Data Analytics: Building Knowledge for Better Health

It's one thing to have data and another to use it. And we're on the cusp of a huge surge­—from adoption rates in the 10 percent range in 2011 to 50 percent by 2016, according to Frost & Sullivan. Analytics is the next level of pushing information from data to knowledge and is a complex proposition for most healthcare organizations. The best advice: start with the basics, hire and educate analysts and build your expertise because demand will increase for sure. 

Too much too soon?

Organizations that try to take on too much too soon cause risk, says Peter Aiken, PhD, associate professor of information systems at Virginia Commonwealth University and founder of Data Blueprint. “We have really good, well-intentioned people trying to make decisions about something they are not familiar with.” Part of the problem is that the data analytics taught to people in IT and the healthcare business “can fill a thimble.”

To meet data analytics needs, healthcare organizations are creating data scientist roles, Aiken says. He’s enthusiastic about the job but says it’s comparable to telling somebody to open a Starbucks without the proper tools. “The situation you’re handing folks is a criminal waste of their time. People are growing coffee beans and trying to figure out how to turn a radiator into a latte machine. It isn’t working.”

Data as an asset should be treated in the same manner and with the same professionalism as other organizational assets, Aiken says. The finance department doesn’t do bookkeeping and the chief risk officer doesn’t test software. Asking IT people to work with data only leads to failure.

“In my experience, everyone throws up their hands after three years when an analytics initiative fails.” Had they used those three years to concentrate on the basics, they would have created enough savings to pay for more sophisticated analytics. “They needed to learn to crawl and then walk so that running is much less dangerous from a risk perspective.”

Aiken cites an example of a hospital that found their palliative care unit was not producing significant savings. It was seen as a cost and an outside consultant recommended cancelling the program. “By taking a step back and looking holistically, they were able to find literally ten times the amount of savings they had anticipated. From an analytics perspective, there is no way either IT or business would have determined that on its own.” Now, palliative care is one of the facility’s leading programs.

Building a program

UnityPoint Health, based in West Des Moines and delivering care in Iowa, Illinois and Wisconsin, has annual revenues of $2.7 billion and provides more than 4 million patients visits a year. Data analytics has been and continues to be a top strategic initiative, says Betsy McVay, executive director of analytics. “We’ve been working very hard to better  integrate different data sources.”

That integration of business intelligence, payer and clinical data will help the organization position itself for the future of value-based care and risk stratification.

“UnityPoint is refining its analytics program and building more robust capabilities,” McVay says. For example, several performance reporting tools are in use across service lines and value-based contracts. 

The organization also is implementing predictive analytics to support decision-making and strategic planning through relevant and actionable data, she says. Last spring, they reorganized to consolidate the corporate analysts into one team under the chief strategy officer “to ensure that our work is propelling forward and supporting our top initiatives.”

Integrating data and working with standard definitions are other hurdles for effective data analytics. The complex organization believes in local regional automony, “which doesn’t always make standardization easy,” McVay says. 

UnityPoint just finished implementing an EHR in the majority of its inpatient settings and has begun installation of the system in ambulatory care. This work will help with standardizing the IT platform and determining which data analytics tools are useful, which aren’t needed and optimize them going forward.

“Our ability to make data driven decisions, measure performance and implement changes quickly is becoming more and more important. We need to be nimble, innovative and as educated about our environment as possible,” McVay says.

Targeted initiatives

Allina Health, a Pioneer ACO that covers Minnesota and western Wisconsin, has developed a more mature analytics program over the past several years. The organization uses predictive modeling for several initiatives that focus on finding less obvious patients at risk for a preventable event before they occur, in response to measured evidence for many hospital admissions. 

These analytics initiatives are the result of Allina’s leadership realizing “we need to be in a position where we have data and informatics to help us share our care models in the new reality of healthcare,” says Ross Gustafson, vice president for performance resources.

“Instead of high utilizers, we target preventable events, so that we can intervene before the high utilization spike begins instead of in the middle of the spike, when patients have already incurred a significant medical event,” he explains. Allina also expands the scope of data from diagnostics and services to include clinical values such as lab values and blood pressure, and enhanced demographics.

