Skip to content

Is Facebook a Predictor of Your Health?

March 27, 2013


I’ve written before about the power of analyzing what you write in order to predict things about your health.  As more and more ‘big data’ companies and projects get publicity, it is fascinating to see how rapidly the field is growing.  We are at the point where each one of us experiences the web differently when we open our browser,  as our clicks and other data available on our habits and preferences are constantly analyzed. It’s all designed to get us to click or tap the ‘buy’ button.  I’d like to see some efforts made to use the same approach to motivate us to click/tap the ‘get healthier’ button.

With that in mind, I was fascinated to read a recent article in the Wall Street Journal (the news appeared in other outlets as well) citing a paper published in the Proceedings of the National Academy of the Sciences that studied how using the ‘like’ function on Facebook can reveal details about your personality.  Researchers first collected a lot of information from participants using standard personality inventories and psychological tests.  They then looked at patterns of clicking ‘like’ on Facebook to see if they could predict any personality traits or other identifiers.

The results were startling.  The researchers found, for example, that ‘likes’ for Austin, Texas, “Big Momma” movies, and the statement “Relationships Should Be Between Two People Not the Whole Universe” predicted drug use.  But “likes” for swimming, chocolate-chip cookie-dough ice cream and “Sliding On Floors with Your Socks On” were part of a pattern predicting that a person didn’t use drugs.  It gets better.  Patterns of using ‘like’ accurately distinguished between democrats and republicans 85% of the time, between black and white people in 95% of cases and between homosexual and heterosexual individuals 88% of the time.

You might be thinking that this is both technology and science run amuck.  “Why can’t you just ask me?” you might say.  It is worth pointing out the preponderance of evidence that, if asked about your health, you will predictably exaggerate those facts that make you look healthier and minimize those facts that make you look less healthy (this is predictable for some other types of questions too).  This is called “social desirability bias” and the phenomenon explains so much of what goes on both in the doctor’s office and when we talk casually to our friends and family about our health.

At the Center for Connected Health, we’re conducting a study now where we ask participants to fill out a short questionnaire on their readiness to move their behavior to a healthier state, followed by a period where we track their actual success (as measured by activity level).  Almost to a one, we’re finding participants overestimate where they are at the beginning of the study and need to be downgraded after one month of tracking.

There are many other examples of this phenomenon in the literature.  I find it liberating to think that there is a future (not too distant) where we’ll be able to calculate a health profile indirectly from records of your online and mobile behaviors (not just likes, but texts, emails, GPS data, etc.).  The ‘likes’ study is a great example of how that might work.

I can just hear the voices of the privacy advocates, the volume of their objections growing as they read this.  Rest assured, privacy will be preserved.  This health profile might be shared with you only, before it leaks to anyone else, even your doctor or your loved ones.  And if you object because it feels threatening to have someone tell you a version of the truth that might be too hard to hear, rest assured again.  A message that falls on deaf ears is of no real value in improving your health.  The message has to be customized to your state of readiness.  That is something we’re working on too.

At some point, there may be implications for your healthcare premium costs in this sort of model.  Exercising your libertarian right to privacy may lead you to have to choose higher insurance premium costs.  That is if we can show conclusively that this objectively-derived healthcare profile is accurate and predictive of your costs to the healthcare system.  There is a growing tension between ‘big data’ analytics and privacy.  When it’s designed to induce you to hit the ‘buy’ button, I think the argument to guard privacy is an easy one.  When we get it so that the design will be to induce you to improve your health, both for your own sake and for the sake of society, I’m not so sure.

What do you think?

27 Comments leave one →
  1. Sarah S. permalink
    March 27, 2013 4:23 pm

    Interesting data, Dr. K, thanks for sharing! I’m curious about other trends coming out of FB and Twitter. Are these published in a centralized database somewhere?

    In regards to your last statement about a libertarian right to privacy being penalized: perhaps a system which rewards good behavior (ie positive lifestyle and health changes) as opposed to punishing the withholders would elicit greater support and quell dissenters. A reward could be a reduced premium, for example, but would only be accessible to people who allowed you to track them.

