The bird’s eye view

With another bout of flu inevitably on its way – and a history of ‘non-starter’ epidemics in the recent past – a more efficient way of predicting outbreaks is needed as the world’s population continues to grow exponentially and bacteria become more resistant. Fortunately, Nicholas Christakis has found a way – by mapping the social network.

With a seeming dependence on the internet and social media sites in recent years, you could be excused for thinking that we've become meshed with our self-created, virtual world of binary - a world where the only virus you're likely to contract is on your hard-drive; where the volume of Facebook friends dictates your popularity; and where you can harvest fame and fortune with little more than a camera and a few clicks of a mouse.

Fortunately, we're not at the levels of this dystopian future quite yet. So for the mean time, continuing real-life interactions, building social networks and meeting people face-to-face will have to suffice. And while it does, our human social networks will continue to divulge information, make and break relationships, nurture ideas - and spread germs.

In the current state of play, the translation of germs and bacteria to potential epidemics is monitored and assessed by centralized bodies such as the Centers for Disease Control and Prevention (CDC). And while they have traditionally done all in their power to detect and warn the general public about potential epidemics, the time lag involved between detection and warning usually sees them turning up late to the party. Obviously a more efficient solution is needed, but how do you ensure that you not only have the ability to detect potential epidemics early, but also have the right knowledge in place to warn the relative social network that one is on the horizon?

Well, Nicholas Christakis, Professor of Medical Sociology in the Department of Health Care Policy at Harvard Medical School, along with his long-time collaborator James Fowler, has put forward a potential solution - and from his subsequent experiments, it's certainly proved its potential. Working out of his laboratory, Christakis' research engages two types of phenomena: the social, mathematical and biological rules governing how social networks form - dubbed "connections" - and the biological and social implications of how they operate to influence thoughts, feelings and behaviours, known as "contagion".

Test-run

In the tail-end of 2009, having realized that individuals near the center of a social network are more likely to be infected with a germ sooner during the course of an outbreak than those on the periphery, Christakis decided to work against the grain of traditional belief, which presumes that you require an understanding of a global network structure to predict an epidemic or outbreak. Instead, he took a far simpler route by monitoring the friends of randomly selected individuals, as they are known to be far more central within their network. To test the theory, Christakis and his team studied a flu outbreak at Harvard College.

Following 744 students who were either members of a group of randomly chosen individuals or a group of their friends, Christakis identified that the progression of the epidemic in the friend group occurred almost 14 days in advance of the population as a whole. More importantly, the friend group also showed a significant 'lead time' on day 16 of the epidemic - a whole 46 days before the "peak in daily incidence" in the population as a whole. In identifying the time lag between the two groups, Christakis and his crew had identified the difference between predicting an epidemic early and sweeping up the aftermath.

"The gist of the idea is that people don't just live in groups, they live in networks," begins Christakis in explaining the logic behind the theory. "This pattern of ties that we have with each other - real face-to-face ties, not online ties - means that some of us are located in the middle of a network and some of us are located on the edge of the network. People located in the middle of a network, on average, are more likely to get whatever's spreading through the network - and are more likely to get it sooner in the course of whatever it is that's spreading."

What Christakis was able to do was to show that it's possible to identify central people in the network without mapping the whole network, by using a phenomenon known as the 'friendship paradox'. In its simplest form, the friendship paradox explains that you are more likely to be friends with someone who has more friends and less likely to be friends with someone who has fewer friends. Whilst it sounds rather harsh, the reality is that we're socially wired to do just that, whether we like it or not.

"With that in mind, if you pick a random group of people and ask them to nominate their friends, their friends have more friends than they do. So your friends have more friends than you do, meaning that the friends of a random group of people are more central in the network than the random people themselves. This allows us to study and monitor a group of friends either actively or passively. And if you do that, you can gain an earlier detection about whatever it is that's spreading through the network, whether that be a germ, an idea or a behavior."

Alternative routes

The beauty of mapping the social network in this context is that it's not exclusive to a one-way input system; neither does it discriminate against non-epidemic topics. Surveillance can be either active or passive, which again will have a change in the input and output results of the network, allowing you to influence as much or as little as desired. To ensure that every possible angle is being covered, Christakis works in collaboration with MedNetworks, a company who deals solely with working on data interpretation and social network mapping from Christakis' findings.

"They're commercializing and have licensed some of our stuff from Harvard and they have a number of applications in both the sensor network ideas and influence processes," he explains. "They have a number of examples outside the realms of regular populations of people for other networks, a good example being doctors. So, if you want to look at the diffusion of innovation or you want to know the relative adoption of a drug in a population of doctors, then that's something that we could facilitate."

