" they could get an accurate diagnosis between 6 and 32 per cent of the time"
I'll stick with tossing a coin.
Search engine results can be a useful predictor for cancer and can even beat doctors to the mark, according to new research from Microsoft. In a paper published in the Journal of Oncology Practice, a Redmond team detailed how, by analyzing data from Bing users who looked up symptoms of pancreatic cancer, they could get an …
Yes... the amazing revelation we can draw from this is:
For some people who's searches relate to $disease it may be that they are exhibiting symptoms for $disease and might therefore have $disease and have yet to be diagnosed, or even checked for $disease. On the other hand, they might not.
Exactly! And how do they discriminate between multiple users of the same computer? eg two or more family members looking up separate symptoms which together might indicate a cancer.
There's lies, damned lies and statistics. And then there's statistics derived from data with poor or no provenance.
Have you ever seen US TV?
It is full of adverts for all sorts of drugs. Some have really serious side effects but that have to tell you in the adverts about ALL of them so the voice runs at about 500 word/min.
Becuse Healthcare in the USA is really, really, really big business with billions and billions of USD at stake, even a bit of extra data about increasing (or decreasing) levels of a particular disease could be worth a huge amount of money to Big Pharma.
Sadly the error range in this study makes it as accurate as me putting £10 on the 16:30 at Haydock and expecting my choice of old nag to win.
"get an accurate diagnosis between 6 and 32 per cent of the time, with an error rate of 0.00001 per cent."
So if 6-32% of diagnoses were accurate, it follows that 94-68% of diagnoses were inaccurate. Interesting how this suddenly boils down to an error rate of 0.00001%. Some magic is going on here.
And how come they give a range (6-32%) for the accuracy? Did they split their cases based on some magic mystery factor to find that magic group 1 had a 32% accurate diagnosis while magic group 3217 had only a 6% accurate diagnosis?
I am well aware that 70% of all statistics are made up on the spot, but the numbers should still make sense!
Normally when analysing data like this you would split the dates into a series of sets si they may be referring to each set. Alternatively they may be using multiple algorithms. The accuracy refers to the false positive rate. So based on search results that aren't for the cancer they can give between 6 and 32% of people a go see your doctor now warning with only a false rate of 1 in 100000.
Yes, lung cancer is one of the things that a shoulder pain can be a symptom of. Deep joy. As a fully paid up hypochondriac I went to see the doctor to check it wasn't that. My null hypothesis (and the physio's when I went to see one) was that it was caused by the dog (large, strong, stupid) straining after hedgehogs.
It's worth noting that this is the same sort of bushwa that NSA & GCHQ are involved in - hoovering up gazillions of data points and pretending they can derive useful information from it.
Difference is, with this piece of nonsense, we can see the figures and draw our own conclusions.