Slashdot Mirror
Using Google To Help Predict Side Effects of Mixing Drugs
sciencehabit writes "Pharmaceuticals often have side effects that go unnoticed until they're already available to the public. Doctors and even the FDA have a hard time predicting what drug combinations will lead to serious problems. But thanks to people scouring the web for the side effects of the drugs they're taking, researchers have now shown that Google and other search engines can be mined for dangerous drug combinations. In a new study, scientists tried the approach out on predicting hypoglycemia, or low blood sugar. They found that the data-mining procedure correctly predicted whether a drug combo did or did not cause hypoglycemia about 81% of the time."
...seeding Google with false comments on drugs to produce something similar to the Santorum effect.
Donte Alistair Anderson Roberts - hi son!
Karma: Chameleon
81% is better than a coin flip, but 19% chance of dying (or worse) using Google Doctor(TM) advice is maybe not as good as a Real Doctor(TM).
Maybe they should try BigPharma(TM) commercials instead and find a reverse-cross-correlation?
It was hyperglycemia, not hypoglycemia, the exact opposite. Also, the research was helped by Microsoft, while Google is plastered all over the summary with nary a mention of Microsoft.
specifically data from IE about search queries to all the search engines
If correlated correctly, the entire sum of human knowledge on the internet is more likely to come up with a correct answer than an average doctor using the database within their brain. Gee, who would have thought it?
Of course the correct methodology would be to have doctors doing the Googling. I would hope that professional pride would not prevent physicians from using an established "expert system" to improve the outcomes for their patients.
(Googles "microsoft windows, caffeine, death". )
About 7,950,000 results (0.33 seconds)
Seems legit.
Koans and fables for the software engineer
...were results like drug x + drug y = anal sex
...the dumbest thing you can do. I can just see the guys at 4chan flooding forums with "omg i took 10 viagra and 5 xanax and my tumour is gone!!" to skew this data mining. There is no substitute for proper one on one medical advice.
I don't know how many people here have elderly parents but my mother is on over ten prescription medications and I've strongly suspected for years that some issues she's having are undocumented drug interactions. I've talked to several of her doctors and the best that they can come up with is "Sorry, there's no real data to support that theory / claim; nothing I can do".
There really needs to be a system in place to gather data on people who are on this level of medication and try and figure out if some of their problems that started after they were on combinations of medication are related.
So this is their Google scraping again, the one they pretended was an 'innovative search algo', but basically amounted to watched what their users typed into Google and recording the URL they went to from Google.
http://www.webproworld.com/webmaster-forum/threads/106763-Bing-Sting!-Caught-Scraping-Google!
I thought they'd stopped that after threats of a privacy lawsuit!? Seriously are you saying their recording medical conditions of their users and making that data public? Even if aggregated?
I thought medical data had Federal Criminal protection, not just privacy protection. As in, it would be a crime for Microsoft to intercept that data.
The only way stuff could be on google, is if said stuff already happened. ie: People having already suffered the consequences of dangerous drug interactions.
While it's certainly useful to have this knowledge, I fail to see how this translates to 'predicting' drug interaction. If people are googling for a combination of drug + a specific side effect, then they either are trying to find out if such an effect exists, or they have reason to suspect that such an effect exists, which *somebody* likely (and inadvertently) discovered that interaction the hard way.
Presumably the idea is that people experiencing reactions to drug combinations will go in the web and look for advice. And in so doing, we can use Google to pick up on this chatter and therefore determine adverse effects. So, the drugs are already on market and out there being consumed. Surely it would be far better to just have good old fashioned surveillance through clinic staff... less noise, etc. i don't see the advantage of using Google to reveal interactions in drugs that are already out there.
This research (AFAICT from readying the original article, but not the scientific papers behind it) is not about prediction at all, it is about discovery. Google and Bing are not "A Logic Named Joe," about to infer subtle connections between different medical research papers and predict troubles when two drugs are combined. They just uncover when people are searching on (and thus, potentially having) multiple sets of symptoms. The difference is rather crucial - if two drugs have lethal side effects when combined, Google and Bing could only uncover that fact once people started dying.
A corollary is that a medically themed advice engine could discover these connections even faster if it had the same critical mass (for example, once it started to see a pattern, it could ask people inquiring about symptom X if they also have symptom Y). Now, if you could only get that past the malpractice attorneys...
The sad, pathetic fact is you need to mine the data to find relevant information. Humans themselves don't have the time or patience to do this type of research manually, and it's an unrealistic burden that we should all be equipped with data mining and analytical tools just so we can use google's otherwise useless map of the web to find out, for example, whether prednisone can cause intense suicidal ideation when given in combination with fluoxetine. (Which is semi-standard FDA packaging warning for prednisone, but good luck finding official citations to this fact among your first three google search results pages...) if your results were like mine you got a dozen auto-gen'ed pagecopies full of chat bots in floating div's. The other dozen results, google lists strong matches in its page summary, but text containing the match is nowhere to be found on the linked result. How productive!
Trying to use google to do your own drug interaction homework is a total nightmare.
IUTBAP (I used to be a pharmacist).
One of the things we studied were drug interactions.
Basically pharmacists are more qualified to know these interactions, because they studied pharmacology, as well as pharmaceutics (how drugs are presented, what each form's method of spreading to the affected area(s) are, ...etc.)
The clinical pharmacy movement was started back several decades ago, so doctors would diagnose, then pharmacists would prescribe together with the doctors, taking into account many things, like other patient health issues, drug interactions, ...etc.
But, this will not happen, because big pharma markets to doctors so they prescribe their brand over the competition.
Sad to see the state of affairs status quo as they were in the early 80s.
2bits.com, Inc: Drupal, WordPress, and LAMP performance tuning.
Impairs your navigational ability because you're always pointing North.
Presenting now comments from actual drug users on this story:
"Uh ah uh uh ah sorry that your car? Cos I just clicked cos theres drugs in title I mean says drugs in title know what I mean."
Read the full article
http://www.nytimes.com/2013/03/07/science/unreported-side-effects-of-drugs-found-using-internet-data-study-finds.html?_r=0