Aetna: Big data in healthcare – Michael Palmer
Rising healthcare costs present a critical challenge for corporations and governments, in developed and emerging economies, and many are looking to big data as a means of moderating healthcare expenditure. Michael Palmer, head of innovation at Aetna, explains how insurers are starting to use analytics for this purpose.
How can big data be used in the world of medicine?
Michael Palmer: Healthcare costs are growing, and that's a huge challenge for us. In the US, they rise by around 5-8 % every year, and there's at least $800bn of waste in the system. We can get rid of a huge amount of that waste through better targeting of care to patients at the time of need.
At Aetna, we currently help manage around 19 million members, soon to rise to 22 million. Each person interacts with the healthcare system an average of four to six times a year, and every time that happens we get information like laboratory or test results. In many cases, these patients are also taking medication.
As we look at all those events across the continuum of care, we can collect that data and begin to find patterns - not just historical trends, but also the ability to predict what will happen in the future. This will help us determine where people are going to have challenges, intervening earlier in that process and helping them avoid potential health problems.
What's the biggest challenge in using big data to improve people's health and, ultimately, lower the costs of healthcare?
It's an intimate dance between being able to help people understand their risk factors and having them take accountability for their part in the play. So, if you think about some of the tough conditions such as cardiovascular disease, diabetes and stroke, patients can have an impact on those through managing their own health even better; for example, let's take a patient who's a bit overweight and has high blood pressure. For many people, if not most, those risk factors for metabolic syndrome are manageable with lifestyle changes. We're talking about what you eat and how much you exercise; that's the hardest part. We need to help them understand their risk factors and deal with them on a daily basis.
Other challenges arise as we look at the sheer volume of information that's coming through our monitoring systems. If you think of the imaging data that comes from MRI scans, those are huge files. To the extent that we're dealing with structured data - things like healthcare visits, procedures and medications - it is relatively easy, but doctor's notes are increasingly becoming free text electronic notes, which makes it harder to perform a query.
We have to somehow put structure round that unstructured data, and you've got to use new technology to dive into this world. It's often called natural language processing - taking the written word, and turning it into real and useful information.
You recently oversaw an important big data project on metabolic syndrome. What is metabolic syndrome and why was it chosen for the study?
The metabolic syndrome is a collection of five factors: large waist circumference, high glucose, high triglycerides, high blood pressure and low HDL cholesterol. When you're out of normal range for three or more of these factors, you're considered as having metabolic syndrome. Essentially, this puts you at a much higher risk for things like stroke, diabetes and heart disease, and the end result of that usually ends up being a shortened lifespan. It costs about 1.6 times more each year to treat a patient who has metabolic syndrome than one who does not.
That's part of the cost challenge, and then there are lifetime costs if you end up with diabetes - imagine having to give yourself a shot in the leg or arm every time you take a meal. The worst-case scenario is chronic kidney failure, which happens if you ignore your high blood pressure for too long.
Stroke is a wholly different debilitating problem, but you never know what the result of that could be - possibly brain damage, depending on where the clot ends up. I use these as the scare message for people who have these risk factors and don't understand what the overall implication can be.
Almost a third of the US population has metabolic syndrome, and it goes hand in glove with the obesity problem. Other parts of the world are experiencing the same thing, especially in the Middle East, where the figure is close to 50%. It's a significant problem for the medical community to solve.
Tell us about the research project and its potential impact.
We took 36,000 patients who had undergone metabolic screening for two years in a row. In metabolic screening, you get a blood test and a tape measure, and it tells you about where you fall in the ranges. We took that data and ran it through an analytics engine from the company GNS Healthcare, and by performing analytics we began to see some trends. This helped us identify predictive patterns at the individual level.
Let's say Mrs Jones is 46 years old with a high waist circumference and high glucose levels. We can look at her data and say, 'In the coming year, you are 92% likely to get metabolic syndrome and your next factor is likely to be high blood pressure'. This allows us to put an individualised intervention programme in place, telling her where she should focus her energies and helping her manage her risk factors with her doctor.
For a lot of people thinking about metabolic syndrome, it's a little bit daunting, wondering how they should keep these factors in check. Obviously, you should try to keep them all in check, but if you've got two or more, you need to know how to avoid acquiring a third and how to get rid of the two you currently have.
In the case of Mrs Jones' high blood pressure, that's a very manageable component of metabolic syndrome. It'd be better if she lost weight - in the research we found that the number-one most important factor is high waist circumference (over 40 inches for males and over 35 inches for females). It's also the most difficult to reduce. The weight-loss industry is worth $66bn and rising every year, so it's a huge problem.
The second-most costly factor is glucose. People don't realise that they have high glucose and it begins to cause all kinds of trouble with their metabolism. Those two turn out to be the worst in terms of driving the other factors, although all five are certainly important.
What opportunities do you see for big data in the years ahead?
The healthcare world is making a transition to population health management. At the moment, patients go to the doctor when they feel bad, so we've really got a sickness system, not a healthcare system. Predictive models can help doctors isolate at-risk individuals and work with them to manage their risk factors. As we gather more genomic data, that will further drive our ability to make predictions and therefore intervene in people's healthcare.
This really is a new opportunity. We're beginning to see a lot more of this data captured electronically, and that allows us to perform analytics and doctors to practise even better medicine. We're going to be able to use this data to change the healthcare cost curve in the future.