Comment by legostormtroopr | Hacker News Reader

Comment by legostormtroopr | Hacker News Reader

Parent Full thread

legostormtroopr

The article quotes a doctor speaking about nursing care. This is as accurate as asking an automotive engineer about automobile repair. Same field, different skill sets.

Nurses are trained to identify trends across many (sometimes too many) patients. Sure its inconvenient if a nurse monitors your blood pressure or oxygen levels every 2 hours after surgery, especially when you are sleeping, and especially if you recover perfectly.

But if you have a post-op internal bleed that occurs during the night, when you are asleep and unable to let someone know you are feeling woozy (because you are asleep), that's when trouble occurs. Frequent observations mean nurses can identify when things are going south before you get to an emergency situation.

Blood pressure drops sharply over 2 hours, monitor every 15 minutes, see if it stabilises. If it doesn't raise the alarm quickly.

Same with audio on machines, that constant beep is annoying to you, but it means the machines are operating and you are getting your prescribed fluids - silence on a night ward is a sign something is wrong, and quickly precedes an emergency alarm.

> Same with audio on machines, that constant beep is annoying to you, but it means the machines are operating and you are getting your prescribed fluids - silence on a night ward is a sign something is wrong, and quickly precedes an emergency alarm.

Sure, if there is one source of beeps, then no beeps would be conspicuous, but when there are many things all beeping, having one of them stop beeping may not be so obvious at all.

A simple technological solution seems appropriate here. Have each machine that needs to work check in every couple seconds with each of two centralized monitoring machines. Have each monitoring machine alert the nurse desk if a monitored machine stops checking in. This is silent and has no single point of failure.

To me the beeping also seemed like a quality backwards solution where much better, seemingly obvious solutions exist. Reality is probably much more complicated though. In a solution like you described and I had in my head we'd need a agreed upon protocol between IV drip and monitoring device. Probably the same standard should be used by other, similar devices. If that doesn't happen we get vendor lock in. In addition we are taking about medical devices so everything has to be fool proof, audited, double checked and certified. So what might be cheap in a consumer device now probably made the IV drip an order of magnitude more expensive. That is, if we agreed on a protocol in the first place.

I’m sure you’re right, but being unable to sleep when you need to heal is bad. "It’s complicated" is not a get-out-of-jail-free card.

May I weigh in with an outside perspective? My impression is this is a cultural issue, i.e. Americans are very insensitive to noisy disturbances during sleep hours. Biggest examples for me are your freight trains. Deadly accidents? Sure, the only solution is to disturb sveryone in a 5 mile radius around each crossing. In my home country this is unthinkable. People get chewed up, sometimes called the police on, for even running the washing machine past 10pm.

I'll never forget the night I was visited by police in northern Norway because we committed the crime of ... haymaking at 11 pm.

To be fair - its a very suspicious time to be haymaking. Threshing maybe yes - but haymaking... A word of advice: Don’t burn the candle in both ends - you need your strenght to work on the jarls estate (unless you are on a tenth or headtax plan of course).

I don't get this? What is haymaking?

sex

Do you know what parts of the world is hay used as a slang for sex?

I don't know the reach, but many agricultural activities can be a euphemism if you put a significant pause in front of them. "Where are alice and bob? Oh they're out... (gathering fruit|sowing barley|haymaking|looking for lost sheep|etc)"

No, no. Literal haymaking. Cutting grass for hay.

I'm surprised you managed to fit the combine harvester in the bedroom

Ironically, in the US we have a saying: "Make hay while the sun shines."

This was in northern Norway, in June. The sun was defininetly shining, even though it was at 11 pm ;)

Since someone already guessed that you live in Switzerland, may I also ask if you are in an apartment building with neighbors above/under/around you? We have similar neighborly complaints in Austria where (at least in cities) almost everyone lives in apartment buildings and thus when someone runs the washing machine, you not only hear it, you also feel it. In the USA, I think most people in cities are living in individual houses and therefore things like running a washing machine at 10pm would probably not even be audible by any neighbors.

We do, and yes, it's the most common form of housing, including upper scale units.

You must be from Switzerland... :-)

bingo ;)

Don't many Swiss buildings have shared washing machines in the basement?

> In a solution like you described and I had in my head we'd need a agreed upon protocol between IV drip and monitoring device. Probably the same standard should be used by other, similar devices. If that doesn't happen we get vendor lock in.

And a vendor lock-in is the goal of the vendor, both in medical and consumer space! Which is why such protocols rarely happen in either.

The medical space has a surprising number of widely used protocols. If you’ve ever gotten a CD with the raw data from some scan, it probably had a crappy proprietary viewer and the raw data. The raw data is almost invariably DICOM, and there are plenty of open-source viewers.

I suspect there are strong economic factors here. When a hospital buys a $2M imaging machine, the vendor is making money on the machine, not on ads, and that $2M machine had better interoperate with the HIPAA-compliant image archiving and distribution system the hospital already has from some other vendor.

