Analytically-minded and data-focused, doctors and their clinical colleagues are keen to know as much as they can about their patients. But there’s no formula for getting clinicians to use predictive or prescriptive analytics in patient care.
“Using health care data for decision making is a priority and certainly an ambition and a strategy,” says Denny Brennan, executive director with the non-profit Massachusetts Health Data Consortium. But hospitals and medical practices vary widely in the types of patients they treat, the insurance plans that pay them, and the way they’re organized, he adds, all of which create challenges when deploying advanced analytics.
Some challenges derive from the availability and limitations of technology and data. Among them: according to the Centers for Disease Control, as of 2011, only 55 percent of physicians had adopted electronic health records (EHR) to capture patient data and provide decision support. On the other hand, the tools for querying complex datasets for insight into treating individual patients are still maturing, says Marvin Harper, chief medical information officer with Children’s Hospital in Boston. “The car we’re driving in the EHR is more like the 1970s car,” he says. “We don’t have GPS or Bluetooth or traction control or airbags.”
The key challenge in adopting analytics in the health care industry is not all that different from the challenges line of business managers in other industries face: you need to win over and engage the front-line people who can make a real difference in producing new kinds of results. So regardless of technical issues, clinicians must be convinced that analytics can help them deliver better, more cost-effective care. And as with older technologies, like ERP and workflow management tools, there’s more to this than training clinicians how to use new systems. The only way to make them useful is to involve clinicians in developing them.
“There’s an ocean of data we’re swimming in,” says Harper. “It’s not really useful to dredge through it. People have to have a focused question in mind, and they have to be willing to act on that data if it doesn’t confirm their intuition.”
Find the Right Context
There’s a stereotype that doctors resist new technology. It’s true there’s a generation gap: younger doctors, as younger professionals in any field, enter the workforce with computing experience that doctors who have been practicing for decades don’t have.
As a practical matter, however, doctors who want to keep practicing medicine have to learn to use software. “They don’t have a choice,” says Mark Mitchell, senior vice president, provider services with Schumacher Group, which manages emergency rooms and provides other staffing services to hospitals.
Schumacher Group has some 3,000 health care providers working in nearly 200 emergency rooms in 27 states, using every variety of EHR systems. “We have to train providers on these systems, and we have to have continuing training,” Mitchell says. “People are embracing technology at a much more rapid pace than they ever embraced technology.”
Clinicians want the data EHR systems deliver, too. “Over the past seven years, I’ve seen physicians using data more and more in a daily manner and actually asking for more data to come into play,” says Douglas Menefee, CIO with Schumacher. But the functionality of these systems, and the level of analytics they provide varies widely, depending on how long they’ve been in place and the extent to which health care providers affiliated with a particular hospital or network have deployed them in their private practices. It also takes time to enter patient data and program decisions support tools, such as order sets to guide treatment decisions and alerts that warn clinicians about adverse drug interactions.
“Getting to adoption, getting them used in the [patient] encounter and creating a robust dataset that can be applied, that’s where we are now,” says Brennan. In fact, the “smarts” in many EHR systems are more likely to resemble traditional business intelligence reports, delivering pre-determined information in response to a specific event. “We can’t call it analytics in that it’s a pre-built analytical engine delivering an answer, but it is presenting physicians with more useful knowledge, intelligence and history.”
But for clinicians to find even these tools useful, the tools have to be designed in a way that is compatible with how they practice medicine. “We have a tendency to talk about providers as though they’re one type of individual,” says Menefee. However, “the way an emergency medicine physician practices is different from a psychiatrist or a cardiologist.”
“A lot of organizations miss the clinical context when they build out their systems,” agrees Laura Madsen, health care practice leader with Lancet Software and author of Healthcare Business Intelligence. As providers move toward deploying advanced analytics, so they are able to use data for predictions that guide care decisions, rather than react to what has already happened, the clinical context will be just as critical. “It has to be proven for them,” Madsen says.
Ask What Users Need
At the moment, two promising areas for demonstrating the value of analytics include helping providers manage groups of patients and understand their performance against treatment standards.
Centerstone, which provides community-based mental health services to patients in Tennessee and Indiana, established an analytics program five years ago because its leaders saw a need for “multi-data-source, longitudinal” analyses to make its operations more productive, says Russ Galyon. Galyon is director of analytics at the Centerstone Research Institute, a research and analytics organization that is closely affiliated with behavioral health services provider Centerstone.
Projects have focused on how Centerstone manages its patient population and the workload of its clinicians. When Indiana made changes to its Medicaid system recently, among them, capping payments and changing eligibility criteria, Centerstone was able to analyze how those changes would affect the profile of its patient population and whether it had the right mix of licensed professionals on staff to treat them. Galyon says Centerstone lost significantly less money than other providers in Indiana because it was able to adapt in advance to the level of care Medicaid would cover.
He observes that business managers accept analytics more readily than clinicians. “Someone will come up with an algorithm that will improve something clinically, and it’s an extremely hard sell,” he says. One factor, he thinks, is physicians’ “hard belief in the randomized clinical trial,” which makes them cautious about accepting results which aren’t based on rigorous research.
Even researchers that Centerstone has hired to develop it analytics projects need to temper their academic tendencies, Galyon says, so models don’t take months to generate results and require experts to run them. They’ve had to learn to define the limitations of the analysis for end users and build tools that users can run themselves.
There are reasons to be cautious when it comes to using analytics for real-time patient care decisions: individual cases are unique and complex and may be life-threatening; patient data is often spread among multiple providers and a lot of data is yet to be codified. “Algorithms need years and years of data to be really valuable,” notes Madsen. Besides, says Boston Children’s Harper, “I don’t think we fully understand what people would do with the data.”
Nevertheless, providers are beginning to succeed at using analytics to hold themselves accountable for following treatment protocols and helping them focus attention on the neediest patients.
Boston Children’s is in the process of deploying tools for tracking physician performance against guidelines for clinical care, efficiency and cost. For example, a “disease of the month” report in the ER shows how well the unit is performing on a variety of measures (what is measured depends on the disease in question, but may include how quickly antibiotics are administered, use of CT imaging or error rates).
Once, the hospital based such reports on a small percentage of patient records, says Harper. Today clinicians can generate weekly or even daily reports using EHR data about all eligible patients and use the results to improve unit performance. Clinicians can also see when—and why—their individual performance falls outside the norm. Units that underperform on certain quality, compliance or performance measures risk having their budgets cut, says Harper.
Analytics projects, like any IT projects at Children’s, are developed in consultation with physicians and other representatives from six business areas, including research and clinical care delivery. “If anyone bubbles something up, it gets discussed and we decide whether that should become a project and prioritize it,” Harper says.
As is the case with many technology-based innovations, deploying analytics first to practice groups that are enthusiastic about having the data is critical to convincing more skeptical clinicians to adopt it. At Children’s once a group has developed performance metrics, they’re published internally. Groups that don’t have them want to keep up.
“You have to have someone who will look at the data and act,” says Harper. “When you’re solving people’s pain points, they’ll generally jump on board.”
Elana Varon is contributing editor with Data Informed. Tell her your stories about leading and managing data driven organizations at email@example.com. Follow her on Twitter @elanavaron.
Correction, Jan. 4, 2013: This story has been updated to correct the last-name spelling and job title of Russ Galyon, who is director of analytics with Centerstone Research Institute, not Centerstone. The research institute is separate from, though closely affiliated with, Centerstone and provides it with research services.