Skip to Main Content

The Graduate Health & Life Sciences Research Library at Georgetown University Medical Center

Artificial Intelligence, ChatGPT and Blockchain Use in Healthcare

A guide to the use of emerging technologies in healthcare and medicine

Stay Current

Loading ...

Large Language Models

Large Language Models (LLMs). gather data from a variety of sources  - websites,  databases, social media, and user input; however, these sources are not always made public.  The system "learns"  from interactions to improve the experience.   These can also be called "Chatbots".   ChatGPT,  (GPT stands for generative pre-trained transformer) from the company OpenAI,  is an example of a sophisticated  chatbot and is " capable of understanding natural human language and generating impressively detailed human-like written text." (Source: The Guardian).  Other companies offer their own generative chatbots  - some examples are Gemini, Claude and Bing AI/Copilot.

Some concerns about the use of ChatGPT :

  • Currency - Responses may not reflect the most recent developments in healthcare or education.
  • Experience - Lacks qualified clinical judgement.
  • Accuracy - ChatGPT answers that sound plausible but are factually incorrect (hallucinations) and increase in inaccuracy over time (Drift).
  • Bias - Incorrect or incomplete information about groups of people or health conditions could possibly exacerbate healthcare inequities/bias.
  • Privacy - Inquires to ChatGPT are not HIPAA compliant or private.  The conversations with ChatGPT may go into training the system and potentially used in responses in the future.

Searching the Literature

Search the following subject headings/keywords in the databases below to find articles on the use of ChatGPT and Chatbot technologies in healthcare.  As this is a newer technology,  there are very limited subject headings for the concept.  When subject headings are not available, use a keyword search.

  • Chatbots
  • ChatGPT
  • Bard
  • Bing AI
  • Copilot
  • Claude
  • Gemini
  • Large language model
  • Natural language processing
  • Medical Informatics
  • Machine learning
  • Artificial intelligence
  • Generative pre-trained transformer

Search the following subject headings/keywords in the databases below to find content on the use of ChatGPT and chatbot technologies in healthcare.  As this is a newer technology,  there are very limited subject headings for the concept.  When subject headings are not available, use a keyword search.

  • Chatbots
  • ChatGPT
  • Bard
  • Bing AI
  • Copilot
  • Claude
  • Gemini
  • Large language model
  • Natural language processing
  • Medical Informatics
  • Machine learning
  • Artificial intelligence
  • Generative pre-trained transformer
Featured Books

Prompt Engineering

Get the most out of an interaction with ChatGPT/LLMs by using prompt engineering, a practice of formulating questions in specific ways to guide the responses.  The way a question is asked can have a significant impact on the information returned.

  • Specific - Instead of “Tell me about Hypertension," try “Provide a detailed description of the tests used to diagnose Hypertension.”
  • Format - Detail the needed format of the response - bullet points, number of paragraphs, style etc. 
  • Ask for References or Reasoning - For a more detailed response, ask for references and/or the rationale behind the answer.
  • Role Play -  Ask the system to assume an identity.  For example, “As a physician, explain Hypertension to a patient newly diagnosed with this condition.” 
  • Tone- Specify the desired tone of the output - empathetic, casual, professional etc.
  • Include Negative Prompts - Explain what information is not desired.  For example,  " You are studing for the NCLEX exam.  Provide 5 practice questions on pharmacological and parenteral therapies but do not include questions on blood administration."
  • Follow up Questions– Ask the system follow-up questions on content

The CLEAR Framework for Prompt Engineering  provides guidelines and methods that can be used to optimize interactions with AI large language models such as ChatGPT.