markolschesky

Joined 13 years ago

Karma

585

Submissions

112

Member since

Feb 4

I lead up Data and Evangelism at Datica. I know a good chunk about health tech things. mark at datica dot com

View on Hacker News

Recent submissions

Comment · 7 years ago

At least in the case of subscriptions I looked at:

1) It's a continually updated file, so if you didn't subscribe you wouldn't get new data weekly/monthly. Likely the subscription aligns with the data refreshes for most products.

2) It's a one-time fee with a hefty cost attached (I saw some healthcare data sets that were $100K+). You are paying for that in its entirely and just have data rights to it.

Comment · 7 years ago

It's rough. We thought about switching to Microsoft Teams and we realized that so many decisions about our relatively young company were tucked in DMs and private channels in Slack that we couldn't leave.

Email is non-sticky (just take your emails/save them on client and leave) Files storage is non-sticky (just move your files from OneDrive to GDrive) Chat/Collaboration is sticky as mud.

Comment · 7 years ago

I agree that some folks often exaggerate the danger in HIPAA, especially for someone like OP with a relatively small operation. But, for companies with larger operations and reach it's definitely a non-trivial problem. Our relatively small organization has two people dedicated to compliance (and plenty of ancillary support) and goes through hundreds of audits a year. Not having a locked down well thought out solution, both technical and operational, can really put growth at risk in healthcare. Of course, that's not "HIPAA compliance", but it is "what it takes to reach scale in healthcare".

Comment · 7 years ago

Our company had early access to this product and I was impressed by it. Our company had built our own FHIR datastore and I can attest to the fact that it's a more complex endeavor than it seems externally.

The killer feature of the product is its simple connectivity to other Google Cloud products for ML/Analytics purposes. Being able to receive a large quantity of radiology images (DICOM) or clinical data (using tools like Epic Kit/Caboodle) and immediately _do something with it_ is pretty impressive and hopefully lowers the burden for innovators in the space.

Of course, there are other options if you are looking for them, namely:

1) Azure API for FHIR: https://azure.microsoft.com/en-us/services/azure-api-for-fhi... -> Focused a bit more on application-workflows currently than ML/Analytics. Also has an open-source version: https://github.com/Microsoft/fhir-server

2) HAPI FHIR: http://hapifhir.io/doc_intro.html Open-source library from the makers of the most popular HL7v2 parser library. We run a bit of this today and it works smoothly. There's unofficial commercial support (same creators, different effort) from https://smilecdr.com/.

3) Vonk: Made by a company that has focused alot on FHIR based tooling. https://fire.ly/vonk/

Comment · 7 years ago

It's not necessarily liability, but it is shared responsibility. Someone validating that they will take responsibility for parts of your stack can be very important, especially at scale.

Comment · 7 years ago

I love Mirth and the community that's built around it. Thank you very much for all the work you did on the project.

Comment · 7 years ago

HL7v2 was invented in 1989. It was invented to meet the needs of then intra-operability between existing HIS systems. HL7 itself is a open organization comprised of a variety of members from vendors to health systems to life sciences/pharma.

HL7v2's problem is that it is 30 years old and probably should have been replaced by something earlier. Then again, so should most X12, mainframes, COBOL, etc. etc. FHIR is good, but the challenge is that the incentives for interoperability are much lower than those for intra-operability.

Comment · 7 years ago

Is this not the full version? https://itunes.apple.com/us/app/microsoft-onenote/id78480155...

It's not quite as feature complete if you're not using Office, but it's still pretty handy.

Comment · 8 years ago

Of course. I mostly wanted to get something out quickly. Those are all a bunch of small snippets of ROS and Physical Exams which demonstrate many of the things which the article discussed, notably finding Dx, Medications and whether or not they were negative or not.

I didn't know that there was a database of notes like MIMIC! I'll have to check it out.

Comment · 8 years ago

The textbook answer is typically that there is exponentially more data in unstructured data, but I agree with you. I thought about the unmodified use cases for this tooling for awhile and the major use case that I could come up with was finding unknown interactions or side-effects for medication and treatment combination regimens. That's all tracked very poorly in discrete fields today, especially with the nightmare that is cross-organization problem list management. Most drugs are approved by the FDA with the understanding that there will be post-marketing research (read this: https://undark.org/article/fda-drugs-post-marketing-research...), which this could hypothetically streamline.

Comment · 8 years ago

If you want to play with this and you're looking for a decent sample exam note, I grabbed some pieces of standard looking notes (Physical, ROS, Hand exam) and tossed them into a gist here: https://gist.github.com/molsches/32fcec2499e95b5b23bc268800e...

Mostly impressive in how it parses the data and can find conditions and tests in the unstructured data. Handles a few things strangely, but I imagine that gets better over time as it continues to be trained.

I think for it to truly be useful it needs some layer of semantic data mapping to something to standards like IMO/SNOMED/LOINC/RxNorm etc but I could see that being where other companies build their "products" on top of AWS vs. AWS competing with other Healthcare ML vendors in the space.

Comment · 9 years ago

That's actually pretty common, especially for older physicians or high-volume specialists. There are also some companies working on solutions to automate this using in-room speech recognition technologies like https://iscribes.co/.

Comment · 9 years ago

De-identified data and data contributed by people (read, not Covered Entities) is not covered by HIPAA. Someone could build a database of procedure cost comprised of these data sets, but price transparency isn't so easy due to the nuances of the problem (See Castlight Health).

Comment · 9 years ago

> Sounds like they didn't buy the database.

RxNorm (the codes discussed) is wholly open source. You can download weekly and monthly updates for free. Just pure laziness/avarice. https://www.nlm.nih.gov/research/umls/rxnorm/docs/rxnormfile...

Comment · 9 years ago

The developer loaded the hard-coded data to evade detection and the PM oversaw the process where the data was validated against the government's test harness for Meaningful Use. I suppose you could argue that those people would have pushed back on the lack of ethics in the situation and halted the process to prevent their certification.