Heart disease has an enormous impact on patients, healthcare and society worldwide. Heart failure is usually the result of a prolonged interplay of physiological characteristics, risk factors and behaviour. Rehabilitation following a heart attack, where patients learn to adapt their lifestyle, is the most important type of intervention there is to speed up recovery and prevent a relapse – and yet less than half of patients take part in programmes of this kind, with many participants prematurely dropping out. In addition, patients often fall back into their old behavioural patterns and habits that played a key role in bringing on heart failure. As such, any gains made through rehabilitation evaporate in the long term. 

Rehabilitation currently an overly draining process

Something that may account for the low level of interest in and dropout rate from rehabilitation programmes is that they are very intensive and require patients to make quite a few changes in their daily lives. At the same time, many patients suffer from extreme fatigue and are passive as a result of the traumatic and profound experience they have just had. Artificial Intelligence (AI) may offer a solution in terms of being able to identify behavioural risks more effectively and being able to continually motivate, coach and support patients throughout their journey towards a new healthy lifestyle.

A bespoke solution using AI

The TIMELY project, which is coordinated by the clinical psychologist Jos Bosch of the University of Amsterdam, received a 5.7 million grant from the European Commission to develop an eHealth platform that identifies the risks surrounding the failure and success of cardio rehabilitation and provides heart patients with continuous monitoring and support. To this end, the communication scientist Gert-Jan de Bruijn is leading a central work package, which ensures the development of a chatbot as a digital coach.

AI forms the backbone of the project: ‘First and foremost, we are using AI to create an accurate risk profile and to identify the key predictors of successful and less successful rehabilitation, using a number of large datasets that are already available. Who exactly are the people suffering from heart disease?’, de Bruijn explains. ‘Secondly, AI is used to monitor, coach and manage patients throughout and after their rehabilitation process based on real-time data collection and analysis. This allows us to respond more effectively to the issue of how best to assist patients, by tailoring services to individuals and anticipating their behaviour.’

‘What is unique about this project is that it involves us using and integrating a range of different sources of information, consisting of detailed medical and physiological information, information on behaviour and psychological characteristics, as well as information on the patient’s environment. This knowledge is brought together within a data system that allows rehabilitation to be optimised both specific to a particular individual or context,’ Bosch adds. ‘Because this project builds on an existing technology platform that is already used in healthcare, the reviewers were particularly enthusiastic about the impact it might have.’  

A chatbot: a personal coach available 24/7

The TIMELY project focuses on the concept of ‘personalisation’. As such, a key outcome of the project will be a chatbot (or conversational agent) as a coach. ‘It is well known that a personal coach is much needed by heart patients during their recovery,’ says de Bruijn. ‘Unfortunately, providing each and every patient with a human coach is simply not feasible – particularly once patients return to their day-to-day home situation. With a chatbot, however, we can.’

Instructions tailored to individuals, patterns and context

This chatbot will soon be able to provide monitoring and advice on health and behaviour, on request as well as unsolicited, in the form of written and spoken text, based on all the personal data available. ‘It won’t be a broken record with a fixed set of instructions,’ Bosch explains, ‘but rather the instructions will continually be tailored to the individual, the patterns and the context. On top of that, the system will be able to “learn” – in other words, the system will be able to get to know individual patients better and better based on their actions and responses.’ For example, a patient may be advised to go for a walk now, given that rain is forecast in an hour, or be given a little pep talk at a time of which the system is aware that the patient may be tempted into bad habits, or may be sent a notification when their blood values need extra attention.

Building on knowledge and collaborating with professional practice

‘What is brought together within this project very clearly,’ Bosch explains, ‘is expertise in affecting behaviour, on the one hand, and that of diagnosis on the other hand. We will also be building on data that we collected in other research projects and will be working with both technology companies and practitioners and hospitals. This allows us to translate things into interventions that can actually be used in practice, across Europe, quite quickly.’

TIMELY funding and launch

TIMELY stands for: A patient-centred early risk prediction, prevention, and intervention platform to support the continuum of care in coronary artery disease (CAD) using eHealth and artificial intelligence. The project was launched on 1 January and will run for 45 months. The project was developed for a top-down call in Pillar III of the European Horizon2020 programme, which focuses on societal challenges, and received a 5.7 million grant from this programme. The UvA heads the interdisciplinary consortium, which includes 12 other partners (from Germany, Greece, Spain and Austria).