AI Chatbot Education for Atrial Fibrillation

Overseen By Samir Saba, MD

Age: 18+

Sex: Any

Trial Phase: Academic

Sponsor: University of Pittsburgh

Disqualifiers: Inability to use device, others

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Compensation: Varies

Number of Visits: 4

40 Participants Needed

Trial Summary

Will I have to stop taking my current medications?

The trial information does not specify whether you need to stop taking your current medications. It seems to focus on education about atrial fibrillation rather than medication changes.

What data supports the effectiveness of AI Chatbot Education for Atrial Fibrillation?

Research shows that computer-based patient education, like AI chatbots, can effectively teach patients about their health conditions by allowing them to learn at their own pace and track their understanding. This approach has been shown to motivate behavior change and improve patient education in other health areas, suggesting potential benefits for atrial fibrillation education as well. ^{1 2 3 4 5}

How is the AI Chatbot Education for Atrial Fibrillation treatment different from other treatments for this condition?

The AI Chatbot Education for Atrial Fibrillation is unique because it uses an internet-based educational program to provide patients with reliable information about their condition, which is not a standard approach in AFib management. This treatment focuses on patient education rather than direct medical intervention, aiming to empower patients with knowledge to better manage their condition. ^{6 7 8 9 10}

What is the purpose of this trial?

Atrial Fibrillation is a chronic disease with significant health consequences like increased risk of stroke, heart failure, heart attack and death. Educating patients about the disease is important for them to be able to understand the condition better, feel empowered and take an active part in their care plan. AI technology can potentially be used to impart such education. However, doing so with care and empathy is equally important. Therefore, it is necessary to ensure when AI technology is used to impart education about atrial fibrillation to patients, the humane aspects of the interaction are rigorously tested. This study examines a way to impart atrial fibrillation education through interaction with an AI chatbot, that uses text and links to educational videos. To participate in this study, people need to be age 18 or older and have a history of newly diagnosed atrial fibrillation. Approximately 40 individuals will be asked to take part in this study. The first step to the study will be reading through, understanding, and signing an informed consent. People who then agree to join the study will have a one-time interaction with the AI chatbot and structured educational material by using an iPad provided to them for the approximately 1 hour duration of the study. People in the study will obtain atrial fibrillation education by typing one by one on the iPad, up to 10 questions about the disease. Answers will include text and links to videos. Before and after atrial fibrillation education, people who join this study will be asked to fill out a survey. The study team will teach patients how to use the iPad and type in questions.

Research Team

Samir Saba, MD

Principal Investigator

UPMC Heart and Vascular Institute

Eligibility Criteria

This trial is for adults over 18 with a known diagnosis of atrial fibrillation. It's suitable for those who can text and use an electronic device, as well as read English. People unable to interact with technology or who have language barriers are not eligible.

Inclusion Criteria

I am older than 18 years.

I have been diagnosed with atrial fibrillation.

Exclusion Criteria

Inability to text/use an electronic device

Inability to read English

Timeline

Screening

Participants are screened for eligibility to participate in the trial

1 week

1 visit (in-person)

Informed Consent

Participants read, understand, and sign an informed consent form

1 day

1 visit (in-person)

Interaction with AI Chatbot

Participants interact with the AI chatbot for atrial fibrillation education using an iPad

1 hour

1 session (virtual)

Follow-up

Participants complete pre- and post-interaction surveys to assess empathy, knowledge, and trust

1 day

1 session (virtual)

Treatment Details

Interventions

• Atrial Fibrillation Education

Trial Overview The study tests the effectiveness of AI chatbot-delivered education on atrial fibrillation. Participants will use an iPad to ask questions about their condition and receive answers with educational content, including video links, during a one-hour session.

Participant Groups

Group I: Pilot interventional arm Experimental Treatment 1 Intervention

Single arm, wherein all consenting participants undergoing interaction with chatbot for atrial fibrillation education.

Who Is Running the Clinical Trial?

University of Pittsburgh

Lead Sponsor

Trials

1,820

Recruited

16,360,000+

Findings from Research

ChatGPT provided correct answers to 69.2% of questions about thyroid nodules, regardless of the type of prompting used, indicating a generally reliable performance for patient education.

However, while prompting for references resulted in 100% of responses including citations, a significant portion of these references (72.5%) did not accurately report the information, highlighting concerns about the reliability of ChatGPT as a medical educational tool.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluating ChatGPT responses on thyroid nodules for patient education. Campbell, DJ., Estephan, LE., Sina, EM., et al.[2023]

A study evaluating responses to 200 medication-use questions found that a medication database provided more clinically correct, complete, safe, and useful answers than an AI chatbot, with 97% of questions answered by the database compared to 80% by the chatbot.

While the chatbot's clinical correctness was comparable to the database at 85%, it lagged behind in completeness (65%), safety (71%), and usefulness (68%), indicating a need for further improvements in chatbot technology for reliable clinical use.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluation of inpatient medication guidance from an artificial intelligence chatbot. Beavers, J., Schell, RF., VanCleave, H., et al.[2023]

Information about benign paroxysmal positional vertigo (BPPV) obtained from Google search results is more readable and of higher quality compared to responses generated by the conversational AI model ChatGPT, based on a study analyzing 30 sources.

