Skip to main content

Q&A: Analytics-as-a-Service from Deloitte and Intermountain Healthcare

 |  By smace@healthleadersmedia.com  
   July 02, 2013

OutcomesMiner, a software application codeveloped by Deloitte and Intermountain, leverages 40 years of clinical data to help analysts glean the "clinical nuances" of comorbidities and various treatment outcomes. How does it work and who will use it?

Can healthcare's Big Data become less of a mountain to be sifted through by experts, and more like a utility to hook up, like water, power or cable TV? We're about to start finding out. Analytics applied to Big Data offers tantalizing possibilities for improved healthcare, but the complexity is enormous.

I spoke last week with Brett Davis, general manager of Deloitte Health Informatics (DHI), following the release of OutcomesMiner, a service that leverages 40 years of clinical data from Salt Lake City–based Intermountain Healthcare to help analysts throughout healthcare glean insights about the relationship between combinations of comorbidities and various treatment outcomes.

HealthLeaders: Where does this fit in the analytics tool universe?


See Also: Intermountain Launches Analytics Tool for Population Health Data


Brett Davis: In a lot of ways, I wouldn't actually think of it as an analytics tool. One of the big things that we and Intermountain coming together are trying to solve is that the secondary use of healthcare data to understand what works for whom, why and in what context and at what cost, is not really a software problem. There's tons of great analytical software and tools out there, but the challenge is bringing together a combination of insights from health systems and firms that have longitudinal clinical data, combining it with analytical tools that provide nuanced understanding of outcomes in subpopulations, and then actually applying that to clinical change and transformation. That's not going to be solved by software alone or data integration tools alone, and so that's why Intermountain Healthcare and Deloitte really came together with our alliance … with this platform, OutcomesMiner, being the first result. OutcomesMiner is a tool at one level, in that there is software, but it's software meeting analytical and clinical insights from the 40-plus years of experience Intermountain has in becoming a data-driven organization—which … led to them being able to prove that you can take variation out of care and increase quality while at the same time reducing costs.

HealthLeaders: Is this a service that Deloitte is just delivering to customers, so there's nothing a customer need install? Do they have to set up a new data warehouse or anything like that?

Davis: There's really two dimensions to it, depending on the client need. There is a full cloud-based subscription service, where a health system who may be doing comparative effectiveness research on real-world data can access the tool and not implement anything on-site, [They can] glean insights through the tool into Intermountain data in an ethical, secure way, but health systems and even pharma companies also want to do comparative analytics with their own data that they may have. In that case, they can feed the tool, the platform, through an existing warehouse that they may have, or they may ask Deloitte to help them with putting a warehouse in place to feed the OutcomesMiner platform so they can look at their own clinical variation. Traditional benchmarking at system levels is interesting to potentially diagnose that your CHF population or your asthmatic population is not performing as well relative to your peers, but if you're going to start taking on more risk in managing populations, you really need to be able to get granular. You need to start getting down into what I call "clinical nuance," and understand that it might be CHF patients who have these other two comorbidities—psychosis, hypertension, and diabetes—are on these two drugs, etc. Existing platforms and tools on the market don't give you that level of clinical nuance to be able to actually go beyond this and start to identify the sub-stratified populations that you really need to be going after.

HealthLeaders: Are these predictive models, or is that overstating what this is doing? Are you predicting outcomes? Are you modeling outcomes?

Davis: It may be taking a step too far to truly call it predictive, in that we're giving the answer. It's giving you the insights, though, of where to look. With observational data, you always have got to be careful [not] to overstate causality and the associations. It gives you associations between outcomes and the multiple comorbidities to go investigate in the right areas; to use the overused analogy, of making sure that if you drop your keys in the dark, you've got a flashlight that you can search and find the right spots. It's kind of that analogy to do the diagnosis and make sure you're targeting the areas that are going to have high impact to addressing cost issues while not impacting quality.

HealthLeaders: How has this been proven out? What is the evidence you have so far that this combination of insight and services makes a difference?

Davis: We're early in the journey, so we don't have client results that we're talking about yet publicly, but we do have validation from the Intermountain clinical and informatics community, working closely with the Homer Warner Center there, which is a 60-plus person informatics center that were codevelopers in the platform.

HealthLeaders: Is there a double-blinded format that protects PHI [personal health information]? How does that work?

Davis: No PHI data ever makes it into any IT environment where PHI information could be potentially revealed. There's a double-blinded format where there's a blinding step that's done by Intermountain before Deloitte or our platform gets involved, and there's a second double-blinded key—so it's sort of like the old nuclear days where you needed two keys to reidentify, to provide extra protection to make sure that there's never any exposure of PHI through the platform.

HealthLeaders: What kind of customers is this appropriate for? Are we talking about customers who have a large IT staff, who have no IT staff, somewhere in between? Give me a feel for that. It might be analysts, not necessarily IT people per se. Analytics people, data people, quants, or whatever.

Davis: The platform and the mission of our venture together with Intermountain is really to scale the insights that come from the secondary use of data to drive clinical practice improvements. Because if you look across the market broadly, the level of investment that, say, an Intermountain or a Partners Healthcare or UPMC can put behind building out large staffs around analytics is limited, and if you look at the skills out there, even if they had the capital and resources, the skills of people who can actually do this kind of work are extremely limited. It's a very competitive field, so to recruit the kinds of data scientists and informatics folks who can actually do this stuff—and then you put on top of that the fact that most health systems just invested tens of millions or hundreds of millions of dollars in their EMR journey—going in and investing in a big complicated analytics environment, it's just not going to happen. The solution was really designed to democratize healthcare analytics, and bring these insights and learnings to the broader market for those systems that haven't [become] a data-driven organization. A lot of these health systems are flying blind in terms of really being able to understand the nuances of what's going on inside their comorbid populations. This solution is really designed to avoid having to go and invest in a big complex IT project for three or four years before getting value.

HealthLeaders: Does the fact that you've announced it at the Drug Information Association's 2013 Annual Meeting mean that you're going to see a dominant customer base of pharmaceutical companies, or did that just happen to be the convenient place to launch it? What's the mix of customers likely to be on this?

Davis: It's hard to predict exact customer mix, but there's absolutely demand in the pharmaceutical/biotech and medical device space, because they're facing similar pressures to prove value and compare effectiveness in their products. Going forward, and maybe not as widely known, some pharma are already entering into their equivalent of ACO-like contracts, where instead of just getting reimbursed based on pill volumes, they actually are engaging in helping to make sure that their therapies, diagnostics, etc. actually are creating the value that they say they will. So on the pharma/med device/biotech side, there's absolutely a market for insights from real-world longitudinal data. Similarly, where we're seeing demand on the health systems side is really around systems that are starting down the journey of moving to value-based care, whether that be in the context of formal ACOs (the CMS shared savings program), or whether that is via the context of just trying to become more clinically integrated, recognizing that management of complex, comorbid populations is going to be a differentiator and a necessity in the coming years. We're seeing demand on both sides.

Scott Mace is the former senior technology editor for HealthLeaders Media. He is now the senior editor, custom content at H3.Group.

Tagged Under:


Get the latest on healthcare leadership in your inbox.