The word “innovation” is tossed around a lot these days, but when your organization is located at “the nexus of IT” — a neighborhood in Boston that houses some of the most cutting-edge IT shops in the country — it’s not just a term; it’s a way of life. It means having a long history of development while also being willing to utilize (and customize) commercial products; it means developing an innovation program to help bring ideas to life; and it means partnering with other organizations when the right tools aren’t available. In this interview, Dan Nigrin talks about what it’s like to be a Cerner-Epic shop, his organization’s data warehousing and analytics strategy, the other “CIO” at Children’s, and the unique collaboration among children’s hospital leaders.
- Custom development with vendors
- Outgrowing the homegrown data warehouse — “Queries would often take days to come back.”
- Data tools from IBM, Informatica & MicroStrategy
- Teaming up with Partners to let clinicians “dig a little deeper”
- Getting IT “out of the middle”
- Data governance steering committee
LISTEN NOW USING THE PLAYER BELOW OR CLICK HERE TO SUBSCRIBE TO OUR iTUNES PODCAST FEED
There’s gold in this data. By simply going to the data with a specific question in mind you’re just scratching the surface, because there’s lots of hidden knowledge within the data that is ripe for exploration.
We didn’t want to have the IT folks serving as the rate-limiting step in getting at that data. With a tool like i2b2, because it allows you to surf this data without compromising patient confidentiality, you can just hand it over to the researchers and allow them to run the queries. That ability to get IT out of the middle has been really beneficial.
We try to disintermediate ourselves, but there are clearly some instances where the complexity of the types of queries are things that outstrip the ability of our end users to be able to do on their own, and so for those, that’s where our business intelligence team steps in.
The demand still outstrips our ability to supply, and so you end up having to prioritize what initiatives you’re going to work on first. And that’s tough, because all of these things are important.
Gamble: It seems like you kind of have a best of both worlds situation where you have the established vendor products, but then you’re also able to do customization, which is something that’s important to you guys.
Nigrin: Absolutely, on the Cerner front especially, we’re using a lot of the tools that they provide. We use something called MPages, which allows you to deploy custom content essentially directly within the outer shell of their EHR PowerChart. The other thing that I’d like to mention is that when we’ve had custom development like that, both of our vendors have been quite open to looking to see what we’ve done and why we’ve done it. And when it’s made sense to them to consider that kind of functionality in their broader product for use by other organizations, they’ve been very collaborative with us in thinking about those things and trying to incorporate the same type of functionality directly within the product.
Gamble: That certainly makes sense on their end too. If someone’s found a better way to use something or added a different functionality, it makes all the sense in the world for them to want to leverage that.
Gamble: Let’s talk about data. You have this large research enterprise, so I can only imagine the amount of data you’re dealing with. I’m sure that having a strategy for analytics is something that’s got to be high on your list at this point.
Nigrin: Absolutely. We’ve been an organization that historically has had a data warehouse and analytical tools that we’ve used going against that data warehouse. But what I will say is that up until a few years ago, the warehouse had sort of grown organically over the years, and frankly just wasn’t optimized for the kinds of analytics and queries that we were finding we wanted to do. Just from a pure performance point of view, there were queries that clinicians or administrators or researchers wanted to run. These queries would often take days to come back with a result — when you’re going against millions and millions of rows and doing complex joins and so on, you can start to understand that. But that was clearly not a tenable long‑term position to have queries consuming that amount of resource and being so slow to come up with a result. Oftentimes, the machines would just run out of memory, and essentially you would not get any result. The query would just fail.
And so a few years ago, we evaluated some of the more up-to-date and current technologies. Basically we’re now on data warehouse 2.0 where we’ve implemented a really nice cutting-edge database infrastructure. We’re using a product called Netezza, which IBM bought a few years ago, and layering that together with tools from Informatica to do our ETL processes. And then for the presentation layer where folks can visualize the data and do their analytics, we’re using a tool from a company called MicroStrategy. That trio of products comprises the tools that we’re using for our analytics work. Most of those tools are used for work that’s done for operational queries, financial analysis, and quality improvement types of analyses.
