Effect of a mobile health intervention optimised with artificial intelligence on cardiovascular risk reduction: The My Intelligent Cardiac Assistant (MICArdiac) Study

Participants who are randomised to the intervention group of the study will receive the MICArdiac mobile health program for 6 months, consisting of the MICArdiac smartphone application (app) connected to two wireless monitoring devices (a blood pressure monitor and a physical activity tracker). The MICArdiac app delivers health messages via in-app notifications, utilising data collected by the wireless monitoring devices to customise messages and using a machine learning algorithm and natural language processing (artificial intelligence) to optimise the content and sequencing of the messages to promote participant engagement with the intervention. The MICArdiac mobile health program aims to improve cardiovascular risk factors in individuals with high blood pressure and high LDL cholesterol. The program was co-designed with researchers, clinicians (GP and Cardiologists) and consumers, and developed in collaboration with the Sydney Informatic Hub (SIH). MICArdiac app: The MICArdiac app will deliver the messages and display data collected from the wireless monitoring devices (step counts, exercise duration, blood pressure, heart rate). The MICArdiac app securely collects the participant’s blood pressure measures and physical activity measures from the participant’s GoogleFit app and Apple Healthkit app. Wireless monitoring devices: Devices will include a wireless blood pressure monitor and a physical activity tracker (wearable bracelet) with heart rate monitoring ability. Participants will be encouraged to wear the activity tracker at all times, removing the bracelet only when required to charge it. Participants will be instructed to measure their blood pressure using the wireless blood pressure device up to three times a week, with a minimum of one blood pressure recording per week. Participants will be encouraged to measure their blood pressure at different times of the day. Messages: The frequency of messages will vary between 4 and 7 messages per week. Messages will be of three main types: 1) educational messages customised using three baseline variables (smoking, vegetarian diet, health literacy) to encourage behaviour change and self-management; 2) weekly adaptive step goals based on data from the physical activity tracker; 3) messages triggered by thresholds in monitored health parameters (blood pressure, heart rate, exercise duration). 1) Educational messages will be drawn from a message bank from 5 topic streams (general cardiovascular health, diet, physical activity, smoking, and medication adherence). The content and sequence of these messages will be optimised by a machine learning algorithm to reduce repetitiveness (message tone and intent, sematic similarity, lexical similarity, message readability) and promote participant engagement with the intervention. The algorithm was developed based on natural language processing analysis of messages and responses from our previous RCTs. 2) Messages containing weekly adaptive step goals will be based on an adaptive algorithm, which first monitors the participant’s step count for one week at baseline then it generates a goal based on this. The algorithm then generates a new goal each week based on the participant’s actual step count and their ability to achieve the previous week’s goal. The aim is to gradually increase participants’ step goals in achievable increments. 3) Threshold-based messages will be of three types: 1) indicating values of health parameters (exercise duration, blood pressure, heart rate) are within the normal/recommended range and encouraging maintenance of health behaviours and management; 2) indicating values are out of the normal/recommended range and encouraging improvement of health behaviours; 3) reminding participants to use wireless monitoring devices, triggered by a lack of data received in the previous week. Extreme values of blood pressure and heart rate will trigger a phone call from a health counsellor. Intervention usage metrics will be automatically collected via an electronic log recording the time messages are sent and the status of the delivery (i.e. delivered or not delivered). A log of responses received from participants and when participants contact the study team, including the reason for contact and the method used will also be kept. At the 6-month follow up, Intervention participants will be asked to complete a structured questionnaire and explore participants’ acceptability of the program (what percentage of messages did they read, which messages they remember, liked or disliked, what they did with the message, i.e. kept or deleted, whether messages were shown to family/friends and/or forwarded, whether the messages prompted some sort of action or behaviour change, whether messages were well understood or confusing). Use of the devices will be monitored using a rule-based threshold approach. If no data is received in the week period, a message reminding the participants to use the device will be sent. Finally, the log history of the devices digital user dashboard will be used to measure overall engagement in the intervention group.