ways that ChatGPT can overcome them by providing examples using the authors’ experience.

Results

Overall, the authors saw that ChatGPT demonstrated some useful qualities, particularly with regards to assistance with communication skills and individualised patient responses.

Conclusion

Although ChatGPT cannot replace traditional mentorship and practical surgical experience, it can serve as an invaluable supplementary resource to education in otolaryngology.

Type
Main Article
Copyright
Copyright © The Author(s), 2023. Published by Cambridge University Press on behalf of J.L.O. (1984) LIMITED

Introduction

ChatGPT has been developed by Open AI (November 2022) as a powerful artificial intelligence (AI) language model based on the GPT-4 (version 4) architecture.Reference Biswas1 Artificial intelligence is the term used to describe the ability of a computer system to mimic human cognitive functions, and GPT is defined as ‘generative pre-trained transformer’. ‘Generative’ refers to its ability to produce unique creative outputs that are not predefined. It was ‘pre-trained’ from large datasets to build a model that is now fine-tuned for specific tasks. ‘Transformer’ is part of the system that uses mechanisms to understand and generate human language by looking at the context of words and the relationships of these words.2 As such, ChatGPT has been designed to produce human-like text pulled from enormous datasets. Text is produced based on user-written prompts that ask the AI chatbot to generate specific information that the user requires within seconds from its vast neural network. Its applications are extensive and include essay writing, artwork and explanation of complex concepts tailored to the user's needs.Reference Patel and Lam3,Reference Eysenbach4

Within the literature, ChatGPT has not been seen to be used in the field of otolaryngology (ENT). It is a specialist field which does not feature prevalently in the UK undergraduate medical curriculum.Reference Mayer, Smith, Carrie and Authors5 Even though such a short time is dedicated to education in otolaryngology, ENT-related conditions are a common patient presentation to both primary and secondary care facilities across the UK.Reference Hayois and Dunsmore6,7 Therefore, education to medical students, general practitioners and non-specialists is especially important. In addition, during the coronavirus disease 2019 (Covid-19) pandemic, otolaryngology trainees saw a significant drop in their elective operating time,8 meaning less exposure to common procedures, and reduced time spent with experienced mentors and ability to maintain surgical skills.

Medical and surgical education has always seen these challenges, but the Covid-19 pandemic saw several new obstacles.Reference Kaul, de Moraes A, Khateeb, Greenstein, Winter and Chae9 Trainers had to adapt quickly to continue providing trainee surgeons with up to date and relevant education. This saw many surgical educational resources take a turn to virtual or simulation-based methods.Reference Khan and Mian10 Therefore, any medical professional looking to accelerate learning in the otolaryngology field may find that ChatGPT provides them with detailed contextual answers, with a myriad of examples, to help them understand complex material in a more simplified manner. The diversity of ChatGPT opens up the possibility of providing a reliable, quick and applicable educational resource for surgical trainees as the world of medical education quickly turns to a virtual format.Reference Eysenbach4

Although ChatGPT is a potential invaluable training resource, to gain the most from this system it is important to be very specific and detailed in the information that is provided to the AI chatbot. It is also imperative that this is also given in context to the situation. It is the authors’ experience that including details, such as what background the AI assumes to provide the information, the audience it is teaching and what level to pitch the information, can all be very useful. The more context and detail that is provided the more the AI chatbot can provide a tailored approach.11,Reference Balel12 In addition, errors can be generated by ChatGPT, but this can be highlighted and corrected by the chatbot. In this article, we look at these various barriers to medical education in otolaryngology training and suggest ways that ChatGPT can overcome them.

Materials and methods

Authors used the ChatGPT system to suggest the most important barriers after prompting it for such a topic.

You have been tasked to find all the ways in which ChatGPT has a role in medical education for surgical trainees in the ENT specialty. Please highlight the barriers that surgical trainees experience during their training and for each one how ChatGPT can help.

Example prompts for each barrier were then created by the authors and inputted into the GPT-4 version of ChatGPT to generate responses. Responses were then trialled by two of the authors, both at different stages of training (core training 1 and post CCT-Fellow of the Royal College of Surgeons specialty trainee level 8 ENT registrar) to provide a review from junior and senior trainee perspectives. The responses were also validated by the ENT consultant author as a third independent reviewer for accuracy of information. Therefore, examples provided used the authors’ experience to further highlight the practicalities.

