The complexity, diversity, and rich context of data being generated in healthcare are driving the development of big data for health [ 1]. Volume, velocity, variety, veracity, variability, and value are the “V’s” of big data, and these are encapsulated in the inherent challenges of biomedical and health informatics. Effective ways of tackling these challenges would pave the way for more intelligent healthcare systems focused on prevention, early detection, and personalized treatments.

The electronic patient health record (EHR) is a source of big data containing information regarding socio-demographics, medical conditions, genetics, and treatments; yet the human ability to process this data without effective decision support is finite. For healthcare, the goal is to provide a continually learning infrastructure with real-time knowledge production and to develop a system that is preventative, predictive, and participatory [ 2]. In order to achieve this goal, computer models are required to help clinicians organize the data, recognize patterns, interpret results, and set thresholds for actions. Examples of big data analytics for new knowledge generation, improved clinical care, and streamlined public health surveillance are already apparent. For example, the EHR has been successfully mined for post-market surveillance of medications and improved pharmacovigilance.

In the United Kingdom, the National Health Service intends to be paperless by 2018. The EHR will provide an integral resource for future clinical decision support systems (CDSSs) that may overcome human limitations in data comprehension and multitasking. CDSSs are already being developed to assess and improve protocol adherence [ 3], for medication reminders [ 4], to improve screening [ 5], and to predict hospital re-admission [ 6]. Certain conditions have multiple treatment options and CDSSs are being developed to help clinicians optimize strategies. For example, problems associated with anterior cruciate ligament (ACL) may be treated by physiotherapy, medicine, or surgery. In this regard, hierarchical learning for EHR data, along with information regarding occupational, recreational, and musculoskeletal data, has been used to classify treatment options [ 7].

It is expected that big data for health can play an important role in pharmacogenetics and stratified healthcare. Patients with a similar cancer subtype often respond differently when challenged with the same chemotherapeutics. For example, CYP2D6 is a polymorphic gene associated with response to Tamoxifen [ 8], BRAF mutations (Y472C) have been linked to Dasatinib response in non-small cell cancer of the lung [ 9], and multiple gene signatures have been recently associated with the response of rectal cancer to chemoradiotherapy [ 10]. Genomic instability is believed to be responsible for the observed diversity of drug response. Recent efforts have focused on exposing the complex interplay of genomics and chemotherapeutic sensitivity, resistance, and toxicity [ 11− 13]. For example, the Cancer Genome Atlas research network has launched the Pan-Cancer project [ 11] to analyze multiple tumor types and molecular aberrations in cancer types, and to enable scientists to discover new aberrations. Similarly, several projects such as the Cancer Cell Line Encyclopedia [ 12] and the Genomics of Drug Sensitivity in Cancer [ 13] are generating large genomic databases to specifically interrogate links between genomic biomarkers and drug sensitivity in hundreds of cancer cell lines. As evidence of the ability to leverage large pharmacogenetic databases to predict drug sensitivity, recent data suggests that computational algorithms for predicting drugs for individual cell lines can be improved based on genomic profiles and drug-response data [ 14]. Future work will involve testing the results of such algorithmic predictions on tumor response and toxicity in patients undergoing chemotherapy.

One of the major contributors of dramatic change in healthcare is the upsurge in communication instigated by social media. One recent estimate suggests over one billion healthcare tweets, reflecting the enormity of dialogue between healthcare providers, patients, organizations, and third parties. Approximately 20% of patients with chronic healthcare conditions such as diabetes, cardiovascular disease, and cancer, go online to actively seek others and share experiences of related conditions; such actions are creating patient communities on social media sites such as Twitter and Facebook. Social media is providing new avenues for investigators to enroll patients in research, and for patients to engage in sharing their health data. For example, in TuAnalyze [ 15], a joint initiative between TuDiabetes and the Boston Children’s Hospital, diabetics can monitor, evaluate, and share their results while actively participating in research on diabetes. Arguably one of the most interesting applications of big data analytics is in the ability to predict and track major outbreaks in order to improve public healthcare resources and the dissemination of healthcare messages to victims using social media. Predictions of serious healthcare emergencies such as exacerbations of asthma can be better predicted in models that combine social media analysis with environmental data. Unlike conventional models that base predictions with a two-week lag, Ram et al. [ 16] were able to create accurate prediction models of the volume of daily emergency-department visits for acute asthma (volume defined as low, moderate, or high) using Twitter activity, Google searches, and air-quality data [ 16]. Similarly, for major public epidemics such as Ebola, data can be collected and analyzed to support early warning systems for epidemic trends and to deliver messages for healthcare education interventions [ 17]. For example, Odlum et al. [ 17] demonstrated that analyzing tweet activity around Ebola virus detection (EVD) captured progressive increases in the number of tweets discussing EVD case identification in Nigeria occurring at least three days prior to the news alert and seven days before the official Centre for Disease Control warnings. Finally, several investigators are exploiting the potential of social media toward behavioral change and improvements in healthcare, including targeted interventions for developing countries for the purpose of global health promotion [ 18].

Another important driver for big data is the widespread deployment of sensing technologies. There has been increasing interest in wearable and implantable sensing owing to improvements in technologies, including sensors with enhanced wireless communications that have increased bandwidth and improved microelectronics. As a result of these developments, continuous, multimodal, and context-aware sensing is now feasible [ 19]. Simultaneously, advances in sensor miniaturization, embodiment, bio-fouling mitigation, and microelectronic fabrication schemes have improved the versatility and reliability of implantable biosensors. Episodic monitoring is being replaced, therefore, with continuous sensing and parallel improvements in integrated care, toward personalized and stratified healthcare. Patients at high risk of critical events, such as arrhythmias, following myocardial infarction may benefit from continuous monitoring of blood pressure, pulse, and cardiac rhythm such that arrhythmias can be detected in near real time and signals sent to a smartphone for ulterior processing [ 20]. Similarly, patients with chronic health conditions such as diabetes may benefit from implantable microsystems for continuous monitoring [ 21], toward improved long-term glucose control. Neuro-modulation, such as with implantable spinal stimulators, is having an impact in the management of chronic back pain [ 22]. Implantable sensors that monitor the axial load on an individual subject’s spine [ 23] may lead to personalized orthopedic prostheses. In future, it is hoped that sensors applied in close proximity to the site of operative intervention, which monitor local hemodynamics or tissue ionic content, will enable wound infections to be detected and treated at a sub-clinical stage, preventing fulminant sepsis. White blood cells and neutrophil counts could be continually monitored in patients undergoing chemotherapy cycles such that the earliest sign of neutropenia could be coupled with granulocyte stimulation to mitigate sepsis.

In conclusion, the sources and computational techniques for big data are rapidly increasing and the possible uses for improving health and well-being are manifold. For healthcare, the goal is to provide a continually learning infrastructure with real-time knowledge production and to develop a system that is preventative, predictive, and participatory. Big data analysis clearly has tremendous potential to improve healthcare and transform the health of populations. However, successfully exploiting this potential will depend on solving challenges associated with data privacy, security, ownership, and governance.

Daniel Richard Leff and Guang-Zhong Yang declare that they have no conflict of interest or financial conflicts to disclose.