4 ms·

> The problems are aversion to technology in places, unwillingness to adopt new (human) protocol/processes, entrenched proprietary technology, that does not adh

> The problems are aversion to technology in places, unwillingness to adopt new (human) protocol/processes, entrenched proprietary technology, that does not adhere to standardized formats. It is not. This is a classic example of assuming the problem exists within the realm of your current understanding. The stereotype of people in medical roles not wanting/liking/understanding technology is wrong. What we don't like is the god awful technology we are handed because the work hasn't been done to understand how we work. THAT technology is not worth learning or caring about. As someone who has rounded patients over many times, the problem is the extensive amount of information that is ingested into establishing a cohesive clinical picture of a patient, then continually adjusting the degree of confidence in numerous facets and levels of the case. Then it all gets dumped into the heads of the next group of people. So people in tech develop these programs to help track the information. Yet they fail to grasp that there are many situations where the conclusions are based on a web of contingencies that operate on disparate time series, are affected by a wide net of things not directly reflected in empirical data, and it is completely impractical for me to continuously update the intermediate changes to my conclusions just so your software can know what's going on. The more that people try to construct a UI that captures the complexity of what can be noted in a case, the less usable it becomes. Even then, they still undershoot the level of depth needed to reflect what doctors are tracking in their heads. You may think it's enough to have someone input the hematologic data as the results and their ranges. You fail to track that my interpretation of those results is contextualized by the fact that nurse Martha was on the floor and likely dropped the ball on how promptly the ABG was run, and that certain staffing dynamics affected how much weight I give certain subjective criteria. In rounds, a simple eyebrow raise can be enough to convey to the next shift a mutually known dynamic that affects how information should be interpreted. Asking us to spell all of that out in your "new technology solution" won't happen. It is a waste of everyone's time and social dynamics will prohibit it. I could go on and on with a variety of other examples, but I don't have the time or desire to do so. My point is that the tech hubris is wasted effort. While you sit there thinking you need to teach the "tech illiterate doctors" how to follow matters of process, we see you as someone that has been asked to develop a more effective tool, and you hand us a fisher-price screwdriver and insult us for not finding it useful. Like the other comment said, https://xkcd.com/1831/ https://xkcd.com/1831/

yoyonamite 3y ago
If you don't mind me asking: Do you see a solution, technical or otherwise, to this problem? My guess is there are limits to knowledge transfer that's difficult, if not impossible, to overcome.
- Zenzero 3y ago
  You are correct that the problem is an information bandwidth issue. The number of variables that must be considered for any given case are extensive and multi-layered. I don't have the time to get much in depth on this but off the top of my head: - There can be errors in our lab tests - The errors are test-dependent in degree and significance - Less than ideal handling or sampling affects different analytes differently - Errors in one analyte can affect others - The absence of deviations from normal in some tests can itself be an abnormal finding - Sometimes abnormal findings can be due to error or genuine in an unknowable way - The confidence of what constitutes "normal limits" is dependent on the underlying epidemiology of what is being tested for - There is no realistic way to have a unified data model for everything we assess - Much of what is interpeted/assessed gets reported as paragraphs of text - Missing data can either be untested or unavailable and it is not always clear which it is - What is communicated to you may be dishonest (such as drug seeking behavior) - As much as we have tried to standardize fundamentally subjective assessments (such as murmurs) they are still subjective - The MO is to push providers to the limit of what is safe in terms of number of cases and aggregate complexity - Putting humans in the loop of anything opens up to emotional/social factors that can influence interpretations/trust in various factors I could go on, but the point is that as a result of all of this, the most efficient route by far is passing information directly from one caregiver to the next since they are often operating on a platform of context (much from their training) that is difficult to replicate in software. The solution in my mind is lifting providers off of the low level data and activities. If we can close the loops on automating decisions in a generally agreed upon and reliable manner, it removes the need to discuss or think on the information fed into that system. Clearly that is a big task, but the realistic path forward in my mind. Harping on matters of process at the same level of detail is a waste of time and effort.
zelphirkalt 3y ago

yoyonamite 3y ago

If you don't mind me asking: Do you see a solution, technical or otherwise, to this problem? My guess is there are limits to knowledge transfer that's difficult, if not impossible, to overcome.

Zenzero 3y ago
You are correct that the problem is an information bandwidth issue. The number of variables that must be considered for any given case are extensive and multi-layered. I don't have the time to get much in depth on this but off the top of my head: - There can be errors in our lab tests - The errors are test-dependent in degree and significance - Less than ideal handling or sampling affects different analytes differently - Errors in one analyte can affect others - The absence of deviations from normal in some tests can itself be an abnormal finding - Sometimes abnormal findings can be due to error or genuine in an unknowable way - The confidence of what constitutes "normal limits" is dependent on the underlying epidemiology of what is being tested for - There is no realistic way to have a unified data model for everything we assess - Much of what is interpeted/assessed gets reported as paragraphs of text - Missing data can either be untested or unavailable and it is not always clear which it is - What is communicated to you may be dishonest (such as drug seeking behavior) - As much as we have tried to standardize fundamentally subjective assessments (such as murmurs) they are still subjective - The MO is to push providers to the limit of what is safe in terms of number of cases and aggregate complexity - Putting humans in the loop of anything opens up to emotional/social factors that can influence interpretations/trust in various factors I could go on, but the point is that as a result of all of this, the most efficient route by far is passing information directly from one caregiver to the next since they are often operating on a platform of context (much from their training) that is difficult to replicate in software. The solution in my mind is lifting providers off of the low level data and activities. If we can close the loops on automating decisions in a generally agreed upon and reliable manner, it removes the need to discuss or think on the information fed into that system. Clearly that is a big task, but the realistic path forward in my mind. Harping on matters of process at the same level of detail is a waste of time and effort.

zelphirkalt 3y ago