Information technology innovation: the power and perils of big data.

The global health-care system is consistently under tremendous pressure to lower health-care costs, maintain high efficiency and quality of care, and remain up-to-date technologically in an era of instantaneous information exchange. In the UK, around 8.4% of the gross domestic product is spent on health care (approximately 0.19 trillion GBP). In the USA, this number is 17.9% of gross domestic product, or 2.7 trillion USD. With the introduction of health-care reform and a shift in payment structure to pay-forperformance, further pressure has been placed on the health-care system to reduce costs and increase health-care quality. Additionally, a shift in patient characteristics to an ageing population and improved access to care have increased the number of patients seeking care. Compounding the situation is a shortage of key practitioners, including nursing staff, in the medical workforce. 4 As a result of staff shortage and external pressure and regulations fromgovernment agencies to reduce costs, the health-care system must find away to improve the quality of patient care formore patients with fewer resources. With these difficulties in mind, and the additional challenges that lie ahead, the health-care system, including anaesthetists, must continue to use innovative medical technologies and becomemore efficient in the collection and analysis of this information to drive cost-effective clinical practice. As discussed in the article by Simpao and colleagues, technological advancements in health care have led to an explosion in data collection, increasing storage and analysis needs. In 2011, there were 1.8 zettabytes of data created globally. In the same year, it was estimated that data from the US health-care system reached 150 exabytes. 7 This number will continue to grow to reach zettabyte (10 gigabytes) followed by yottabyte (10 gigabytes) levels over time. Data of this magnitude are known as ‘big data,’ defined as electronic data sets so large and complex that they are difficult or impossible to manage with traditional software, hardware, or both; nor can they be easilymanagedwith traditional or common data-management tools and methods. There are three primary characteristics of big data: volume (the amount of data generated by organizations, individuals, or machines), variety (data in all forms; structured, unstructured, and semistructured), and velocity (the speed of data generation, delivery, or processing). The creation of these massive data sets with varying formats is a result of the proliferation of electronic health records (EHRs). The EHRs have vastly improved the maintenance of health information and have promoted the collection and sharing of information among providers across all health-care disciplines, leading to a more collaborative approach to patient care. In thefield of anaesthesia, the EHRs, also knownasAnaesthesia Information Management Systems (AIMS) or Anaesthesia Information Systems (AIS), have decreased inaccuracies, incompleteness, biases, and inherent errors. However, implementation of these EHRs has created data sets that can be difficult to analyse for quality control or research purposes. In one study of AIMS, event recording dependent on user input can have a low sensitivity (38%), leading to under-reporting of key clinical events, demonstrating that EHR systems are in need of improvements and analytics to identify issues. Further adding to the predicament of using big data in health care is the development and use of low-cost, non-invasive, wearable health-monitoring systems that allow for continuous monitoring of patients’ vital signs and mobility from external locations rather than the traditional approach of hardwired equipment. These devices, along with EHRs and existing clinical monitoring technologies, have challenged the health-care industry in the storage and analysis of this big data. Additionally, such data can be collected in systems that do not communicate, and datamight not be collected in a structured format, further confounding analyses. There are several systems that have been developed to overcome structural and analytical data issues. Themost popular system at this time is the open-source distributed data processing platform, Hadoop (Apache platform). Initially, Hadoop was developed as a platform toaggregateWeb search indexes. Usingnumerous servers, known as nodes, Hadoop has the potential to store and process extremely large amounts of data by allocating partitioned data sets to each node. An analysis request in Hadoop (a Map Reduce request) is allocated to each node and each data set, and executed at the data level in parallel (the Map process), and the results are integrated and aggregated for the final result (the Reduce process). Hadoop has the ability to analyse unstructured, semi-structured, and structured data. As a data-protection method, Hadoop maintains redundant data sets in different nodes to protect the data and analyses from system crashes. If a node becomes unusable, an additional node will be used to continue the requested analyses. 9 Therefore, Hadoop is structured to serve dual roles, namely the ability to store massive amounts of data