“My Health Gain”, development of an online tool for effective health coaching
Noncommunicable diseases (NCDs), including heart disease, stroke, cancer, diabetes and chronic lung disease, are collectively responsible for almost 70% of all deaths worldwide WHO, 2018). The epidemic of NCDs poses devastating health consequences for individuals, families and communities, and threatens to overwhelm health systems. The socioeconomic costs associated with NCDs make the prevention and control of these diseases a major development imperative for the 21st century.
Currently, however, huge improvements in reduction of NCDs are hampered due to two perspectives dominating our vision on health and healthcare. First, we often focus on disease and symptoms studying detailed processes and linear cause-effect relations within human physiology. In doing this, emerging technologies has aid us in the pursuit of this reductionistic vision and allow more and more precise measurement of such processes and relations. Second, we study the effect of interventions on processes in human physiology, using well-designed random clinical trials and clearly defined intervention studies.
While both perspectives have resulted in a wealth of mechanistic understanding of NCDs and continuous improvements in treatment, this approach appears to be not sufficient to prevent lifestyle related diseases, and currently it has been proposed that an additional more complex perspective is needed (Rutter et al., 2017; van Wietmarschen et al., 2018). In this view, health is studied as a complex adaptive system with high resilience in a healthy state. When resilience is decreased, chronic lifestyle-related diseases may evolve over time including interactions of many biological and psychosocial factors to which the individual’s physiology adapts. Moreover, everyone follows their own trajectory in his/her context, resulting in various and different emerging health states over time. For an early prevention and optimal health advice, understanding and determining the “personal psychosocial and biological health state” and possible health gain at a specific time is therefore crucial.
In 2015, TNO developed a methodology to identify the “real age” based on the above-mentioned two dominating perspectives at that time, found at https://www.actify.nl/echte-leeftijd/. The “Your real age” of an individual was calculated based on the influence of several factors on life expectancy. ‘Bad’ factors reduce life expectancy thus heightening the ‘real age’, while ‘good’ factors increase life expectancy thus lowering the ‘real age’. As such, an evaluation was made based on Exercise, Smoking, Alcohol intake, Eating behaviour and Relaxation. While this technique made first steps towards a comprehensive health evaluation, new insights on especially nutrition, wellbeing, resilience and computational science, allow us a more sophisticated and nuanced approach, taking into account more factors and a more individualized approach.
In the meantime, human cohort data and intervention studies using emerging technologies such as wearable technology and ecological measurement assessment (EMA) to define well-being, lifestyle and context over time (van Os et al., 2017; Myin-Germeys et al., 2018) are providing a wealth of personal health data that can be combined with biological and genetic data. At the same time, researchers are challenged to interpret this data in an integrated and meaningful manner for both the individual and healthcare provider. Multiple health parameters describe processes and interactions between processes, where the total outcome of information becomes complicated. This means that while we know a lot and the information and data are present, there is little system-based methodology yet available to show the interconnectedness and ‘overall’ health state of an individual in a simple to communicate ‘health-marker’.
As a first step towards such technology, TNO developed a systems dynamic view on bio-psycho-social health and wellbeing (van Wietmarschen et al., 2018; Veldhuis et al, in prep, 2019), and developed an interconnected biological marker analysis which provides a visualisation of an individuals’ health against a healthy and unhealthy reference group (Bouwman et al., 2012; van den Broek et al., 2017), which is discussed below in more detail (Workpackage 1).
An important aspect is the importance to identify the change in health state over time. Both healthcare professionals and clients are interested to see changes and preferably progress over time in overall health for early prevention and optimal health advices. Ideally, a circle as shown in Figure 1 is present, allowing for continues feedback and information on the health-state and the effect of interventions over time. However, when comparing results of tests, it is hard to see what is more relevant: the change, for example, in blood pressure, or the change in mental health. If the blood pressure rises, but a client improves his/her mental fitness - which may improve awareness of unhealthy behaviour -, is this a positive shift? Even trickier might be the communication: how can we communicate this in such a way that both client and professional understand the same concept? And how are we able to communicate this in such a persuasive manner that clients actually change their behaviour? To create a clear road through this forest of health information, knowledge and methodologies on complex data modelling, persuasive communication, and visualisation are becoming available. These tools can help us to indicate causality, giving us the opportunity to focus on the most relevant factors which may determine sustainable physical and behavioural change.
To this end, many multivariate visualization and classification tools have been developed in data science, both unsupervised and supervised, such as Principle Component Analysis (an unsupervised method), Random Forest Analyses (Breiman, 2001), or Support Vector Machines (Cortes & Vapnik, 1995). These methods aim to combine several sources of information, in a more concise output giving one or more composite results on the data. However, all these statistical methods are hard to report back in a simple, obvious way for those not used to working with these methods while retaining the interpretation and information necessary.
As a solution to this issue, TNO developed a supervised method to interpret complex effects of nutritional and behavioural interventions in an intuitive way. This method has been dubbed the ‘health space’ and it combines knowledge and data to provide an intuitive abstraction of complex data (Bouwman et al, 2012; Van den Broek, 2017). Separating, classifying and visualizing data based on one or more meaningful scales resulting from the ‘health space’ facilitates interpretation of the data. However, the current health space analyses are primarily focussed on metabolic health parameters and at this moment includes only a very limited set of psychosocial parameters. As such, it is important to investigate how we can extend this approach to include a more comprehensive view of ‘overall’ health, incorporating lifestyle and relevant social factors as well. Furthermore, only recently a temporal component has been added to the ‘health space’, i.e. the health state before and after a personalized nutrition intervention was visualized, including individual health gain or loss (de Hoogh et al., unpublished results). Whereas these were only two measurement timepoints, this approach can be expanded to multiple timepoints to truly visualize and evaluate how an individual’s health changes over time. To study this innovative approach, we foresee that appropriate data pooling from available cohort data could be a useful approach to develop the methodology and a first prototype from a mathematical point of view. Iterative discussions of the approach with knowledge experts and partners are needed to develop a first prototype useful in real life.