ChatGPT has taken the world by storm, but what does it mean for improving health and nutrition and for healthcare professionals?

By Stephanie Tucker 

What is ChatGPT?

ChatGPT is a large-language model chatbot developed by OpenAI that launched in 2022. It allows users to write prompts and receive responses in a human-like manner.

ChatGPT market overview

• ChatGPT gained 57 million users in the first month
• ChatGPT is backed by $10 billion USD investment from Microsoft
• The newly released GPT-4 is 100x more powerful than GPT-3
• As of August 2023, ChatGPT has over 180.5 million users 

What does ChatGPT technology entail?

ChatGPT uses generative pre-trained transformer (GPT) architecture (a neural network model) to process text based on human language data that it has been trained on. It provides detailed answers based on publicly available information. In the words of ChatGPT itself, it “aims to stimulate natural conversation and provide assistance, information or entertainment in a human-like manner” 

ChatGPT in personalized healthcare

For the public, ChatGPT can be used to provide quick answers to questions such as “How can I lower my blood pressure?” where it will provide generic advice on good practice. The accessibility of ChatGPT also enables it to provide a 24/7 second opinion on healthcare queries. However, medical information given by ChatGPT does not undergo expert review and should be regarded with caution.

To date ChatGPT has managed to pass the medical board exam , with knowledge equivalent to a third-year medical student. It provided context and logic for most answers, presenting a strong case or utilizing ChatGPT in medical education.

Ada Health utilizes AI similar to ChatGPT for hospital triage, to maximize face to face time with patients amidst staff shortages and rising hospital costs. They provide their chatbot with a continuously updated medical knowledge base and because of this can provide accurate condition suggestions to the healthcare provider prior to the first appointment.
 

With growing restrictions on the amount of time healthcare providers have with every patient, ChatGPT has potential to reduce time spent on recording symptoms and support doctors with diagnosis.

When it comes to wellness Google has recently infused Fitbit with generative AI, which allows an individual to receive personalized recommendations and advice based on their health data.

ChatGPT in Personalized Nutrition

ChatGPT can assist in providing unique nutritional guidance and develop meal plans, tailored to specific needs and goals. It has potential to be used in obesity treatment, creating meal and exercise plans. ChatGPT is also capable of more general analysis of dietary intake and calculation of energy expenditure when given detailed input. This technology opens up the opportunity to provide 24/7 advice and support, continuous feedback and personalized recommendations based on the latest data collected from the user.

ChatGPT potential to drive research

The impact of our diet on the gut microbiome is a growing area of interest for researchers. ChatGPT has been evaluated for its use in identifying novel research questions in IBD and the gut microbiome. ChatGPT provided clear, relevant questions that identified priorities for future research, but the questions were not thought to be original when evaluated by a team of researchers in the field. Showing how ChatGPT can be used to guide research, but further development would be needed to generate novel topics.

State of the science

Studies on the applications of ChatGPT in health show both limitations and potential benefits in personalized nutrition. The consensus remains that with further development, AI like ChatGPT will become part of everyday practice.
In a recent study ChatGPT was shown to have high accuracy when answering nutrition questions in comparison to dietitians. ChatGPT scored higher in five out of eight questions than the dieticians, with higher average scores in scientific correctness and comprehensibility.

However, it is important to consider where the questions required answers that left out complexities, ChatGPT scored higher than the dietitians who had a more careful and cautious approach considering their lack of specific information. This research shows the huge potential of chatbots to provide accurate nutrition support and accurately answer nutrition questions, arguably as well as human dieticians.
Interestingly ChatGPT does not have the same accuracy when it comes to providing references. When asked to provide information on IBS, ChatGPT had an accuracy of 80%, however, when asked to provide references, only 33% were suitable and shockingly, 13% were nonexistent. AI “hallucinations” where it has access to incomplete information and tries to fill in the gaps, causing complete fabrication, is a driving factor behind the push for regulation for AI to make it more reliable and reduce the infiltration of fictitious material into literature.

