Building the Data Analytics Team

Leaders are finding that recruitment from the outside can be useful but difficult, and that internally, talent can be developed to support analytics efforts.

This article appears in the October 2015 issue of HealthLeaders magazine.

Healthcare's many imperatives, including population health, require keen analytical resources coupled with a deep understanding of clinical systems and the obstacles that face healthcare's march to achieve the triple aim.

Finding and training talented staff—from line analysts all the way up to visionary chief medical information officers and chief information officers—remains a tall order. By and large, however, analytics team leaders find it is easier to train clinicians in the art and science of analytics than it is to find, hire, and train data analysts from outside of healthcare.

"It's not a skill that you can learn going through some training program outside of this industry," says Selvan Ekambaram, director of business intelligence and data warehouse at MemorialCare Health System, a six-hospital integrated delivery system in California's Orange and Los Angeles counties, with between 650 and 700 affiliated physicians. "You pretty much need to have healthcare-specific industry experience for you to call yourself a practitioner."

Often, providers such as MemorialCare must contend with job candidates who prefer to remain independent consultants, instead of coming onboard as full-time employees, "because they know the demand for this particular skill set is so high," Ekambaram says. "Recruiting and retaining these specific skill sets that are needed for running a program like a data analytics shop like we're trying to do here is not easy."

MemorialCare has had a couple of full-time positions open for nearly a year. The kind of skills required generally fall into two categories: data integration and content visualization. Data integration involves procuring and consolidating different data sources such as EHR data, claims data, and patient satisfaction data. Then that data must be visualized and deployed, Ekambaram says. All told, the MemorialCare analytics team engages 21 workers. A second branch, which serves the organization's independent physicians associations and medical groups, contains another 25 workers involved in pay-for-performance reporting and quality metrics for ambulatory settings. Six other analysts joined MemorialCare when it acquired Seaside Health Plan.

One successful strategy: Tap into talent at healthcare payers. "We recently rolled in one full-time equivalent who came in from the payer industry, although we would prefer recruits from the provider space," Ekambaram says.

The team uses an iterative process by which it presents data to clinicians and determines if that data is as useful as possible, if there might be ways to better present the data, or if it should enable clinicians to further explore the data presented—an aspect known broadly as self-service business intelligence analytics.

"Some of my colleagues are more toward the self-service side than I am," says Harris Stutman MD, chief medical information officer at MemorialCare. "But we have, within the last couple of months, put together the idea of a physician analytics program, a kind of a mini-fellowship, to take our physician leadership, certainly the folks that are responsible for leading our medical group and leading the best practice initiative within our hospitals—clinical documentation improvement, population health, those sorts of things—and train them in what data sets we have available and what tools we have available, and help them understand that we really need to ask answerable questions," he says.

"We invited 13 physicians to be in the initial mini-fellowship class, and all 13 accepted—which really surprised me, because we're not paying them to spend 2 or 3 hours a week in a classroom, learning this material. But they all felt it was sufficiently relevant and important to what they do to make time to attend those programs, and we'll see what positive outcomes develop," says Stutman.

"Vendor partners train us in the tools and the technology," Ekambaram says.

Among the quality improvement programs being driven by the analytics team are quality measures and core measures for sepsis management for MemorialCare's inpatient and pediatric populations. MemorialCare's vice president of population health is also working with the analytics team to gather palliative care metrics.

"It's very important for people to understand what sorts of business problems they're trying to solve, and to make sure that when they start really mining the data, they're doing it with specific goals in mind and not just fooling around to produce some pretty pie charts or bar graphs," Stutman says.

Another provider that has tried dual approaches to recruiting analysts from inside or outside of healthcare is Pittsburgh-based UPMC, which operates more than 20 hospitals and employs more than 3,500 physicians. "We merge the skill sets of both" externally recruited analysts and internal analysts, says Vivek Reddy, MD, chief medical information officer for UPMC's health services division. A major push has been to provide analytics training for UPMC clinicians who "we thought were up-and-comers in our organization that had an aptitude to learn a new skill set," he says. Toward that end, UPMC built a training program in conjunction with nearby Carnegie Mellon University. So far, UPMC has put 75 workers through the program, a mix of clinicians, nurses, and financial employees.

