IBM Research: Health is a complex system -- and understanding it requires a complex system too

Tomorrow, IBM Research will kick off a new multi-year Grand Challenge project. The challenge is to build a platform that integrates multiple scientific models to create meaningful connections among different domains -- such as agriculture, transportation, urban planning and much more -- to better understand the ecosystem of influences on health, the health of individuals and the health of populations. The project will initially focus on the problem of childhood obesity.

The idea is that health -- not just healthcare -- is actually a complex system of systems in which agriculture, transportation, economics, family life, how we eat and exercise, medical practices and health insurance, advertising, and many other variables lead to interactions and outcomes that can have profound effects on populations. Yet today data and models from all these systems exist in different formats, coded in different ways, using different languages and different conventions. They are the domains of different people, experts in different areas.

It is straightforward to make 1:1 connections between these areas. For example, earlier this week a new study was published in Archives of Pediatrics and Adolescent Medicine concluding that the state of Oregon has the lowest rate of childhood obesity and that Mississippi has the highest. But they are not sure why. What are the underlying reasons for this difference? I mean, we know some of them -- we know there are connections between childhood obesity and processed food and too much TV. But that's not enough. There are many more influences and many more interactions -- and so solutions are largely a guessing game.

If one lever in the system is pulled, how will it affect the others? If one contributing factor is squashed, will obesity rates diminish or will the problem just pop up somewhere else in the system? A colleague of mine Dr. John Sterman recently had a fantastic example of these "unintended consequences." He told the story of an anti-smoking campaign in an African nation in which experts were pretty sure what they'd get when they showed teenage girls simulations of what they'd look like in 20 years if they did not quit smoking. Instead of quitting , the girls simply bought more makeup. This is a perfect example of complex systems of systems, of how human behavior and unintended consequences collide to create big questions when policy makers are trying to determine how best to spend a dollar. Should it be spent on a new grocery store or a new park for children? How will we get the most healthy result?

That's where modelling and simulation comes in. Last week at the 10th annual Almaden Institute, leaders in their fields gathered from science, medicine, education, government, food management and more to discuss the topic of Smarter Health through Modeling and Simulation, beginning to build the community necessary to create better health solutions. It was illuminating and encouraging to get the calibre and breadth of people in one room, sharing ideas and creating ad hoc alliances. But as exciting as the event was, building the community is only half the challenge. The other half is what my partner on this project, IBM Fellow Pat Selinger calls, "one of the biggest data problems out there." And it is more than data, it is models, simulations and visualizations -- ultimately integration of all these to improve our understanding of complex system interactions -- anticipating how actions in one affect outcomes in another.

So we at IBM Research are setting out to solve this problem of complex model integration. We're excited to work on it because we think we can make a difference. But we're not experts in all the individual domains and areas that relate to health. Our goal is to create a platform and capability for integrating data, models, and simulations by experts to address real problems and provide insights for real decision makers, ultimately building a community of stakeholders on all sides. At IBM, we have substantial capability in massive scale analytics, data integration, and computer science -- we think this puts us in a perfect position to solve the problem of integrating complex system models.