I agree, I don’t think that is a good use of funds. Until we have some actual interest from the smaller universities, etc we should think about other paths. The NLP thing is still very interesting, but I’m not sure it has practical application in a production physician practice. At least not until it got some testing, might be worth spending some time on that as an experimental module?
–Tony
Seems like we need a consortium of universities who would use LibreHealth EHR, probably initially just for EHR training. At that point, we might be able to branch out with other avenues such as analytics, FHIR apps and NLP.
What do you think about sending out an invitation to universities to subscribe ($) to LibreHealth for the Winter/Spring semester? It would give us a reality check as to the actual interest level. I could offer a $500 scholarship to universities who lack funding. In case there is more interest than expected, how many universities could you host comfortably at one time?
What will it take to include the new reporting and activity log functionality into the NHANES version of LibreHealth EHR? More funding?
That sounds like a plan that can’t hurt. No additional funding is needed to get the reports online, just me having a bit of time to focus on it. Which I will apply.
Status Update LibreHealth EHR September 18.docx (17.8 KB) I updated the status of the project. Please look over before I send out and hopefully Sunbiz will get a chance to respond shorlty
Document looks fine.
Dr. Purkayastha emailed me separately about going for an interesting NSF grant to enhance a platform or ecosystem that largely revolves around LibreHealth EHR. I think I will send out a newsletter to the roughly 475 informatics instructors to see how many are interested in using/testing LibreHealth for the Winter 2019 semester. I’ll offer limited $500 scholarships for smaller schools. The goal would be to continue to gather important insights into the IT and EHR needs of informatics instructors. Does this make sense going forward?
Yes, sounds like a solid plan
I think my solicitation via newsletter will be a reality check to determine whether an educational version of LibreHealth is viable or not. The instructions are for schools who can pay to directly contact Tony. Those who seek grant support contact me. One concern is that some may wait till closer to the Winter/Spring semester before responding. Time will tell what the true interest level is and their price point. In the meantime, we should pursue the NSF grant.
I agree. it will be a very interesting experiment to see whether the small universities and colleges need this kind of support or want it.
Yes, I agree, this will be a good thing to understand demand. But like you mention, there is a caveat that there might not be an urgent need since the Fall semester has already begun.
It will still help us narrow down our partner institutions for the NSF proposal that we are planning to submit.
It would be helpful to know where we stand with the Toolkit, C-takes and other areas that rely on the EHR as a sandbox for education, analytics and R&D. The Dean at UNMC did take a very short blurb from me and gave it to a healthcare foundation at UNMC. They seem to have more money than most centers due to Warren Buffet and Berkshire Hathaway being in town. I don’t want to oversell or undersell the potential so any updates about where we stand would be appreciated
Another issue related to the above is whether we can find a better dataset than the NHANES data. I emailed Partners Health twice about the i2b2 data set that has been used for NLP exercises (no responses and the web site seems old). It appears that the data is deidentified from an Epic EHR and includes some of the same type of info we have from NHANES. It did not mention encounter notes but they must be there if they use this for NLP. More information about the database schema is here.
My question is, if I formally request a dataset from them (IRB approved) how hard is it to change horses? In other words, what would it take to upload i2b2 data for educational purposes, instead of NHANES? I do have money to contribute but have no idea as to the complexity. Thanks
It wouldn’t be any harder than the normal process all the vendors have to do to import real data from other EHRs. It depends a lot on the incoming format, but I typically estimate that process as a 10 to 15 hr task, with the usual caveats for unknowns.
In the process of trying to research the i2b2 datasets further I ran into an article suggesting that Docker also helps research. The abstract is here
If there is interest I can get the full text article
From what I can glean from a new email, they are releasing the clinical encounter data purely for NLP researchers to use internally. They will not allow the data to be external or public
I think that’s the same case with MIMIC III. It might not be available to be publicly available.
You’re probably right. Nevertheless, I have a phone call with UAB on Thursday since they are close to me and I have two connections there. Just trying to get some de-identified encounters we can match up with the NHANES patients to make the EHR more realistic and get us ready for possible NLP down the road