The Thorniest Barriers to Robust Data Analytics? Panelists Uncover a Tangle of Them

Aug. 19, 2015
Panelists at iHT2-Seattle tackle a complex set of thorny issues that must be addressed in order to overcome barriers to robust data analytics for clinical transformation and quality and efficiency improvement in U.S. healthcare

How can data really be made useful to efforts to improve patient care outcomes and engage in population health initiatives? Panelists participating in a discussion around data analytics plunged into some very thorny issues in healthcare, during an afternoon panel discussion on Aug. 18 at the Health IT Summit in Seattle on Aug. 18, being held at the Seattle Marriott Waterfront, and sponsored by the Institute for Health Technology Transformation (iHT2, a sister organization to Healthcare Informatics, under the corporate umbrella of parent organization Vendome Group, LLC).

The panel, entitled “Analytics: Integration, Standards, and Workflow,” ended up tackling some of the most vexing issues facing healthcare leaders who are attempting to fully leverage data analytics for clinical performance improvement, cost reduction, population health management, and other purposes.

Zachery Jiwa, former innovation fellow at the U.S. Department of Health and Human Services, led the discussion. The other panelists were David Chou, M.D., chief technology officer at UW Medicine (Seattle); Dean Field, M.D., vice president for informatics and operations at the Tacoma-based CHI Franciscan Health; Steve Weiss, R.N., CNIO for the Seattle-based Swedish Region of Providence Health & Services; Sean Kelly, M.D., vice president and chief medical officer at the Lexington, Mass.-based Imprivata and a practicing emergency physician at Beth Israel Deaconess Hospital in Boston; and Erik Giesa, senior vice president for informatics and operations at the Seattle-based ExtraHop Networks.

panelists (l. to 4r.) Geisa, Chou, Field, Kelly, Weiss, and Jiwa

Among the problems inherent in the current struggle to leverage data for analytics purposes, Weiss noted, is the fact that such efforts have been relatively recent overall, and have followed a number of years focused on electronic health record implementation and on the creation of some informatics foundations, including the creation of data warehouses. “Early on,” Weiss noted, “we were really working on the EHR, and we weren’t necessarily capturing data discretely; instead, we were focusing on getting people on board. And as we progressed, we focused on moving onto enterprise data warehouses and registries, and beginning to work on data definitions. That’s about where we are now,” he said. “It would be great to move into data definitions in communities,” he added. “We want to continue to work on population health. The problem is that definitions in medicine are difficult.”

We’ve been live on our current EHR for two years now,” Field reported. “And while we’re still in our infancy on our implementations, now suddenly, we’re realizing we need to be able to pull data out of it. And now that we’re in this adolescent phase, we’re still very much reactive, reacting to CMS [policy mandates from the federal Centers for Medicare & Medicaid Services], reacting to other external pressures, and not necessarily following our own vision.”

“Two things are necessary” to begin to leverage data analytics robustly, Chou asserted : “a useful vocabulary, and understanding data context. I don’t think either of them are at a satisfactory level yet,” he said. “And the consistent practice of medicine isn’t there yet, either.” In fact, Chou said, one fact that should sober any leaders attempting to move forward to robustly leverage data analytics, is this one: “There are something like 690 definitions of glucose” in EHRs and other clinical information systems, he noted. “And that’s a disaster. And that’s assuming that they all mean the same thing, which they don’t. So you have to decide what you’re going to map to and map to. And every time I go through an interface, I lose information. And with regard to, for example, blood pressure, I don’t even know what the information is around the blood pressure, I don’t have the context. So,” he said, looking at an analytics landscape that encompasses clinical, technological, policy, and practice challenges, “you have to understand the practice of medicine, and you have to have the context. And eventually, without that, you’re going to drive the clinician crazy.”

EHRs never designed for analytics work

A very simple reality is also very important to keep in mind, Giesa said. “When you look back at the design, from as much as 30 years ago, of the EHR, and you look at how we’re now trying to apply it to analytics, it’s like trying to turn a square into a wheel now. And when I hear terms like data mapping, I want to note that the practice of medicine isn’t standardized or structured,” he emphasized. “When they built these applications, they did not anticipate using them for analytics or informatics. So there’s a new paradigm emerging now around structuring data in unstructured data stores, giving you the flexibility to not necessarily have to do data mapping.”

It might seem like a stretch to apply such informatics concepts to patient care, Giesa said but he noted that “That’s something that applications like LinkedIn and Facebook do: the same principles apply. You have one user who might be doing five, 20 different things, interacting with all sorts of different applications, but at the end of the day, that user wants to see what they want to see. All of that relies on structured data being put into unstructured contexts for end user use. I don’t believe that the structures around EHRs were designed to do what we’re trying to accomplish.”

