Georgia Tech will study data models using IBM analytics technology to better understand how to improve pediatric care.
IBM and the Georgia Institute of Technology have launched a program
called One Million Healthy Children to use data modeling to analyze the
health information of children suffering from asthma, autism and
The goal of 1MHC is to analyze the health data of 1 million children in the state of Georgia, according to IBM.
1MHC will analyze models of fee-for-service health care, in which
patients pay providers for specific services. They'll also study how
transportation, health services, socio-economic status, education
history and food resources affect the health of children.
For the project, IBM is contributing financial support as well as the data storage capabilities of DB2 database software
and data analytics from SPSS and Cognos.
The tools will allow Georgia Tech to plug in various factors from
unstructured data to see what the outcomes might be, IBM reports.
"We've developed some extraordinary ways of taking data from a given
field and parsing that data and building models around it using
intelligence," Bernard Meyerson, IBM fellow and vice president of
innovation, told eWEEK.
IBM and Georgia Tech announced the big data initiative on Oct. 27.
To adhere to privacy measures, the data will be "anonymized" and
untraceable back to patients. "We're extremely careful of this aspect,"
Meyerson said. "We don't want any possible way to work backwards."
Researchers will study data from more than 16,000 health records of
Emory University employees' children in the first stage of the project.
In later stages, analytical models will incorporate data from
Children's Healthcare of Atlanta, Georgia Cancer Coalition and the
Georgia Department of Community Health.
Researchers aim to use the analytic models to enable health care
providers to save time and money and better understand differences in
pediatric health care expenditures and outcomes among various
Using a multi-level model, Georgia Tech's researchers will show
visualizations of a medical history across the provider, patient and
payee. Data models will incorporate information on the medical and
financial aspects and run "correlation analyses" from multiple angles
to see if outcomes may occur that doctors wouldn't anticipate, Meyerson
By analyzing the data, researchers hope to see how to fix the
pediatric health system systemically rather than at one level at a
time, according to William B. Rouse, executive director of the Tennenbaum Institute
at Georgia Tech.
"We're looking at very large data sets and need to portray relationships in those data sets," Rouse told eWEEK
. "We're trying to use the computational power to provide that broader view."
For a bone fracture, researchers might run models to see how a
patient would fare based on location, speed of treatment and whether
they took calcium supplements or not, Meyerson explained.
"If you believe that the healing time of a fracture is related to
some unforeseen medical issue such as the presence of diabetes, how
important is it to tightly control sugar levels?" Meyerson asked. "It
should show you when children have well-controlled diabetes."
For conditions such as asthma, autism and diabetes, researchers will
create models based on claims and diagnoses as well as drugs
administered, Rouse said.
"Think about health analytics as treatment protocols involving
seven, eight, nine different things that would yield an outcome you
wouldn't find unless you've looked at it together," Meyerson said.
The data analytics work will reveal data incompatibilities in IT systems that need to be fixed, according to Rouse.
"We have the basic machinery to do this well, but we have to get
behind problems of incompatibility of patients' data sources and all
the complicated factors of industrial systems," Rouse said.