Predictive modeling shapes the future at Boston Children's

Hospitals nationwide are turning to informatics to streamline care processes, improving both the efficiency and safety of patient care. One facility that appears to be ahead of the curve, however, is Children's Hospital Boston, which, according to CIO Dan Nigrin saves roughly $1.4 million annually by using informatics in its medication delivery system.

"The pharmacy is made aware of changes much more expeditiously than when we were paper based," Nigrin told FierceHealthIT in an exclusive interview. "The pharmacy has gotten to essentially be much more of a just-in-time delivery model, where they're continuously delivering medications to the floor, every hour or every two hours, as opposed to every 12. If a medication has changed, the amount of potential waste that occurs because the medication was, for instance, discontinued, has dropped dramatically."

Nigrin and colleagues Jonathan Bickel, director of clinical research informatics at the hospital, and Marvin Harper, the facility's chief medical information officer, recently spoke with FierceHealthIT about those and other ways in which they currently are using informatics.

FierceHealthIT: How is Boston Children's currently using predictive models?

Dan Nigrin (right): One of the main ways in which we get at that data is using a technology that's been developed locally here in a joint effort between Children's Hospital and Partners HealthCare on a platform called I2B2 (Informatics for Integrating Biology and the Bedside). In short, I2B2 is an application that allows end users to be able to look for cohorts of patients that meet a certain criteria. For example, I could be looking for all of the patients seen within the last year at Children's who have diabetes and who have hypertension, and this thing crunches away for a few seconds and then I'm given that de-identified cohort of patients.

Jonathan Bickel (left): We've had the technology around for the last four years, and it's been in constant development for at least seven or eight. It's also installed at about 60 academic medical centers across the country.

The thing that's most innovative is we're starting to connect up individual I2B2 instances together in a network called the Shared Health Research Information Network (SHRINE). It's essentially a group of I2B2 instances where the data lives still within the firewall of the home institution, but yet we here at Children's can initiate a query and ask those same kind of cohort questions previously mentioned across this network and get the same information; we're really starting to develop national cohorts.

FHIT: How fast can you get results from SHRINE compared to I2B2?

JB: It maybe takes 150 seconds as opposed to 20 seconds within your own hospital to get your number. We're talking you don't have time to get up and get a cup of coffee.

DN: With respect to these tools, though, they're not yet to the point where if you've got a new patient in front of you, you can query the database to find out how other similar patients have been managed or how they fared on a certain system; the system is not yet essentially real time for us. It's not informing care at the point of care. Currently, we're using the system in a research mode.

I think over the next few years, however, that could change.

Marvin Harper (right): When Dan and Jonathan say that we've used I2B2 or our data warehouse for being able to do research, that research is applied, pragmatic research. We've had publications in recent years that stemmed from being able to search our database that help us to predict significant bacterial infections in children that range from bloodstream or urine infections to lime infections or septic arthritis. While it can't be in real time searched and applied to that patient individually, I think that this data has been very powerful already in the care of children nationwide.

It used to be if you were trying to assess the quality of care for a specific type of patient, you might audit manually four or five patients; now you can very quickly review the care of 70 patients, which makes a difference when you're trying to find small things in your care that might make a difference, but might otherwise go unnoticed.

FHIT: Have there been any privacy concerns with the system?

MH: The system is very secure. However, there are concerns of, if we have such easy access to the data, do we end up researching some subgroups of patients too heavily.

DN: Imagine a very obscure disorder--and we see our share of these here at Children's--where there's literally only handful of patients known to have that disorder. If I search I2B2 for all patients who have this rare disorder, I will get back all one or two of them. However, I2B2 is smart enough to know that when a search results in a cohort of patients that's under a certain small number, it automatically adds a fudge factor to the results. So it won't tell you exactly two, for instance, but it will tell you that the answer is four or five, but that it's plus or minus three. It prevents you from being able to identify a single patient, simply because you know that they have a particularly rare disorder.

FHIT: You talked about predictive modeling from more of a research perspective. How is it helping to improve care at the bedside?

JB: Our electronic whiteboard system--ALICE (Aggregated Local Information Collected Electronically)--allows clinicians to constantly keep track of patient information. Every patient is assigned a standardized score every few hours dubbed CHEWS (Children's Hospital Early Warning Scores)--this helps us to assess how sick or in need a patient might be. A validated algorithm connects to the CHEWS score to notify clinicians when a patient's score signifies they need extra care.

Doctors can look, not only on a per patient basis, but also across the unit, to get a sense of how sick patients are, and if nurses need extra help.

Editor's note: This interview has been edited for length and clarity.