Promise and pitfalls in the application of big data to occupational and environmental health

Promise The potential power of big data to inform public health decision-making has been widely recognized [1, 2]. However, there is a paucity of published primary research employing these methods in this journal and elsewhere [3, 4]. The American Journal of Public Health encouraged new research in this area and recently appointed an inaugural associate editor for digital health [3]. Big data are typically defined in relation to the “three Vs”, volume, velocity and variety (and more recently, variability, veracity and value) [5]. Other defining characteristics include the emergence of new data sources and providers such as social media, mobile applications and wearable technology such as fitness trackers (the “quantified self” [6]), the need for new analytical methods such as machine learning, non-traditional multi-disciplinary partnerships and real-time analysis and forecasting [7]. Along similar lines, sharing of clinical trial and other study data has also been advocated as a means of broadening access to and more fully exploiting the collective power of data. In addition to increasing statistical power, which could potentially facilitate detecting small signals earlier, which may be particularly important in environmental health, advantages of pooling data include enhanced ability to examine heterogeneity between diverse populations, and consideration of novel hypotheses not tested by the original investigators [8]. Data sharing initiatives must overcome barriers including providing protections for original investigators, particularly those in low-resource countries [9], and issues related to data ownership, privacy and security [8]. The Healthy Birth, Growth, and Development–Knowledge Integration initiative is an example of a data sharing initiative which has navigated many of these issues [8]. A need has also been identified to address barriers to the international sharing of routinely collected public health data, including technical, motivational, economic, political, legal and ethical factors [10]. Exposure analysis is the keystone of occupational and environmental health. As a result, the concept of big data in this context is linked closely to that of the exposome, the totality of human environmental, occupational and other exposures from conception to death [11]. These exposures interact with other determinants of internal dose and health effects characterized by their own data-rich “omes” – the genome, metabolome, lipidome, transcriptome and proteome, among others, analysis of all of which requires novel data analysis methods [11–14]. The exposome may be characterized using a vast array of methods including measurement of both exogenous and endogenous biomarkers in biological specimens, direct environmental monitoring using dedicated sensors, and indirect sources such as operational data from metering and energy use, and facilities management data [12, 15–17].