Ai And The Evolution Of Medical Writing: Improving Accuracy And Reducing Workloads

  • Daniel Reidy Immunology, Immunology, University of Glasgow
  • Gabriella Bryant MSci Bioveterinary Sciences, Royal Veterinary College
  • Saira Loane Master's of Toxicology, Institute of Biomedical Research, University of Birmingham

Get health & wellness advice into your inbox

Your privacy is important to us. Any information you provide to us via this website may be placed by us on servers. If you do not agree to these placements, please do not provide the information.

Best Milk Alternative

Introduction

Medical writing is a field of work dedicated to the communication of scientific research, ideas, and observations to a wider audience, and is essential for scientific and medical advancements. Since the public introduction of ChatGPT (an AI assistant trained on an enormous amount of online data) in late 2022, multiple fields across scientific and non-scientific disciplines have rapidly evolved – including medical writing. What was once a time-consuming task to draft, edit, and publish manuscripts, articles, journals, letters, etc. has suddenly become significantly quicker and easier, thanks to the explosion of AI tools.

When used well, AI tools can greatly reduce the workloads involved with medical writing. Not only that, but the accuracy of the statements being made can be improved too, as the “clumsiness” of human authorship is ironed out. Medical writing is required to be factual and credible, so authors must be careful not to misuse AI tools. As these tools are relatively new, the writing community’s understanding of how they work and how best to use them is still being developed. There is no agreed-upon method for using AI in medical writing, which has led to much interest and experimentation, but also concern. AI is here to stay, so the field of medical writing must learn its strengths, navigate its weaknesses, and carve out clear guidelines and policies to minimise its misuse and maximise accuracy, efficiency, and credibility to advance medical science.

What is medical writing?

Medical writing1 is the written communication of scientific research, ideas, and observations. There is a wide array of types of medical writing, including publishing data in scientific journals, writing and editing regulatory documentation, preparing presentations and posters for conferences, and translating complicated scientific concepts into digestible language for marketing purposes or educating lay audiences. This multidisciplinary field bridges the gap between hard science and a range of audiences and is essential for making scientific and medical advancements, such as drug development and healthcare regulation.

What is AI?

Artificial intelligence,2 AI for short, is a technology that allows computers to learn and solve problems in ways that simulate human behaviour. Although AI has existed since the 1950s (although primitively), throughout the decades it has risen to prominence in mathematical fields. More recently, we can find it in the background of everyday websites, apps, and technology such as predictive text and advertising algorithms. However, the introduction of ChatGPT in late 2022 catapulted the use of AI into the spotlight.

What is ChatGPT?

ChatGPT4 is an AI tool created by the company OpenAI that comes under the category of Large Language Models (LLMs). The “GPT” in its name stands for Generative Pretrained Transformer, which means it was trained on an enormous quantity of online data using Neural Network algorithms (which analyse complex datasets and patterns) including media, articles, and books.5 Using this data, ChatGPT (and other such LLMs) can interpret and generate human language – in a sense, it “learns” to talk like a human.

ChatGPT became publicly available on 30th November 2022 and is free-to-use requiring minimal setup or hardware prerequisites. With such global reach and accessibility, the use of ChatGPT caught the attention of people in every field of work. Millions have and are engaging with this powerful tool, with no sign of slowing down. Therefore, its use in medical writing needs to be examined and carefully considered.

Improving accuracy and reducing workloads – how AI is used in medical writing

AI tools, particularly LLMs, are a natural fit when it comes to being utilised by medical writers. The subheadings below outline a few of the main ways in which AI is being used to reduce workloads and improve the accuracy of scientific and medical documentation and written communication.

Overcoming “blank-sheet syndrome”

LLMs like ChatGPT can generate large amounts of coherent text within seconds, which helps to overcome Blank-Sheet Syndrome (also known as Writer’s Block).6 In fact, these AI tools are especially good at generating high-quality blocks of text using only short notes and ideas. One study from 2024 that examined the use of ChatGPT in medical writing likened this to assembling a jigsaw puzzle.6

Not only does AI help medical writers generate first drafts and bring their ideas to life on the page rapidly and in a structured manner, but it can also be used to manage these ideas, which is especially useful for citations. Scientific literature can become lengthy and AI tools are a way for medical writers to accurately track their use of citations and reference them correctly to fit publishing and regulatory guidelines.7