ChatGPT responses were found to be more difficult to read and comprehend, with lower scores in readability metrics and quality assessments, indicating that traditional search engines may provide more accessible patient education materials.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

BPPV Information on Google Versus AI (ChatGPT). Bellinger, JR., De La Chapa, JS., Kwak, MW., et al.[2023]

Computer-based patient education can provide personalized instruction that adapts to individual learning paces, which is crucial as healthcare becomes more patient-centered and hospital stays shorten.

This method not only helps in teaching patients about their health and treatment options but also tracks their understanding, which is important for informed consent and potential reimbursement, highlighting the need for more research in this area.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Computer-based patient education revisited. Gillispie, MA., Ellis, LB.[2019]

ChatGPT provided correct answers to 71.9% of questions about obstructive sleep apnea, indicating it can be a useful tool for patient education, regardless of the type of prompting used.

Prompting for statistics and references significantly improved the quality of responses, with 96.1% of those answers including references, while patient-friendly prompts resulted in simpler language suitable for patients, maintaining an acceptable grade level for medical information.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluating ChatGPT responses on obstructive sleep apnea for patient education. Campbell, DJ., Estephan, LE., Mastrolonardo, EV., et al.[2023]

The AI algorithm demonstrated a moderate ability to predict new cases of atrial fibrillation (AF) in a cohort of 1403 patients presenting with palpitations, achieving an area under the curve (AUC) of 0.74.

With a sensitivity of 79.1% and specificity of 66.1%, the algorithm can effectively identify patients at risk for AF, although its clinical utility may be limited due to the low incidence of AF (3.1%) in this emergency department population.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluating atrial fibrillation artificial intelligence for the ED: statistical and clinical implications. Kaminski, AE., Albus, ML., Ball, CT., et al.[2022]

The ASK FOR IT educational program, designed for patients with atrial fibrillation, showed promising results in a pilot study with 15 participants, leading to significant improvements in symptoms and mental health-related quality of life after 3 months.

Participants found the program easy to use and understand, indicating its potential as a reliable source of information for patients, which will be further evaluated in a larger randomized controlled trial involving 200 patients.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

ASK FOR IT: An Internet-based educational program for patients with atrial fibrillation - Results from a pilot study and design of the randomized, controlled, multicenter ASK FOR IT study. Walfridsson, U., Brohede, D., Walfridsson, H.[2022]

In a study of 52,547 patients with newly diagnosed atrial fibrillation, rhythm-control strategies were found to be more effective than rate-control strategies in improving patient outcomes, including lower mortality rates.

Using artificial intelligence and Q-learning, researchers identified optimal treatment strategies for different patient clusters, showing that aligning provider-selected treatments with AI recommendations led to significantly better outcomes, including fewer deaths.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluation of Quantitative Decision-Making for Rhythm Management of Atrial Fibrillation Using Tabular Q-Learning. Barrett, CD., Suzuki, Y., Hussein, S., et al.[2023]

Recent AI models have been developed that can accurately assess the risk of atrial fibrillation (AF) by analyzing electrocardiogram waveforms, providing additional predictive information beyond traditional clinical risk factors.

By identifying individuals at higher risk for AF, these AI-based models could enhance preventive strategies, such as targeted screening and risk factor modification, ultimately aiming to reduce the incidence of AF and its related health issues.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

The Use of Artificial Intelligence to Predict the Development of Atrial Fibrillation. Pipilas, D., Friedman, SF., Khurshid, S.[2023]

Machine learning, particularly artificial neural networks, has significant potential to enhance the detection and management of atrial fibrillation, which affects one in three people and can lead to serious complications like stroke.

These technologies could improve patient care by enabling better screening for asymptomatic atrial fibrillation and helping cardiologists identify patients at high risk of recurrence after treatment, thus potentially improving clinical outcomes.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Machine learning in the detection and management of atrial fibrillation. Wegner, FK., Plagwitz, L., Doldi, F., et al.[2022]

References

1. United States pubmed.ncbi.nlm.nih.gov

Evaluating ChatGPT responses on thyroid nodules for patient education. [2023]

2. England pubmed.ncbi.nlm.nih.gov

Evaluation of inpatient medication guidance from an artificial intelligence chatbot. [2023]

3. England pubmed.ncbi.nlm.nih.gov

BPPV Information on Google Versus AI (ChatGPT). [2023]

4. United States pubmed.ncbi.nlm.nih.gov

Computer-based patient education revisited. [2019]

5. United States pubmed.ncbi.nlm.nih.gov

Evaluating ChatGPT responses on obstructive sleep apnea for patient education. [2023]

6. United States pubmed.ncbi.nlm.nih.gov

Evaluating atrial fibrillation artificial intelligence for the ED: statistical and clinical implications. [2022]

7. United States pubmed.ncbi.nlm.nih.gov

ASK FOR IT: An Internet-based educational program for patients with atrial fibrillation - Results from a pilot study and design of the randomized, controlled, multicenter ASK FOR IT study. [2022]

8. England pubmed.ncbi.nlm.nih.gov

Evaluation of Quantitative Decision-Making for Rhythm Management of Atrial Fibrillation Using Tabular Q-Learning. [2023]

9. United States pubmed.ncbi.nlm.nih.gov

The Use of Artificial Intelligence to Predict the Development of Atrial Fibrillation. [2023]

10. Germany pubmed.ncbi.nlm.nih.gov

Machine learning in the detection and management of atrial fibrillation. [2022]