By Christina Jones, Policy Analyst
The use of AI in health care is rapidly expanding in many areas from documentation, diagnostics, to disease prediction. Trending artificial intelligence (AI) tool ChatGPT (Chat Generative Pre-trained Transformer) is a complex machine learning tool based on a large language model consisting of a neural network with many parameters, trained on large quantities of text[i]. This type of AI trains mass amounts of data to carry specific-language oriented tasks including[ii]:
- providing answers to questions
- completing a sentence
- writing content based on given prompts
- producing human-like chatbot responses
- performing calculations
Possible applications of ChatGPT in health care
Clinical workflow. Can be used to generate automated summaries of patient interactions, discharge summaries, medical histories and streamline the medical recordkeeping process[iii]. Clinicians can dictate their notes, and the model can automatically summarize key details, including symptoms, diagnoses, and treatments. Depending on the data inputs from sources like the EMR, lab reports, images, manually entered clinic notes, voice dictation, ChatGPT can generate a holistic report or textual “picture” of a patient’s health.
Clinical decision making (CDS). Can be prompted to generate treatment options, highlight drug interactions, and provide guidelines based on a CDS alert title and description. Although the clinician will be the final decision maker, as the technology evolves and new iterations come out, it can serve to support by generating initial CDS including diagnoses, treatment plans, and teaching materials.
Patient education. Can sort and summarize large amounts of data into natural language text based on a given instruction prompt. Given the language-specific capacities ChatGPT offers, it may lead to great advances or head starts in health care research, writing and data collection[iv]. Crafting written material with prompts can assist providers in producing customized materials for patients such as educational pamphlets, teaching modules, and checklists.
Patient attitudes. A study conducted to investigate patient reactions to current and future AI technologies with a mixed method approach of quantitative and qualitative via survey questions and structured interviews. Overall, a majority of respondents (52.9%) opted for a human physician versus 47.1% of respondents selected an AI clinic. Variations in respondent attitudes changed more in favour of AI technologies when conditions presented such as “my primary care provider recommends AI technology”, and “when the AI system is personalized to listen to the patient”[v].
Potential benefits & drawbacks. OntarioMD Patient Leader Anna Foat suggests starting with automating administrative functions in ChatGPT, where there is a low risk of information leakage and data can be anonymized. However, Patient Leader Sydney Graham emphasizes that primary care should focus on personal interactions and understanding beyond just the medical aspect of a patient. While she acknowledges the value of these technologies, Sydney believes that automation can lead to the loss of important elements of care from a patient’s perspective.
Potential risks. Anna warns that current versions of ChatGPT are not HIPAA compliant, posing a risk of patient health information being accessible to OpenAI employees and used for training future iterations. Moreover, both Anna and Sydney highlight that large language models lack context and nuance, essential elements in health care delivery. Sydney emphasizes that each patient’s unique needs cannot be adequately captured within a broad set of data statistics.
ChatGPT and similar AI tools can be harnessed by health care providers to assist in many aspects of providing patient care, however as the technology and its capabilities continue to evolve to mimic actions and tasks humans do, it should be a complement to a clinician and their expertise, not a replacement.
[i] Goled, S. Self-Supervised Learning Vs Semi-Supervised Learning: How They Differ. Analytics India Magazine. May 7, 2021.
[iii] Marr, B. Revolutionizing Healthcare: The Top 14 Uses Of ChatGPT In Medicine And Wellness. Forbes. March 2, 2023.
[iv] Moons, P and Bulck, L. ChatGPT: can artificial intelligence language models be of value for cardiovascular nurses and allied health professionals. European Journal of Cardiovascular Nursing. February 8, 2023.
[v] Robertson, C et al. Diverse patients’ attitudes towards Artificial Intelligence (AI) in diagnosis. PLOS Digital Health. May 19, 2023.