The Most Interesting Man In The World

25 thoughts on “The Most Interesting Man In The World”

1. This stuff is extremely difficult to get right, even with competent people and a centralised structure like the English NHS. The VA has a working EHR system it developed in-house, cutting out the large fees and rents paid to people like Cerner. Trump administration policy aims to maximise these.

   1. Just so that we're clear, I assume that the "these" you refer to are "the large fees and rents paid to people like Cerner."

      1. Of course. How could I possibly have been referring to "competent people"?

         BTW, The research benefits from mining large-scale computerised health records are not imaginary. I cited here one Ontario study showing a link between air pollution and Alzheimer's. Here's a bigger one, also on air pollution, using the entire Medicare population. These studies, and medical science generally, are not helpful to the Trump policies for fossil fuels; but the cat is out of the bag and Pruitt will not be able to credibly deny in court that air pollution creates very large health burdens.

         1. The Feds have been doing their damndest to divest themselves of those "competent people" for going on 40 years now. Those faceless bureaucrats who wrote the National Airspace System for air traffic control have all been replaced by contractors, and VA is well on the way to getting rid of the people who could have modernized their EHR system.

            There's a lot of talk about using "commercial off-the-shelf" systems instead of in-house-developed custom software, but once a problem reaches a certain size, even so-called COTS software becomes a custom system once it's implemented. If you already have a big custom system, it's not clear that there's any advantage to using a customized 3rd-party solution instead of just using your in-house expertise to upgrade what you've already got. Somehow this message seems never to get through to the people who sign off on plans for management, no matter how many times the commercial vendors fall flat on their butts.

            1. I spent 19yr working at a well-known extremely-large enterprise I/T software/hardware/service company. For much of that time, I was a troubleshooter of last-resort for large enterprise I/T system failures. I got to know my employer's software intimately. And I can state, without fear of contradiction, and categorically that binary-code-only enterprise software (COTS software) is the shittiest software on earth. We used to joke that one well-known product made a server-quality UNIX box as reliable as a Window 3.11 PC. There are many reasons for this lack of quality, and one of this is the one you state: "by the time an enterprise gets thru customizing COTS s/w for their needs, it's pretty damn custom". But that's not the only reason. For example almost all of this software is sold by relationship-sales. So even if the software sucks, as long as the salesman got the sale, he gets his commission. BTW promotion at these companies is usually from the ranks of salesmen, so …. well, you can see how the incentives work. Bugs are fixed by shipping "fixpacks", and those are installed by customers — so the cost to the vendor is minimal. [contrast with hardware, where a big error can mean {m,b}illions of dollars of machines being returned for refund.] The madness goes on and on — I'm just scratching the surface of the bad incentives at every level. Every level.

               So basically, COTS is no solution. OTOH, custom internally-sourced development really isn't either — the internal org that develops it, ends up as "a vendor with one customer, but customer with irreplaceable one vendor", and that means the vendor has the customer by the balls. The VA's system (VISTA) was apparently quite good, and it's a shame that it wasn't developed into an open-source product that could be used all over the health-care sector. My belief is, something like that is where things should go. B/c EHR isn't a profit-center, not a differentiator, for health-care. It's a cost-center, and should be ripe for commoditization. In software, such commoditization in the 21st century invariably leads to an open-source system becoming the de facto winner.

            2. VA VistA is in fact open-source and there are entities other than the VA that are using it. It's just that it's big enough (and in some parts, specialized enough) that you can't just plonk it down in a community hospital and start using it: you have to have developer expertise to customize it.

               I think the fascination with COTS by management types is they hear "off the shelf" and think of things like MS Word. There are a lot of applications for computers where everybody is best-off with some standard stuff (OSs and things like Office apps and even SQL servers and web servers are good examples) but once you're dealing with problems that nobody else has, you should hire people to write and maintain code for you. People who work for you that is, not people who work for some contractor you've hired, or for some software corp with a sales staff who work on commission.

            3. [ETA: Oh wait, I take the VISTA part back! I just searched, and found it. Nice! I'll have a look around.]

               Two responses:

               (1) I remember when VISTA got OSSed (via the FOIA! Ha! Ha!) and a couple of years later I went looking for it. Couldn't find it. Found several companies claiming (probably truthfully) to use it, but no place to download it, and certainly nothing describing how to setup and run.

               (2) I completely take your point, that you'll need a dev team to apply it in a hospital. My point is different: over time, and esp. with the network-ish nature of health care, you'd find that there's no good reason for differentiation. So a collection of hospitals could eventually fund a team to build an EMR system that they all could use, and that eventually they would find advantageous to get others to use.

               My understanding is that this is effectively how things work with EPIC (the big gorilla in the space) — EPIC installations are pretty interoperable (e.g. to move patient records from one to another) but not with non-EPIC systems. Of course, EPIC is doing this for profit, but there's no good reason it couldn't happen in open-source.

               Again, the thing that most surprised me, was that I searched pretty hard, and could find no shred of VISTA source out there. I'd be interested if you could point me to that, for sure.

            4. The FOIA VistA files are here.. That's just the FOIA sources, not productized at all I think. VxVistA, OSEHRA, OpenVistA, and WorldVistA are the projects I know about that use it.

               The go-to site for information on VistA internals (outside the VA) is hardhats.

