The man is brilliant, but he does make my head hurt!
Read this if you are up for your comfortable concepts of interoperability and standards being challenged, and be prepared to go away and do some thinking about it:
And as a pdf:
This looks very interesting. We are currently covering standards in DIGH701.
Iāll work to understand them and then visit this article!
Heās right that we need to adapt continuously to changing circumstances. And yetānot one mention of continuous quality improvement in the whole paper! Weāve known how to measure and assess variation in processes since 1931, and how to manage it since the 1950s. We mostly just havenāt got around to doing this in Medicine quite yet!
I think heās fundamentally incorrect in so blithely asserting that standards are āa workaround where entities cannot adaptā. There are layers upon layers. There are standards that we have nailed to the mastāquite appropriately; there are standards that change very slowly; and there are standards that are almost conveniences on the road to progress. We have standard physical constants that we would be insane to fiddle with; we have standards used for traceable calibration of things like mass and pressure measurement that might undergo minuscule changes (for example, the kilogram was redefined on 20 May, 2019), and then there are standards that are much more transient like how to care for a disease based on a model that describes the condition and its management.
In Medicine, if youāve finally set a āstandardā, then youāve established what is comfortably mediocre at present, and will be quaintly obsolete in a decade or so. Just a few years ago, the neurologist would have seen someone with a stroke the next day if they were luckyāand done no more than try to rehabilitate them or confine them to private hospital care; now we do clot retrievals and make the paretic person walk and the mute person speak again.
He saysā¦
Yet in open worlds, the common model between any 2 entities will likely include standardized and nonstandardized components. When faced with highly heterogeneous and dynamic environments, co-operating entities may need to find ways to interoperate with little standardization available to support them. In the complete absence of standards, interacting entities would be free to interoperate in any manner they wish.
ā¦ and seems to find this somehow desirableāperhaps mediated by āAIā. I think this is (a) almost always unnecessary; and (b) a recipe for disaster, especially if you invoke confabulating, sycophantic LLMs. After all, we are not fighting asymmetrical warfare here, we are simply trying to continuously improve Medicine! I think the following statement in his conclusion is entirely (and dangerously) speculative:
Machine learning should have a major role in adaptive fitting.
Finally, not all competing āstandardsā are of equal value. Once weāre measuring and comparing adequately, rather than brokering communication using complex and unreliable methods, we often need to admit that someone elsewhere is doing a better job, and change our practice and the associated āstandardsā. We need to embrace the scientific method, and the resulting natural selection of better processes and standards.
I think he may have missed more than one trick.
My 2c, Dr Jo.
2 Likes
Agree with your comments. He definitely seems to be developing an argument for AI/ML to replace standards. I also found his claim that we can conceptually ādecouple the concepts of interoperability and standardization.ā He cites Chomsky and the hypothesis of linguistic innateness to argue that intelligent entities can sort things out between themselves. This ignores the fact that this often results in misunderstandings and poor outcomes, and more fundamentally, that human linguistic innateness in Chomskyās work is essentially a biological standard that we are all born with.
5 Likes
A post was split to a new topic: āConverge or collide?ā A call to embrace 3 complementary standards - editorial in JMIR
Iāve done a huge variety of trades in my life, my second attempt at a career pathway was in mechanical engineering. Heavy Fabrication as a Fitter Welder initially, followed by a number of years working in fastenings warehousing.
Ideally, when you need something to stay where it is, you use a fastener and standards for these come in a series of very heavy textbooks, but a critical to keeping the big stuff from falling apart when you donāt, and in some cases to fail in a way that causes easily repairable way (Bolts are cheap, everything else is often not).
There are two main systems for bolt measurement, Imperial and Metric, and to shorten the long political controversies, Metric was build on a foundation of scientific understanding of the physical constraints of physics and chemistry, and the other, the length of 3 Barleycorns.
Standards used in health service measurement are equally either well grounded in research, or based on the political whims of the ruling sovereign at the time of creation. Or as becomes quickly evident in the example of Performance measurement used, the dubious attempts at applying Scientific Management Principles that attempt to take complex systems and view these as a series of Inputs and outputs that can be manipulated at a high level.
Put simply, this paper in its tone sees health workers as cogs in a machine to be measured, a bit yuk really.
The next section deals with āAdaptorsāā¦ 2.54cm in an inch, roughly, and as any engineer knows well, adaptors are often the most complex and frequent parts of a system to break. It is a compromise made where communication breaks down, and these become very costly. a great example of this is the Mars climate Orbiter launched in 1998, where due to a conversion error led to a complete mission failure (193.1 Million USD).
Conferance services and tools are then presented as a straw man on a slippery slope where we land on āits just too complex to get it perfectly right but maybe an AI can do it for usā. and the concluding remarks summarise this sentiment āJust because standards are everywhere and commonplace
does not mean we understand themā.
To draw back on my time in Heavy industry, āHealth and safety rules are written in bloodā, and just because you donāt understand why a system or process requires a standardised approach, does not mean that it is not important, or critical.
likewise, we cannot expect our poorly documented methodologies used on equally dubious datasets structured for abacus use will be accurately identified as critical, and subsequently culled due to a low inference during AI model training and compiling.
Perhaps, the better solution, would be for those pesky reviews of measure fit and performance get done and updated accordingly, but that costs money and timeā¦ two things in increasingly short supply.
Put simply, we need to move away from a patchwork of ill fitting measures based on the distance of the kings nose to his thumb, and bring our aging health systems into the 21st centuryā¦ and i donāt believe letting our future robot overlords do that is wise.
1 Like
Assuming that we are discussing healthcare information standards, interoperability is an outcome created by the conformant and consistent implementation of (preferably international) standards. To have any chance of widespread adoption, those standards need to be created and maintained by a large community representing all stakeholders. To quote a recent article by FHIR Product Director Grahame Grieve - āinteroperability is a team sportā. Therefore, why Iām sure that some architects will place the process (standardisation) and the outcome (interoperability) in separate boxes, they are tightly-coupled concerns.
4 Likes