Using digital clinical records and text mining to search for words or phrases, medical professionals at the University of Pittsburgh Medical Centre (UPMC) can assess the likelihood of certain patients needing emergency care.
When “wheeled walker” (a walking frame) occurs in clinical notes, for example, the likelihood rises. When the word “mother” appears, it falls.
Of course, it is not that everyone with a walking frame will end up in emergency care or that everyone with a mother will not, explains Pamela Peele, chief analytics officer for UPMC’s insurance services division. But when included in clinical notes and compared with vast numbers of electronic records, certain words provide strong signals of what may happen to a patient.
“You could never do that if all you had at your disposal were millions of pages of clinical notes,” says Professor Peele, who is also a University of Pittsburgh faculty member. “But when we’ve got it in electronic format, this is the kind of new knowledge we can drive – that’s the promise of big data.”
Clinical notes are just one source of the big data (complex information produced in vast amounts) that many believe could transform healthcare. Others include insurance claims, which can highlight, for example, where and when use of certain procedures is spiking.
Meanwhile, rich data streams are generated by personal health technology – from the mobile devices people use to monitor their conditions to the wristbands a growing number of consumers are using to track their fitness.
GPS-enabled asthma inhalers, which record the date, time and location of each inhalation, not only track an individual’s condition but also generate data that can be analysed to assess how asthma affects certain segments of the population.
Cost control is another potential use of data analytics. “There are significant pricing differentials across regions and procedures,” says Steven Van Kuiken, a McKinsey director and head of the firm’s work on healthcare information technology. “One thing big data can do is highlight those variances and allow patients or payers to act on that.”
Fraud detection is facilitated by data analytics. In the US, LexisNexis, the information provider, has helped public sector clients identify abuse of Medicaid, the government insurance programme for the poor. In one state, data analytics uncovered fraudulent Medicaid claims being made by a group of people living in a million-dollar condominium.
Big data can also improve consumer choices on healthcare, says Paul Bleicher, chief executive of Optum Labs, a research centre that is part of UnitedHealth Group, the US healthcare company. He points to an app developed by the group’s insurance company, UnitedHealthcare.
The app, he explains, allows users to identify doctors in their area, see how much they charge for certain procedures, assess them based on objective quality measures and work out what the procedure will cost, given the benefits in their insurance plan.
“Being able to bring together data from disparate sources to drive specific actions is the big data vision,” says Dr Bleicher.
But while the benefits of big data are starting to emerge, the fragmented nature of the healthcare industry, combined with privacy concerns about electronic health records, mean that choice and efficiency may take longer to spread than in other industries.
“What’s standing in the way is that the current clinical data systems are not interoperable,” says Regina Herzlinger, a Harvard Business School professor known for her research in healthcare. She points out that someone with congestive heart failure may have at least 34 related disorders.
“So it’s likely that 34 or more providers have dealt with them,” she says. “And they have systems that don’t talk to each other.”
In a January report, the Rand Corporation, think-tank, found that sluggish adoption of health information technology, lack of connectivity and the fact that the technology was not easy to use all prevented health IT systems from generating widespread cost savings.
Yet, given the financial pressures on healthcare providers, some believe the adoption of data analytics will accelerate. Part of this is being driven by demographics. As populations age rapidly, individuals may live for many years with several conditions and so need greater care co-ordination, which can be expensive.
Here, predictive modelling can play a role. “The ability to target scarce resources around care co-ordination to the people most likely to need them is an extremely good example of the efficiencies of big data,” says Prof Peele.
Meanwhile, to cut costs and improve care for everyone, healthcare systems are shifting their focus from individual procedures to end-to-end care, also requiring greater co-ordination between different medical professionals.
“The economics are too compelling,” says Mr Van Kuiken at McKinsey. Using predictive analytics, he says, health systems can deliver “much better care at a lower cost than anyone can operating in the old world of fragmented decision making and indirect incentives”.
Get alerts on Innovation when a new story is published