Monday, April 22, 2013

Ignorance Isn't the Problem; Preconception Is

The dumbest thing you could do if you were trying to find the answer to a question is to ask someone who’s clueless, right?

Nope – it turns out you could do worse.  Hans Rosling of the Karolinska Institute in Sweden explains how.

Asked to teach a course on Global Health to Sweden’s top undergrads, Dr. Rosling administered a pre-test to see what the class already knew.  Five questions in the pre-test consisted of a pair of countries, and students were asked to select which of the two countries had a higher infant mortality rate.  To keep things sporting, he made sure that, in each pair, one country had at least double the infant mortality rate of the other. 

On average, the students scored 1.8 out of 5.  (In an amusing bit of self-deprecation, he confessed that this made him happy because now he had a job and his course had a place in the institute.)  The problem, he observed, wasn’t ignorance, it was preconception.  These top students knew statistically significantly less about world health than chimpanzees, who, he noted, would have scored a 2.5 answering randomly for banana bonuses.

Just so you don’t think this world-renowned researcher was picking on his students, he also performed what he called an “unethical” study on the professors at the Karolinska Institute.  These are the folks who hand out the Nobel Prize for medicine.  They scored on par with the chimpanzees with 2.4 correct answers.

Dr. Rosling told this story as an introduction to his TED talk, “Stats That Reshape Your Worldview”, which he delivered in 2006.  Yes, that’s right – 2006.  Why did I bother to blow the dust off of a 7 year-old presentation?  Because the chimp comparison is comic and the insight is brilliant.  As you read the news and watch world events unfold, you may begin to notice, as I have, how often false preconception, rather than ignorance, explains the ill-advised behavior that seems to be everywhere.

The other reason I chose to write about this TED talk is because current efforts to advance healthcare using big data reminded me of the kind of data wrangling I watched Dr. Rosling perform.  Out of the information mined from millions of patient EMRs, registries, and insurance claims, GNS Healthcare has created enormous datasets from an impressive list of collaborator companies and institutions.  The data is run through analysis algorithms that can then be used to predict whether a given treatment is likely to work for a given individual – one of a given gender, age, ethnicity, symptomatology, and genetic mutation.  PatientsLikeMe also performs data analysis to help patients assess likely outcomes of particular therapies.  The data they use come from subscribers of, a health data-sharing platform of 200,000 members suffering from over 1000 different diseases.  Both organizations are among a growing number in the industry that believe the way to less expensive, more effective healthcare can be found in data.  In fact, as members of Orion Bionetworks, a new alliance of private industry, non-profits, and research hospitals, they will be collaborating to develop predictive models for Multiple Sclerosis and other chronic diseases.

Dr. Rosling would be proud.  His TED talk has over 5,000,000 views, so maybe you saw it.  If not, here’s the link: . It’s excellent, and as far from a boring PowerPoint presentation as you can get. 

Jamie Heywood, the founder of PatientsLikeMe, also delivered a TED talk in 2009: .  An MIT grad, Jamie talks about the sophisticated algorithms used by the investment industry, and asks, “Wouldn’t it be great if the technology we use to take care of ourselves was as good as the technology we use to make money?”

Yes, yes it would.

by Laurie Meehan