Results

Limited access to experienced mentors

For otolaryngology trainees, an experienced mentor, for example a consultant surgeon, may not always be available to guide them through procedures or answer difficult clinical questions. ChatGPT can serve as a supplementary resource for otolaryngology trainees by answering questions or providing insights based on its extensive training data.11 This could potentially provide a constant ability to ‘ask the expert’ and therefore accelerate the trainees learning.Reference Sallam13

Example prompt: ‘You are a senior ENT consultant surgeon working in a university hospital in the UK, who has extensive experience in the tonsillectomy procedure. You undertake these using the bipolar electrocautery technique. You are skilled in the ability to explain complex concepts in a simple way. What are the key steps in performing a tonsillectomy?’

Interestingly, the chatbot did provide an initial disclaimer stating that it is a ‘simplified explanation, and the actual procedure is more complex and requires specialised training and expertise’. It then went on to provide a step-by-step explanation which was at a level that a medical student could understand as seen in Figure 1a. A more detailed explanation when prompted was then provided (Figure 1b).

Figure 1. (a) Snapshot of example response generated for key steps in tonsillectomy. Figure from chat.openAI.com. (b) Snapshot of example response generated for altered prompt for key steps in tonsillectomy. Figure from chat.openAI.com

Time constraints

The busy lifestyle as a surgical trainee means that time to read, study and analyse the vast amount of medical literature relevant to their speciality is limited. ChatGPT can assist in summarising medical articles or identifying key points to provide an up to date and concise resource saving hours of valuable time.Reference Kumar14

Example prompt (Figure 2): ‘You are a professor in rhinology and are up to date with all the current guidelines and new discoveries in the current specialty. You are able to critically appraise the literature and summarise detailed studies into bullet point key findings that are easy to understand by junior trainees. Summarise the main findings of the European Position Paper on Rhinosinusitis and Nasal Polyps 2012 (EPOS 2012)’.

Figure 2. Example response generated when chatbot notified of error. Figure from chat.openAI.com.

This example was chosen for ease of verifying the information provided from the AI chatbot with the original source. The AI language model does provide a simplified overview of the EPOS 2012 guidelines. The AI chatbot does however emulate a discussion with an expert in the field, allowing for opportunities to ask questions for clarification. It can also be prompted to ‘provide several examples’ to further consolidate the concepts that are mentioned.

Knowledge retention

Retaining and recalling essential information during high-pressure situations can be challenging for surgical trainees. Trainees can use ChatGPT to create customised quizzes, flashcards and revision resources to help reinforce their knowledge, providing access through their smartphone at any time.Reference Gilson, Safranek, Huang, Socrates, Chi and Taylor15

Example prompt: ‘You are an experienced ENT surgeon and expert anatomist who has been tasked with creating flash card style questions to help junior ENT trainees learn and retain knowledge on the topic of clinical anatomy of the skull base at the level of what an ENT surgeon undertaking skull base surgery should know. Please ask questions one at a time which have very specific answers and only move on the next question once answered’.

This interactive session with appropriate questions and answers gives students the ability to constantly examine themselves (Figure 3). This allows for more focused reading leading to a more efficient learning process.Reference Kung, Cheatham, Medenilla, Sillos, Leon and Elepaño16

Figure 3. Example response generated when user inputted incorrect answer to revision question. Figure from chat.openAI.com. CN = cranial nerve.

Surgical technique improvement

Developing and refining surgical skills requires practice and feedback, which may not always be readily available. Trainees can describe specific surgical techniques or scenarios to ChatGPT and receive feedback or suggestions on areas for improvement.

Example prompt: ‘I'm an ENT trainee who is having difficulty with visualisation during endoscopic sinus surgery. Can you as an expert rhinologist and excellent trainer suggest any tips to improve my technique using the 0-degree endoscope to visualise the nasal cavity during sinus surgery. Please ask any questions to clarify before starting?’

The ChatGPT can provide tips for surgical procedures using its extensive data pool of surgical literature. The prompt in this example allows the chatbot to ask more questions which can further tailor the information that is given (Figure 4).

Figure 4. Figure example response generated where chatbot establishes users’ needs. Figure from chat.openAI.com.

