Big Data in mental health: a challenging fragmented future

Big Data has been a buzzword in almost every possible field during the last few years. The rapid integration of massive amounts of information from diverse sources fed the hope for a new era also in health sciences. Following its impact on other fields (i.e., marketing and commerce), many authors hypothesized that, by dynamically merging diverse datasets and a mining process, groundbreaking conclusions would be obtained in almost every medical specialty. This would, allegedly, represent a paradigm shift on how research is performed and, as a consequence, a dramatic change in clinical practice1.

Neurosciences were not out of this wave of – somewhat grandiose – expectations over the Big Data potential, given the increasing need to bridge the gap between brain structure/function and behavior. This complexity requires a comprehensive, holistic perspective in order to fully understand the course of an illness2. However, one of the most important limitations in brain research is that it has so far yielded partial, diverse and not generalizable results, which can hardly be directly transferred into clinical practice. Moreover, psychiatry research has not been able to link the current taxonomy and brain functioning3, so that psychiatry is apparently doomed to remain one of the few medical branches in which nosology does not relate to etiology. Therefore, the Big Data promise seemed like a hand in a glove to get the full picture of psychiatric disorders and fill the gap between biomedical and behavioral data. Yet, as years went by, Big Data stood still as it was first born: a promise failing to bring significant integrative answers to neurosciences.

While Big Data relies on numerous and different sources of information, the wide availability of mobile technologies is, out of doubt, amongst the most significant factors that boosted its potential. Mobile devices with hundreds of sensors and powerful processors are carried ubiquitously all day long by more and more people, for multiple and seamless purposes. This is gradually surpassing personal computers usage as a source of information for Big Data. In addition, there is an increasing preference of consumers to integrate mobile technologies and the data they can offer into their own personal health care. Physical activity, sleep patterns and location tracking data sets are easily obtained from either increasingly cheaper smartphones or newer, discrete and affordable wearables. Big technological corporations did not miss this opportunity, offering a number of devices and cloud services which could store and integrate all the health data generated (e.g., Apple's Health, Google Fit). Likewise, some of these companies grew progressively interested in health research through their promising platforms (Apple Research kit and Google Study kit, respectively).

These technologies can bring many advantages over traditional research methods in mental health. For instance, ecological momentary assessments allow a continuous and real‐time data collection in the subject's own environment4. Additionally, users’ interactions with their mobile devices and their respective sensors could provide passive, objective information about their behavior patterns5. In theory, this kind of information has the potential to help designing newer prophylactic strategies as well as allowing personalized treatments6. Moreover, integrating behavioral and biomedical data (i.e., genetics, biomarkers, neuroimaging, etc.) and analyzing these datasets could ultimately allow the development of new predictive models, and the identification of previously unsuspected etiopathogenic factors and possibly new treatment targets7.

However, in practice, big challenges lie ahead in the pathway of translating the above promises into reality alongside the expected barriers of age and education level. One problem affecting not only behavioral sciences, but all mobile technologies which could provide an essential component of Big Data, is that of fragmentation. The market fragmentation of current mobile operating systems (Android, iOS, Windows Phone, Blackberry, etc.) and smartphones ownership represents a still unaddressed and growing issue in the field. According to the International Telecommunications Union, in 2015, around 95% of the worldwide population had a subscription for a mobile‐cellular telephone, but less than 50% of these devices were connected to the Internet8. Among smartphones users, Android and iOS together have 96.4% of market share according to the International Data Corporation, with 78% of this amount pertaining to Android9. The fragmentation in the market of wearable devices is much more problematic, with more than 20 companies offering these products, which are continuously feeding information to independent databases.

Hence, the promising future of the components of Big Data provided by mobile devices might be severely hampered by the companies’ aim of selling their own products with data collected through non‐open platforms. In other words, any smartphone or wearable provides relevant information to databases owned by these companies, not available for being exported or integrated with clinical information10.

In order to obtain significant results from the promised potential of Big Data, it may seem that fragmentation is a small technical problem in comparison to other relevant issues, such as ethical and sociological aspects. However, keeping in mind that the ultimate goal of the Big Data promise is to integrate data from multiple sources and that mobile technologies play a key role in this process, fragmentation should not be underestimated. This problem can be overcome on the long run, providing there is a true effort of both public and private sectors to altruistically collaborate towards an open health science, which could ultimately improve health research and as a result have a significant impact on public health.

Diego Hidalgo‐Mazzei, Andrea Murru, María Reinares, Eduard Vieta, Francesc Colom
Bipolar Disorder Program, Department of Psychiatry and Psychology, Institute of Neuroscience, Hospital Clinic, University of Barcelona, IDIBAPS, CIBERSAM, Barcelona, Catalonia, Spain

D. Hidalgo‐Mazzei is supported by an Emili Letang grant from the Hospital Clínic of Barcelona. F. Colom is funded by the Spanish Ministry of Economy and Competitiveness, Instituto Carlos III, through a “Miguel Servet II” postdoctoral contract and a FIS (PI 12/00910).