Data prescription for better healthcare
We’ll send you a myFT Daily Digest email rounding up the latest Anthem Inc news every morning.
Machines beep and drone throughout the neonatal intensive care unit at Toronto’s Hospital for Sick Children. Intimidated parents stand by as nurses scurry between the glass cases.
Among their various checks, nurses chart a baby’s heart rate once an hour. This is standard practice at most hospitals, but Carolyn McGregor, a professor of health informatics at the University of Ontario, Institute of Technology, says it leaves a lot of critical data to waste.
While a baby’s heart beats around 120 times a minute, it is the pattern of those beats over time that that can give early warning if something is wrong.
With the help of Watson, IBM’s supercomputer that trounced the top Jeopardy! champions last year, Ms McGregor analysed live streams of every heartbeat in a project nicknamed “data baby”. She found patterns that revealed signs of infection 24 hours before the baby showed any visible symptoms. In premature babies, advancing treatment by even an hour can be life-saving.
“At the moment, all nurses have are some very poor alarms that go off all the time, so everyone ignores them,” Ms McGregor says. “We’re trying to watch more streams of data that will give more intelligent alarms.”
The healthcare industry is under extreme pressure from regulators and market forces to reduce costs and produce better outcomes. The US, in particular, spends 30 per cent more on healthcare per capita than other developed countries – about $750bn in total. But despite this higher level of spending, its life expectancy or infant mortality rates are not significantly better, according to a McKinsey report.
The US healthcare industry has therefore become the target of sweeping reforms aimed at reducing tests and procedures that bulk up the bottom line but show minimal to no impact on making patients better. Hospitals and insurance companies must demonstrate efficiency and the improved health of its customers to remain compliant with government policy changes and stay competitive in the free market.
Harnessing vast troves of data is increasingly seen as the solution. From medical devices and insurance claims, to scribbled doctors’ prescriptions and social media sites, the data in healthcare is massive and messy.
One of the promises of organising and analysing all those data is the ability to predict the future, with the goal of early intervention preventing heart attacks or hip fractures from happening in the first place.
“Historically, quality in medicine has been a retrospective affair,” says Stephan Fihn, a doctor and director of analytics and business intelligence at the US Veterans’ Health Administration. “The Holy Grail right now is what would be termed ‘predictive analytics’.”
One of the earliest health systems to begin keeping health records electronically, the VA now has data on 20m patients, including 2bn text notes, 16.2m X-rays, and 1.5bn drug prescriptions. From analysing that data, it can, for example, identify characteristics of patients who suffered renal failure after getting a certain drug, then use that to predict who else is likely to have the same reaction.
Such measures are now of supreme interest to private hospitals in the US. Under new health reform laws, hospitals now face financial penalties if certain patients are readmitted to the hospital within 30 days of being discharged. To maintain profit margins over the long term, they must invest now to develop new protocols that conserve resources and improve performance.
Several hospitals are running data analyses to determine which patients are at the highest risk of being readmitted, for example after a heart attack or pneumonia, then prescribing in-home monitoring devices and outreach programmes to keep them on track.
“We’re moving away from pay-per-pill and pay-per-procedure, and going into a business model that is pay-per-person,” says David Dimond, healthcare strategist at EMC.
Health insurance companies are just as invested in correcting the inefficiencies of the system. Premiums have risen in step with the overspending, and customers, whether employers that sponsor health plans for their employees, or individual consumers, are at a breaking point of what they are willing to pay. Under US reforms, insurers cannot deny coverage based on pre-existing conditions, so they must manage the risk instead of avoiding it.
WellPoint, a health insurer that covers more than 33m patients, has contracts with dozens of technology suppliers, including several to manage and analyse its data.
One of the main goals is to use computers to make the human decision-making process more efficient, says Elizabeth Bigham, vice-president of health IT strategy for WellPoint. For example, the current process for pre-authorising claims can take up to two weeks and involve input from several nurses and physicians. Once a computer is trained on how the company decides to pay or reject claims, most of these humans could be replaced and the patient could have an answer almost instantaneously.
It is even exploring what it can do with all the data generated by medical tracking devices, such as glucometers and heart rate monitors, and even health tracking devices such as the FitBit, a pedometer that tracks the wearer’s activity, calorie intake and sleep patterns. It uses those data to inform diet and lifestyle coaching programmes for patients with diabetes, obesity and hypertension with an eye on preventing a catastrophic – and expensive – heart attack or amputation.
John Edwards, director of the health advisory at PwC, expects health insurers to offer lower premiums in the next one to two years to customers willing to wear a mobile health tracking device, similar to car insurance companies that give discounts to drivers willing to install a GPS device on their cars that tracks speed.
“It’s not far-fetched at all that tracking exercise could become part of an insurance design that says ‘you’re doing the right thing’ in order to control healthcare costs,” he says.
Wireless sensors are also seen as a way to control costs associated with the ageing population. Motion sensors in beds, walls and floors of elderly people’s homes can track patterns in activity and movement to predict devastating falls. Researchers at the Sinclair School of Nursing at the University of Missouri are identifying patterns of restlessness during sleep and changes in physical activity that precede a devastating fall or visit to the emergency room.
Bonnie Wakefield, a research professor at the University of Missouri School of Nursing, says such data patterns and predictive algorithms could soon aid nurses making critical decisions at all levels of healthcare.
“It’s quantifying intuition,” she says.