Drawback – lack of scientific depth

Most LLMs, like ChatGPT, are currently trained on conversational databases from public sources. This allows them to imitate natural human language, but it comes at the cost of scientific depth. One study, which assessed ChatGPT’s ability to write articles on cutting-edge research in Neurosurgery, determined that whilst the quality of writing was excellent, the nuance of scientific understanding was lacking. This was exemplified in the way that ChatGPT repeatedly used the word “backbone” instead of the more technical name “spinal column”.8

Additionally, it has been shown that when asked the same questions multiple times, ChatGPT may provide differing and contradictory answers.6 This brings into question the accuracy of its statements, particularly technical scientific ones. However, it is suggested that AI tools of the future may be created solely for use in scientific fields, which are trained on scientific sources and can provide accurate statements based on peer-reviewed data. As of today, AI tools must be used alongside the human supervision of the expert medical writers themselves, to ensure that scientific accuracy is maintained.

Rapid analysis and summarisation of large datasets

Literature reviews are an important part of the academic process which takes science from theory and benchwork to useful inventions and drug discovery. Literature reviews analyse and summarise the current state of published research on a particular subject, to collate global ideas and identify areas where more research is needed. AI tools can be of great benefit in performing literature reviews, as they can analyse vast quantities of data and text in a very short space of time, generating short, digestible summaries.9

Drawback – poor citation generation

The integrity and benefit of medical writing hinges on the credibility and accuracy of the work that is written. Whilst summaries of large quantities of data are certainly made quicker by AI tools, the accuracy must always be scrutinised.

Additionally, current AI tools are bad at backing up their statements with relevant or even real scientific citations. Whilst AI tools are fantastic for citation management, as mentioned previously, they falter when left to generate citations independently. One study found that ChatGPT made a confident statement about the role of the human liver in a rare disease, even though the liver’s role in this particular disease is not known or understood scientifically. When the team asked ChatGPT to provide citations for its claims, some of the references provided were outdated, unrelated, or entirely non-existent. It also struggled to provide full titles and author names.7

Overcoming language barriers

English is the predominant language in scientific literature, but scientific research and collaboration are carried out on a global scale. For the huge number of scientists who are not fluent in English, AI tools can allow high-quality English text and translation from their native tongue. Not only can ChatGPT generate human-like English text with very little input needed, but the ability to highlight and correct errors and introduce consistency of speech cannot be understated.10

In fact, AI-generated English text is so good that one study comparing ChatGPT-generated scientific abstracts versus human-written ones demonstrated that blinded (unbiased) reviewers found it difficult to distinguish between them, although they did correctly identify 68% of the AI-generated abstracts as such. The telltale signs of ChatGPT’s text are that it often sounded vague and formulaic.5 However, this is sure to become less noticeable as AI technology improves over time.

Drawback – false experts

With the barrier of language overcome, coupled with the increased ease and efficiency of the writing process, AI tools allow medical writers to greatly increase their output rate of completed work. However, if the field is not careful, this could cause long-term negative effects on the credibility of authors globally. Specifically, the worry is that the number of publications under a writer’s belt may no longer be indicative of their expertise or trustworthiness. This could lead to an influx of “false experts” which would cause real damage to the scientific process, as well as open up ethical and legal concerns.6

Summary

The use of AI tools such as the Large Language Model ChatGPT in medical writing offers significant advantages in terms of efficiency and accuracy. It can allow writers to quickly generate well-written and structured scientific text from short notes, analyse and summarise large amounts of data, and bridge language barriers. However, as with all recent technological advancements, there are drawbacks to AI tools. Scientific depth and nuance are lacking from the generated text, and AI struggles to offer entirely accurate claims with solid citations to back them up.

The future of AI in medical writing

AI is still in its infancy. The potential gains in efficiency and therefore scientific advancements are exciting to consider, but medical writers and publishers must be cautious. With such ease of access, the creation of “false experts” through increased literature output at a lower quality is a concern, which would diminish the credibility and usefulness of medical writing. Scientific journals, conferences, and any other avenues of publishing need to establish clear guidelines on the use of AI, and agree upon how to evaluate submitted literature that has used it. By embracing AI tools with consideration and forethought, medical writing can be kept to its necessarily meticulous standards and evolve alongside AI.