But surely the compliance of the most 'central' node will affect, and to a certain extent dictate, the resulting predictions from the relative social network? Well, not according to Christakis. "You could use similar ideas just using passive data collection - for example, mapping out email networks to identify people who are more centralized in a company. Again, the focus is very much on the sensor network idea, which is fine, but the important thing to realize is that, because human beings are interconnected in networks in mathematically sensible ways, it means that we can both deliver information to the network, trying to foster diffusion of innovation, workplace wellness, safety practices, innovation and creativity. We can intervene in the network to foster diffusion of desirable properties, and we can extract information from the network. They're two sides of the same coin. The sensor network idea is the obverse of the intervention idea or the flow of influence idea."

However, despite its potential in mapping and predicting disease in social networks, many critics are extremely weary of the latent dangers of putting such power in the wrong hands - and with good reason. As Christakis himself explains, understanding the dynamics of social networks and their implications when mapping them allows you to interact and influence the network itself, leaving it open to exponential abuse in the wrong hands. But as Christakis asserts, any technological advance has to face the same hurdle.

"Pick any kind of human technology that's been invented and it has the potential for evil, so I think the solution is some type of democratic control. You usually want the populace to be able to express its will as to how a certain technology is being used. So my feeling is that this new information technology has the promise of improving our society. But it also, unfortunately, has the peril of harming our society. The challenge is working out how we get the benefits without paying the cost. I believe that transparency is important and I think that a collective expression of will is important.

"We have laws about wiretapping and privacy. We have all kinds of laws that regulate what private corporations can know or sell about you and what other private individuals can do, so I obviously think we need a regulation of information that's available in this new computational era. You have to keep in mind that companies use this type of information all the time. Visa tracks who's purchasing what and where and with which bundles. Supermarkets know that you buy peanut butter and jelly, so they put them together on their shelves. The question is, how do we use the same kinds of processes to invent or deploy methods that facilitate the diffusion of desirable health behaviours or other desirable public policies."

And with the introduction of EMRs into the healthcare setting, it looks like Christakis could have just found his answer. As hospitals edge further down the road towards complete EMR implementation, security and standardization will become the norm, bringing with it the key to nurturing Christakis' social networking model. Where and when it will be able to jump on the back of the EMR is still unclear, but affording it the security it needs to work proactively should allow it a safe environment to do so.

"To the extent that we make anything electronic, we facilitate the ability to process data to learn and to acquire knowledge," concludes Christakis. "So, to the extent that the EMR becomes more prevalent, I think it will facilitate the use of data in the fashion that we're describing." And the potential for the same model in hospitals to map out HAIs? "Absolutely. The same kind of ideas could be deployed in hospitals and amongst other kinds of communities to get an earlier detection of nosocomial infections or other institutional outbursts. Absolutely."

For all the critics who claim that social mapping and subsequent influences on the network is another step in the direction of a 'big brother' state, there's little to say that will change their minds. But for the proponents of Christakis' work, realizing that having a tool that could predict an epidemic or outbreak significantly quicker than the current standard is just too strong an opportunity to ignore. And for Christakis, it's just too promising to drop.

MedNetworks

Ensuring that the work of Christakis and his team at Harvard is tested in real-time, MedNetworks was formed by Larry Miller, President and CEO of MedNetworks, and Nicholas Christakis to apply highly validated and fully functional technology to map, analyze and activate social networks across a broad range of constituencies including consumers, patients, health plan members, physicians and hospitals.

Using MedNetworks technology, networks, whether existing or created, can be mapped using either single or integrated multiple data sources - from administrative data such as medical and pharmacy claims data, to communications data relating to emails and phone traffic. MedNetworks thrives on a mantra of unleashing the tremendous power of existing social networks by using the "multiplier effect" to create changes in behavior out to three degrees of separation.

To do this, the company takes existing data from their customers, who already have extensive information on providers, patients and plan members, and maps out their relative networks. Once completed, MedNetworks then situates individuals within their social context, allowing for more efficient and effective targeting for information flows and behavioural changes.

In essence, MedNetworks is the commercial side of Christakis' experiments, with both working as part of a reciprocal relationship. Christakis' findings translate into solutions for companies and industries alike, while MedNetworks field work reports back to straighten out kinks in the system that Christakis and his team can then work to improve efficiency. What MedNetworks has proved is that understanding social networks through mapping them not only improves an understanding of one's business, but also the effects of intervening within that system - for both good and bad.