Communication and collaboration

Effective communication and collaboration with colleagues and other healthcare professionals are crucial but can be challenging for trainees to master. Trainees can use ChatGPT to practice various communication scenarios, such as presenting cases, discussing treatment plans or explaining procedures to patients.17

Example prompt: ‘For the duration of this conversation, please fully immerse yourself in the persona of a 30-year-old female professional singer who has a hoarse voice. Your responses should be consistent with her beliefs, knowledge and experiences of such a life. Additionally, please convey their speaking style and anxious nature during this difficult period whilst not being able to sing. You have attended an ENT outpatient clinic to see me to discuss your diagnosis. As I ask questions, please respond appropriately, providing me with a truly engaging interaction’.

The virtual patient would increase the trainee's confidence in various aspects of patient interaction, including breaking bad news and discussing management plans. The ChatGPT system generates these dialogues in a varied manner (Figure 5), meaning the same prompts can lead to unique conversations and so can provide further authenticity to the interactions.Reference Khan, Jawaid, Khan and Sajjad18

Figure 5. Example response generated from chatbot responding as a patient. Figure from chat.openAI.com.

Discussion

Overall, all authors felt that ChatGPT demonstrated some useful qualities, particularly with regards to assistance with communication skills and individualised patient responses (Table 1). ChatGPT as a revision tool was also highlighted as a main strength; the AI chatbot provided a stream of varied questions related to the prompted topic and was appropriate for all stages of training. Authors saw however that when prompted for more specific answers relating to surgical skills in otolaryngology and otolaryngological guidelines that ChatGPT was lacking. Specific concerns include the need for the user to notice mistakes in order to ask an alternative prompt and generate a corrected response, as well as the lack of detail needed for a more senior trainee learning higher surgical skills.

Table 1. Author experiences of responses generated

Author A – core trainee 1, Author B – ENT speciality registrar, Author C – ENT consultant. AI = artificial intelligence; GA = General Anaesthetic; EPOS = European Position Paper on Rhinosinusitis and Nasal Polyps.

The reliability and validity of data analysed and provided with ChatGPT has not yet been assessed. Therefore, the production of incorrect or inadequate information may be seen, especially with insufficient prompts. However, OpenAI have a clear disclaimer prior to beginning a chat warning that this may be the case. During author reviews of information generated from the same prompts we saw some variability in responses generated. Core facts remained the same overall, but wording was sometimes changed. This may be advantageous when using ChatGPT for communication skills but could be less appropriate for surgical techniques and guideline summaries. As such, with no current review boards or guidelines there is no accountability for the information provided to students.

  • Artificial intelligence and ChatGPT are still in the early stages of development, but already offer exciting promise for the development of medical education

  • ChatGPT has already been found to be useful in enhancing surgical training through summaries of large quantities of information (e.g. guidelines or literature), patient consultation simulation, and revision resources

  • So far however there are no reviews of how ChatGPT could be a potential adjunct for education in Otolaryngology

  • This article found that specific prompts currently must be written in order to generate the most useful responses

  • However, some aspects of the ChatGPT system for education in Otolaryngology are already beginning to show to be an invaluable supplementary resource

Additionally, medical education requires cost effective and economically viable resources in order to be accessible to everyone who requires it. Although ChatGPT has a free package at present, there may be additional costs in the future as it develops and is used on a commercial basis.

In summary, although the potential of this software offers exciting developments in the field of medical education there are of course limitations. In its present stage, ChatGPT requires detailed input of salient information under specific circumstances within the prompts to generate a worthwhile response. This is likely to continue to develop quickly, but in its current state would require careful expert review and possibly the development of guidance to students who are using it.

Conclusion

Even in its first 18 months of development ChatGPT has already shown the enormous potential for improving medical education by addressing common barriers to education in otolaryngology. By leveraging this powerful AI tool, trainees could enhance their knowledge, refine surgical techniques and improve communication skills, ultimately leading to better patient care and outcomes, as well as enhanced trainee satisfaction of training. It is important to stress however that much like other online resources the quality and accuracy of the information provided by the AI chatbot are variable and sometimes incorrect. Therefore, we would advise caution when undertaking such exercises. While ChatGPT cannot replace traditional mentorship and practical surgical experience, it can serve as an invaluable supplementary resource to education in otolaryngology.

