Google Used To Diagnose Disease

dptalia writes "About 20% of all diseases are misdiagnosed, a percentage that has remained steady since the 1930s. However, scientists have discovered that by inputting the key symptoms into Google they can get the correct diagnosis about 58% of the time. For rare and unusual diseases, this provides doctors the information they need to get a correct cure. Of course, Google is only as good as its knowledge base, and its users, so this isn't a cure for everything."

20% error compared to 42% error of Google?

[epsalon]

What the blurb doesn't say, how much of the 58% google gets right overlaps with the 20% doctors get wrong, if at all.

Since we're using famous websites

[mnmn]

There should be a global wiki for medical professionals searchable by symptoms.

The contribution weight of better/senior/more respected doctors should be higher compared to new graduates. The wide open public should not be allowed to write, but should be allowed to read it.

This way better healthcare will be available in poor countries with Internet access, people will be able to double-check their diagnosis online and better doctors will be able to make a name for themselves the way CowboyNeal has.

I would prefer

[Timesprout]

If there was a publically available performance/competency grade for doctors online so I could just google for a good doctor in my area rather than hoping some med student hits paydirt with an 'I feel lucky search'

misgivings...

[MollyB]

I can think of two somewhat trivial reasons why this could be bad medicine:

Latent Hypochondriacs will type in some general symptoms and find that they have the dreaded newest and hippest malady. I foresee needless worrying and driven-up-the-wall family members.

If Google Bombs are still extant, what's to stop a special interest group from planting links to "cures" for wildly improbable scenarios?

"Caveat, surf-or" is never out of style, I s'pose...

Re:Gives you ideas

[kentrel]

Considering that the majority, if not all of Osteopathy is a pseudoscience and treatments like Bowen technique are unproven it's no surprise your doctor wouldn't recommend it. I'd question any doctor who would.

The article is full of hype

[badfish99]

The article says that the researchers found the correct diagnosis amongst the top 3 found by google in 15 cases out of 26.

In other words, they took a very tiny sample, and then cherry-picked the good results from the bad ones. There's no mention of any serious statistical analysis (why pick 26 as a sample size? why pick 3 results instead of 4 or 5?). And there's no mention of any "control" experiment (e.g. guessing the answer, or perhaps looking it up in a medical textbook). This is a classic example of how to fit the facts to the desired conclusion.

Re:20% error compared to 42% error of Google?

[Anonymous Coward]

What the blurb doesn't say, how much of the 58% google gets right overlaps with the 20% doctors get wrong, if at all.

The blurb isn't much to begin with - it is only 28 cases that were difficult to diagnose.

Even so, there isn't much information about the 28 cases. Were those 28 cases all misdiagnosed at one point, or were only 20% of them were issues? Also, how accurate are search engines on correctly diagnosed diseases?

The internet is useful in picking up diseases with a unique symptom, but is less effective if the disease's most prominant symptom matches with anthoer common disease. As an example, Vomiting and Diarrhea [google.com] may seem like something simple that can be waited out for a days. However, I turned out to have something a bit more serious - IIRC, it was Gastroenteritis [wrongdiagnosis.com], but it was a long time since I had it.

I'd post using my nick, but this is a bit into my medical history.

No...

[tkrotchko]

The write-up is a bit funny and misleading.

It's saying of the 20% that's mis-diagnosed, Google correctly identified 58% of those.

However, what no one has brought up is that when something is misdiagnosed, no one knows until they do the autopsy, so you can't just do simple math to lower the error rate to 8%. As you suggest, while google does better when the doctor is wrong, Google is worse than the doctor when he's correct. I'm not sure it's even correct to assume that if the doctor used Google the diagnoses would be better or worse, since there is an element of human judgment in medical practice.

What is does suggest is that doctors and patients should consider using Google to do a check on their patients and themselves for diagnosis and treatment options.

Re:20% error compared to 42% error of Google?

[i kan reed]

As long as it was a random sample, that still is perfectly valid with a sufficient margin of error and certainty interval. Bigger samples are necessary when you want to prove a relationship, not for demonstrating that there might be one.

This is a very, very important topic, by the way

[viewtouch]

This idea, that people can list their symptoms and answer questions about how they feel, and receive a diagnosis from a database, is an idea whose time has come. It offers the opportunity, for the first time, for people to have access to knowledge that they need to understand what might be causing ill health, pain and suffering. This opportunity also allows for people to provide information back to the database that can be used to improve it.

It is too often the case that our search for information about alleviating our ill health, diseases, disorders and pain is limited by the amount of money that we have to give to doctors and hospitals. It is too often that the doctors themselves are wrong when diagnosing the causes of our symptoms. It is too often that doctors fail to learn from the mistakes they make when attempting to diagnose ill health and diseases.

It is time for people to be given a mechanism to empower them in the search for good health, a mechanism that does not depend upon how much money they have with which to purchase the opinions of doctors, one which can be improved as it is used.

In virtually every area of human knowledge we recognize that software and databases are used to do jobs that no single person could possibly be able to do, be expected to do, to do these jobs better than, faster than, and at far less cost than any single person could do them. It's time to accept that this is also true of assisting us in understanding the meaning of the symptoms of our ill health, ill nutrition, pain and suffering.

There should be no objection by anyone to the idea that it is anyone's basic right to such knowledge, and that the Internet is the ideal method of providing this.

Could be a step in the right direction

[GrumpyGeek]

I read an article about eight years ago about an expert system written in Prolog that allowed doctors to select a list of symptoms, it would then ask about additional symptoms and then return a list of likely conditions. It apparently had a very high degree of accuracy (I think it was in the low 90's), and was vehemently rejected by the majority of doctors.

Hopefully the next generation of doctors will be so use to using internet search engines, that they won't feel threatened by a tool designed to help them diagnose a patient, not to replace doctors.

Re:Since we're using famous websites

[InternationalCow]

A mistake that patients and other laypeople commonly make is to think that their search is just as good as the doctor's. It isn't. An untrained individual (patient, curious person, whoever) using Internet resources to gather information about their real or perceived diagnosis usually ends up barking up the wrong tree. I see it all the time in my patients. I warn them about it and still they make this mistake. I deal with rare diseases, the 20% that are usually diagnosed wrong. Trust me, Google by itself or any other internet resource doesn't do you any good if you don't know EXACTLY what the key symptoms are. And selecting those is not something an amateur can do. So you can go on and tell me about how you corrected your doctor and beat the medical establishment and crap like that, but at least for the kind of disorder discussed here (IPEX and family) you as an amateur would not arrive at that diagnosis. And oh, Google is not the best resource for medical type searches. Try Pubmed, or OMIM, or if you're really serious (IPEX is an X-linked disorder caused by FOXP3 mutations) use the London Dysmorphology Database (LDDB). Amateurs should NOT, I repeat NOT, try to diagnose their own diseases. They simply lack the background to judge their own symptoms.