A new study published in the journal Nutrients (Ponzo et al, 2024) investigated the potential of ChatGPT in providing nutritional guidance for non-communicable diseases (NCDs) including hypertension, type 2 diabetes, sarcopenia, obesity and non-alcoholic fatty liver disease (NAFLD). The researchers found that the AI’s advice was generally accurate, with appropriateness rates varying from 55.5% for sarcopenia to 73.3% for NAFLD. These results suggest that ChatGPT has potential utility in providing dietary advice, however, the study also highlighted drawbacks of using ChatGPT including responses which were contradictory to guidelines, unsupported or not fully matched. These findings underscore the importance of human expertise in health-related matters, especially when dealing with complex and individualized scenarios.

In another study (Agne et al, 2024) researchers investigated whether ChatGPT, could offer a reliable solution for widening access to personalized dietary recommendations. By comparing ChatGPT's advice with that of the Food4Me algorithm for 20 obese subjects, they found that ChatGPT can provide suitable PN advice in line with that given by Food4Me such as reduction of salt, sugar and saturated fats. Despite this it was shown to produce inconsistent advice, particularly concerning macro- and micronutrient intakes. It is not yet recommended for non-experts to rely on ChatGPT for personalized dietary advice however, this research suggests that ChatGPT shows promise and may evolve into a reliable tool for personalized nutrition advice in the future.

Another recent study assessed ChatGPT’s performance in providing nutritional guidance to college students. 30 dietitians were recruited to assess the quality of ChatGPT’s dietary advice, including its nutrition literacy (NL) achievement and response quality. The results indicate that ChatGPT’s performance varies across scenarios and is suboptimal for achieving NL with full achievement rates from 7.50% to 37.56%. While the responses excelled in readability, they lacked understandability, practicality, and completeness. The top concern among the dietitians was that the “response information lacks thoroughness or rigor, leading to misunderstandings or misuse”. Despite the potential of ChatGPT as a supplementary educational tool, significant gaps must be addressed, especially in detailed dietary inquiries.

Innovators in leveraging ChatGPT for personalised nutrition

Whoop – Whoop is an optical heart rate sensor that tracks data on sleep, strain, recovery and health, to optimize performance. It now has a ChatGPT powered coach designed to provide personalized advice based on the Whoop fitness tracker. It allows you to choose specific goals and provides individual advice such as workout recommendations, nutrition coaching and fitness plans, allowing you to ask any question on your whoop data and get answers in seconds. By tracking macronutrients on the WHOOP app, the coach will provide information on how to optimize intake for energy and performance improvement.

January AI – Is an app that uses a continuous glucose monitor and activity tracker, to predict the impact of diet on your metabolic health and use this information to help you reduce excessive change in blood glucose levels to improve health and help users lose weight. AI-powered insights allow users access to real time glucose monitoring and predictions.

Prevess – Developed to support companies wanting to provide evidence-based nutrition and wellness advice, it uses AI driven technology and a team of performance nutritionists and athletes to ensure the data is accurate. It takes data from food products and connects it to user health profiles to provide personalized insights for clients.

Research shown huge potential of chatbots to provide accurate nutrition support  and acurately answer nutrition questions, arguably as well as human dieticians

Current challenges in using ChatGPT in health

Ethical considerations of ChatGPT in healthcare include misinformation, racial bias and stereotyping. Inadvertently enabling racial bias and reinforcing stereotypes in a society of increasing inequality can deepen socioeconomic divides and hinder progress towards equal healthcare. Misinformation creates a lack of trust in solutions and is worsened by poor quality tools that haven’t undergone expert review. Fast spread misinformation is already a problem within the media and can significantly impact public health, caution is needed when taking information from AI.

