> the odds of the monitor detecting something in the intervening time and the intervention being correct are lower than the odds of the intervention causing some other unwanted side effect.

Huh. So that means, when the doctor decides to intervene based on what the continuous monitor comes up with, the interventions have negative expected value? Which means the doctors are making bad decisions about what interventions to make based on the data they have? I'll believe this is possible, but I want to ask to be sure.

I also would wonder about other explanations. You say "a nurse installs it, takes a reading, and removes it about once an hour"; presumably the nurse also glances at the patient and, if anything seems off, might ask the patient questions or take other appropriate actions. Could that be a significant effect? (In other words, to eliminate this potential difference, the better comparison for "continuous monitoring" would be for a nurse to come by once per hour and give the patient the same level of attention, perhaps going through the same motions that are involved in the monitor process.) Incidentally, as I read your comment, I expected it to conclude that the monitor itself or the process of repeatedly installing it and removing it was harmful (although that would point in the opposite direction).

It somewhat makes sense to me. Measuring all the time, you're susceptible to odd readings that are essentially false positives. If your device has a 1/1000 false positive rate, but you're usually only making one reading with it at a time, that's maybe acceptable. If you're making a million readings over a week because you're checking every second, than you're almost definitely going to have it show the same false positive enough that it looks real.

But the situation you describe only makes it look real if you were trained incorrectly, so that's a very important factor.

Yep, it's a form of the Base Rate Fallacy:

https://en.wikipedia.org/wiki/Base_rate_fallacy

No, they're saying that given a true positive, the intervention has positive expected value and given a false positive, the intervention has negative expected value (as there is nothing for the intervention to address in this case, the expected value cannot be positive).

The additional monitoring leads increase the number of false positives such that the overall expected value becomes negative.

I just went and checked and I was slightly wrong about the methodology. The comparison was actually between continuous electronic monitoring and manual listening with a stethoscope. I would hope/expect that a nurse is coming to visually inspect a laboring woman at least once per hour regardless.

This [0] is one of the studies cited by the book where I learned about this phenomenon. Continous monitoring was found to have a higher incidence of c-section and forceps deliveries. The only condition where continuous monitoring came out ahead was in detecting seizures.

[0] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD...

Is everything else really equal here? I'd think "higher-risk" situations would be more likely to have a higher level of monitoring. In other words, was this randomised?

Also from my experience the "left in place" part is technically very challenging, because a woman in labor tends to move around quite a bit, whereas the nurse holding something against the patient for a moment is much more workable.

..and the woman can hear the heartbeat, and when it gets different it can have a closed loop affect that causes unnecessary stress.

> Common sense would dictate that negative outcomes are reduced with continuous monitoring

This is also how people see constant glucose monitoring (CGM) and Type 2 diabetes. The idea is that knowing your glucose all the time would lead to better management, but that doesn't tend to be the case for populations of people.

The issue is that CGM gives information that the wearer can use to make a different choice in the future, but many Type 2 patients lack the mental framework/ability/experience to actually make those decisions.

CGMs are very useful tools (I wear one for Type 1) but they don't actually address the underlying issues for many.

A lot of negative comments on this and there are no references provided, however some human traits are well-understood:

* people irrationally tend to add rather than remove complexity to solve problems

* experts' decision accuracy improves faster with more information than their personal evaluation of / comfort with those decisions

* people asked to evaluate the contents of a picture repeatedly while being shown increasingly faithful renderings who are shown worse (unusable) renderings of the picture to begin with are slower to reach an accurate evaluation overall

As a general observation on medical care, Team USA has been slow to adopt the practice of weighing absorbent materials during procedures including childbirth to measure blood / fluid loss and this has had measurable, negative outcomes for patients.

I think we have too much cultural habits, marketing suggestions, etc etc around the whole notion of "more is better". So more complexity, more gadgets, more yada yada without ever going back and checking the premises and accuracy of these things and their claims along the way. We get stuck in these "dwell points of the ideal" where we assume because something is deployed that it's in some ideal operating state and take a bunch of stuff for granted at that point.

