Health care companies turn to 'big data'
Hospitals, insurers purchase intimate details about patients
growing number of health care companies are purchasing intimate details about their
patients -- such as their income and buying habits -- and using it to make
predictions about their health.
marketers, many of the health care providers and health insurance companies
that collect this kind of intelligence aren't trying to sell more goods or
services. Instead, they're trying to identify which patients need extra help
managing their health care -- and keep them out of the emergency room.
trying to] understand the health care needs within our community and within our
health care and hospital community and improve outcomes," says Dr. Michael
clinical officer for analytics and outcomes research at Carolinas HealthCare. The
multistate health care network operates hospitals and clinics all over North
and South Carolina and has just started looking at consumer data it purchased
from a data broker.
A growing market
researchers have been collecting and analyzing huge amounts of data for years.
In just the past decade, the "data exhaust" left by clinics and
hospitals -- including electronic medical records, physician notes and claims
data -- has fueled numerous research projects and led to some potentially life-changing
example, researchers have used "big data" analytics to cut down on hospital readmissions, combat sepsis and take better care of newborns. Now, some health care companies
are taking data collection efforts one step further and incorporating patients'
personal and financial details into their projects.
to a review by CreditCards.com of white papers, PowerPoint presentations and
other business-to-business documents, everyone from the health insurer Blue Cross Blue Shield to the major analytics company SAS to the data giant FICO is experimenting with or at
least talking about using consumer data for health care purposes.
for example, made waves in 2011 when it introduced the Medication Adherence Score,
which uses consumer information and claims data to predict whether or not
you'll take your medication on time.
smaller analytics companies, such as the Massachusetts-based startup
Predilytics, are also getting into the game and using consumer purchase data and financial
records to make other kinds of health predictions.
a whole new thing that's happening with health data and predictive analytics,"
says Pam Dixon, executive director of the World Privacy Forum. Earlier this
year, Dixon looked at a number of health care analytics projects for an April 2014 report on the hundreds of consumer scores data scientists have developed for different purposes. For example, Johns
Hopkins University's Hopkins Frailty Score uses health data to predict how
likely you are to experience serious complications after a surgery.
a whole new thing that's happening with health data and predictive analytics. There
are entire conferences about it. They're just going
crazy with this stuff.
World Policy Forum
then, she says she's heard from numerous people about health-related scoring. "After
publication, a lot of people came forward and said this is happening," she
are entire conferences about it," adds Dixon. "They're just going
crazy with this stuff."
say incorporating consumer data into predictive models for health care is an
especially hot topic now that the Affordable Care Act has brought major changes
to the health system. Health insurance companies are no longer allowed to make
underwriting decisions based on your health status to manage costs. So they're
looking for other ways to drive down expenses and prevent people from over-using
costly health care services.
hospitals are being penalized by the Centers for Medicare and Medicaid Services
for too many hospital readmissions, and so researchers are furiously trying to
come up with fresh ways to care for patients.
Varied consumer data available
that goes into these research projects varies, but health care researchers who contract
with a data broker have access to a startling amount of personal information. According
to a May 2014 report by the Federal Trade Commission,
data brokers collect information about your court records and criminal history,
where you shop, what you buy, how much you make, what you post online, whether
or not you vote and what kind of car you drive.
cases, the data collection is so extensive that researchers can figure out your
personal habits, based entirely on the trail of information you leave behind
when you go about your day.
though none of us have realized it, for the past five years all of our
financial transactions, what we purchase with debit cards or credit cards, is
being tracked," says the World Privacy Forum's Dixon. And the information
isn't anonymous, she adds. "We know for sure from the FTC investigation
that data brokers are purchasing retail transactions attached to your name,"
consumer advocates to cry foul when marketers and health care companies try to
use consumer data purchased from third-party data brokers to target and
research ordinary consumers. The problem, they say, is that consumers don't
know this information is being shared with other companies and have no say over
how their personal details are being used by other entities.
don't realize that information from these relatively innocent events is being
collected. It's being sliced and diced and combined to form big data dossiers
about them that are then sold to companies," says Paul Stephens, director
of privacy and advocacy at Privacy Rights Clearinghouse.
