44123_en-US

liliconcom

Medical Open Text Data

Does anyone have demo dataset for looking at medical records in text explorer? The customer would like to mine free form medical record text to describe and predict Medicaid fraud.  Any datasets with medical open text data could work even if not for fraud. Thanks for any suggestions!

 

Olivia

Re: Medical Open Text Data

Olivia: Two other thoughts for you...do you have any contacts in your former role in SLG that might be able to provide some text? I'm thinking since Medicaid is a state run program at an operational level, maybe you have a contact in that space? They could maybe provide some text or at least maybe some survey forms where you could 'fake' some text that's a mix of fraudulant and legit claims.

 

I did find some resources on the CMS.gov (Centers for Medicare and Medicaid Services) site where there is survey forms that claimants can fill out for various aspects of their health care journey. Most is focused on providers. Unfortunately the surveys don't offer for free form text...but the questions might help guide 'answers' to fictional claims if you really wanted to create something. For example, here's a survey for hospital services:

 

http://www.hcahpsonline.org/files/2017-08_%20Survey%20Instruments_Mail_English.pdf

 

 

 

 

Re: Medical Open Text Data

Thanks to all for the ideas. 

Re: Medical Open Text Data

It is true that JMP Clinical can generate word documents that contain text on adverse events, medical history, conmeds, lab data, etc. However, getting the free text into JMP would be a bit tedious. A subject ID column is easily extracted out of JMP Clinical exmaple data set (Nicardipine), but it would be a copy and paste excercise for each page that is written for all of the subjects you would like to have text. So it is possible to generate with some time investment.

 

There is a domain called Comments (or CO to follow SDTM naming convention) that contains free text, but the example one I have is for animal data and there is not much in there (and could be a bit uncomfortble to show since it deals with rat reproductive parts).

Re: Medical Open Text Data

Olivia:

 

I'm pretty sure Chris Kirchberg trolls the SE space...but in case he misses your thread...you may want to reach out to him. JMP Clinical composes what are called patient narratives, which essentially are free form text documents autowritten by JMP C that contains all sorts of medical treatment, cause/effect, adverse events, etc. information. Perhaps he has some dummy/example narratives in exactly this fraud space because fraud detection is a huge issue in clinical trials. Not exactly Medicare/Medicaid fraud...but fraud in the medical space nevertheless. Or maybe something could be constructed as well?

 

Additionally, if you come up with something that works can you let me know? We have some big medical insurers in our patch that look for Medicare/Medicad fraud (mostly they focus on spend/treatment data...not words) as well and the text explorer angle is one I'd like to incorporate into demos if applicable.

Re: Medical Open Text Data

Unfortunately, there are no free text fields in there.

Re: Medical Open Text DataDid you look at Don's data from NY dept of health? The database might have free text fields.