Another modeling project focuses on minimizing readmissions. To develop a readmissions predictive model, Allina uses data on about 200,000 inpatients stored in the health system’s enterprise data warehouse (EDW). Hundreds of variables that might influence a patient’s readmission risk are analyzed on a continuous basis. To make this information useful, it is uploaded onto the health system’s patient census dashboard, providing Allina staff with a one-stop place to obtain a list of inpatients with a high risk of readmission. 

A combination of Allina’s EDW, business intelligence and the analytic piece of predictive modeling is attached to the census dashboard. “That is how our social workers and care managers are figuring out which patients to round on. Do they have a transition conference in place? Do we have recommendations for an outpatient provider?” Ensuring these steps has resulted in a decrease in the likelihood of readmission by double digits, Gustafson says. 

“Using the data at times will drive opportunities on the back end to see whether interventions are playing out so we know if we should tweak them, add more or leave them alone.”

The organization began building its EDW in 2007. Gustafson’s team includes data architects, enterprise reporters, clinical data analysts and a performance improvement group. Over several years, Allina implemented a full EHR at its hospitals, clinics and ancillary facilities. They also built specialized data marts to organize the information. “We build on our analytics horsepower to further the strategic initiative. We are getting more and more mature with what we do with our data every day.” Reports record past data, dashboards detail what’s happening now and the predictive component helps with efforts to learn what the patient population is going to look like, what services Allina should be offering and how they can efficiently hit the triple aim goals. 

The EDW has yielded a very positive ROI, Gustafson says, financially but also through extensive buy in and engagement of nurses, pharmacy employees and all physicians. “I can’t even begin to say how powerful that is and how quickly we’ve been able to move the ball on some very tricky clinical opportunities. Because we had meaningful data the physicians bought into, it was easy to tie to the triple aim. Where there is a win for clinicians, there is a win for the organization and, more importantly, wins for individual patients.”

Despite their successes, Allina has room for improvement, Gustafson says. “One thing we lack is a stronger, overarching governance process” for everything from the intake process to follow-up care decisions to data access. They began the data governance process slowly, he says, drafting policies and procedures in “key areas where we had physician traction.” 

As they continue to grow, he says, the administration overhead to oversee everything has lagged a bit. However, he points out that other organizations that spent a lot of time on governance are further behind on the analytic piece. “By and large, the way we set up is how we’d do it again. As we've started to grow, we’re making sure we have the structure to support pent-up demand for more analytics.” 

Looking ahead, Gustafson says Allina is interested in finding ways to better link up EHR data and developing better partnerships with payers and the government “so we can look at total episodes of care.” Even with the Affordable Care Act, patients will continue to use other healthcare services that Allina will want to digest, he says, “to see how that information could help shape our care strategies and approaches to population health management.” 

In particular, Gustafson says he hopes to use existing system information to help identify the lowest cost of care setting for patients. “Over the next one or two years, we will have intense focus on that space.”

Worthwhile advice

For those organizations not yet focusing on data analytics, Gustafson recommends starting on a small scale. “Begin to engage people within your own organization or an external partner to grow that capability and capacity.” Without some of this information, it will be incredibly difficult to effectively manage and operate. “You need to have a good sense of what it is you do, what it is you do well, what kind of patients you take care of today and tomorrow and how you can catch the subset of patients who are just on the cusp of becoming chronic patients. Without access to that data, you’re just taking guesses.” 

“In IT, we have death by 1,000 cuts—the little, insignificant things here and there that add up,” says Aiken. “Virtually every industry can benefit from better data management practices as a precursor to data analytics.” Data provenance is important, he says, because if you can’t even agree on the numbers, how can you make intelligent decisions? “It’s really a matter of concentrating on the basics.” 

Aiken also agrees with others who have predicted a flood of healthcare data. However, “as organizations break through, we’re going to see not just a tsunami of data but in the ability to use data.” 

[[{"fid":"18553","view_mode":"media_original","type":"media","attributes":{"height":259,"width":640,"style":"width: 460px; height: 186px; float: left; border-width: 0px; border-style: solid; margin: 10px 0px;","alt":" - CIT_Jan_2014_CoverChart","class":"media-element file-media-original"}}]]