Big Data: Let's Temper Our Expectations

By Viet Nguyen, MD and Dan Ramunda
01:20 PM

“Big data” is one of the hottest topics in healthcare these days. Indeed, there is a belief in many quarters that big data tools are required to help healthcare organizations take the most effective care of patients. That belief is not necessarily true.

In many instances, traditional analytics tools and technologies remain perfectly capable solutions to business and clinical challenges. Employing big data analytics tools may be like using the proverbial armored tank to kill a fly—in other words, overkill.

Don’t misunderstand; thanks to their ability to collect, manage and analyze huge amounts of varying types of information, big data tools are emerging as highly promising solutions for identifying and tackling a wide range of clinical and operational challenges. The question is: When does it make sense to use a traditional analytics approach instead of a big data analytics approach? In large part, the answer depends on the volume, velocity and variety of data being used.

Big data techniques tend to lend themselves to working with high volumes of variable, unstructured data. They offer a good solution for trying to “uncover” patterns and details that you don’t necessarily know exist. They help break data into smaller units in order to gain new understanding faster. Consider the example of an organization that sends out one million emails with the hope of generating a one percent response rate. Big data tools are needed to help narrow the target audience so that a three percent response rate is garnered from only a half million emails.

Analytics tools, on the other hand, are appropriate with structured, relational databases. They work well when the goal is to evaluate relationships within data to prove out a theory, or just to improve upon current methods being used for assessing the data. While big data tools divide data into smaller units for analysis, traditional analytics tools can be used to identify patients who are most at risk for a certain condition or who could most improve from a certain intervention. Traditional data analytics solutions that improve outcomes by even a small percentage can have a tremendous effect on outcomes.

The key to determining how and when to use a big data approach is to first figure out what you want to accomplish. Then, look at what’s available to you in terms of data. That includes evaluating the type, quantity and quality of data; as you approach large volumes of data that are highly varied, you should consider leveraging a big data approach. With this in mind, analyze which tools and techniques—big data or traditional analytics—can best help you to achieve the desired outcomes.

Also keep in mind that the right people with the right skill sets will be necessary regardless of the approach used. Just because staff members have done a lot of data warehousing and traditional analytics in the past, for example, doesn’t mean they know how to successfully conduct a big data analysis. Big data represents a large model shift in terms of how to approach analytics.

Ultimately, healthcare organizations need to understand what it means to make the most effective use of data. What are the costs and investments? Are the resources and the expertise available to make effective use of big data? In many cases, the answer will be no—and that’s OK.

In healthcare, analytics typically is not about trying to predict or change the behavior of every individual patient. It’s about improving the probability of better overall outcomes; an improvement that seldom is immediately evident.

Big data tools and technologies certainly can be helpful for improving the probability of better outcomes, but so can traditional data analytics tools. For healthcare organizations, investing in big data means investing in additional hardware, software and staff resources. Go there when appropriate, but don’t assume it’s always the answer.