Skip to main content

Conference Program

Subpage Hero

Loading

Learning from Artificial intelligence: Leveraging AI to Modify a Direct-to-Consumer Genetic Testing Point of Care Tool. 

Education and Research Strategies
  • Primary Categories:
    • General Education
  • Secondary Categories:
    • General Education
Introduction:
The Direct-To-Consumer Genetic Testing (DTC-GT) Point of Care (POC) tool is a flow chart that was developed by genetic professionals who are members of the Inter-Society Coordinating Committee for Practitioner Education in Genomics (ISCC-PEG)/ DTC-GT project group, to aid clinicians when presented with DTC-GT related questions in clinical practice.  Chat-GPT is an Artificial Intelligence (AI) software developed by Microsoft and OpenAI which uses machine learning to generate seemingly indistinguishable answers from a human’s critical thinking. Healthcare professionals report encountering DTC-GT results in clinical practice and have questions about how to manage these results.  Artificial Intelligence (AI) use is growing in many facets of medicine and it is anticipated that healthcare professionals and learners may use AI to input questions about a patient during a clinical encounter to determine how to manage genetic test results. The aim of this study is to use DTC-GT case-based scenarios to compare AI responses to the ISCC-PEG DTC-GT POC tool.

Methods:
Clinical vignettes used to validate the ISCC-PEG DTC-GT POC tools were run through an AI bot to compare responses. Narrative analysis was used to characterize themes, and the DTC-GT POC tool versus AI answers were assessed for similarities and differences. The analysis was completed by three medical students and reviewed by one clinical geneticist over several time points between November 2023 and September 2024.

Results:
Themes generated were similar between the AI answers and the POC tool, e.g.; obtain a family history, explain results, evaluation and management steps, and referral to genetics.  The AI answers unique themes included: empathic response, encourage patient education and support, reassurance, and follow up. AI responses were easier to follow and had clearer titles. Errors in AI responses  were only noted by the clinical geneticist; such as performing a limited family history pertaining to the condition in the scenario, providing reassurance for a negative DTC-GT result for breast cancer, using DTC-GT to evaluate for carrier risk in reproductive partner, not noting that third party analysis results maybe false positive and acting on third party analysis generated results.

Conclusion:
There is significant overlap in the responses between AI and the ISCC_PEG DTC-GT POC tool, with subtle differences that may not be clearly apparent to non-genetics professionals. Additionally, AI-generated responses are easier to navigate, with clear titles  and ranked higher in terms of empathy. These differences informed changes to the ISCC-PEG DTC-GT POC tool by adding: empathy, clear titles such as risk assessment and risk management, educational resources for patients and a follow up plan. This comparison also helped streamline the POC tool flow chart and eliminated redundancy. Next step is to input the modified POC tool chart in an AI software, analyze responses, and to train the software to eliminate erroneous responses. 

Agenda

Sponsors