Kelly agreed. “I think you’re absolutely right,” he told Giesa. “The reality is, you need to try, and fail, and try, and fail,” as leaders in patient care organizations beginning to move forward to harness analytics. “And it’s an iterative process.” In fact, he said, “you probably got traction to begin with because someone caring for patients or doing the billing and coding, cared about what you were doing. Some stakeholder in the hospital cared about that data. And all you can do is try a first cut of it and reiterate that over and over again.”

Indeed, Kelly said, “The places that are beginning to succeed fail and fail over and over again, but have multidisciplinary teams working on this. So I ask them, what are you doing to get the right stakeholders together at the same table, and asking the right questions? And some of the most interesting stuff we’re finding” in terms of revelations coming out of analytics, “is actually unexpected, right? It’s not necessarily what we were looking for.”

Achieving early gains—and seeing the light at the end of the tunnel

When Jiwa asked his fellow panelists how far they’d gotten so far in beginning to share data with their communities, and about the interoperability standards that needed to be addressed, Field said, “I don’t know that I would describe our journey as completely successful yet. We’re on a journey,” he stressed. “For us, part of the challenge was creating a unified platform between inpatient and outpatient. We decided to choose a single vendor to create a platform for that. It also requires creating an organizational vision, and lens, for where you want to go. For any organization, you can look at IT infrastructure as an expense. And when you manage that expense, you want to manage the cost of it. But it’s much more important to focus the lens on the community,” he emphasized, “and to have that to support the community, rather than focusing solely on the expense.”

In fact, Chou said, “I can tell you from the personal side, that for me as a clinician and as an informaticist, I can say that there’s a huge gap in terms of discharge of patients into the community. We have no good mechanism” for fully documenting and sharing clinical data around discharges of patients into the community, he said. “Nursing homes don’t have EHRs of any kind, some of them. And they seem to have very little incentive” to implement EHRs, “given low reimbursement, and so on. And the whole discharge at UW—we don’t have a smooth transition to nursing homes.”

Kelly shared his view that “There’s good news and bad news. The bad news is that it’s a mess—the whole area of [documenting] transitions of care [and sharing data around them]. Despite that,” he said, “we’re doing a good job. We’ve created home-grown systems built on our EHR, to manage the transitions. There are electronic things going back and forth, CCDs”—continuity of care documents. Still, he noted that the infamous problem of “note bloat” in physician documentation is only getting worse.  “On the CMIO listserv last month,” he reported, “someone sent out an example of a 337-page CCD, something like that.”

Continuing on now to speak as a practicing emergency physician, Kelly said, “So, let’s say that I’m in the ER and a patient comes in unconscious, and somewhere in that 337 pages is something important that I don’t have time to look at or find, but you can be sure that a malpractice lawyer will find it sometime. So are we helping ourselves and each other, or harming?” he asked, in reference to the over-abundance of data and text points in clinical documentation. “But the good news is that there are a lot of people who care about the patients, and vendors are building things, and there are a lot of things happening out there in smaller places that are very exciting.”

All of this speaks at a fundamental level to how EHRs were originally conceived, of course. “To be honest,” Field said, “EHRs were built for billing and coding, not necessarily for patient care.”

Things are in early stages of maturity around analytics work, as a result of all the factors cited by the other panelists, Weiss said. “When we do risk stratification and look at populations and why they’re coming back, we can look at issues around continuity of care. But we haven’t yet figured out how to share important information across the organization.”

“The data’s actually there, right?” Kelly said. “But what do we do with it? The critical questions are, who needs it, and how do you get it to them? And it takes people with clinical, operational, and technical knowledge, to sort through all of it.”

“I agree with you,” Chou said. “Two years ago, we wouldn’t even have been talking about transitions of care. But we’re at the point where we are transmitting, which really is a big, big improvement over where we were.” In other words, panelists agreed, things remain in early stages around successfully and robustly leveraging analytics for clinical transformation, population health management, and other important purposes in U.S. healthcare. And yet they are also further along than they have been—and moving forward in a landscape of accelerating effort and activity.

Sponsored Recommendations

Healthcare Rankings Report

Adapting in Healthcare: Key Insights and Strategies from Leading Systems As healthcare marketers navigate changes in a volatile industry, they know one thing is certain: we've...

Healthcare Reputation Industry Trends

Navigating the Tipping Point: Strategies for Reputation Management in a Volatile Healthcare Environment As healthcare marketers navigate changes in a volatile industry, they can...

Clinical Evaluation: An AI Assistant for Primary Care

The AAFP's clinical evaluation offers a detailed analysis of how an innovative AI solution can help relieve physicians' administrative burden and aid them in improving health ...

From Chaos to Clarity: How AI Is Making Sense of Clinical Documentation

From Chaos to Clarity dives deep into how AI Is making sense of disorganized patient data and turning it into evidence-based diagnosis suggestions that physicians can trust, leading...