Even cavemen understood the value of using pictures to help tell a story. But telling the story of a woolly mammoth or saber-tooth tiger is one thing; explaining the clinical and financial insights gleaned from crunching millions of data points is another.
Or maybe not. For all the complexities of analytics programs, the goal of user-facing visualization efforts is pretty simple: Boil down the essentials into a visual display that enables executives, clinicians and other end users to quickly grasp what the data is saying, and what actions can be taken to ensure the "story" has a happy ending.
The problem in healthcare is that there are so many stories to tell with visualization that dashboard designers and executives responsible for creating dashboards can get themselves-and users-lost in the forest, hit with so many visualizations and accompanying data that they can't understand what the data is trying to tell them or what they need to do about it.
In this market, there can be no analytics without a cause, says Vincent Emanuele, Ph.D., a data scientist by training who heads the data science group at Wellcentive, an Atlanta-based provider of population health management technology.
Emanuele, who has spent years poring over complex visualizations accompanied by lengthy reports, admits to a certain admiration for "sexy" analytics and elegant visual displays, which are much appreciated by trained scientists and can be works of art on their own. But elegance and sexiness don't carry much weight in healthcare analytics, he says.
"Healthcare users, unlike users in some other fields, have no patience for lengthy reports and dazzling visualizations-they want to see the story right away, and want to see a path of action immediately," he says. "They want to see a minimum number of visualizations, but they want to see the visualizations of the data that matter."
But visualization is not simply about tricking out a dashboard or using the best graphics. To create an environment that provides the best visualization techniques and ensures everyone's seeing the same story is a large-scale undertaking in facilities where reports are done in spreadsheets, text documents and a multitude of other formats and applications.
The need for depth and granularity in data visualization has to be balanced against the same concern over any medium for reporting: Throw too many bar charts, pie charts, heat maps and line graphs at any user, and they all become equally meaningless to end users who are being pelted by information on a minute-by-minute basis.
"When designing dashboards and visual displays, often the biggest roadblock is that organizations don't know what they want to get out of the information-you can draw data from all these different sources, but what's the purpose," says Mike Erickson, a dashboard designer at Lancet Data Sciences, Burnsville, Minn. "What you're trying to do is visualize connections and information that otherwise wouldn't be easy for users to see. The key is to get an understanding of what you're trying to learn from all those numbers before you start visualizing the data."
Line of sight
David Delafield, chief financial officer at Swedish Medical Group in Seattle, says the hospital was in a situation typical for the industry-up to 3,000 different reports were flying around, in various formats using a variety of different data feeds, with no clear line of sight to the enterprise issues that HIT leaders, executives and clinicians needed to collaborate on to address enterprise issues.
Delafield also knew from experience at other facilities that committing to a centralized reporting platform wasn't a solution in itself-in fact, it often ended up being an expensive way of slapping lipstick on the proverbial pig.
"I've seen it happen before-an organization makes a move to enterprise reporting and it ends up being an expensive exercise that yields very little value," he says. "Having standard visualization tools and similar-linking reports doesn't help if you're still generating numerous reports that don't communicate what the analytics are telling you need to be addressed."
What's needed in tandem with a technology solution is a commitment to keep it simple, he says. Swedish did so by standardizing its reporting on a platform from Tableau Software and slimming down to 30 reports that use mostly simple bar and line charts to provide answers to the questions around readmissions, patient satisfaction, claim denials, full-time-employee usage and other top-line issues that Swedish wants to address.
The trick, Delafield says, is to utilize visualization tools to boil down very complex functions and data into very basic results: A hospital could put together five complex reports about revenue cycle management, for example, but at the end of the day it needs to get its claims cleaned up, out the door and accepted by health insurers. Showing claim denial rates-but proving additional layers of data behind the visualization for users who need or want to dig deeper-is all many users need to know to understand to start acting on that information, he notes.
Patient satisfaction is another example of how visualization can be used to boil an operation down to its essence. The hospital sends patient experience surveys to more than 15 percent of its patients each year, and the resulting feedback from tens of thousands of surveys used to be reported in a variety of ways, in a variety of formats, to different segments of internal staff and physicians.
Everyone was looking at different aspects of patient satisfaction; not surprisingly, Delafield says, a lot of staff didn't even look at the data because it didn't make any sense.
So Delafield's team rolled up those 35,000 survey results and created dashboard visualizations that enable users to look at monthly trends in patient satisfaction using a simple line graph that shows the target score and charts scores by individual physician, facility and the delivery system's overall performance, as well as benchmarks against internal and external peers.
Not only does that provide a concise, actionable view of an enormous amount of data, it also frees up time for executives and administrators responsible for raising those scores to focus on that task, instead of having to search for data and generate their own reports on smaller slices of data.
"The goal of visualization is to use those visual elements to tell a story that can help the organization move in a common direction," Delafield says. "We now have very sophisticated tools for data modeling and graphing, but that hasn't led to an expansion of reporting and another layer of reports-it's helped us keep narrowing reports and visualizations on the real critical issues we have in terms of patient safety and financial performance. It's made it simple to grasp the important points, and have everyone looking at that same view of our performance."
