Experts offer big picture on big data

Approaches to big data, data analytics and business intelligence dominated the discussions during Thursday's sessions of the 22nd Annual Physician-Computer Connection Symposium in Ojai, Calif.

Leaders from three healthcare provider organizations and a health information technology vendor, all doing pioneering work in healthcare data analytics, gave updates in a morning session on their experiences and offered advice on how to approach what, for even some of the more experienced of the 200 or so physician informaticists in attendance, looms as a daunting challenge.

Panel moderator Dr. Christopher Longhurst, chief medical information officer at Lucile Packard Children's Hospital, said the question and the challenge posed by big data is, “How are we going to change the system so that we're learning from every patient every time?”

Dr. Dick Gibson, vice president, healthcare intelligence, at Providence Health & Services, Renton, Wash., said his organization is just finishing the installation of an electronic health-record system, but it will take many more tools to be able to do what the organization plans to do in the future. Gibson posted slides of an array of products and systems configurations, including an enterprise data warehouse, Providence is either using or plans to implement.

Gibson said the organization is using these data analytics systems to solve a pair of problems. “One is to improve the encounter, the other is to avoid the encounter all together,” he said, adding that Providence will launch an accountable care organization in January.

“We believe big data will revolutionize healthcare, but it's going to take five to 10 years to get there,” he said. “It's not either or, with regular data, and big data, but both. Meanwhile, we need to master regular data. We need a special effort in population health management analyzing existing and future claims data.” Natural language processing will be another challenge. “We have gigabytes of dictation that we could use to provide better care for our patients.”

The full value with big data will be achieved with the combination of clinical, financial and operational data in a single database across millions of patients, he said. But for now, he described Providence's data analytics program as “pretty basic, just getting started.”

Dr. Jonathan Palm, medical director of analytics at Lucile Packard, said they, too, have a data warehouse receiving information from multiple sources, including an EHR, cost and accounting information, and patient-satisfaction survey results, and uses business intelligence software to produce standardized and customized reports, scorecards and dashboards for clinicians up to C-suite users.

“We don't have a lot of financial data on dashboards, but that's coming in the next couple of months,” Palm said.

Palm said, they, too, are looking at ways to use natural language processing to sift though a trove of physician notes and other text-based information. “How do we learn from every patient, every time?" Palm said. “One of the ways is making sense of this unstructured data.”

How should a hospital or multi-hospital organization stage their approach to big data, asked symposium attendee Dr. Peter Basch, medical director for ambulatory EHR and health IT policy at MedStar Health.

“That's a question I hear frequently,” said Longhurst. “I would challenge every one of us in this room to find one piece of actionable data that's not being acted on today and make a difference. We know there is a gap between what we should be doing and what we're actually doing and this data can make a difference.”

“Go talk to your pathologist,” Longhurst said. Ask him or her for a report “that's being issued today and not being acted on.”

“Don't palpitate,” Longhurst said, responding to comment about the seemingly daunting challenges big data analytics pose, smaller healthcare organizations can do it. Dr. George Reynolds, vice president and CMIO at Omaha's Children's Hospital and Medical Center, started with a Web-based data analytics tool and “went out and started building dashboards.” Both Longhurst and Reynolds were 2010 winners of Modern Healthcare's Top 25 Clinical Informaticists award.

But Reynolds warned that successful data analytics projects must solve local needs. “It's only going to work if it's complementary to local resources to answering local questions. That's where the value comes from.”

Fellow panelist Mark Hoffman, vice president of the Cerner Research Organization, said the vendor has been gathering de-identified data from its clients for 13 years and has 500 providers supplying its research database with information. He talked about genomics, molecular diagnostics and other forms of complex testing

Hoffman said “a constructive first step” for a provider organization is to have a conversation with your clinical community on how to respond when a patient comes in with a genetic profile obtained from an outside lab or other provider. “Evaluate what level of complex testing is being done within your organization so you can start a strategic approach.”

In a separate presentation on business intelligence, David Garets, an executive director of the Advisory Board and head of its healthcare IT research and advisory services, said BI systems will be called on to perform descriptive analytics, telling providers what happened, locating, for example, past capacity bottlenecks. Many “are getting into dashboards,” Garets said, “but that's as far as healthcare organizations are going now.”

They will also be required to do predictive analytics, telling organizations what might happen, for example, identifying in advance a panel of high-risk patients ripe for intervention. At the highest level, BI systems will perform prescriptive analytics, Garets said, advising healthcare leaders on how to optimize their capacity. “We are nowhere near where we want to be in prescriptive analytics,” he said.

In the current market, there is no one single business intelligence software product that is likely to meet all of a healthcare organization's needs, Garets said, which is forcing providers to find “best of breed” solutions for different analytics problems.

Garets advised that organizations that are just starting to plan for their business intelligence futures should include on their planning committees a leader empowered to make final decisions.

“I would argue that IT doesn't drive this bus,” said Garets, a former CIO. “IT would own the infrastructure, but it doesn't own the governance of this. The name of the game is you need an executive steering committee.”

When an organization is trying to define a common set of terms—key to successful data analytics—such as the start of the patient day, whether it's midnight or 6 a.m., the committee needs someone with the authority to say, “The day starts at 6 a.m.,” Garets said. “Next.”

“You need people at the table that can make those kinds of decisions, or you're going to be at this for a long, long time.”

The chief decision maker probably won't be the CEO, Garets said, but “you better have the CMO there. You'll probably have your chief strategy officer involved in it” too, he said.

Garets also advised providers to aim low to start.

Get value from a data analytics project early on “so you can provide air cover” for all the hard work that lies ahead, Garets said.

Follow Joseph Conn on Twitter: @MHJConn