Using Big Data to Improve Healthcare
As in all industries, big data will be revolutionary for healthcare – sparking the development of new treatments, more refined care strategies, and more streamlined systems. However, since big data is still a relatively new concept, healthcare companies are still just beginning to understand how to get the most value out of the information at their fingertips.
Technology consultants Steven Escaravage and Joachim Roski of Booz Allen have experience with almost three dozen healthcare-related big data projects. They recently shared what they believe to be basic rules of thumb healthcare organizations can utilize for initiatives of this type.
Make sure you have access to the best possible data
One common misstep Escaravage and Roski have witnessed is that enterprises don’t always choose their data patiently and strategically. Companies “often quickly gather and use the data that is the easiest to obtain,” the consultants explained, “without considering whether it really goes to the heart of the specific health care problem they’re investigating.”
The lesson to be learned from these companies is that you have to conscientiously design your project to source the best data if you want your efforts to pay off.
Booz Allen was working with a governmental agency to determine if a big data program could help detect healthcare fraud. Officials at the department who commissioned the project had identified the data they wanted to use. It was data that was immediately accessible and was already built into the agency’s auditing process. The consultants wanted to pull data from other locations so that the project would be more far-reaching and generate stronger insights.
Escaravage and Roski realized that there were archive documents and marketing language that could assist in the identification of fraud as well. However, the people who put together the project did not want to feed in that data source since including it was viewed as too time-consuming. The conclusions of the study were stale and redundant.
Make sure your test projects can be expanded
Narrowing your focus to solve a particular problem can sometimes be the right way to go. However, it’s important with your first project to widely showcase its power and get everyone engaged. Many times, healthcare big data projects are so laser-focused on one issue that isn’t easy to see how the project might be expanded organization-wide.
The consultants had experience with this problem when one of their clients asked for two specifically targeted analytic projects – one a bioinformatics algorithm and the other a real-time data storage tool for a research device.
“While each pilot solved a big data analytics challenge,” Escaravage and Roski commented, “the resulting capabilities did not provide examples that would be powerful enough to push transformational change across the organization.”
When you consider data, think in terms of provenance and lineage
Data is the ingredient of a big data project, of course. Sourcing data is similar to sourcing any other supply. You can potentially be misled if you don’t properly assess the provenance of the data – its origins – and lineage – how it’s been handled and manipulated. When you look at data that is incoherent, incomplete, or error-filled, you may think you are finding useful patterns when you are really looking at smoke and mirrors.
In other words, bad data, like any supply, can ruin a project. That’s true even if it’s only one piece. Let’s step outside data to cars. In 2014, the National Highway Traffic Safety Administration (NHTSA) discovered that a metal piece inside Japanese supplier Takata’s airbags could explode when the airbag deployed, sending shrapnel throughout the cabin. As of March 2015, 17 million cars made by 10 manufacturers were recalled. Certainly all those carmakers would prefer, in hindsight, to have chosen a different supplier. Think of those airbags as data. Don’t let bad information make your project undrivable or unsafe.
Escaravage and Roski had an issue related to lineage with one client. It appeared that there was a substantial rise in the quantity of inexperienced agents working within certain treatment segments. The people working on the project thought they had discovered an important pattern that made the organization vulnerable.
“However, when the findings were presented first to the administrator for the data source,” the consultants said, “he suspected that the trends might coincide with the roll-out of new address fields.”
The administrator was right: the address field transition accidentally created duplicate agent files when new addresses were added to the system. Numbers were inflated.
In other words, know your data’s backstory. You never want to let the backstory command the narrative of your project. Instead, you want to be in full control as a storyteller, using your data as material.
If you want to convince people, it’s hard to beat a great story. You can better understand what the data means within the context of the story, and it’s also more immediately exciting than looking at a bunch of numbers.
Tell a story, but know everything about your data. The storyteller shouldn’t be sent off on tangents by unwelcome surprises.
Enable your data scientists and researchers
Organizations now have the ability to give health researchers full access to big data. However, that capacity frequently isn’t used.
One of the consultants’ clients could have given those proficient in certain subjects access to look for unusual trends, but “the agency stayed with the conventional approach, and simply provided canned business-intelligence reports and visualizations to the end-users.”
Frustrated with the lack of insight from this approach, the leaders of the enterprise shifted gears and started giving subject-matter experts more freedom to poke around and develop their ideas. That approach was significantly more successful.
Choose a HIPAA-compliant platform
Get the best data. Design broadly applicable projects. Consider data histories. Enable users.
Beyond these insights from the consultancy executives, you also want the platform you are using to be 100% HIPAA compliant. Try out our advanced cloud computing platform built exclusively for healthcare.