Introduction

Artificial intelligence (AI) is the ability of machines to perform functions requiring human intelligence, such as solving problems, learning, reasoning, decision-making, and understanding languages and pictures. Instead of following basic instructions, AI learns and adapts to performing tasks that require more complex intellectual abilities. With the rapid advancement of AI, many industries are undergoing huge transformations and medical writing, a subset of medical affairs, is not an exception.¹ AI has become an integral tool in various domains, including the healthcare system. Its incorporation into the healthcare system has improved patient outcomes and quality of life.² Theseus was the first AI system to come into existence and was developed by Claude Shannon in 1950.³

Overview of traditional medical writing

Medical writing entails the creation of organised documents pertaining to healthcare, scientific research and medicine. The process of medical writing requires the interplay of the knowledge of medical sciences and writing to transform intricate medical data into a style that is easy to read and comprehend by those who are not specialists in the topic. Medical writing can come in a variety of forms, which include patient learning guides, drug safety reports, research drafts and summaries, pharmaceutical promotional collateral, study design guidelines, patient consent statements, compliance filing documents, and peer-reviewed journals.⁴

Traditional medical writing is the conventional method of writing documents which are strictly related to medicine and healthcare. It involves the human act and science of crafting, reviewing, revising and finalising medical documents in a clear and concise way for the purpose of scientific publication, regulatory submissions, clinical trials, instructional documents and educational materials. This process of writing medical documents requires the human writer to have a deep understanding of the subject matter thoroughly, be able to evaluate and interpret data, and follow strict regulatory requirements and editorial standards. This frequently takes a significant amount of resources and time to complete.⁴

What is chatGPT?

ChatGPT is a generative AI chatbot tool released on the 30th of November, 2022, by a company named OpenAI. It is a Large Language Model (LLM) which generates text based on patterns it learned from tons of data online, a term described as Generative Pre-trained Transformer, or GPT. At the heart of ChatGPT is something known as the neural language model, which is just a way of saying it’s built to have conversations with humans. It uses deep learning (basically, really advanced pattern recognition) to figure out how to respond in a way that feels natural. It can understand the context of your communication and respond in a way that makes sense, which is impressive.^5,6

How is AI revolutionising medical writing?

Currently, AI has emerged as a game-changer in the field of medical writing because it can analyse data, generate text, and handle repetitive tasks, which makes everything faster and easier. It uses tools like natural language processing (NLP) and machine learning (ML)  to understand and create human-like text. For medical writers, this means they can get their work done more efficiently while making sure everything is accurate and follows the necessary guidelines. Basically, AI helps take care of the heavy lifting, so writers can focus more on the important details without worrying as much about time-consuming tasks.¹

Natural language processing in medical writing

Natural language processing (NLP) is an aspect of AI that comprehends, interprets and generates human language using machines. It focuses on the interaction between machines and human languages. NLP has made significant advancements in medical writing by allowing generative AI technologies to assist in creating scientific writing and documentation.^7,8 Not only does NLP work with texts, but it also translates speech to text using a speech-text recognition pattern. 

For instance, enormous volumes of medical literature may be used to train models like OpenAI's GPT-4 to produce language that is logical and human-like. With impressive speed and accuracy, these AI models can generate material for research articles, patient narratives, and clinical trial reports. When you are creating documents that are both scientifically correct and comply with regulations, the use of NLP-driven technologies has resulted in saving time. Because it is good at processing information, AI leveraging NLP helps to reduce the chances of human error, making sure everything is spot on while freeing up more time for other important tasks.⁶

Automated medical content generation

Automation of medical content generation is one of the most substantial applications of artificial intelligence in medical writing. Based on input data and proper prompting, AI-driven tools can generate coherent and structured text. For tasks like writing research summaries, clinical trial reports, or even those super detailed regulatory documents, AI can draft the first version for you. These tools help kickstart the writing process, making everything flow more smoothly from the start.⁸