    The fact that personal information is so available to corporations but not physicians strikes me as unfair, too. Often these companies are trying to promote products that negatively impact our health (degree of capitalism and health seem to be inversely proportional!). Your suggestions of a big data healthcare system is a potential foil.

    • March 27, 2013 8:36 pm

      great thoughts and comments. the most interesting debate is regarding the efficacy of carrots vs. sticks. I have to confess that i tend to think of the ultra-private crowd responding more to sticks than carrots, but it is an empirical question.

      • Jim Hutchinson MD permalink
        March 31, 2013 2:13 pm

        Who is the “ultra private crowd?” I agree with Sarah that incentives outweigh the stick? Statistics are never 100%, so for those in the minority % you list, are then unjustly penalized.

  2. March 27, 2013 9:22 pm

    What I think re: “…induce you to improve your health, both for your own sake and for the sake of society.”

    Sacrificing privacy for the sake of the collective is the essence of Big Brother’s hold on society in Orwell’s 1984. Sacrificing privacy is probably a slippery slope that we don’t always appreciate.

    In the healthcare professions we can probably deal with the risks of this privacy/big data slope if we always maintain, as our guiding ethic, the ethic of guardianship versus commerce ethic. If the healthcare professions use private data/big data guided by the commerce ethic, to the exclusion of the guardian ethic, then we will move ourselves toward Big Brother’s society. I am concerned that the drivers of big data are most often driven by the commerce ethic.

    • March 27, 2013 9:44 pm

      Collective? The boogyman of the Faux News crowd. Sometimes you need a collective force to serve as a countervailing influence in the market. Sadly, America is all too lacking in this area. This is why we are forced to endure runaway costs in health-care that threaten the entire nation.

    • March 28, 2013 1:51 pm

      Dan: I like the way you frame it from an ethical perspective. It doesn’t bring me any closer to clarity on the carrots vs. sticks matter, but the framing is eloquent.

  3. March 28, 2013 1:01 am

    Reblogged this on lava kafle kathmandu nepal.

  4. March 28, 2013 6:56 am

    This has been happening a while eg. store cards predicting customers are pregnant

    http://www.forbes.com/sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl-was-pregnant-before-her-father-did/

  5. John Smith permalink
    March 28, 2013 12:45 pm

    Re: carrots vs. sticks and insurance premiums … The economic reality is the same whether the opt-ins get a “discount” or the opt-outs get an “extra charge”. The same pool of premiums still has to cover the same costs (and profit margin). So you can label it either way but the opt-outs will know they’re being penalized.

    This sort of use of data is a major reason why I don’t use Facebook or social media sites, why I regularly clear browser caches of cookies, why I seldom “register” to post responses like this, and why my supermarket loyalty card points to John Smith at a random address and telephone number. But I’m not kidding myself that I’m anonymous or in control. I’ve paid for a shopping cart full of goods with a debit card enough times that the contents of the cart and the store where I purchase them uniquely identify me even when I pay cash. And with enough computing time and bandwidth, some algorithm somewhere can detect enough similarity among my few web postings to tie them specifically to me and all my public records no matter what junk email address or pseudonym I use. I wonder might show up in my snail mail or the inbox of my junk email account next week in response to this post?

    Perhaps I’m one of the privacy crackpots, but the probability of getting me to opt into such a system even if it’s “good for me” is near zero. Of course, I’m not a member of the Facebook generation. Perhaps the aging crackpots like me will die out and “solve the problem”. But in the meantime, we’ll eat up a disproportionate share of health care costs as we enter our Golden Years.

  6. Paul Dattoli permalink
    March 28, 2013 12:59 pm

    Strangely, I was reading a Gartner paper about Big Data when this arrived. It discusses the pace of technical complexity and how accelerants are introduced from year to year further exacerbating the problem. IT analytics can be applied to help manage complexity by turning big data inward. Considering these new fields of research now that we can mine the data, I began thinking about the applications in healthcare. Often when I hear about mining big data, I think about genetics. Perhaps in the future as we monitor and / or control individuals wants and needs, we will also consider the genetic makeup of the individual so we don’t kill anyone ! Thanks Dr Kvedar – Can always count on you for the most interesting stuff !