            5. Nice! I still -do- stand by my comment about an eventual open-source platform for EHR. It's an obvious win. Unlike with banks, there's enough "network synergies" and lack of "competitive differentiation" that it makes complete sense to standardize the record-keeping. Whereas, I completely expect that there'll be -no- standardization in the data-analytics performed on those records — that's where the real money is.

               I'll have to ask around amongst my friends who work in this area, why VISTA isn't gaining more traction …..

            6. > I'll have to ask around amongst my friends who work in this area, why VISTA isn't gaining more traction …..

               I think the answer to that one is MUMPS.

               I think VistA does not have a good data model for transferring records from one instance of VistA to another. Because MUMPS.

            7. My link on the English NHS is to a short but well-referenced survey by the Parliamentary Office of Science and Technology, a surviving child of the American Congress' OTA murdered by Gingrich. It includes material on open-source projects, supported by the regulator.

               After the collapse of the gigantic top-down IT scheme that McKinsey talked Tony Blair's government into, the policy shifted towards bottom-up diversity with standards for interoperability, as pursued by Obama's HHS. Open-sourcing is encouraged but not required. There is a little information on Scotland, Wales and Northern Ireland, now administratively separate in health. Northern Ireland apparently has a system that works, for which they paid all of £9m.

               The legacy of the failed megaproject includes a universal "spine record" for everyone with an NHS number, issued at birth. There isn't much in it, but it's a basis for development.

2. Skimming the report, I found these gems:
   "Pharmacists, in particular, found the system difficult to use. They were working extended hours due to longer prescription order workflows. Pharmacies averaged fill times of 45 minutes or more for prescriptions that previously averaged 15 to 20 minutes. Pharmacists had to employ manual processes to fill orders due to interface problems. MHS GENESIS does not support National Provider Identification numbers or National Drug Codes, forcing pharmacists to do manual searches for medications to dispense. …
   Users often suffered from unacceptably long MHS GENESIS login times, particularly dental providers. Login times were inconsistent, ranging from 3 to 20 minutes."
   This software is properly described as a piece of s*t. 20 minutes to log into a program that doesn't work! I'm surprised the testers didn't mutiny.

   1. Geez, this is precisely the sort of penny-ante numb-nut idiocy that an open-source system would resolve. This ain't rocket surgery [sic] fer pete's sake.

      1. https://youtu.be/THNPmhBl-8I

3. Hard to tell if this is a Jared screw-up or just a typical government screw-up. The IRS is using code originally written 60 years ago and which few programmers even learn any more. How they update the code for the new tax law is anyone's guess, but they are terrified of the day when it no longer works at all. They could adopt a system that would take income information they have on us from payers and do our taxes for us (as in the Netherlands, I believe) but Intuit and H&R Block have fought that off.

   1. First, I'd say it's hard to tell if this is a Jared screwup or a big bureaucracy screwup - where "big bureaucracy" includes large corporations as well as government. Even if it is a "typical government screw-up" it was certainly aided and abetted by Cerner.

      More important, WTF is Jared Kushner doing recommending software vendors? Does he know anything at all about this stuff? The decision was either corrupt OR stupid.

      1. I don't see anything that says JK himself chose or recommended Cerner. Anyway, Cerner has a 40-year track record of successful installations. But custom enterprise software installation success is very dependent on customer competence. I've been through one or two attempts as a customer with an organization that tried to automate a manual system that itself was unworkable. No way that software can do that. You have to rationalize the procedures before you automate.

         1. Take a look at the first link in the main article for the source of the assertion that Kushner recommended Cerner. And see the full statement of his speech to interns, here: https://www.wired.com/story/jared-kushner-leak-fu… "There are good and talented people in government; they just lack direction."

            1. I reread it. It doesn't suggest that Kushner ever even heard of Cerner. I admit the thing looks fishy, but no evidence of a fix in the article, and if Shulkin and Mattis were actually consulted, it makes me less suspicious.

               “Dr. Shulkin and I talked about this issue, called Secretary Mattis, he sent over his top five people on the EMRs, and we got contract people from the VA on the EMRs, electronic medical records. We said, ‘Guys, we want a solution to some of the systems. It’s absolutely crazy.’ They came back in two weeks with something that made a lot of sense.”

   2. Maybe you should read Daniel Davies' 1-minute MBA.

      1. Indeed, it was worthwhile, though I don't see the relevance to my dilemma.

4. Anyone remember the Obamacare web site rollout?

   1. Yes. I also remember that nobody reportedly died from the glitches, and that they fixed it.

      1. There are many other dissimilarities, right? healthcare.gov performs a specific function that really only happens once a year, so difficult-to-test. That function was new, and had to interface with many, many different partners (states, insurers). By contrast, whatever replaces VISTA will replace a -running- system (testability). There aren't as many partners involved, and one presumes they were all involved in VISTA, so the data-formats and such are well-known.

         1. Oh thought of some more: a rollout of a new patient EMR system can by its very nature be done in stages — one hospital at a time. You can even double-up staff and have 'em enter data into the old system, too. You can run that way until the kinks are worked-out, and then roll out to more hospitals, etc. This is …. *standard practice* in large-scale I/T — find a way to roll out new function/behaviour/bugfixes to customers in waves. Whereas, that's *impossible* when you're offering a new service that pretty much only gets used for 60 days, then goes dormant again until the next year.

Comments are closed.