Of course, getting data standardized and cleaned up to the point where it can be displayed in a straightforward, standard way is more than half the battle when it comes to visualizing analytics.
Executives and physicians don't want to see how analysts and IT staff are making the sausage. But it is important to show them the building blocks behind those simple visualizations, says Luke Shulman, principal consultant with Arcadia Healthcare Solutions.
Just as when you're selling broader analytics efforts, "You have to get people to trust the data behind visualizations," Shulman says. "The mantra is, 'Own your own data.' To get executives and clinicians on board and to trust visualizations, it's often very helpful to show them the building blocks that went into those charts and graphs. All those building blocks add up to measurable performance indicators that users have to trust."
For that reason, the meaningful use EHR incentive program has been a boon to healthcare data visualization, he notes. The program has standard target goals for a number of quality/documentation measures, which yield visualizations that show performance measured against those targets and documentation gaps that need to be addressed. And therein lies a huge opportunity for providers to use visualizations to tell an action tale, Shulman says.
"Each visualization can be viewed as an opportunity for improvement," he says. "When you're developing visualizations, a successful strategy is to start right at the patient-provide a graphic display of what happens during an encounter, and the opportunity right at that point to improve quality and patient satisfaction by documentation at the point of care. That can be directly correlated with readmissions, patient satisfaction and a number of other measurements if you balance the use of raw information-specific dates about encounters and medication, EHR data, claims data and other information-against what a user should glean from all that data.
"There's always that balancing act between throwing as much data into visualizations as you can and visualizing what information is really necessary from a practical standpoint."
Meaningful use and other quality programs, which targeted goals and measurable benefits, are setting the stage for the new analytics frontiers, population health and accountable care, where the questions to "ask" data-and the answers-can be harder to come by, says Emanuele at Wellcentive.
"The challenge for population health is that you're trying to tell so many different stories with visualizations," he says. "And as you move through the three main components of population health-data aggregation and post-aggregation, analysis and then action-the audience and language shift."
For population health/accountable care, data visualization is a little closer to the sausage-making portion of analytics programs. Many initiatives focus on specific disease states, such as diabetes, but visualizations of the data aggregation itself is critical for clinical and financial strategies downstream, Emanuele says.
For example, for a visualization of a diabetic population, a visualization of the code being used to indicate diabetes might reveal that certain physicians or facilities are using the most generic ICD-9 code for diabetes. Using the generic code affects not only reimbursement but blurs the picture of what types of diabetes an enterprise has to manage.
Visualization at the aggregation/post-aggregation stage also offers insights into how prepared an organization is for new financial and care models, he adds. "Many organizations embarking on population health and accountable care don't really know where they stand in terms of the data they need and where their quality gaps are," Emanuele says. "While everyone is rich in data, they're not always rich in terms of analysis of how that data relates to the specific requirements of different quality programs, or which areas of their organizations need to change documentation practices or improve quality. All this is very hard to explain in a report, but fairly simple to express in a visualization."
How CIOs View the Analytics Landscape
Nearly all provider CIO respondents in a recent survey believe data analytics will play a big role in succeeding with accountable care and other value-based healthcare initiatives. But while 42 percent say they have a flexible and scalable analytics plan, more than three-quarters report only moderate or minimal commitment to integrating analytics into practice.
The April survey from eHealth Initiative and the College of Information Management Executives got responses from 98 provider organizations-35 percent delivery systems, 27 percent hospitals, 14 percent academic medical centers and 9 percent community health centers/clinics. Only four respondents were not running analytics at the time of the survey.
Seventy-two percent of responding providers extract data from more than 10 platforms or interfaces-some more than 100-with EHR and billing/financial data still by far the most common. But data also comes from patient-generated sources such as portals and health risk assessments (45 percent), unstructured text (39 percent), remote monitoring devices (29 percent), health information exchanges (22 percent), mobile applications (11 percent) and genomic data (7 percent).
Analytics remain in the early stages of maturity. Traditional common uses of analytics continue at high levels. These include quality improvement (93 percent), revenue cycle management (91 percent), resource utilization (81 percent) and population health management (79 percent). Respondents use descriptive analytics that mine for historical or retrospective analysis at a 94 percent rate. Only 68 percent use predictive analytics to forecast outcomes, trends or performance, and this is done mostly on a monthly or quarterly basis. One-third use prescriptive analytics with sophisticated models to optimize performance and recommend specific actions. Only 20 percent of respondents' analytics operations regularly integrate and coordinate at an institutional level.
Trained staff to collect/process/analyze data, along with interoperability and cost, have been common barriers to implementing an effective analytics program. Survey respondents report new challenges are emerging. These include access to external data beyond proprietary networks; cost-prohibitive work required to clean/validate/integrate external data when available; lack of funding or return on investment; increased regulations on data use; and patient privacy. "These trends suggest the critical need for strategic planning in implementing analytics, no matter how large or small," according to a report of survey results.
Consumer engagement has started to make its mark on data analytics initiatives. Two-thirds of respondents use analytics to support engagement with the primary focus on patient satisfaction. But few organizations also are applying analytics to consumer strategies such as personalized communication and services, acquisition and retention of consumers, or targeted behavioral change programs.
This story originally ran on the Health Data Management web site.