Skip to main content

Healthcare's Big Data Problem

 |  By Philip Betbeze  
   August 01, 2012

This article appears in the July 2012 issue of HealthLeaders magazine.

Decision-making in healthcare can be painfully slow, as any physician will tell you, because of complexity. Patient discharge, for example, can involve a coordination of gears that could make a clockmaker sweat, because of all the information that must be processed to coordinate care outside the walls of the hospital. But thanks to a flurry of innovation in real-time processing of data, many healthcare organizations, including physician group practices, are getting better—and quicker—at dealing with data.


LIVE Webcast

Going Mobile: Defining Strategy and Establishing Metrics
Mobile solutions provide new ways to serve patients, improve efficiency, and deliver greater value. Discover how Swedish Medical Center, St. Elizabeth Healthcare, and Mercy Health have built mobile strategies that advance their business goals and how you can too.

Register Today



They're being spurred on in part because healthcare is beginning to deal with a shift in reimbursement toward one that rewards quality and disincents inefficiency and waste. Refreshingly, most hospitals and health systems have lots of data that can help improve outcomes and cut waste.

The problem is getting that data, which is often unstructured, into a format that allows clinicians to make decisions faster and in a more coordinated fashion. Leaders have long had difficulty with breaking down data silos and finding ways to use fragmented information. They strive to find ways to use leading, instead of trailing, indicators, so that interventions can be made not only in a more timely manner but with more predictive power behind them.

Catching up
Healthcare delivery organizations are 10–15 years behind other industries in managing and capitalizing on the data they own, says Greg Tipsword, healthcare provider practice leader for West Monroe Partners, a Chicago-based healthcare consulting firm. The need to catch up is urgent. Being able to do predictive modeling is critical to risk-based contracting because to reduce waste, labor-intensive interventions need to be used on a distinct group of patients who are most likely to exhibit a complex web of behaviors that, left unchecked, will result in the need for expensive care. The trouble lies in predicting who those patients are before they encounter those complications so that interventions can be made.

Numerous stakeholders need to be involved in defining the type of information required. This depends on the mission of the organization and many other variables, and should begin with the deliberative process, he says.

"The big challenge is getting data out of these systems, integrating it, creating information, and getting that into the hands of people who can do something about it," says Tipsword.

There is an abundance of data in healthcare, but in many cases, it's trapped in silos. That means it's important to develop a common enterprise goal—in other words, a governance platform that allows executives and clinicians to understand the organization's priorities as a whole. Tipsword calls it developing an information management master plan, and it's common, he says, to emerge with literally 100–200 needs for integrated healthcare data.

"How can we cut out nonvalue-add tasks and make sure they stay out? No two organizations are the same," he says, "but the organizations that are most successful at it have the highest levels of executive sponsorship."

Leadership at the top
Chicago-based Rush is a nonprofit healthcare, educational, and research enterprise consisting of Rush University Medical Center, Rush University, Rush Oak Park Hospital, and Rush Health, which is a network of providers including more than 800 physicians.

When the organization also owned a health plan, Rush Health's physician groups were bound together in a physician-hospital organization, and had to deal with capitation and risk on a daily basis. That required a huge amount of data processing power, but for the past 10 years, since it sold the plan, the organization has been operating largely on a basic fee-for-service model. As significant pay-for-performance incentives have been added to all of its payer contracts, says Rush Health President Brent Estes, its perspective on data management has changed significantly.

"We decided we needed to change the rules of our organization such that we required all hospital and physician members to establish data interfaces," he says. For its part, Rush Health built a data warehouse to process information coming from the practices, "so we could look at it on a holistic basis."

By integrating that data into one platform that is usable by all practices, the warehouse allows Rush to look at patient population health across every payer class, and allows it to implement P4P programs that are payer-neutral. It makes extensive use of prompts to help clinicians keep track of patients' needs based on practice-initiated treatment protocols.

"Without going down that path, we couldn't do a patient-centered medical home or get Level 3 NCQA designation for seven of our practices," Estes says. The Web-based patient registry that came from the collaboration among practices identifies populations within those practices and, critically—because of the expense of extra labor and monitoring required—which patients may require a higher level of intervention.

"We thought about a lot of these clinical data points in our master planning process before we spent a dime of money," he says of a process that dates to 2006. "That was an invaluable decision. We took a lot of time to engage people from different segments of the enterprise and [learn] what they would do with the tools."

Many such projects have a tendency to die on the vine in the face of inertia on user engagement. With its medical home and patient registry tools, Estes says, "we mitigated that to some degree because we engaged the clinical people in the design process of the registry tool itself before we did any coding work. The only way it's going to be a useful tool for them comes from asking them what they want it to do."

One of the issues that can affect adoption is the question of data quality, which can be a big problem for an organization like Rush because it includes a wide variety of clinics, some owned by Rush and some not.

