Big data is a term used to describe the enormous IoT datasets made possible for analysis by cloud computing and the more recent increases in processing power and storage. ‘Big data’ is contrasted with regular data by three Vs: volume, velocity and variety—referring to the sheer size of datasets, the speed at which it is added to, and the integration of data types often considered incompatible using regular data techniques.
Big data analytics is the packaging of that data into comprehensible formats for human analysis—often with the use of AI and/or machine learning pre-processing to detect trends and patterns that may offer actionable insights.
By collecting, maintaining and analyzing more information about an individual’s health, doctors can diagnose more holistically, and even proactively. On this level, that would be the use of devices—especially wearables like smart watches and fitness bracelets—that can monitor vital health information like heart rate massively increases the data available.
By collecting, maintaining and analyzing health data across the population, doctors and researchers can get vital context for individual diagnoses which can be used as the basis for risk modeling. In the same way that genetic markers are now used to assess the chance of someone developing, for example, breast cancer, big data analysis opens the opportunities for all sorts of health and lifestyle factors to be taken into account. Again, wearables are set to bolster this type of modeling by supplying ever-greater amounts of data for analysis, on top of existing lab results, questionnaires, and patient outcome data.
Predictive analytics can be used to keep increasingly vital machinery in working order. Using real-time machine health monitoring, remote technicians can detect signs of possible deterioration, enabling them to schedule preventative servicing. In this way, X-ray machines, MRIs, and other lab equipment can be available for crucial medical tests without delay.
The two big uses cases for big data analytics in healthcare are:
Using general population data—comprised of lab results, questionnaires, biometrics, and more—patterns that may lead to particular conditions can be identified. By collecting data on the individual, the correspondence to those patterns can be used to assess the risk of developing that condition. This kind of risk modeling is crucial in an era typified by lifestyle diseases like cancer and diabetes, that are best (and most cheaply) treated by preventing them in the first place.
Similarly, continual monitoring of an individual’s health can lead to proactive interventions. As their vitals and exercise patterns begin to match the pattern for pre-diabetes, for example, they can be called in to see their family doctor for personalized recommendations on how to avert the transition into full diabetes.
Both of these processes become more accurate as more data is fed into the models. Therefore, although wearables can provide crucial information like heart rate, step counts, and even sleep patterns, a more complete picture is built by also including the result of any hospital tests—data made accessible as more medical devices become connected to the Internet of Medical Things (IoMT)—and treatment outcomes. The overall effect is to shift the delivery of medicine from a reactive, episodic model, to a proactive, holistic, and continuous approach. Rather than treating high blood pressure after the fact, patients are given contextual, risk-based recommendations to prevent it from occurring in the first place.
There are three big challenges for big data analytics in healthcare:
For big data analytics to yield any substantial benefits, the datasets need to be large enough. This requires professional medical devices to be connected—a process that is underway, largely fueled by preventative maintenance capabilities, but will take time to fully permeate. And it requires the further proliferation of wearables in the general population. Fitness bracelets and smart watches are present, but they need to become far more popular for big data analytics to have the desired impact.
A key part of big data analytics in healthcare is electronic health records (EHR). These need to contain a persistent digital thread of every piece of data connected to an individual. They also need to be transmittable to centralized repositories for use in larger datasets. Without the population data to reference against, risk modeling is all but ruled out, and predictive analytics are of limited use. The challenge here is a lack of interoperability. EHRs are often not transferrable even between two hospitals in a local area, let alone on a national—or international—level. And each silo has its own standard. However, there is a race to solve this problem, spearheaded by Apple and a number of competing blockchain initiatives.
The biggest concern, of course, is privacy. Health information is treated as ‘sensitive data’ under the European General Data Protection Regulation (EU GDPR). People are rightly concerned about their intimate details being pored over unscrupulously or stolen by hackers. Both scenarios are made more likely by inclusion in large pools of data. Again, great strides are being made in this area—focused on the anonymization and pseudonymization of data before inclusion in large datasets—but it’s an issue that will always loom over big data analytics in healthcare. To get an insider’s view on the now and future of big data analytics in healthcare, read the Axendia IoMT Impact Survey.
All these challenges are surmountable and have potential solutions on the horizon. Once these obstacles are overcome, the IoMT and the big data analytics it serves have the potential to revolutionize healthcare; shifting it from a model of episodic, reactive treatments, to proactive, contextual interventions. That kind of medicine is exactly what our era needs.