Big Data is not just a numbers game
When it comes to hacking through the hype of big data, there are two types of analytics projects: those seismic, boundary-pushing advancements that, where they do exist, are mainly the product of big hospitals and academic medical centers, and humbler, more doable – but sometimes just as valuable – insights that can be gleaned by smaller providers.
"Your demarcation between what's practicable and what's 'Star Wars' is a good one," says John Hoyt, executive vice president of HIMSS Analytics. "When we do our Stage 7 validations, we do not ask for 'Star Wars,'" he says, referring to the seven-step HIMSS Analytics EMR Adoption Model.
Clearly, some hospitals are better prepared than others to make hay with big data. HIMSS Analytics figures show that 51.03 percent of hospitals are automated with financial business intelligence tools, while 45.8 aren't – and don't immediately plan to be. The numbers are roughly similar for data warehousing/mining technology (52.53 and 44.02 percent, respectively).
For clinical BI, the numbers are less encouraging: 29.04 percent have it, but 64.14 don't. More hospitals are making use of clinical data mining tools (42.71 percent) but still more than half (53.69 percent) are not availing themselves of technology that could help make sense of the patient data they have.
The 'Star Wars' stuff exists – genomics, proteomics, metabolomics, for instance – and some clinicians are making big advances in the way treatment is delivered by drilling down into those billions of tiny data points.
"Some organizations are participating in consortiums and submitting genomics data for comparability and analytics and all of that – but that doesn't work with an 80-bed hospital in New Hampshire," says Hoyt.
For more advanced providers, HIMSS Analytics does require some smart use of analytics, however. "That's exactly what we expect in a Stage 7 visit," he says. "Our line is: Use analytics to find problems you did not know existed and use analytics to show that you are making improvements in those problems "We tell them, you must show us at least three case studies," says Hoyt. "You've got data that you've never had before. Are you analyzing it? Are you finding anything 'profound'? And we do use that term when talking to Stage 7 candidates." Indeed, since the great EHR implementation wave of the past five years, that "data that you've never had before" exists everywhere – even in the smallest critical access hospital. It's incumbent on those providers to start making the most of it, says Hoyt.
"We now have the data," he says. "We have the responsibility to begin using it."
Hoyt remembers a recent visit to a modest-sized, 150-bed hospital. "They were analyzing their own data," he says. "They're not participating in the Beijing Genomics Institute, and neither should we expect it. But on the other hand, they should be doing something with all this data. That is the expectation we all should have of each other."
The bigger facilities are already starting to make some interesting inroads with big data. Hoyt points to one academic medical center that "did some research and they decided to write a clinical rule that, 'childbearing-age females, if they have a mother, aunt, cousin, grandmother, great-grandmother with the following diagnoses, fire a rule that says we recommend genomics testing for BRCA1 BRCA2,' DNA markers that have a high propensity for breast cancer."
That rule, he says, "fired 1,373 time in the month of November. Stunning. Now, they only got 22 visits out of that. They've got a long way to go. But they recognized that they've found something, and now they've got work to do to enable and teach the consumer population to get their genetic testing done. That's profound."
As for smaller organizations, Hoyt mentions a one-hospital system in Georgia that did some analysis of its surgeons: "'Who are my cheaper physicians, and who is more expensive? Who's using all the OR time and all the expensive implements?'"
Upon doing so, "They found their less expensive neurosurgeons and orthopods and they tracked their patients for two years – and found that they had a higher infection rate and higher surgical revision rate. Oh my."
Clearly, we've a sort of tipping point, with widespread access to "piles of data" that's rarely been available before. Now it's time to develop the will – and the skills – to start doing something constructive with it.
"We have a responsibility to now use that data to improve quality, safety and efficiency," says Hoyt.
And that means having a strategy. "That's one of the things that's clearly a weakness," he says. "No strategy for analytics – just (reacting to) whatever problems roll up to the surface."
"I won't buy that excuse for this one," he says. "Because if you look inside a health system, those are different skill sets. The CIO and nursing and medical staff are responsible for helping select and implement. But somebody else entirely probably does the analytics. It's not in the CIO's group; it's over in medical staff, quality control, or something. They've had this responsibility for years – now they just have new tools."
Becoming a "data-driven culture" is crucial, says Hoyt. And an essential component of that is making use of claims data to give and extra dimension of insight to the clinical data in hospital systems.
"If you're responsible, like an ACO, for a defined population, if you're not getting claims data, you're missing the boat," says Hoyt. "You can buy it. You can go out and buy Medicare claims data."
As an example, he points to Boston-based Pioneer ACO Atrius Health. "They buy claims data, and it tells them several things," he says. "It tells them leakage, where their marketing is not working: 'I've got a bunch of docs up here in Winchester and they're all going someplace else.'
"It also tells them, because they have history, and they have suspicions, some positions who may not have the highest quality, and when they see patients going there a lot they can expect to maybe see them again later somewhere else with the same issues.
"If you're responsible, like an ACO, of if you're a per-member, per-month, you have got to be getting claims data or you're missing, y'know, 30 percent of the picture."
Looking toward the future, Hoyt sees more and more organizations taking advantage of these tools. But on the way, a couple things have to happen.
"First, we have to bring this downstream to the smaller hospitals," says Hoyt. "I do believe we're going to get there. They're the later ones implementing the systems so they don't have the piles of data that a Geisinger has, that's for sure. Second of all, the place we really need to get to is predictive alerting: 'I see six things have occurred; seven and eight will likely occur because our data says so. Here's a warning, doctor.'
"When a patient is discharged, you should know what's the probability of that patient coming back in 30 days with this diagnosis, and should be on the lookout," says Hoyt. "And anybody who's got a probability about 45 percent should be turned over to case management so they're on that case all the time and you don't get an unexpected high readmission rate – because that's penalized."