References

  1. ‘What Is Medical Writing? A Comprehensive Guide’. CareerInPharma, 25 Aug. 2023, https://careerinpharma.com/what-is-medical-writing/
  2. ‘What Is AI, How Does It Work and What Can It Be Used For?’ BBC News, 14 June 2023. www.bbc.co.uk, https://www.bbc.com/news/technology-65855333
  3. Roser, Max. ‘The Brief History of Artificial Intelligence: The World Has Changed Fast — What Might Be Next?’ Our World in Data, Jan. 2024. ourworldindata.org, https://ourworldindata.org/brief-history-of-ai
  4. Haleem, Abid, et al. ‘An Era of ChatGPT as a Significant Futuristic Support Tool: A Study on Features, Abilities, and Challenges’. BenchCouncil Transactions on Benchmarks, Standards and Evaluations, vol. 2, no. 4, Oct. 2022, p. 100089. ScienceDirect, https://doi.org/10.1016/j.tbench.2023.100089
  5. Gao, Catherine A., et al. ‘Comparing Scientific Abstracts Generated by ChatGPT to Real Abstracts with Detectors and Blinded Human Reviewers’. NPJ Digital Medicine, vol. 6, Apr. 2023, p. 75. PubMed Central, https://doi.org/10.1038/s41746-023-00819-6
  6. Khalifa, Ahmed A., and Mariam A. Ibrahim. ‘Artificial Intelligence (AI) and ChatGPT Involvement in Scientific and Medical Writing, a New Concern for Researchers. A Scoping Review’. Arab Gulf Journal of Scientific Research, vol. ahead-of-print, no. ahead-of-print, Jan. 2024. Emerald Insight, https://doi.org/10.1108/AGJSR-09-2023-0423
  7. Alkaissi, Hussam, and Samy I. McFarlane. ‘Artificial Hallucinations in ChatGPT: Implications in Scientific Writing’. Cureus, vol. 15, no. 2, p. e35179. PubMed Central, https://doi.org/10.7759/cureus.35179. Accessed 14 July 2024. 
  8. Alattar, Ali A., and Michael M. McDowell. ‘The Rise of AI Co-Authors: Navigating the Future of Scientific Writing with ChatGPT’. Journal of Neurosurgery, vol. 139, no. 5, May 2023, pp. 1485–86. thejns.org, https://doi.org/10.3171/2023.3.JNS23254
  9. Biswas, Som. ‘ChatGPT and the Future of Medical Writing’. Radiology, vol. 307, no. 2, Apr. 2023, p. e223312. DOI.org (Crossref), https://doi.org/10.1148/radiol.223312
  10. Chen, Tzeng-Ji. ‘ChatGPT and Other Artificial Intelligence Applications Speed up Scientific Writing’. Journal of the Chinese Medical Association, vol. 86, no. 4, Apr. 2023, pp. 351–53. DOI.org (Crossref), https://doi.org/10.1097/JCMA.0000000000000900

Get health & wellness advice into your inbox

Your privacy is important to us. Any information you provide to us via this website may be placed by us on servers. If you do not agree to these placements, please do not provide the information.

Best Milk Alternative
[optin-monster-inline slug="yw0fgpzdy6fjeb0bbekx"]
This content is purely informational and isn’t medical guidance. It shouldn’t replace professional medical counsel. Always consult your physician regarding treatment risks and benefits. See our editorial standards for more details.

Get our health newsletter

Get daily health and wellness advice from our medical team.
Your privacy is important to us. Any information you provide to this website may be placed by us on our servers. If you do not agree do not provide the information.
my.klarity.health presents all health information in line with our terms and conditions. It is essential to understand that the medical information available on our platform is not intended to substitute the relationship between a patient and their physician or doctor, as well as any medical guidance they offer. Always consult with a healthcare professional before making any decisions based on the information found on our website.
Klarity is a citizen-centric health data management platform that enables citizens to securely access, control and share their own health data. Klarity Health Library aims to provide clear and evidence-based health and wellness related informative articles. 
Email:
Klarity / Managed Self Ltd
Alum House
5 Alum Chine Road
Westbourne Bournemouth BH4 8DT
VAT Number: 362 5758 74
Company Number: 10696687

Phone Number:

 +44 20 3239 9818