Short-Term Efficacy of the Artificial Intelligence HeartBot II in Increasing Awareness and Knowledge of Heart Attack in Women: Protocol for a Randomized Controlled Trial With a Waitlist Control

Abstract

Background Heart disease remains a leading cause of death for women in the United States. Despite this burden, awareness that heart disease is the leading cause of death among women declined from 65% in 2009 to 44% in 2019, with the largest declines observed among Hispanic, Black, and younger women. Thus, innovative, scalable, and cost-effective educational strategies are needed to improve women’s awareness of heart attack symptoms and appropriate care-seeking behaviors. Objective This study aims to evaluate the short-term efficacy of the artificial intelligence (AI) HeartBot II, a chatbot-based educational intervention, in improving women’s awareness and knowledge of heart attack symptoms and care-seeking behavior compared with a waitlist control group. Methods This randomized controlled clinical trial (RCT) with a waitlist control will enroll 200 women aged 25 or older, who will be randomized using a 1:1 allocation ratio. The intervention group will download the AI HeartBot II app and complete the 4 modules (including information on heart attack symptoms, risk factors, and calling 911) over 12 weeks. The waitlist control group will start receiving an identical intervention at 12 weeks. The primary outcomes will be change from baseline to 12 weeks in a 4-item heart attack response preparedness score, calculated as the mean of 4 self-reported items assessing confidence in recognizing signs and symptoms of a heart attack, distinguishing heart attack symptoms from other medical problems, calling 911 or an ambulance if a heart attack is suspected, and reaching an emergency room within 60 minutes of symptom onset. The primary analysis will estimate the intervention effect using constrained longitudinal data analysis implemented with linear mixed models, including fixed effects for time and time-by-treatment group interaction. Sensitivity analyses for the individual ordinal items will use ordinal logistic mixed-effects models. Results We received approval from the University of California, San Francisco, Institutional Review Board (No.