NEPHROLOGY •
CONTINUED FROM PREVIOUS PAGE
Making the links The data wranglers
Working with mouse models of AKI, Drs. Soranno and Faubel are
finding it affects almost everything: the lungs, heart, liver, even the
spleen. One of its more disastrous effects is immune dysfunction. If the process of combining hundreds of data elements from three different systems into one de-identified, longitudinal
record sounds complicated, it’s even more complicated than it sounds. Every institution codes its EMR data differently.
Finding common variables is difficult. Combining common patient data into one continuous record even more so.
Kids who undergo the Norwood and develop AKI are three and a half
times more likely to get a post-surgical infection, independent of
other factors, Drs. Gist, Soranno and Faubel found. They published
their results in Pediatric Nephrology last fall. In Children’s Colorado’s
Pediatric Kidney Injury and Disease Stewardship program, or PKIDS,
they collaborate with neonatologist Jason Gien, MD, to study AKI
in intensive care through a large bank of blood and urine samples
collected from every consenting patient across Children’s Colorado’s
three ICUs. “We take care of the data-wrangling to provide data in a format that works for researchers so they can just focus on analyzing
the data,” says Davis. “We can slice and dice it many different ways.”
The effort, however, is campus-wide. Wrapping in pediatric and adult
specialists from neonatology, cardiology, surgery, anesthesiology,
pulmonology, infectious disease and critical care, the Multidisciplinary
Translational Research in Acute Kidney Injury Collaborative, or
M-TRAC, acts as a research umbrella across institutions. And their
retrospective studies of AKI suggest the long-term consequences
may be even farther flung: stroke, blood clots, fractures, GI bleeding
and infection.
“If you’re going to make these associations, they should work in mice,
in neonates, in kids, in adults,” says Dr. Faubel. “We’re trying to make
those links among all these different populations.”
“We’ve probably seen just the tip of the iceberg,” says Dr. Soranno.
“We haven’t even looked at the big picture data.”
Here’s what that process looks like:
Children’s
Colorado
UCHealth
“Five years ago, if a researcher wanted to compare EMR data from
UCHealth and Children’s Colorado, they’d have to submit two
different requests to both institutions, manually curate the data and
then do the analysis,” says Sarah Davis, principal informatics analyst
for Health Data Compass, a company set up on the Anschutz Campus
to do just that.
University of
Colorado School
of Public Health
Public health
records
EXTRACT, TRANSFORM, LOAD
Data quality automation
makes sure large volumes move
from one database to another
without lost data, inversions or
changes of value.
That’s not because the data isn’t there. Continuous renal
replacement therapy (CRRT) machines, standard in ICUs, offer
vast, largely unexplored repositories of information. The electronic
medical record, too, offers reams of longitudinal data just waiting to
be parsed. The question is how to parse it.
Variables and variables
University of
Colorado School
of Medicine
Managing master patient identity means
finding common patients to create a
long record — essential in cross-linking a
pediatric and adult institution.
Transforming source data
without loss or corruption
is essential for creating a
master model.
CENTRAL DATA WAREHOUSE
Business Intelligence codes against the data warehouse to
pull data needed for smaller, research-friendly sets.
Algorithms find patients that meet
the criteria and provide all the needed
data elements within that set.
Cohort
Diagnosis
Other inclusion and
exclusion criteria
Pediatric cardiac intensivist Katja Gist, DO, adult nephrologist
Sarah Faubel, MD, and pediatric nephrologist Danielle Soranno, MD,
collaborate across the institutions of the Anschutz Medical Campus.
The database of acute kidney injury they’re building will be the most
comprehensive of its kind.
34
RESEARCH DASHBOARD
NEW CONSTELLATIONS
35