Researchers at the University of California are investigating whether an AI-powered chatbot can help women recognize the symptoms and risks of cardiovascular disease. Developed at UC San Francisco, HeartBot provides personalized information about heart health and warning signs of a heart attack. A randomized clinical trial is now examining whether conversations with the chatbot improve women's knowledge and ability to recognize potentially life-threatening symptoms.
Heart disease remains the leading cause of death among women in the United States, affecting more than 60 million women. Despite public health campaigns, awareness of cardiovascular risks has declined. According to figures cited by the researchers, the proportion of women who recognized heart disease as their leading cause of death fell from 65 percent in 2009 to 44 percent in 2019. The decline was particularly pronounced among younger women and Black and Hispanic populations.
These findings motivated Yoshimi Fukuoka, professor and nurse scientist at UCSF School of Nursing, to explore whether conversational AI could provide a more accessible approach to cardiovascular education.
Personalized information
HeartBot was originally developed in 2021 by an interdisciplinary team from UCSF, UCLA and UC Davis. Researchers combined expertise in cardiovascular nursing, communication, linguistics, computer science and biostatistics. The team compiled approximately 600 potential questions about heart disease, together with evidence-based answers. Particular attention was given to how women describe symptoms and discuss cardiovascular health.
The chatbot explains common risk factors, preventive measures and symptoms that may indicate a heart attack. Although chest pain is common in both men and women, women may also experience symptoms such as shortness of breath, nausea, vomiting and discomfort in the back or jaw. These symptoms can be mistaken for other conditions, potentially delaying medical attention.
HeartBot can also refer to information shared during previous conversations, making interactions more personalized. The objective is not to diagnose cardiovascular disease, but to improve awareness and encourage appropriate medical attention when warning signs occur.
LLMs with controlled medical information
The latest version, HeartBot II, was enhanced in 2025 with educational modules covering cardiovascular disease, symptoms, risk factors and prevention. The system also evolved from an SMS-based service into a smartphone application using a large language model to support more natural conversations. To reduce the risk of inaccurate medical advice, its responses are grounded in a curated knowledge base developed by the research team.
Accessibility was another consideration. Researchers found that terms such as risk factors, although familiar to healthcare professionals, were not always clearly understood by participants. Subsequent versions therefore adopted more conversational language and simpler explanations. The application is currently available to study participants through the Apple App Store and Google Play Store. Wider availability is planned only after its effectiveness and safety have been established.
Clinical trial
Fukuoka's team is conducting a six-month randomized controlled trial involving 201 female UCSF Health patients. Approximately half receive immediate access to HeartBot II educational sessions, while the remaining participants form a control group and receive access later. Researchers will assess whether using the chatbot improves participants' understanding of cardiovascular disease and their ability to recognize heart attack symptoms over time.
The trial moves beyond measuring chatbot engagement to examining whether information is retained. Fukuoka ultimately hopes to evaluate future versions among several thousand women and make the technology more widely accessible, potentially through QR codes in clinics and public spaces. However, the current research has not yet established whether HeartBot improves emergency care-seeking behaviour or reduces cardiovascular complications.
AI identifies cardiovascular risks in mammograms
The HeartBot project reflects growing interest in using AI to address gaps in women's cardiovascular care. Another approach involves identifying previously overlooked cardiovascular risk indicators in routine breast cancer screening. As researchers at Penn State College of Medicine investigated whether breast arterial calcification visible on mammograms could help predict cardiovascular events.
Their study examined mammograms from 10,348 women. An AI system assessed the presence, severity and progression of calcium deposits in breast arteries. Women with worsening calcification showed an increased risk of cardiovascular events. Those in the highest severity category faced up to twice the risk, according to the researchers.
Unlike HeartBot, which focuses on education and symptom recognition, mammogram analysis aims to identify cardiovascular risk before symptoms develop. Both approaches remain under investigation. The mammography findings require further clinical validation, including evidence that identifying calcification improves patient outcomes. Together, these developments illustrate two potential applications of AI in women's cardiovascular health: improving recognition of urgent symptoms and identifying opportunities for earlier preventive care.
References
Add ICT&health on Google
Show more content from ICT&health in Google Search.