    • March 28, 2013 1:52 pm

      Thank you, Paul. I appreciate your kind words of support.

  7. David Goldsmith permalink
    March 28, 2013 1:47 pm

    Regarding the shift in carrot/stick dynamics, I recommend checking out Jane Sarasohn-Kahn’s recent post on how CVS is now navigating this choice http://bit.ly/XvKMiX.

    CVS’s approach has been widely reported in the press and roundly criticized as too heavy-handed. For some, the company’s tactics are considered tantamount to punishing employees for not doing more to improve their health. As Jane points out, it’s a logical if not twisted extension of a health care system that places the burden on employers to bear the costs of health care for a highly disengaged employee population. And as she further points out, while the stick approach may not be the friendliest or least intrusive from a privacy standpoint, it is unfortunately all-too-rationale from the employer’s perspective.

    In this paradigm, we can all expect to see a barrage of new tactics and technologies aimed at predicting and monitoring our unhealthy behaviors and doing whatever can be done to get us to either change course or pay the price.

    • March 28, 2013 1:57 pm

      I agree, but is that such a bad thing?

      • David Goldsmith permalink
        March 28, 2013 10:31 pm

        No, I didn’t mean to imply it was a bad thing. On the contrary, I see a lot of these forces driving positive change — in terms of personal health decisions, workplace initiatives, systemic change, etc.

  8. March 28, 2013 6:33 pm

    Great post! Correlating the messages based on an individual’s social graph to a PAM score would be ideal. Is there a way to gather metrics on readiness to change other than a standardized PAM survey?

    • March 28, 2013 9:10 pm

      I don’t know of an automated way to gather, but it would be most interesting.

  9. Larry Diamond permalink
    March 28, 2013 11:37 pm

    PAM is not a measure of readiness to change but that of a person’s knowledge, skills, and confidence to engage and participate in health behaviors. Proper use of the PAM enables clinicians and coaches to meet a patient where they are at to begin a process of activation. Unlike readiness to change, PAM has demonstrated that all people are ready to change if you meet them at their level.

    Mobile health is no different. A Harris Interactive study uncovered that all of the PAM levels use interactive and mobile technology at equal rates, but uncovered the low activated do not use interactive and mobile technologies for health care. Why? the content is written to a single level. And it is chance to find the content appropriate for you.

    If we want to be able to use interactive and mobile technologies effectively, we must write content to the different levels of activation to meet people at their level just as Medica Health Plan of Minnesota has done on their MyMedica web site.

    PAM also will inform us prospectively if a person can be successful using telehealth technology. The PAM research has mapped over 240 health behaviors and the probability of each activation level being successful completing said behavior. Said a different way, asking low activated people to use telehealth technology with success is a bad bet. We need to activate people first and then give then the tools to build their knowledge, skills and confidence around self-management.

    Two factors we cannot get away from in health care, 1) 5% of the population consume 55% of the resources, & 2) low activated people account for the majority of the readmission to hospital. Our biggest problem in health care is low activation. As Bill Hewlitt said” You cannot manage what you cannot measure. If we can uncover via Facebook what PAM has already uncovered, great. But, PAM delivers the measure to patient engagement now, all we need to do is use it. If we do, the effectiveness of mobile health will be everything we all know it can be.

    • Catherine McNair permalink
      March 29, 2013 7:19 pm

      Hurrah! The key to changing “low activation” is first finding out what the individuals beleifs and attitudes are that are causing their actions in regard to health. Obesity isn’t declining one bit in this country even though “sticks” have long been applied to them i.e. social stigma (which in may ways is more painful than the level premiums could ever be realistically raised).

    • March 30, 2013 10:22 am

      Thanks so much for clarifying the PAM for our readers.

  10. March 29, 2013 8:33 am

    Data and healthcare – the last frontier?