We also have a tool that’s specifically geared toward research. One of the things that we’ve long recognized at our institution is basically that there’s gold in this data. By simply going to the data with a specific question in mind you’re just scratching the surface, because there’s lots of hidden knowledge within the data that is ripe for exploration. Together with folks at Partners HealthCare, we’ve developed over the years a tool called i2b2, which is used now by over a hundred organizations worldwide for these types of exploratory analytics going against large clinical databases. The key point is that i2b2 allows for clinicians and researchers to be able to do these exploratory queries without a detailed knowledge of how the databases underlying are set up. The terms used are clinical terms. You also don’t need a degree in computer science to be able to use the tool. It’s a drag-and-drop tool that allows you to basically identify patient cohorts that are defined by specific clinical attributes. If you’ve got a question about whether drug X, for example, could be associated with adverse outcome Y where outcome Y might not be well defined in the pharmacologic data as a known side effect, we can actually quickly query out entire patient population to see if we’ve got a large number of patients who are on drug X and who have side effect Y. And if that number is big, then we can start to dig a little bit deeper to see if this is a problem.
Importantly, from the research point of view, we do this while preserving patient confidentiality. We only allow this exploratory type of query in a de-identified manner where aggregate numbers are returned to the investigator. So in that example that I gave where you would run a query to look for intersection of drug X and side effect Y, your answer is only going to be, ‘100 patients in the last year have had those two symptoms’ — not, ‘here are the 100 patients and their names and dates of birth.’ Only if you’ve gotten IRB approval and so on will you be able to drill down into that level of detail. So it’s a nice tool that allows clinical investigators and researchers at our organization to be able to use the data and mine the data for interesting patient cohorts while preserving patient confidentiality.
Gamble: Right, of course that’s key. But in those unique cases you see where the patients really need some kind of intervention quickly, I’m sure that it’s really critical for clinicians to be able to access the data quickly.
Nigrin: Absolutely. Most importantly, we didn’t want to have the IT folks serving as the rate-limiting step in getting at that data. With a tool like i2b2, because it allows you to surf this data, if you will, without compromising patient confidentiality, you can just hand it over to the researchers and allow them to run the queries as they please. That ability to get IT out of the middle and acting as a middleman has been really beneficial.
Gamble: I can imagine. Obviously data management and data analytics are such huge topics right now. When you are dealing with absolutely huge amounts of data, have you put together some sort of data governance team? Who is charged with taking this data and turning it into useful information?
Nigrin: We have a multifaceted approach. Around the same time where we started to invest in building that warehouse 2.0 that I described, we also recognized that we needed a better governance process. We do have a data governance steering committee in place. It’s composed of very senior level people — C-level people — from across the entire organization. The role of this committee is a) to prioritize some of the analytic efforts that we take on as an organization, and b) to also define a little bit more clearly who should have access to what data, beyond HIPAA kinds of questions. There’s often sensitive data in these repositories. From an external point of view, and even from an internal point of view, you can imagine that you may not want to disclose individual physician performance to anyone other than the chief of their particular department. And obviously financial performance data has to be limited to the appropriate owners of that particular department, or in the institution’s case, the right people who oversee the institution’s financial efforts. That governance role is important, and frankly we didn’t have it in place up until a few years ago, so that’s been a step forward for us.
Beyond that though, within our IT organization we’ve definitely bolstered our data analytics group, and we now have an entire business intelligence team that’s focused on just the nuts and bolts of making all of those systems work that I described before, and also who are in charge of running the analytics and getting the results to the folks that need them when it’s not in a self-service mode.
We do try and distribute, as much as we can, the ability to run these queries to appropriate folks within individual departments. So again, we try to disintermediate ourselves, but there are clearly some instances where the complexity of the types of queries that need to be run are things that outstrip the ability of our end users to be able to do on their own, and so for those, that’s where our business intelligence team steps in.
Gamble: So obviously that’s something you’re doing a whole lot with at this point, and will be going forward.
Nigrin: We are definitely really focused on this, like everyone now. We’ve worked so hard as an industry to get the data in through automation and implementation of EHRs, and although we clearly still have a long way to go, I think many organizations now have gotten to the point where they’ve done it. They’ve gotten a good portion of the data that comprises the care that we provide in digital form and tucked away in databases. Now we’ve really got to work harder to realize the value of that data, whether it’s for an individual patient’s perspective or for doing population based analytics, financial modeling, etc. The list is endless.
Gamble: That’s got to be a little bit overwhelming, the idea of how much you can possibly do with the data if you just have all the resources you need and the time.
Nigrin: It is. That’s why that prioritization that I mentioned before is important. Clearly, we’ve got more resources than we had before to be able to do these analytics, but we still don’t have enough. The demand still outstrips our ability to supply, and so you end up having to prioritize what initiatives you’re going to work on first. And that’s tough, because all of these things are important. That’s why we do need some of the highest level folks within the organization to comprise that governance group to make some of those tough decisions.