Looking at this from the other direction is interesting. Suppose hospitals had a way to monitor fetal heart rate with no possible side effects (e.g. no possible placebo or psychological effects from wearing the monitor). And they discover that intervening based on the results leads to poor outcomes, and they do better if they completely ignore the data except for one sample taken every hour roughly on the hour.

One solution would be to say “great! Let’s sample once an hour!” But surely one could come up with much better filters and heuristics.

Of course, doing this in an ethical way might be complicated.

I work in software A/B testing. We have similar problems.

The better filter and heuristics is using an algorithm that’s fit for anytime valid testing. The issue here is a high false positive rate. If you come up with all your treatment guidelines in an environment where you check every few hours and change it to every second, you’ve increased your samples by 3,600. Something like heart rate isn’t independent, but that’s still a lot.

You need to use a different algorithm (or algo at all) and/or relax the thresholds. Just random numbers, but if 45bpm was an alert before maybe it needs to be 50bpm.

Any references for this? I would like to read further.

Posted it above but I read about it in the book Expecting Better by Emily Oster. She cites this study https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD...

We're going through this as I type this comment.

My wife started having contractions and showing early signs of labor. Our nurse midwife gave her a tocolytic to relax the uterus, and recommended we go to the hospital to make sure that everything was okay if labor started again because she was 4cm dilated. The contractions had stopped by the time we arrived. The hospital hooked her up to continuous monitoring and determined that the contraction monitor (which is simply a pressure sensor on her belly) showed repeated contractions enough for many medical interventions that ended up doing significant harm (making a 95# woman carrying a baby not eat or drink for over 24 hours, take large doses of medication that gave her horrible side effects, and taking preventative antibiotics prior to test results coming in because of a fear of the baby's imminent delivery).

Her insistence that she wasn't having any contractions were ignored, and the doctors were visibly upset that we would contradict them and their equipment.

Later she told me she was about to fart again and to clear the area, and I noticed that her monitoring was showing a contraction. We called a nurse in and asked them to confirm, to which they did confirm that they thought it was a contraction. My wife let them know that it was actually a fart and they said they'd never heard of that before... Because I guess what patient is going to say their doctor is wrong because they have been gassy? It was a stroke of luck that I was watching the monitor when she told me that, otherwise she may have had an emergency c section as she was coerced into signing consent for that one. Not a good experience at all, and it made me completely disillusioned at the whole process.

The primary function of fetal heart monitoring is for pseudoscientific use by ambulance-chasing lawyers to generate $1M-$40M (!!) malpractice payouts from ignorant juries for children born with cerebral palsy.

The saying goes: "The only thing that can prevent a birth injury is a expert witness".

It's an open secret that the readouts from the monitoring during labor don't really mean anything by themselves. They are mostly used to make sure something is going on and that the baby is doing okay. Theoretically you can calibrate the monitoring to have continuously comparable results but in practice there's no point. It's less about the values or the patterns and more about a general trend. But in terms of measuring the actual process, dilation trumps every other metric.

Also contractions stopping when you go to the hospital is a well-known thing literally caused by the change of scenery. This is why some midwives recommend taking the stairs: not because the elevator might get stuck but because the pelvic motion of going up stairs while very pregnant can (re-)induce the contractions. Also while being stuck in one place because of continuous monitoring can mess with labor (which is why midwives recommend pacing, squatting, etc rather than lying down), the periodic checks can also be counterproductive because they can "pause" the process.

It's good that we have modern pre-, post- and natal care available nowadays but a lot of the medicalization of the birthing process actually makes it more difficult to give birth and doctors (and lawyers tbh) tend to err on quantifiable metrics even when everything is fine and the measuring is counterproductive.

(this was written as a reply to hospitalhusband's reply which has since been deleted)