Predictive analytics, otherwise known as data mining or advanced analytics, has become an area of increased interest in many industries. Predictive analytics leverages existing data sets and statistical algorithms to create models that predict future outcomes. With the advent of Big Data, large data sets combined with predictive analytics have led to jaw-dropping results.

To understand why, imagine a game show where you are asked to identify what picture is being slowly revealed square by square. As each piece of the larger picture is revealed, the chances of you accurately identifying the full picture increase. Similarly, the new ubiquity of data in fields – like healthcare – has increased the utility of predictive analytics.


For instance, in a recent pilot program at Roanoke, Va.-based Carilion Clinic in partnership with IBM and Epic, the team was able to create a predictive model that could identify patients at risk of developing congestive heart failure within the next year with 85 percent accuracy.

Of the data examined, it was predicted that 8,500 patients were at risk that previously had not been identified. Not only does this improve the likelihood of positive patient outcomes by enabling early intervention, but it also enables Carilion to reduce costs by proactively working with patients before high-cost emergency care is needed.

Although we now have a plethora of data, one challenge is that it is often located in multiple data silos. Another is that as much as 80 percent of this data is in unstructured formats, such as doctor notes and patient interviews. Here is where tools like IBM’s SPSS Modeler really show their worth.

Modeler is IBM’s desktop software for predictive analytics. It includes powerful tools for bringing data sets together for analysis regardless of what silo they may originate in. It also includes Natural Language Processing, or NLP capabilities, that enable organizations to tap into their unstructured data.


This is critical, as IBM work at another healthcare provider demonstrated that key predictive data was more frequently found in unstructured sources, and often this data was far more accurate than corresponding data from structured sources. See the table below.

Predictor Analysis % EncountersStructured Data % Encounters Unstructured Data
Ejection Fraction (LVEF) 2% 74%
Smoking Indicator 35%(65% Accurate) 81%(95% Accurate)
Living Arrangements <1% 73%(100% Accurate)
Drug and Alcohol Abuse 16% 81%
Assisted Living 0% 13%

Given Modeler’s advanced capabilities, I’ve had customers assume there is no way they could afford it. In fact, since you can get started with Modeler with just a single desktop license, it actually provides a very economical way to delve into predictive analytics without large investments in other analytics infrastructure. The low cost makes it easy to test by confronting  a specific population health challenge, like limiting readmissions of congestive heart failure or CHF patients, and thereby proving out the potential savings from a larger organizational analytics effort.

If you’d like to learn more, contact your CDW account manager for all the details about IBM SPSS Modeler and the QuickStart services we can provide.

One thought on “Predictive Analytics and Healthcare: I Can See the Future

Comments are closed.