Using conversational AI technology to support patients with atrial fibrillation

Atrial fibrillation (Afib) is the most common heart rhythm disorder, affecting more than half a million Australians, and 37.5 million people globally. Its prevalence is increasing rapidly and is projected to double by the year 2050. This is of concern as Afib significantly increases risk of stroke, heart failure, cardiovascular events, and all-cause mortality. Afib costs the health system more than any other heart rhythm disorder due to hospitalisations, ongoing follow-up care and remote monitoring support required by patients. Recent clinical guidelines have recommended the use of digital tools to help support Afib patients. These tools need to be effective, scalable, and cost efficient so that the needs of a growing population and an at-capacity health system can be addressed.

Conversational technologies, ‘chatbots’, have become of particular interest in the field of healthcare, and have shown effectiveness in improving education, medication adherence, physical activity in various patient populations. These technologies simulate human text or verbal conversations. Conversational agents that use artificial intelligence (AI) are more advanced as they can interact with humans via speech input and output (using natural language processing), closely mimicking conversations with a human. Conversational AI technology has not been well studied in healthcare, however, provides potential as it is interactive, ease to use and scalable.

At the Westmead Applied Research Centre, University of Sydney, researchers are evaluating the effectiveness of a multicomponent digital health solution, which involves automated phone calls (with conversational AI), text messages and an educational website to support patients with Afib to self-manage their health in the community. The program reaches out to patients via automated scheduled phone calls, and does an assessment of the patient’s health whilst also providing educational information. The AI component means that the patient can respond to calls with their voice, as they would when talking to a relative on the phone. Patients that are identified as high risk (through ‘alert’ systems) are then triaged and directed to further medical attention if required.

This program is being evaluated through a pilot randomised controlled trial, which is now completed, and the results are being analysed. As part of this research, qualitative interviews were conducted with the participants that received the program to understand the barriers, enablers and other

user perspectives of using this novel technology. Findings from this research will be valuable in understanding if digital health tools can be used to assist with self-management of Afib patients and will also provide insight into user perspectives of conversational AI in healthcare.

Read Ritu’s full research paper here