Participants Must be Women Aged 25 Years or Older, Participants Should Have no Self-reported History of Heart Disease or Stroke, Participants Should Have no Terminal Illness or Diagnosed Cognitive Impairment, Including Alzheimer's Disease, Participants Should Not be a Healthcare Professionals or Healthcare Trainees, Participants Should Not be Employed in the Healthcare Field, Participants Should Possess a Smartphone With an Active Data Plan or Access to Wi-Fi, Participants Should Reside in the United States and be a University of California, San Francisco Health Patient
Conditions
Keywords
Artificial Intelligence, Intelligent systems, Chatbot, Natural language processing, Large language models, Machine learning, Heart disease, Mobile applications, Women, Randomized controlled trial
Brief summary
The goal of this clinical trial is to find out whether an app-based program called HeartBot II, which uses an artificial intelligence (AI) chatbot, can help improve women's awareness and knowledge of heart attacks in women. This is an online study with about 200 women taking part. Participants will be randomly assigned (by chance) to one of two groups: an intervention group or a wait-list control group. Participants in the intervention group will begin using the HeartBot II program right away. Participants in the wait-list control group will wait 12 weeks before starting the HeartBot II program. The HeartBot II program includes four short modules. In each module, participants will interact with a chatbot and spend about 10 to 15 minutes completing the content. Participants will be asked to complete an online screening and baseline survey at the start of the study, as well as two follow-up surveys at 12 weeks and 24 weeks.
Detailed description
We conducted a series of studies to evaluate the feasibility, acceptability, and potential efficacy of an AI chatbot (HeartBot I) in increasing women's awareness and knowledge of heart attack symptoms and appropriate care-seeking behavior. The results of those studies and HeartBot I design details were published elsewhere. In brief, HeartBot I was feasible (i.e., no withdrawal from HeartBot I conversation) and accepted by women, and its interactions were significantly associated with improvements in awareness and knowledge of all heart attack outcomes. However, since HeartBot I was a rule-based chatbot, its behavior and responses were limited to content authored specifically for HeartBot I conversations. To improve HeartBot I's capacity, we incorporated all women's heart attack and relevant questions collected in our previous studies and expanded its knowledge bank to enable more educational conversations. To achieve more human-like natural and personalized conversations, we implemented a new version of HeartBot (HeartBot II) powered by an LLM, specifically gemini-2.5-flash through the Google Conversational Agents platform.
Interventions
OTHERHeartBot II Program
Participants assigned to the intervention group will receive the HeartBot II intervention during the first 12 weeks. Participants in the intervention group will download the HeartBot II application on their smartphone, register their account with the access code provided by the research team, and will be asked to start Module 1 immediately after randomization. The HeartBot II application consists of 4 module and each module is scheduled approximately 3 weeks apart. It will take about 13 minutes to complete each modules.
OTHERWaitlist Control
The waitlist control group will not receive the HeartBot II intervention during the initial 12-week period, but will receive a text message once week for 12 weeks for attention control. These text messages do not include any educational content related to heart attack or heart health and are limited to neutral study reminders and general self-report prompts. Following completion survey at week 12, participants in the waitlist control group will initiate the HeartBot II program, delivered identically to the intervention group.
Sponsors
University of California, San Francisco
University of California, Davis
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
PREVENTION
Masking
NONE
Eligibility
Sex/Gender
FEMALE
Age
25 Years to No maximum
Healthy volunteers
Yes
Inclusion criteria
* Women * 25 years or older * Living in the US * University of California, San Francisco Health patient * Possessing a smartphone with a data plan or Wi-Fi access
Exclusion criteria
* Individuals who identify as male * Having self-reported cognitive impairment * Having history of heart disease or stroke * Healthcare professional/trainee * Working in the healthcare field
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Awareness and knowledge of symptoms and response to a heart attack | Baseline, week 12, week 24 | Participants will answer the following four questions assessing participants' awareness and knowledge of symptoms and response to a heart attack on a 4-point scale, ranging from 1, in which 1 indicates "not sure", to 4, which indicates "sure": (1) How sure are you that you could recognize the signs and symptoms of a heart attack in yourself, (2) How sure are you that you could tell the difference between the signs or symptoms of a heart attack and other medical problems?, (3) How sure are you that you could call an ambulance or dial 911 if you thought you were having a heart attack?, (4) How sure are you that you could get to an emergency room within 60 minutes after onset of your symptoms? |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| HeartBot II evaluation | Baseline, week 12, week 24 | We measure chatbot communication quality by looking at 1) message effectiveness using the validated Effectiveness scale consisting of five semantic differential items (e.g., effective-ineffective) on a 7-point scale, 2) impression of chatbot messages using the validated Anthropomorphism scale consisting of five bipolar adjective pairs (e.g., fake-natural) on a 7-point scale, and 3) conversational naturalness and coherence using two single-items on a 5-point scale (e.g., "Overall, how would you rate the conversations with your texting partner?" (1=Very unnatural, 5=Very natural). Chatbot identity perception is measured by asking "Do you think you texted human or artificial intelligent chatbot during your conversation?" with response options: (1) Human and (2) AI chatbot. To explore perceptions of AI in healthcare, participants will be asked to rate agreement with the three statements using a 5-point scale (e.g., "The use of artificial intelligence will result in better healthcare") |
| Perceived risk of heart attack | Baseline, week 12, week 24 | Participants are asked 3 items asking to compare their risk of experiencing a heart attack to other women their age over different time frames (next 5 years, next 10 years, lifetime) |
Countries
United States
Contacts
CONTACTYoshimi Fukuoka, Ph.D., RN, FAAN
CONTACTDiane Dagyong Kim
PRINCIPAL_INVESTIGATORYoshimi Fukuoka, Ph.D., RN, FAAN
University of California, San Francisco
Outcome results
None listed