Can ChatGPT be used in healthcare?


Share this article

Artificial intelligence language model ChatGPT has garnered much buzz and attention for its ability to understand and answer open-ended questions about any topic. Choong Zi Chao, Lead Specialist, and Alvin Yuen, Senior Specialist from Synapxe Data aNalytics & AI team, look at how ChatGPT can be used in healthcare and its limitations.

What is ChatGPT?

ChatGPT is a large language artificial intelligence (AI) model developed by OpenAI, a US-based AI research company. A variant of the GPT (Generative Pre-trained Transformer) model, it uses deep learning to generate human-like responses to natural language prompts. Since it was released to the public in November 2022, it has generated much interest for its high coherent and contextually relevant replies in wide-ranging domains.

How can ChatGPT be used in healthcare?

The release of ChatGPT has prompted discussions regarding its capabilities in the healthcare industry. As ChatGPT can be used for different natural language processing (NLP) tasks, such as text summarisation, question answering and classification, it could be useful for potential applications such as chatbots and virtual assistants.

Tools such as ChatGPT could possibly help to automate repetitive tasks such as answering frequently asked questions from patients, generate clinical notes and write medical reports, triage patients based on criticality of their condition, thereby freeing up resources for healthcare professionals to focus on providing primary clinical care. On top of that, it can aid in clinical decision support by extracting and analysing clinical notes to identify trends and patterns based on patients' medical history in electronic medical records.

Patients can receive personalised information and assistance in a timely manner. The complementary clinical expertise with predictive analytics can also lead to better health outcomes, as the treatment plan can be tailored to every patient's unique health condition.

For administrators, it can be a cost-effective solution for healthcare providers to offer virtual assistance and clinical decision support at scale while relieving the strain on our healthcare professionals, especially during manpower crunches such as during the Covid-19 pandemic.

Limitations

ChatGPT is not intended to be used as a tool to retrieve real time information as its data is limited up to 2021. Hence, it is unable to produce accurate answers for recent events such as the latest status of Covid-19 in Singapore or the winner of the 2022 FIFA World Cup.

ChatGPT is not intended to be used as a tool to retrieve real time information as its data is limited up to 2021. Hence, it is unable to produce accurate answers for recent events such as the latest status of Covid-19 in Singapore or the winner of the 2022 FIFA World Cup.

Due to the popularity of ChatGPT, users may occasionally face accessibility issues due to the high volume of traffic. In such instances, users would have to wait for access which can be an issue when prompt responses are needed.

As ChatGPT is trained on a large dataset of text from the Internet, it may unintentionally reflect the bias that is present in the training data. There may be examples of stereotypes and assumptions for gender or ethnicity that can be reflected in the responses based on the training data. The chatbot may not always generate responses which are relevant or accurate as well, and users can overwrite the response given by ChatGPT with an inaccurate one.

Additionally, ChatGPT may also have limited knowledge in expertise areas such as science or medicine. For example, if a user asked about a disease/drug which was discovered recently, ChatGPT may not have information about it in the training data and hence, is unable to provide an accurate answer.

Quick tips for developer

ChatGPT has yet to be made accessible via an official API, hence limiting its usage in commercial products and services. However, the web version of ChatGPT is currently free to use and is available via OpenAI's ChatGPT webpage (ChatGPT: Optimizing Language Models for Dialogue (openai.com)). All you need to set up an account is an email address and phone number to complete the registration.

Once an account has been setup successfully, you can ask ChatGPT just about anything you want. Some of the prompts include:

  • Asking of information, example "What are the symptoms of Covid-19?"
  • Asking for help with a problem, example "How to write a Python function to print odd numbers from 0 to 100?"
  • Generating text, example "Write a short story about a sunny day"
  • Translating text, example "Translate this sentence from English to Chinese: The quick brown fox jumps over a log"
  • Summarising text, example "Summarise this article"

 

For developers, there are some potential areas in which ChatGPT can be useful in their work. However, the responses from ChatGPT will have to be tested and validated by human developers as there can still be possibility of misinformation.

Developers can use ChatGPT to write out segments of code or use it as an assisted tool in debugging of codes, as illustrated in the examples below:

 

  • ChatGPT can also be used as an assisted tool in understanding codes or in script documentation as it is able to comprehend and provide explanation of the code when prompted, as illustrated in the example below:

OpenAI is also exploring ways to pilot a paid version of the tool known as ChatGPT Plus, features of the paid version may include constant availability, faster responses and priority access to new updates to the tool. The paid version of the tool is currently only available to selected users in the United States and OpenAI has shared that there are future plans to expand the subscription to additional countries.

Conclusion

Given some of the limitations, it remains to be seen whether ChatGPT will eventually gain widespread adoption in healthcare. More effort will need to go into its customisation (e.g. training on healthcare data, inclusion of more recent data etc.) to cater to the needs of healthcare professionals, ensuring accuracy of the data, as well as making sure that its usage conforms to the MOH Artificial Intelligence in Healthcare Guidelines (AIHGle)  However, one thing is for sure - ChatGPT has an immense potential to make a positive impact within healthcare, and it's unlikely to go away anytime soon.

 
The opinions expressed in this article are those of the authors. They may not reflect the opinions or views of Synapxe.

 

Related articles

X

By continuing to use and navigate this website, you consent to the use of cookies in accordance with our Privacy Policy.

Confirm