Patient Databases For Symptom/Disease Mappings

Question

[aed83]

Hi all -

I'm trying to get a mapping between diseases and symptoms (with probabilities, i.e. given a disease how often does each symptom occur and vice-versa) by extracting visit information some medical outpatient databases. I found such a database on the CDC FTP website and managed to get something going by using Bayes models.

However, the entire dataset was about 200,000 entries, and only about 50,000 of them had both diagnoses and symptoms data - so for most symptoms the data is not significant. After a couple days of searching, I found a lot of inpatient databases that only had diagnoses but without symptoms.

I was wondering if you were aware of any other (public or fee-based) patient record databases that include both symptoms/reason for visit and diagnoses.

Thanks! AED

[sohum2002]

Can you please provide me with the database URL for the 200,000 data you have gotten? That would be great! Thanks in advanced.

[rajmayank771]

hello can you please provide me with your medical database?...

[agostino.forestiero]

Hi, can I have, please, the database URL for the 200,000 data you have gotten? Thanks in advance

[ATpoint]

Answer 1

[Alex Paciorkowski]

I suspect -- if you are in the US -- the only way to get the type of data you are after is to collaborate with clinicians who have access to their medical center's local electronic medical record db's for this sort of purpose (and have IRB approval to mine the data). I would be doubtful as to the accuracy of any mapping between disease and symptoms done in this way, though. Most user interfaces for patient record databases I've seen are so horrible that most clinicians log out as quickly as possible and enter the bare minimum of data required for billing. Better would be to work with clinical researchers who have an interest in a particular disease group, and are collecting high-quality phenotyping data for purposes of research.

[aed83]

Understood, but I would think there'd be some sampled data at least at the national or state level (in the US or in other countries).

Answer 2

[Pawel Szczesny]

Have you tried to contact vendors for Clinical Decision Support Systems? I believe they have the data you need. My impression is (although I haven't spend much time in this area) that medical records are too noisy to do any meaningful analysis.

[aed83]

Thanks Pawel - this is a good start!