None listed
Conditions
Brief summary
Online symptom checkers are digital health tools allowing health consumers to input symptoms to receive triage or diagnosis advice. If appropriate, another important feature is to provide consumers with self-care advice for managing their symptoms at home. Integrating Generative Artificial Intelligence (AI) into symptom checkers may help address unmet needs of diverse users, including those with lower health literacy. A retrieval-augmented generation (RAG) framework may be particularly useful by addressing concerns around accuracy via greater control of the quality of information informing the AI output. However, it is unknown how to best communicate the use of RAG generative AI in symptom checkers, nor how to present the advice. This project aims to 1) evaluate the effects of different presentations of generative AI symptom checker self-care advice on intentions, trustworthiness and understanding of the advice, and 2) explore in-depth user responses and perspectives on AI generated triage and self-care advice in an online symptom checker.
Interventions
2 (acuity level) x5 (framing group) parallel group design of generative AI self-care advice from an online Symptom Checker after participants are told to imagine they are sick (vomiting and fever). Generative AI self-care advice from an online Symptom checker involves inputing the symptoms one is experiencing into the symptom checker to receive tailored care advice about how to best manage the symptoms at home. The generative AI component is used to collate the relevant self-care information for
2 (acuity level) x5 (framing group) parallel group design of generative AI self-care advice from an online Symptom Checker after participants are told to imagine they are sick (vomiting and fever). Generative AI self-care advice from an online Symptom checker involves inputing the symptoms one is experiencing into the symptom checker to receive tailored care advice about how to best manage the symptoms at home. The generative AI component is used to collate the relevant self-care information for the symptoms and present the relevant information/care to the user in a way that produces plausible human-like output. Acuity levels: Arm 1 = Triage advice given to participants will be 'self-care at home' Arm 2 = Triage advice given to participants will be 'see the GP within 24 hours' Framing group: Arm A = Generative AI enhanced version without enhancements shown in Arms B to D Arm B = Step-by-step care advice - this advice emphasizes the most important next step, in a logical and numbered sequence. Arm C = Multi-media content - this advice also provides videos and images to increase user engagement with the health advice (for example, a video about how to manage a fever, and a graphic for signs of dehydration). These specific multi-media content are readily available resources. (Image: Healthdirect 2024 https://www.facebook.com/photo.php?fbid=870444085101447&id=100064075893726&set=a.160444812768048&locale=ga_IE Video: MyDr 2024 https://mydr.com.au/first-aid-self-care/how-do-i-manage-a-fever-dr-norman-swan/) Arm D = Detailed information about AI used in the model The the duration of the intervention is 2-5 minutes (this reading a vignette the specific health symptoms and reading the symptom checker advice). Adherence to the intervention will be assessed by submission of the survey. The online survey will be designed that request participants to view the symptom checker advice for a minimum of 2 minutes before moving forward through the survey.
Sponsors
Kirsten McCaffery - Sydney Health Literacy Lab, The University of Sydney
Study design
Allocation
Randomised controlled trial
Intervention model
Parallel
Primary purpose
Educational / counselling / training
Masking
Blinded (masking used) (Subject, Investigator, Outcomes Assessor)
Eligibility
Sex/Gender
All
Age
18 Years to No maximum
Healthy volunteers
Yes
Inclusion criteria
Inclusion criteria for participation in the trial include: a) Reside in Australia b) English proficiency
Exclusion criteria
No exclusion criteria
Outcome results
None listed