BIG DATA
the Library of Congress.
Velocity is the speed at which the data
accumulates—very quickly in the case of a
hospital, which adds reams of test results,
images, vital signs, and clinician notes every day.
But McGlynn says the third defining
quality of Big Data—variety—is the one
most people overlook.
“There’s not only the data we create,
but what’s on social media sites, or data
about people’s shopping or driving habits,
what’s happening demographically in a
given neighborhood,” McGlynn says. For
example, hospitals trying to reduce their
readmission rates might want to know
what social support networks are available in the patient’s community. Maybe
the neighborhood block association has
a Facebook page where neighbors can
check in on each other or volunteer to
bring dinner to a patient who’s been discharged. Combining all the available information can create a picture of which
patients are at greatest risk for readmission and what interventions are most effective at keeping them healthy and out of
the hospital.
As usual, health care is playing catchup in the Big Data game. Brad Putnam,
executive director of HealthShare Montana, a growing health information exchange, came from the financial industry
and says banks were in the dark about
their customers up until about 10 years
ago. “They had all their information on
slave dummy terminals, and they’d send
out a survey occasionally and think that
they were giving great service,” he says.
Once banks started tracking how customers actually behaved, they saw a gap
between the services offered and the services needed. That’s why today you can
deposit checks via your smartphone and
get pinged when a stock hits your desired
price point. “It was a painful shift, but now
people can actively manage their financial
life instead of reacting when the monthly
statement comes,” Putnam says. “When
we’re able to look at patient populations
and measure how well we’re doing, we
can create benchmarks and help patients
change their behavior. That’s when it gets
fun and fascinating.“
The average hospital I.T. department
may feel it’s drowning in data already and
is not inclined to deal with more. But here
are three Big Data areas to watch:
Harnessing free
clinical text
Jonathan Handler, M.D., chief
medical information officer for transcrip-
tion and speech recognition software
company M*Modal, sent his two sons off
to summer camp with pre-stamped post-
cards to send home so their parents knew
they were still alive. One set of cards was
blank, but a second set had a pre-printed
summer-camp-specific checklist: (Camp
is: fun/not fun; Today I did: swimming/
archery/hiking/crafts/campfire, etc.). One
son chose the checkbox format, while the
other wrote on a blank card. It’s easy for
Handler to say which will be the more cher-
ished keepsake. “The check-boxes were al-
most worse than not having any informa-
tion at all,” he says. “The one with the free
text had 18,000 times more information—
what his nickname was, how exciting it was
to win the climbing wall competition.”
Much EHR work has focused on get-
ting clinicians to stop scribbling free-text
notes and instead use drop-down menus,
checkboxes, and other forms of structured
data that are easy to dump into databases.
Not only have they been reluctant to make
that radical change, but some Big Data
proponents now say it isn’t necessary—or
even desirable—to take structured data
too much farther than it’s already gone.
“The more advanced thinking is that
free text is where the action is, and we
have the tools to use it,” says Handler,
who’s somewhat biased because his com-
pany is in the business of natural language
processing, one of the tools that unlocks
free text. However, he points out that the
early adopters of Big Data analysis, such
as Google and Amazon, don’t need heav-
ily structured data to work their magic.
Google plows through billions of pages of
free text routinely, guided only by its algo-
rithms, and usually comes up with some-
thing approximating the right answer to
almost any inquiry.