Competing interests

None declared

Footnotes

Laura Brennan takes responsibility for the integrity of the content of the paper

References

Biswas, S. ChatGPT and the future of medical writing. Radiology 2023;307:e22331210.1148/radiol.223312CrossRefGoogle ScholarPubMed
ChatGPT. In: https://chat.openai.com [3 May 2023]Google Scholar
Patel, SB, Lam, K. ChatGPT: the future of discharge summaries? Lancet Digit Health 2023;5:e107–810.1016/S2589-7500(23)00021-3CrossRefGoogle ScholarPubMed
Eysenbach, G. The role of ChatGPT, Generative language models, and artificial intelligence in medical education: a conversation with ChatGPT and a call for papers. JMIR Med Educ 2023;9:e46885CrossRefGoogle Scholar
Mayer, AW, Smith, KA, Carrie, S, Authors, C. A survey of ENT undergraduate teaching in the UK. J Laryngol Otol 2020;134:553–7CrossRefGoogle ScholarPubMed
Hayois, L, Dunsmore, A. Common and serious ENT presentations in primary care. InnovAiT 2023;16:7986CrossRefGoogle Scholar
The UK otolaryngology trainees’ lived experience during the COVID-19 pandemic. In: ENT & Audiology News. https://www.entandaudiologynews.com/development/trainee-matters/post/the-uk-otolaryngology-trainees-lived-experience-during-the-covid-19-pandemic [16 April 2023]Google Scholar
Kaul, V, de Moraes A, Gallo, Khateeb, D, Greenstein, Y, Winter, G, Chae, J et al. Medical education during the Covid-19 pandemic. Chest 2021;159:1949–60CrossRefGoogle ScholarPubMed
Khan, S, Mian, A. Medical education: Covid-19 and surgery. Br J Surg 2020;107:e269CrossRefGoogle ScholarPubMed
The role of an open artificial intelligence platform in modern neurosurgical education: a preliminary study. In: https://link.springer.com/article/10.1007/s10143-023-01998-2 [17 April 2023]Google Scholar
Balel, Y. Can ChatGPT be used in oral and maxillofacial surgery? J Stomatol Oral Maxillofac Surg 2023;101471CrossRefGoogle ScholarPubMed
Sallam, M. ChatGPT utility in healthcare education, research, and practice: systematic review on the promising perspectives and valid concerns. Healthcare 2023;11:887CrossRefGoogle ScholarPubMed
Kumar, AH. Analysis of ChatGPT tool to assess the potential of its utility for academic writing in biomedical domain. BEMS Reports 2023;9:2430CrossRefGoogle Scholar
Gilson, A, Safranek, CW, Huang, T, Socrates, V, Chi, L, Taylor, RA et al. How does ChatGPT perform on the United States medical licensing examination? The implications of large language models for medical education and knowledge assessment. JMIR Medical Education 2023;9:e45312CrossRefGoogle ScholarPubMed
Kung, TH, Cheatham, M, Medenilla, A, Sillos, C, Leon, LD, Elepaño, C et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. PLOS Digital Health 2023;2:e0000198CrossRefGoogle ScholarPubMed
ChatGPT for clinical vignette generation, revision, and evaluation. In: https://www.medrxiv.org/content/10.1101/2023.02.04.23285478v1 [2 May 2023]Google Scholar
Khan, RA, Jawaid, M, Khan, AR, Sajjad, M. ChatGPT - reshaping medical education and clinical management. Pak J Med Sci 2023;39:605–7CrossRefGoogle ScholarPubMed
Figure 0

Figure 1. (a) Snapshot of example response generated for key steps in tonsillectomy. Figure from chat.openAI.com. (b) Snapshot of example response generated for altered prompt for key steps in tonsillectomy. Figure from chat.openAI.com

Figure 1

Figure 2. Example response generated when chatbot notified of error. Figure from chat.openAI.com.

Figure 2

Figure 3. Example response generated when user inputted incorrect answer to revision question. Figure from chat.openAI.com. CN = cranial nerve.

Figure 3

Figure 4. Figure example response generated where chatbot establishes users’ needs. Figure from chat.openAI.com.

Figure 4

Figure 5. Example response generated from chatbot responding as a patient. Figure from chat.openAI.com.

Figure 5

Table 1. Author experiences of responses generated