Diagnosis percents

Right now, WebMD and Wikipedia are no substitute for visiting a doctor, but why? What’s missing? Either way, you describe symptoms and see/hear a list of potential diagnosis. Why do you even visit a doctor for diagnoses if he has no more information than you can find?

Doctors have one set of information that you can’t find on the web, and that makes all the difference. They know the probabilities of diagnoses given certain symptoms. If your symptoms are headache and dry mouth, he knows that almost all of the people who present with these have an ear infection and only 1% have a brain tumor. This is something you will not find online. Doctors acquire this information not from medical texts, but from collecting data in the field.

For some reason, this data is never written down. No self-diagnosis site you visit will offer it or let you use it. Instead, the best you can do is see a list of differential diagnoses, equally weighted, for your symptoms. They always start with something ridiculously benign like allergies and work their way up to a rare cancer. Maybe everyone is too afraid of the risk of deterring those who may have something serious from seeking help.

The information presented now is incomplete to the point of being misleading in suggesting that all the listed differentials are of equal probability. Who hasn’t looked up their symptoms and been at least a little frightened at having some strange tumor or rare disease? It’s also possible that more people are deterred from visiting a doctor by the grave differentials than those who are encouraged to go after searching and finding something frightening.

I would like a real Bayesian database created of symptoms and diseases which showed probabilities for each differential diagnosis as one checked or unchecked symptoms. For example, I may choose dry mouth and headache and see “65% - ear infection, 10% - pregnancy, 15% - tension headache, etc.” However, if I also add speech difficulties, the percentages would shift to “20% - brain tumor, 20% depression, 1% - ear infection.”

The database could even be improved by allowing entry of age, sex, and even previous ailments to further refine the percentages. The obstacle in creating such software is only getting a reasonable sample of doctors to make notes of symptoms, diagnoses, and a follow-up to determine if the diagnosis was correct.