Giermanski of Belmont, North Carolina, agrees. He says he doesn't mind if a health
care provider uses his personal information if he agrees to it. But if they buy
the information from a third-party data broker, "that's something
different," he says. "I don't support anything like that without the
knowledge and permission of the individual involved."
researchers say they're interested in patients' personal and lifestyle
information because it can help shed light on behaviors and environmental
influences that don't necessarily show up at a 20-minute visit to the doctor's
very careful. We understand that using the data is not
without risk. We really
have put a lot of thought around the governance of the data itself,
Dr. Michael Dulin
you think about health care, a lot of it is related to just our everyday
activities," explains Carolinas HealthCare's Dulin. For example, if you
smoke, eat poorly or spend most of your day sitting down, you're more likely to
develop a chronic disease, such as diabetes or high blood pressure. "Those are
really the things that are most important, that drive your health
outcomes," he says.
this information, health care providers hope doctors and nurses will be able to
give patients more personalized advice. In some cases, providers may even be
able to help patients overcome some of the personal barriers preventing them
from living a healthier lifestyle.
consumer advocates say the quality of the information that's being used for
these projects may not be reliable. "Data broker information is
notoriously inaccurate," says Privacy Rights ClearingHouse's Stephens. So
providers may be relying on erroneous information.
advocates also worry the data collection could get out of hand. While some organizations
may be using the data simply to advance their research and make smarter
decisions about care, consumer advocates worry that it will be used to unfairly
also worries that consumers' health scores may be shared with other companies. Health care
providers, such as Carolinas HealthCare, are regulated by the Health Insurance
Accountability and Portability Act (HIPAA) and aren't allowed to share your
personal information with third parties. But if an analytics company, such as
FICO, generates a health score using alternative data, such as your financial
information, that's different.
it's not using health care data, it's not covered by HIPAA. It can be
sold," says Dixon. "That's our guess. That's our legal guess." There's
no telling then what other companies may do with that information, she says.
"Nothing's off the table yet."
Using consumer data for 'the
Dulin of Carolinas HealthCare says that in his group's case, researchers take
special care with the data and don't let anyone outside of a select group of
people see it.
very careful," says Dulin. "We understand that using the data is not
without risk." That's why the hospital network has hired a data governance
expert to help guide the researchers on privacy and data security and has put
systems in place to make sure the data is being used properly. "We really
have put a lot of thought around the governance of the data itself," he
rewards, in turn, are worth it, Dulin says. "There's a huge potential
impact when we do this right."
Carolinas HealthCare has used community and clinical data to figure out a wide
range of interventions for its patients. For example, by using population-level
data about the communities served by Carolinas HealthCare, the research team
recently discovered that a number of patients were having a hard time seeing a
doctor because they lived in a remote area of town with spotty public transportation.
"Our intervention in that case was to embed a primary care physician in
the neighborhood itself in order to overcome that transportation barrier,"
team is thinking about using household data it purchased from a third-party
data broker to pinpoint additional interventions. "The consumer data that
we're bringing in now, we're still in the process of looking at that data and
understanding it," says Dulin.
group hopes it will eventually be able to use the marketing data it purchased --
ranging from a patient's income to how many people live in a particular
household -- to prevent problems before they start. But they're not yet able to
give specific details about how exactly they'll use the data for prevention since
they're just now starting to look at it.
and his team are currently experimenting with using patient risk scores based partially
on consumer data to help determine whether a patient is at higher risk for
developing a chronic health condition, such as heart disease or asthma.
is being selective about the data it uses. For example, Bloomberg BusinessWeek reported in July 2014 that Dulin's analytics
team is using credit card purchase data in its forecasting model. But Carolinas
HealthCare disputes that characterization and says it's not collecting specific
transaction information. So if you purchased cigarettes or pizza, that
information isn't going to be used in your health risk score -- at least for
now. (Dulin has voiced interest in this kind of data, however. For example, in
2014 interview posted by Smart Data Collective, Dulin said that consumer information
"like the data that you provide when you swipe
your membership card at the gym and buy doughnuts" may be incorporated into
and nurses also won't be able to see what goes into a patient's risk score. Instead,
they'll only be given an estimate that predicts how likely you are to develop a
certain ailment, which they can then use to tailor their recommendations to
to Dulin, "big data" projects such as the risk score model that his
team is currently developing can have a transformative impact on people's
primary care physician, he says that's especially rewarding because it means he
can potentially help an enormous amount of people. "When we leverage the
data, when we start to think about populations as a whole, we can really have
huge impacts on patients' care," he says. "That kind of thing really
See related: Scared of Big Brother? Too late, says 'Big Data' co-author
, 12 creepy details data collectors know about you
Published: August 14, 2014