Every day, smartwatches and smartphones record huge amounts of data: how many hours we sleep, how many steps we take, how our heart rate varies throughout the day and night, and even air quality and weather conditions.
In a previous article, we discussed intelligent textiles, another frontier in the field of monitoring health via wearable devices.
Could a wearable device that we use daily thus also be useful for monitoring brain health? This is the question at the centre of a study conducted at the University of Geneva, where researchers from the Quality of Life Technologies and the Cognitive Aging labs attempted to understand whether wearable devices, which many people routinely use, could provide useful information for monitoring fluctuations in cognitive functioning and emotional states over time.
Continuous monitoring of these two aspects is essential for understanding brain health, which is in constant flux, affected by physiological, behavioural, and environmental variables, and is assessed using tools like neuropsychological tests and questionnaires – which, by their very natures, can only provide a snapshot, and are unable to record fluctuations over time.
The study, published in the journal npj Digital Medicine, was carried out as part of the Providemus alz project, a larger project that combined continuous information collection via wearable devices with quarterly assessments of cognitive functioning and emotional states.
The objective was to verify whether the passively recorded data was able to predict the results of periodic cognitive and emotional assessments.
For ten months, the researchers tracked 82 cognitively healthy adults living in Switzerland and neighbouring regions of France. All of the participants used smartwatches and an application installed on their smartphones, which allowed for the collection of information on physical activity, sleep, heart rate, weather conditions, air quality, and exposure to atmospheric pollutants, for a total of 40 variables.
At the same time, every three months, the participants completed validated questionnaires and cognitive tests that made it possible to measure 21 brain health indicators, including memory, attention, information processing speed, cognitive flexibility, inhibitory control, reasoning, and
additionally, anxiety, depression, stress, nervousness, irritability, hostility, and positive and negative emotional states.
In order to assess the value of the passively recorded data, the researchers developed machine learning models capable of predicting the results of the cognitive tests and questionnaires. A further objective was to test the models’ predictive capability for people never before observed, as well as to monitor the changes in given individuals over time.
Overall, the models thus developed were able to predict, with low margins of error, numerous cognitive and emotional indicators. Subjective results, self-reported by the participants via questionnaire (such as anxiety levels or perceived mood), were found to be easier to predict than objective ones. The latter – consisting of the results of the cognitive tests – were more difficult to predict, because a given individual's performance might vary greatly from one assessment to the next.
Analyses also showed that environmental variables, like weather conditions and atmospheric pollutants, helped most of all to explain differences between individuals. Conversely, when the goal was to monitor changes in a single individual over time, physiological and behavioural variables, and especially sleep (which we discussed in a previous article about smart pyjamas), heart rate, and biological rhythms, proved to be the most important factors.
Which information contributed most to the models’ predictive capabilities? One result that the authors found especially interesting showed how, if we consider cognition and emotional states separately, the types of information collected take on different importance. Cognitive aspects are most strongly linked to weather conditions, atmospheric pollutants, and heart-rate over the previous 24 hours. Emotional states, on the other hand, are more noticeably associated with heart rate during sleep, in addition to weather and overall heart rate.
According to the researchers, the study’s principal contribution has been to show how data recorded passively by widely used devices can be used to predict changes in emotional and cognitive wellbeing in our daily lives. The authors make clear that, given that this is an observational study, it cannot establish causal, but merely associational, relationships. Nonetheless, the study has the distinction of proposing a new possibility for monitoring brain health, which is usually done by means of check-ups and tests carried out months or even years apart.
These outcomes allow us to imagine an unbroken record, based on the analysis of digital data collected daily, and provide a hint of a future where passive monitoring via smartwatch and smartphone has a place alongside clinical assessment. If these results are confirmed by larger-scale studies, wearable devices could become simple, minimally-invasive tools for monitoring changes in cognitive and emotional wellbeing over time.