    You buy a car, you get an owner’s manual. You have an array of indicators, lights, dealer follow up notices, all intended to help you maintain the “health” of your car.

    You have a baby, perhaps you get a congratulatory card in the mail from friends and family.
    The point, there is no “owner’s manual”, no guidebook, no indicators to alert the parents, the child, the adult as they “use their body” and it ages. The vast majority of the population simply make it up as they go. Figuring out along the way, when to engage healthcare, or a catastrophic event occurs that lands them in the ER.

    Yet, aren’t there a multitude of “on-board” indicators that can assist in managing health? Trended over a lifetime, these “indicators” can provide a reasonable alert for potential intervention. Unfortunately, the trend is the best indicator, not the occasional physician office visit a fraction of the population may make for an annual, or less, physical.

    An example. My wife’s BP has averaged 108/56 for years. If she were to present with 135/95 that would likely result in a suggestion to perhaps watch her diet, exercise a bit more, “we’ll keep an eye on that”. In reality, for her that is a significant shift from her lifetime average and might be an indicator that something has/is happening that may indicate further analysis is warranted. As with most things – the trend is key.

    Unfortunately, we have a passive culture relative to healthcare. We tend to take it for granted until something “breaks”. We need to transition to an active culture. We need individuals to become far more proactive in monitoring their health, trending indicators and engaging healthcare at the onset of an issue, not at the point of crisis.

    Two things are required, (at least). First, given we’re dealing with human behavior, incentives. Money talks. What CVS has done and many companies are doing is to make a financial connection between being “conscious of health” and the amount paid for health insurance. The second requirement falls to “us” in the industry. That is, we need to make it easy and convenient for people to routinely monitor their basic health indicators. With 30mm or so new insureds coming into the system, we certainly can’t have everyone begin visiting their PCP with greater frequency. So we need new points of “care” that are easily accessible and convenient for consumers. Health needs to integrate with life.

    This of course implies the aggressive use of personal health information. For some, that may be an invasion of privacy, a “big brother” concern. But I suspect as the trend of aggressively utilizing personal health indicators results in better quality of life, possibly longer life, (w/quality) and lower costs not just for the “person”, but for that baby delivered at the start of this post – the vast majority of people will enthusiastically engage and we will accomplish the cultural transition. (Of course, the “side effect” of aggressively managing data across the population is the very strong potential to better understand disease, refine treatments and develop new treatments/cures.)

    So I applaud the actions of CVS in their efforts to encourage proactive health management. We have a long way to go, but the path is actually quite clear.

    • March 30, 2013 10:20 am

      First, thanks for such a well thought out comment. I’m honored that you spent the required time and thought to add it. My only quibble is with the blanket statement that ‘money talks’. We have seen evidence (and I know of other evidence, i.e., PhD thesis work) that shows that monetary incentives are not always powerful and not for everyone. I suppose ‘everyone has their price’, but within the confines of an ROI that makes sense, I don’ think its a panacea.

      • March 30, 2013 1:56 pm

        Agreed – definitely an over generalization. Different motivations drive different people and some people will never be motivated/engaged. Ultimately we’re talking about a fundamental cultural shift – from a passive view of “health” to active engagement. The combination of an appropriate incentive and a simplified means of engagement/access that fits a consumers day to day routine has the best chance of addressing the greatest number of people.

        Long term, I believe we want to align the “level of care” with the level of need. This might traverse everything from organized self-service health monitoring, to retail based clinics, better use of PAs/nurses, PCPs and up the value chain as the level of care warrants.

  11. March 31, 2013 9:21 am

    Well stated. Again, thanks for your contributions.

  12. April 7, 2013 6:50 pm

    Reblogged this on HERO Network LLC and commented:
    Is Facebook a Predictor of Your Health?

Trackbacks

  1. Is Facebook a Predictor of Your Health? - The Doctor Weighs In | The Doctor Weighs In
  2. Is Facebook a Predictor of Your Health?
  3. What "Facebook likes" tell about you as an individual personality | TheFemaleGene

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: