With the onslaught of data created by the internet of things (IoT) and wearables, data sets are now appearing everywhere. In the next few years, device manufacturers will continue to create all kinds of highly specialized gadgetry – from sensors that note changes in glucose levels and dispense insulin like an actual pancreas, to facial masks that help users voluntarily move weakened cheek muscles – that will significantly contribute to the generation of up to 2 terabytes of health data per person.
But what sort of systems can actually handle this quantity of data? Certainly not the same systems that handle the relatively modest amount of IoT and wearables data being generated today. Some of this new data will be serialized, streamed, or unstructured, while some new data sets, like those generated by IoT devices and wearables, will be truly enormous.
A provider can try to navigate this new world as much as he or she wants, but once 2 terabytes of variable data are plopped in front of the clinician, how is he or she ever supposed to base a meaningful decision on that data in the 15 minutes allotted to see a patient? It’s just not possible.
What is possible is the use of a real-time healthcare analytics platform that will crunch that data and enable the clinician to have the cognitive support to make an informed decision right then and there.
Yet, despite the feasibility of such a system, most healthcare analytics platforms aren’t built that way today. Instead, they are built to accommodate an unremarkable sequence:
- A transactional system dumps data into an operational store.
- That operational store dumps the data into a data warehouse.
- Some analytics are performed within that data warehouse.
- The results are dumped into a workflow or engagement engine.
That sequence won’t fit into the data-rich future of healthcare. Tomorrow’s platforms will need to:
- Answer all the integration challenges that arise in the healthcare environment;
- Scale to handle all that data;
- Be flexible enough to keep pace with society’s ever-evolving quality-measurement needs;
- Integrate directly into care-delivery processes and workflows; and
- Leverage machine-learning techniques to more accurately predict outcomes in a healthcare environment.
So, with all of that in mind, here are the two essential steps to take to guarantee a platform will meet the unprecedented challenges tomorrow will surely bring.
1. Eliminate the problems created by a lack of interoperability.
Despite the best intentions of regulatory incentives and standards consortia, inconsistencies and quality issues continue to plague many health information exchange (HIE) interfaces, which can inhibit the effectiveness of most analytics platforms. If a platform is going to handle all those terabytes of data generated by the IoT and wearables, it will have to support high-quality integrations, too. This is a job for tools that:
- Understand the message formats of source systems;
- Address the specific challenges presented by event-based integration;
- Support and analyze real-time message feeds to identify variable data and data-quality issues;
- Accommodate message loading and output analysis while offering both robust monitoring and an error-handling infrastructure;
- Offer a way to store virtually unlimited data in a single repository;
- Allow acquired data to be processed and mapped to ever-evolving models;
- Provide an infrastructure that handles message re-ordering and variable data update modes; and
- Offer a set of prebuilt, standards-based models that cover core clinical-, claims-, and device-data domains.
It’s worth noting that HL7’s Fast Healthcare Interoperability Resources (FHIR) standard has rapidly evolved to provide a quality data model set to satisfy the demands of healthcare integration, so it’s the obvious model for the platform described here.
2. Consider the importance of accommodating IoT-device and wearables data at scale.
As organizations look to integrate all of this new data, healthcare analytics platforms will naturally encounter ever-increasing scalability challenges.
Other sectors have already begun addressing these challenges, and technologies such as distributed databases and computing engines are striving to meet the demand. Databases built for speed, such as Cassandra and Elasticsearch – and engines for big-data processing, such as Apache Spark – allow storage and processing capacity to be distributed over a number of servers, which makes it possible to incrementally ramp up capacity by increasing the size of the server clusters supporting the deployment. There are already many examples of these technologies being deployed across countless clusters, managing petabytes of data, and processing millions of transactions in an instant.
For software developers, creating these technologies requires a large research-and-development investment and dedicated integration/deployment teams. Waiting until the current technology reaches a breaking point will leave many developers without sufficient time to accommodate the transition.
When selecting an analytics platform, look for one that has the ability to deploy and scale solutions on demand by its explicit use of technologies such as Cassandra, Elasticsearch, and Apache Spark, and make sure it’s tied to one of the big cloud providers, like Amazon Web Services.
The anxieties that the onslaught of IoT devices and wearable data is causing are understandable, so it seems absolutely reasonable that users of today’s healthcare analytics platforms are wondering how they are expected to make sense of the tsunami of self-reported data headed toward their organizations’ doorsteps.
Those worries, however, are actually unfounded.
The technology already exists to make short work of literally any amount of data the IoT devices and wearables can and will throw at the industry. By following the steps outlined here, the onslaught of IoT devices and wearables data that continues to roll in will be both manageable and useful.