There are also concerns around security, where exposure of sensitive information could compromise confidentiality. The lack of verified database means it is reliant on publicly available data that was available at the time of the last update. When asked to provide references for information on IBS, ChatGPT gave a significant number of nonexistent papers, likely due to limitations in the current design of ChatGPT where it may have access to incomplete information.

Furthermore, the human aspects of healthcare, such as physical examination, non-verbal cues from patients and emotional support are unlikely to be replaced by technology. Nutrition experts are uniquely trained to take dietary needs, cultural considerations and health conditions into consideration when giving advice. Unlike ChatGPT or AI tools, they provide empathetic and sensitive guidance, considering emotional and psychological aspects linked to eating habits.

In recent news, the National Eating Disorders Association (NEDA) faced a significant setback when it had to remove its chatbot, "Tessa," from its website. The chatbot, which was intended to provide support and resources for those struggling with eating disorders, was found to offer harmful and unrelated advice. This incident highlights ongoing concerns around generative AI tools, where the potential for misinformation and unintended consequences remains a challenge. As AI continues to advance and as companies are increasingly looking to provide chatbots, ensuring ethical use, accurate guidance, and safety in sensitive areas is critical (CNN 2023).
 

Opportunities for companies and brands

Opportunities exist for companies to use ChatGPT to provide technology that can give personalized insights into individual health and wellbeing on a large scale. If companies can provide a verified database in the shape of a small LLM, ChatGPT could be harnessed to give more reliable advice within companies and also be harnessed to maximise the time of healthcare professionals.

Here at Qina we have integrated a chatbot in the QinaVer search engine, via a Magic Prompt feature, allowing users to find specific market information in seconds. Quality and accuracy with our Magic Prompt feature are ensured through a dual strategy, when generating responses, the model references the text it was trained on, citing sources to enhance accuracy and transparency. If inaccuracies are identified, we collaborate closely with our tech team to address and rectify them as quickly as possible.

 

Conclusion

We believe that ChatGPT will never fully replace dietitians and nutritional professionals, as the intricacies of human interaction and emotional support will always remain a huge factor in healthcare support. However, the potential of ChatGPT to provide advice, provide a second opinion and maximize face-to-face time with patients would be hugely beneficial to the changing landscape of healthcare delivery. Healthcare professionals will need to keep up with how technology like ChatGPT will change their roles and learn how to harness it. In addition, companies will need to stay abreast on how consumers are engaging with the technology to discover and research brands.
 

References

1. Microsoft | https://azure.microsoft.com/en-us/blog/chatgpt-is-now-available-in-azure-openai-service/ 
2. ChatGPT Statistics, Gitnux | https://blog.gitnux.com/chat-gpt-statistics/
3. ChatGPT | https://chat.openai.com 
4. Gilson et al. 2023 | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9947764/ 
5. Ada Health | https://ada.com/editorial/what-llms-mean-for-healthcare/ 
6. Lahat et al. 2023 | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10011374/ 
7. Kirk et al. 2023 | https://www.hindawi.com/journals/jnme/2023/5548684/ 
8. Healio.com | https://www.healio.com/news/gastroenterology/20231026/chatgpt-generally-accurate-in-answering-questions-providing-references-on-ibs 
9. Emsley 2023 | https://www.nature.com/articles/s41537-023-00379-4 
10. Whoop | https://www.whoop.com/gb/en/ 
11. JanuaryAI | https://www.january.ai/ 
12. Prevess | https://www.prevess.com/ 
13. Wang et al. 2023 | https://www.jmir.org/2023/1/e48009 
14. Ponzo et al 2024 | https://www.mdpi.com/2072-6643/16/4/469
15. Agne et al. 2024 | https://www.sciencedirect.com/science/article/pii/S2667268524000512
16. Liao et al, 2024 | https://doi.org/10.3390/nu16121939
17. NEDA CNN | https://edition.cnn.com/2023/06/01/tech/eating-disorder-chatbot/index.html