Project Update EDIC: Exceptional and Deep Intelligent Coach

Measurement brings knowledge

Can we create an artificial coach that is perceived by the user as truly convincing and reliable, supporting the user in his/her self-management and supporting the development and maintenance of a healthy lifestyle?

This is the big research question within the EDIC project.

Chronic conditions like diabetes cannot be cured, but a person's lifestyle has a major impact on the course of the condition, quality of life and to what extent co-morbidities occur. Here, the aim is to get a holistic picture of a person's behaviour, looking not only at the physiological elements but certainly also at a person's well-being.

By using various sensors and data sources that allow important disease-related aspects as well as physical, cognitive, social and emotional components of health and lifestyle in daily life, an individual map of behaviour is created for this purpose. Of particular interest is what correlations might occur between these different components, how do you get the flywheel for a healthier lifestyle going?

The consortium

The consortium is an excellent mix of two university groups with extensive experience in the field of intelligent coaching systems to influence behaviour in chronic diseases (UTwente) and knowledge of and experience with advanced data analysis techniques to analyse complex streaming datasets and detect changes in them over time (TU/e).

From the private sector, Philips (monitoring systems), Holst (innovative sensors) , Roessingh Research (telemedicine services) and IMEC (digital technology research centre) are participating.

To demonstrate the acceptance and usability of the coaches, two pilot studies are being conducted in key chronic conditions: diabetes and (morbid) obesity; this in close cooperation with a hospital (ZGT) where many of these patients are seen. We spoke to Hermie Hermens (Professor of Telemedicine at University of Twente) together with two of the five PhD students at UTwente, Carlijn Braem and Ellen Kuipers, in late September 2022.

Pilot study on prevention

Within the project, Carlijn is organising a study using a wearable sensor from IMEC 'Chill Plus' in which lifestyle change is measured. The aim of this study is to gain more insight into which factors are most decisive for successful lifestyle change and then to be able to give people more guidance.

This pilot will start soon with a group of about 50 participants. "By measuring continuously, we expect up to a million data points," says Carlijn. The participants in this study are people who have already been guided by a GLI coach (combined lifestyle intervention coach) for eight months. They aim to change their behaviour and achieve lasting weight change.

"That is of course always the most difficult thing in behaviour change, to make these changes permanent, which is why we are going to measure a lot," he said. "Participants wear this sensor prior to the programme, shortly after the start and after the eight months to see how lifestyle changes have gone during this period.
We also put out several short daily questionnaires (via an app) to collect more contextual data such as 'how do you feel, how do you think you moved'. This is very difficult to capture yet, but of course we do know this also has a big influence on someone's behaviour and motivation."

Pilot study predicting 'unsuccessful outcomes'

Ellen is engaged in another study, namely predicting 'unsuccessful outcomes' in weight gain or insufficient weight loss after surgery, with about 350 participants from the bariatric* population so far.

"We started with a literature review where we concluded that little is known about what predictors of unsuccessful outcomes after surgery.
Many of these patients (20-30%) regain weight after surgery or do not achieve the desired outcome. This can have adverse consequences, for example the return of obesity-related diseases (e.g. diabetes mellitus) or insufficient improvement of these diseases. And that is often where the health gains lie. In this study, participants are followed and several questionnaires are administered on the various lifestyle-related factors. We also monitored physical activity with fit bits. What is 'new' is that in our study we combined different forms of physical behaviour (e.g. number of steps/day, moderate-to-intensive physical activity, sitting behaviour, etc) and thus created a 'holistic' profile. Based on this, we then created an exercise profile."

"It is fantastic how Ellen managed to implement this research in the hospital because we desperately need this data!" emphasises Hermie.

Biomedical analysis and computer science together

"What you want is a data-driven intelligent as well as personal coach," says Hermie. "We are developing a new method and we want it tested against reality, hence the two big studies above."

Data mining across many and with high granularity sampled data, will result in new knowledge about behaviour and the interrelationship of all these components. In addition, new time-series analysis techniques make it possible to detect timely changes in the continuously measured variables in their interrelationships. This new knowledge will be integrated with medical knowledge and the knowledge from psychology around behaviour change models to make more effective, persuasive interventions.

The ideal end result of EDIC?

"There is still quite a bit of work ahead of us, with EDIC we are already making good steps towards creating such a coach. At the end of the research, I expect that we will have algorithms to monitor behaviour in a holistic way, that we will have analysis techniques to objectively detect changes in behaviour and that we will have a kind of framework of a coach to manage with. We very much want to be able to show that we are able to perform holistic monitoring AND that it works well."

* Bariatric surgery refers to all operations aimed at reducing weight.

More information on the Personalised eHealth Technology Strategic Programme