"We put them through a fairly intensive, six-hour-a-week course, with homework, set around analytics and analytics principles," he says. "We've had very strong and powerful results with a lot of our clinical folks that didn't really have a lot of exposure to the idea of using data to drive decision-making really effectively. They were good at receiving reports, but they didn't really understand how you get insight from reports and how you can actually look at data differently, so our clinical folks that go into the program do very well and learn a lot of different skills."

UPMC's next challenge has been to get these newly trained clinicians to spread this data-centric view across the rest of the organization, Reddy says.

"Our firm belief is, this is just a fundamental change in the way people think about their day jobs," he says. Even registration clerks need to understand how what they are doing on a day-to-day basis ties into UPMC's analytics program, he says.

"What we're doing by taking some of our up-and-comers and making them go through this is allowing them to then take that message and help our day-to-day operations start to embrace the idea that if we're going to become a data-driven organization, we really have to look at data integrity and data quality and really be concerned about these things and not just say, 'Oh well, sometimes it's not right' and I just move on to the next problem, or I call someone else to get a new report if I don't like the data that I see," Reddy says.

In the latest cohort of analytics trainees, UPMC is giving its program participants a series of business problems to tackle. "It's a much more practical, hands-on approach," he says. "We'll see how that goes."

At many organizations, analytics resources exist in scattered pockets. This is true for the University of Missouri Health Care, which employs 6,000 physicians, nurses, and healthcare professionals and operates 50 outpatient clinics in central Missouri that receive outpatient visits exceeding 500,000 annually.

Now, UM Health Care is in the process of consolidating such resources under its newly hired chief financial officer. "We're partnering on the academic side with our HIM department and with our master of health administration program to formulate a certificate in healthcare analytics, as well as a degree program," says Bryan Bliven, UM Health Care's chief information officer. To lead this team, UM Health Care wants to hire a registered nurse to act as its director of clinical analytics, he says.

A key objective is for this analytics team at UM Health to "step up our game" to produce more engaged conversations between UM Health leaders during bimonthly enterprisewide operating reviews, when metrics for each department and service line are reviewed and interventions planned as needed, Bliven says.

"We have a population health view, so we'll look at the different contracts we're working on, and look at our performance to that, and it's a similar approach that really helps us kind of see across the organization and catch trends and then get an idea of each organization's ability to engage on the data and where we might need to have some oversight or assistance," he says.

"All of population health management is essentially cohort analytics," says Joe Kimura, MD, MPH, deputy chief medical officer at Atrius Health, an alliance of nonprofit community-based physician groups in eastern Massachusetts and Boston that serves 675,000 patients via 750 physicians, with more than 50% annual revenue coming from risk-based contracts. Atrius is a Pioneer ACO, a pilot CMS program recently found to be saving Medicare $400 million over its first two years.

"If it's a payer-designed cohort like Pioneer, you have a set of patients that is defined by an attribution model or something along those lines, but it could be a disease-based model, just anyone with diabetes, and the question is, how are you defining diabetes in your population?" Kimura says. "Once we actually define that population, we run a whole series of analyses to help
characterize that population, not only in terms of its utilization patterns, disease characteristics, and comorbidities, but also in terms of what we generally understand as care metrics or quality metrics."

Atrius employs four levels of SQL programmers to perform data analytics queries, Kimura says. "Those folks basically take about 16 to 18 months of training to get to a place where they can sit with a business owner and generate a specific ad hoc report," he says.

Such results could also be delivered as discrete information elements in Atrius' EHR, as well as reports in Crystal Reports or Microsoft Excel. More recently, Atrius has turned to SAP Universe and Dashboard technology to provide self-service business intelligence analytics to clinicians.

"We've pivoted there because SAP has single sign-on with Epic, which made it easy to utilize that capability," Kimura says. Although a number of Atrius employees already have statistical training, its academic partnerships in the Boston area allow it to engage with consultants to build BI models in-house and deploy them. If such consultants come from academia, "it's a lot easier for us to recruit them."

Over the course of a year, Atrius currently runs three or four such BI research initiatives, and so continues to engage consultants. "We don't think we're going to invest a lot around bringing in full-time staff to do some of the advanced stuff until we get bigger," Kimura says.

One payoff for Atrius: being named the number two Pioneer ACO in the United States in terms of quality reports in fall 2014, and the top such organization in New England. "We've been able to maintain the highest quality scores in New England and be number two among all Pioneer programs in 2014 and have begun to realize financial savings from CMS," Kimura says.

Reprint HLR1015-8