The work started by performing a systematic review of models that predict mortality, hospitalisations and quality of life of CHF patients. We also analysed clinical guidelines, which proved the best starting point for the development of the HeartMan decision support system (DSS).
We collected user requirements via human-centred design, involving CHF patients from the beginning to the end of the project. The patients first participated in a diary and interview study to understand their everyday life, problems, needs and wishes. We continued with paper prototypes, through which we developed an outline of the mobile application. After that, we developed mock-ups of the HeartMan application, and the last round of prototyping used the actual HeartMan application.
In parallel with the human-centred design, technical work was carried out. We developed a sensing wristband to monitor the patients, with sensors for heart activity (PPG), sweating, temperature and movement. The movement and heart-rate data from the wristband was used to develop machine-learning models that can recognise the patients’ physical activity and its intensity, and trigger actions in appropriate contexts (e.g. a message on diet during eating). The PPG data was used to build a machine-learning model that can continuously estimate the patients’ blood pressure. While the method is experimental, its accuracy – when calibrated for a particular user – is close to the standards for regular blood-pressure monitors. Finally, we developed a machine-learning model that can accurately detect the patients’ psychological profile based on voice analysis performed during weekly conversations with a friend or relative.
The outputs of all these models are used by the HeartMan DSS. The HeartMan DSS provides personalised advice regarding physical exercise, diet, medication and self-monitoring. We developed methods that use a predictive model to advise regarding ambient temperature and humidity. The DSS also assesses the patients’ cognitive dissonance – a conflict between their desire to be healthy and practicing unhealthy behaviours – and provides cognitive behavioural therapy messages to align the patients’ actions with their desires. The final element of the DSS are mindfulness exercises for relaxation and general psychological wellbeing.
The sensing wristband and the methods for monitoring the patients were integrated in the mobile application using the IoTool sensing framework, which allows easily connecting sensing devices. On top of the framework, the user interface was implemented based on the human-centred design. We developed a backend where medical data retrieved from hospital information systems is stored in the standard HL7 FHIR format. The HeartMan system also features a web application where these data and other information coming from the mobile application are shown to medical professionals.
To validate the effectiveness and usability of the HeartMan system, we conducted a proof-of-concept clinical trial in two countries: Belgium and Italy. As mentioned before, the results showed several positive effects of the HeartMan system on self-reported and clinical outcomes.
The main result of the project is the HeartMan system itself. The consortium has prepared a business plan for its commercial exploitation, which considers providing the whole solution to end-users, or providing HeartMan components to other developers and software integrators. The main exploitable components are the IoTool framework; HeartMan wristband; activity monitoring and blood pressure estimation methods; psychological monitoring, cognitive behavioural therapy and mindfulness; DSS; and HL7 FHIR clinical data repository.
The scientific results have been published or submitted for publication in 11 papers in scientific journals and 18 papers in conference. The consortium has also reached out to other stakeholders via the project website, social media, newsletters, brochures, project videos, non-scientific publications and events. Of particular note is the conference organised in April 2019 in Brussels.
