> But overall the deviation from the statistically recommended procedure will lead to worse outcomes in aggregate. It has to, because if it lead to better outcomes, the doctors 'personalized' process should become the statistically validated process with time.
Yes, but not for the reasons you're implying.
Following the impersonal process leads to better outcomes in aggregate because it's less labor-intensive; and so doctors end up seeing more patients; and so more patients end up being seen.
But "number of doctors per patient" is only a bottleneck because we make it one, with medieval-guild-like limited-acceptance-per-year medical-bar licensing.
(And the medical bar does what it does, IIRC, because doctors really like being paid as much as they are; and having a more efficient market for doctors would mean doctors would be paid less. Even if it'd also mean that e.g. hospital doctors could work 8-hour shifts rather than 24-hour shifts.)
There's also the fact that flowchart-based diagnosis is liability-minimizing from a medical-malpractice-insurance perspective. With the flowchart, a hospital being sued can always use the defense that the doctor took the lowest-risk, best-practice next step they possibly could at each juncture. (Mind you, it's the best practice because it's the lowest-liability-risk — not for any other reason.)
> Measurement in medicine is also in many cases not particularly quantitative. E.g., fatigue, pain, cognitive dysfunction etc. So taking more measurements doesn't necessarily give you any useful information.
You only need quantitative instruments if your goal is to aggregate data to achieve statistical significance. An individual patient can be evaluated perfectly well ("clinical significance") with qualitative instruments. There are tons of objective-yet-qualitative instruments — e.g. observation of signs (rather than symptoms) of a syndrome, where there's no clear weightings to give to any given sign, but where one can always track each sign "dimension" of the syndrome separately, and then observe whether a given intervention improves outcomes along the "dimensions" you're tracking.
Following the impersonal process leads to better outcomes in aggregate because it's less labor-intensive; and so doctors end up seeing more patients; and so more patients end up being seen.
If the personalized process performs better on medical outcomes. Then over time doctors should be pushed to include more personalized steps in their treatment plans. I actually think that in many cases, personalized steps are already included in treatment plans that contain many branching paths.
It's also quantitatively true that in some cases (I believe most cases, but I don't have a good data set in front of me), a well validated flowchart leads to much better diagnoses and healthcare outcomes.
A good example: When a standardized checklist for the diagnosis of a heart attack was first introduced, doctors were upset about it, claiming that it took away professional judgement from clinical assessments.
Over time though, the checklist proved itself to be much more effective at diagnosing heart attacks than the vast majority of doctors. And most people who were having a heart attack were diagnosed correctly.
Of course some people will still be misdiagnosed (and unfortunately the misdiagnosed ones often fall into minority groups). But the statistically validated processes are not aiming for perfect. They're aiming for better than a human can do, most of the time.
And research will improve the flowcharts over time.
It also means that we can begin to think about delegating routine assessments where outcomes are very well understood to nurses, leaving doctors free to think about more complicated cases.
Yes, but not for the reasons you're implying.
Following the impersonal process leads to better outcomes in aggregate because it's less labor-intensive; and so doctors end up seeing more patients; and so more patients end up being seen.
But "number of doctors per patient" is only a bottleneck because we make it one, with medieval-guild-like limited-acceptance-per-year medical-bar licensing.
(And the medical bar does what it does, IIRC, because doctors really like being paid as much as they are; and having a more efficient market for doctors would mean doctors would be paid less. Even if it'd also mean that e.g. hospital doctors could work 8-hour shifts rather than 24-hour shifts.)
There's also the fact that flowchart-based diagnosis is liability-minimizing from a medical-malpractice-insurance perspective. With the flowchart, a hospital being sued can always use the defense that the doctor took the lowest-risk, best-practice next step they possibly could at each juncture. (Mind you, it's the best practice because it's the lowest-liability-risk — not for any other reason.)
> Measurement in medicine is also in many cases not particularly quantitative. E.g., fatigue, pain, cognitive dysfunction etc. So taking more measurements doesn't necessarily give you any useful information.
You only need quantitative instruments if your goal is to aggregate data to achieve statistical significance. An individual patient can be evaluated perfectly well ("clinical significance") with qualitative instruments. There are tons of objective-yet-qualitative instruments — e.g. observation of signs (rather than symptoms) of a syndrome, where there's no clear weightings to give to any given sign, but where one can always track each sign "dimension" of the syndrome separately, and then observe whether a given intervention improves outcomes along the "dimensions" you're tracking.