Predictive Modeling and Big Data – A Match Made In Cyber-Heaven

Back in 2014, HealthInfoNet, which runs Maine’s health care exchange, sought to find a way it could, short of a better term, read customers’ minds. Specifically, the health care network wanted to study market share, clinical performance, population health and patient risk, all in real-time.

The network had its work cut out on the mission, with 1.3 million customers and clinical data culled from 31 acute care hospitals and 326 ambulatory locaters state wide, along with a host of specialty care, family care, and long-term care practices.

The answer came from a real-time blend of data analytics and predictive data modeling.

More specifically, using predictive data analytics from HBI Solutions in Palo Alto, Cal., analysts and customers at HealthInfoNet can now easily pull up “timely assessments of patient risk that can be leveraged for clinical treatment and care coordination before high-cost events occur.” Additionally, a patient’s risk score is updated in real-time each time the patient has a new clinical event at a participating site.

One hospital – St. Joseph’s HealthCare in Bangor – was an “early user” of the network’s predictive analytics tool, leveraging it to study whether their use has resulted in a reduction in hospital readmissions.

“They took a team approach to testing the new tool, setting up super users in the hospital and the outpatient setting,” says Devore Culver, HealthInfoNet’s chief executive officer. “And they are seeing results.”

All that has lead to a lower rate of readmissions at the hospital – a clear win for a health care center that wants to keep people healthy and better manage patient inflows on a day-to-day basis.

 

Predictive Data Defined

By and large, predictive data analytics is defined as pulling data from existing data sets to determine business and consumer habits and patterns, and use that data to forecast future usage habits and trends. Nobody says predictive modeling is fool-proof – but what it can do is reliably provide what-if scenarios and risk patterns, so companies and end users can make good, solid decisions on user (i.e., consumer) behavior.

The tools that data analysts use in predictive modeling are myriad, and primarily include:

– Data mining

– Statistical modeling

– Machine learning

It’s the blending of predictive models with big data that really give data scientists a turbo boost.

As companies accumulate a high volume of real time customer data, predictive data models are able to take all that historical data to forecast future behaviors and buying choices. In a word or so, predictive models enable companies like HealthInfoNet to transition from historical, reactionary data analysis to future, forward-based analysis of consumer behavior.

In real world terms (much like how HealthInfoNet clients use predictive data to lower hospital readmission flows), predictive modeling can tell companies what works and what doesn’t in driving consumer behavior. Think of a clothing retailer who needs to know what coupons and discounting programs resonate with consumers, and compel them to make buying choices.

Predictive modeling also contributes to the bottom in other ways.

At St. Josephs Healthcare, William Wood, a physician at the hospital, says a company-wide initiative to curb costs is well underway. Historically, Wood says cost-cutting campaigns weren’t effective, as the health care center had to rely on cost predictions based on risk that were formed by stale, outdated claims data. Now, the hospital can merge claims data with real-time, predictive clinical data.

“That enables us to sit down and negotiate with payers, using data more current that what they’re using,” he explains. Wood adds that as the health care sector heads toward value based healthcare design and risk-based contracts, hospitals and other health care providers must “tightly manage their patient populations and get a firm handle on their data in order to succeed.”

Companies in any industry can also use predictive analysis online, to better figure out customer browsing behaviors and steer them toward optimal buying decisions. An especially intriguing subset of predictive modeling is the use of Internet of Things, where data scientists can easily collect useful consumer behavior from data streams from millions of smart technology devices to get a firmer grip on consumer patterns.

 

Hindsight To Insight

Maybe the best result of data prediction models is that now, companies can shift from understanding why they lost a customer to a new level of awareness – one where predictive data prevents firms from losing customers before he or she walks away.

That shift in moving from hindsight to insight is at the heart of the predictive data revolution. Going forward, the mark of an effective consumer out reach program is defined by how companies capture, create and leverage data that can steer resources to likely consumer buying and usage decisions.

In a nutshell, it’s real-time, predictive data that is changing the customer experience – and changing company strategies in ways few CEO’s thought possible only 10 years ago.

Share this post

Leave a Reply

Your email address will not be published. Required fields are marked *