Googling yourself fitter?
Medical study needs health warning
Can you do science with just 26 case studies? Two doctors in Australia seem to think so.
"Searching with Google may help doctors to formulate a differential diagnosis in difficult diagnostic cases," they conclude. The study, published in the British Medical Journal today, appears to give the search engine a clean bill of health - and their cheerful conclusion has been gleefully reported in the popular press today.
On closer examination, however, we discover doctors Hangwi Tang and Jennifer Hwee Kwoon Ng used just 26 case studies. And it gets worse, the closer you look. Google only found the correct diagnosis 58 per cent of the time.
The "researchers" were also remarkably generous with their definition of a correct diagnosis. If one of the top three results returned by Google was correct, it was considered a success.
So Google was returning false diagnosis up to 80 per cent of the time. You might as well throw darts at a spinning dartboard, tied to the back of a drunken horse.
Yes, it's another of those pieces of research which start with a conclusion, works back to a premise, and then tries to pad out the bit in between (what used to get called "evidence") with garbage.
Still - a success rate like that must keep those waiting lists down. As the old joke goes, Doctors can always bury their mistakes.
(For each case study three terms that an expert medic might use were entered into Google - the researchers concede that the general public is likely to get much worse results from Google, not knowing the medicine.)
(Google is remarkably pleased with itself as MD to the world. Co-founder Sergey Brin likes to tell a story where Google saved a heart-attack victim's life. In fact the American Heart Foundation warns that the first thing you should do when you have a heart-attack is call the emergency services - not go wiki-fiddling).
The two authors of the "study" practice at the Princess Alexandra Hospital in Brisbane.
A place to avoid if you value your own health, we suggest. ®