"That has been a significant challenge to our physician practices, many of which are private and have already made decisions independently on EHR or practice management systems," Estes says. "They all have different functionalities and different usage of the functionalities. Just getting the data out and figuring out how usable it is has been a significant barrier."

Estes and his team are addressing that problem in a couple of different ways. The Rush hospitals in the PHO and all physicians employed by the hospitals are using Epic for clinical documentation, scheduling, and patient billing. That's not necessarily true of some of the other practices, but Rush is attempting to solve that problem through PHO structure, which allows the rest to get on Epic for essentially no cost.

"We will absorb the cost, but there's been a slow uptake for a lot of reasons," Estes says.

Some are suspicious, some are too busy, and Estes thinks that third-party billing companies can also be agitators against the change.

"We're not mandating you use the same business office and you can continue to do billing on your own," he says. "We'll even train the third-party billing company to do it. But the big thing is standardizing and shared clinical data. We're trying to show them the cool things we can do, but in many cases, the data they have is not in the state it needs to be."

There are opportunities for immediate financial benefits as well. Rush has built a surveillance application that allows executives to look at all claims across the network and see if they were paid as expected. "It allows us to calculate an expected payment," Estes explains. "This application allows us to aggregate all the issues and work with payers on behalf of all our members at the same time. For instance, here are the 200 claims we found that were not paid in the right way. That is a big benefit to any physician practice because it doesn't force them to chase underpays one by one."

Such underpays are small individually but big in the aggregate, he stresses. In addition, executives can look into productivity and outcomes both organizationwide and on the individual level. More important over time will be its ability to allow executives to monitor growth in terms of new patients over time, and per capita cost.

"Those will be two things important for me to focus on," Estes says. "If we want to take population management and get ready for

ACOs, whatever those end up being, or direct contracting or direct-to-consumer work, we need to focus on clinical data and change how care is delivered to reduce costs and improve outcomes."

Graduating from process to prevention
Ryan Leslie says success with population health strategies will hinge on enabling effective decision-making by clinicians, both in real time and for planning purposes. Leslie is vice president of analytics and health economics at Seton Healthcare Family, a 10-hospital system based in Austin, Texas, which has been working recently with IBM on the problem. Using the same software components that run the famous Watson computer of Jeopardy! fame, Seton is helping its clinicians identify patients who would benefit most from extra attention following discharge. The program started with congestive heart failure patients, and Leslie hopes to expand to other disease states.

"A lot of it is about enabling decision-making," he says. "It's taking the whole universe of information we have and cutting out what's extraneous and giving clinicians the information they need to make decisions."

Taking unstructured clinical information and connecting that with billing or administrative information and social demographic information, "you start connecting all those things together and you get a more complete picture of the patient as a person, rather than as a recipient of a bill," he says. "That's been the exciting thing recently. You realize that a patients' success or failure may not have to do with the care plan details or the clinical attributes of the patient as much as the social attributes."

The program is the backbone of the Seton Total Health Partners program, which, as Leslie explains, is an "extensivist" program, under which a physician outside the hospital works with a team of social workers, nurses, and others to visit patient homes and figure out what's keeping a patient from effectively following treatment protocols that will likely keep them out of the hospital. The problem is, as Leslie says, "you'll never have the resources to do that with every patient."

The software helps determine based on a host of combined data which patients are most likely to be rehospitalized within 30 days. Targeting the patients is like looking into a crystal ball. "If you target the wrong patient, you get all the cost and accrue none of the benefits. That's where we're taking this sea of information and filtering it to make it relevant, predictive, and actionable."

It's a task that's easier described than done. Much of the magic comes from natural language processing technology that integrates clinician notes from the patient's EMR—which are too often ignored by other clinicians because they are difficult to review in a timely manner—to be used to determine social or other difficulties that might result in a readmission. Combined with statistical analysis and data mining, such tools can provide a powerful picture of the patient's needs outside the hospital.

Seton chose to focus on CHF patients first because it already had clinical programs to address such patients. The disease is prevalent among the large number of uninsured patients it treats in Central Texas.

"Untreated, it steadily gets worse, but it's very treatable," says Leslie. "But like a ratchet, when it gets worse, you can't get people back to where they were before. We're trying to prevent people from getting sicker."

Next, Seton might look at using the technology to take better care of patients with diabetes or other chronic maladies.

"Now we have a number of these modular programs set up, and we're refining the process of using the predictive information," he says. "This particular work with IBM was a proof of concept of the value of our unstructured data and the value of the technology."

As one of a handful of organizations that are part of CMS' Pioneer ACO program, the work holds big promise for cutting treatment costs, he says.

"Pioneer is really putting a lot of this stuff on the line for us," he says. "We have to do this to bend the cost curve."


This article appears in the July 2012 issue of HealthLeaders magazine.

Reprint HLR0712-4

Philip Betbeze is the senior leadership editor at HealthLeaders.

Tagged Under:


Get the latest on healthcare leadership in your inbox.