AI in data analysis for medical writing

Interestingly, AI is not just good at generating texts; it can also help out with analysing datasets. AI-powered algorithms can take really complex data, from sources like clinical trials, and break it down into understandable chunks that can be easily added to reports. For example, the whole process of analysing clinical trial results can be automated by AI, including running the statistical analysis, and then presenting the findings in a way that is basically ready to go for publication. It takes a lot of the heavy lifting out of data analysis, saving time and making sure the results are accurate and clear.⁹

AI in generating bibliographies and references

AI has become a great and handy tool in generating citations and references. It makes the task much easier. Instead of going through the manual source and formatting everything yourself, generativeAI (genAI) can create references and bibliographies automatically in the correct format it is prompted to respond. Though the citation and references generated by AI based on LLMs may not always be 100% accurate, a study conducted on the citation accuracy of ChatGPT-4 reported an accuracy of about 80%, while ChatGPT-3.5 has an accuracy of about 58%.^9,10

Improving compliance with regulatory requirements

The ability of AI to process large datasets and detect pertinent regulatory requirements promises that medical documentation is constantly in compliance with updated regulations and guidance.⁹

AI-assisted manuscript writing and editing

A substantial contribution of AI in scientific publication has been seen in the writing and editing of scientific manuscripts. This is an aspect of medical writing that consumes medical writers’ significant amounts of time. It is interesting to say that some AI tools have been designed to aid humans in the task of manuscript writing, the editing process, plagiarism check, and ensuring manuscript compliance with linguistic standards and free of grammatical errors and inconsistency.⁹  

Potential issues in AI-driven scientific publication and documentation

Loss of human expertise and interpretation 

As the use of AI in medical writing advances, there is a risk of reduced human touch in documentation, which invariably gives rise to reduced human critical thinking, domain-specific judgement and editorial judgement among writers.⁵ 

Ethical and legal concerns

Many ethical concerns, including authorship and intellectual property, have arisen due to the increasing use of AI in writing medical documentation. Questions like who should be the rightful author of a publication whose portion was written by AI are yet to be answered. Presently, there is no stipulated law guiding the use of AI in medical writing.⁵

Data privacy and security risks

With the use of AI in medical writing, there is a rising concern about data privacy and security infringement as these documents contain confidential research discoveries, sensitive patient information and clinical trial results.⁵

Does AI introduce human bias?

The use of AI in medical writing can introduce biases, as it learn from human-generated data. AI can pick up on biases in such information, such as unjust language or preferring one group over another. People who speak different languages may find it more difficult if AI is mostly trained in one language. Alternatively, AI may perpetuate unjust prejudices found in historical data.¹¹

AI transparency and accountability

According to research, the decision on accountability and authorship in the use of AI for writing a manuscript is undergoing changes. The responsibility of an author is far more than simply writing; it entails taking ownership of the work that goes into the publication.¹¹ According to the  International Committee of Medical Journal Editors (ICMJE) 2024, an author must satisfy the following 4 criteria, which AI-driven tools do not, before claiming authorship:

1. Substantial contributions to the conception or design of the work, or the acquisition, analysis, or interpretation of data for the work
2. Drafting the work or reviewing it critically for important intellectual content
3. Final approval of the version to be published
4. Consent to take responsibility for every facet of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved⁵

Instead, to give readers an idea of how the AI tool was utilised in the preparation of any manuscript, the ‘Acknowledgements’ section can include a description of any instances in which the LLM was used for writing aid.^5,11

AI tools used in medical writing

There are many LLMs that have been invented to aid medical writers. Some examples include: ⁵

• SciSpace
• ScholarAI
• Scite_
• Elicit
• Connected Papers
• Research Rabbit
• SciSummary
• Aether Brain
• Quillbot
• ChatPDF
• Jasper AI
• Copy.ai
• ChatGPT
• Julius AI

Summary 

The integration of AI into medical documentation and publication has aided human writers with tasks like automation of medical content generation, data analysis, generation of citations and references, and writing and editing of manuscripts. However, there exist potential issues with the use of AI in medical writing. These issues include ethical and legal concerns, loss of human expertise, data privacy and security risks, and introduction of human biases.