> The medical space has a surprising number of widely used protocols.

There are also a wide range of interpretations of a single protocol. See the horror of HL7.

In terms of DICOM viewers, I’ll be the one who mentions Horos. If it doesn’t do what you want with a DICOM or have a plug-in that does, you are doing something quite unusual. Before the psych department come at me - I work in radiology. Functional imaging is best served with other tools, and I don’t understand them.

Can we not repurpose CAN for medical purposes?

Technically yes but seriously just don't. CAN is a horror show for security that can't handle a radio without being a security liability and medical privacy violations are expensive let alone wrongful deaths.

absence of signal is not a good indicator of operation. it is the same signal as total failure of your system.

Interestingly that reflects Airbus vs Boeing cockpit philosophies.

In the early 1980s Airbus adopted Dark Cockpit where the default state of all annunciator lights is off. If one illuminates then it indicates something worthy of attention. The colour of illumination indicates functional state; blue is good, amber is malfunctional.

When would something illuminated blue? When it's both working smoothly and worthy of attention?

I assume that's for indicators of abnormal, but safe, state. One example could be the undercarriage being down, a situation that would usually not be expected, but when you're in the process of landing doesn't warrant attracting your attention as an error state.

According to [1] for the wing anti-ice system, Blue = Turned on, Unlit = Turned off, Amber = Malfunction

Of course, one might very well argue that 'turned on' is not 'worthy of attention'

[1] http://www.smartcockpit.com/docs/A330-A340_Overhead_Pushbutt...

The nurse desk is a SPoF. If something goes wrong with that area, then how can you tell each monitoring machine is working?

The nurse desk is a desk. If the desk literally fails, you have serious issues, and your average multiple person desk is well enough engineered that there isn’t a single piece that can fail and take out the entire desk. But there are generally multiple nurses at the nurse desk, and there are often multiple desks, too. Also, you could hook the monitoring system up to the PA system.

edit: when I say hooking it to the PA system, I mean that, if no one acknowledges the light or computer message or whatever, then an announcement could play after a set delay.

Network failure in the desk room sounds like a complete outage for that.

How do you know if each machine is reporting in a consistent manner without being there 24/7 and making sure they beep all the time vs %40 of the time?

Failover to beeps?

Central monitors are quite common. There are even entire units based on it (telemetry).

> Same with audio on machines, that constant beep is annoying to you, but it means the machines are operating and you are getting your prescribed fluids

This article is a terrible treatment of the issue. The NYT article [1] or actual research papers are far better.

To try and sum up the actual issue that's being discussed: it's a tragedy of the commons situation, where the commons is a patient's sleep.

Every device and procedure priorities derisking above ALL else. In aggregate, this results in a huge negative impact on sleep.

The suggested change in behavior is to simply prioritize quality sleep somewhere about "irrelevant."

From the original research, a few points (from memory):

- Outcomes for non-critical patients are not substantially improved by the "every 2(?) hour vital check" regime. It's followed mostly because of legacy medical inertia

- There is little effort to batch interactions

- A substantial amount of sleep-disrupting noise is a modern hospital is non-functional

From personal experience at a good cancer research hospital in a major city, one omnipresent alarm was a low battery warning... for a device that was plugged into mains power. No one knew how to turn it off or fix it.

[1] https://www.nytimes.com/2018/12/03/upshot/why-hospitals-shou...

> There is little effort to batch interactions

The nurses are highly incentivized to batch interactions, just like waitstaff. What you don't see is that if the nurse has is working on the batch for patient A, and patient B crashes, then the nurse is too tied down to respond and calls in a colleague, who shows up to crashing patient B and tries to get up to speed.

Can't speak for all arrangments, but with my mother's post-surgical care, it was typically different draws for different purposes.

E.g. This or that specialist wanted to run a test, so they sent someone or had a floor nurse draw. But none of this seemed to be sync'd to the regular interaction schedule.

  > Same with audio on machines, that constant beep is annoying to you,
  > but it means the machines are operating and you are getting your
  > prescribed fluids - silence on a night ward is a sign something is
  > wrong, and quickly precedes an emergency alarm.

Sure, but that stuff should be monitored from the nurses' station, not by patients who are trying to get some sleep.

Seems like it’s backward to me. The machine should be silent if everything is ok and start beeping if things are going wrong.

That was my first thought too.

However, the downside of that solution is that the failure signal relies on correctness of the machine. In the current system, the failure signal is guaranteed to occur during failure with a zero false negative rate.

Monitoring over the network exists, and should be the preferred way to detect failure.

> In the current system, the failure signal is guaranteed to occur during failure with a zero false negative rate.

I can easily imagine software bugs that make the machine continue beeping even if the normal function doesn't work anymore.

Somehow, this sounds like an excuse for lazy design.

It sounds like an excuse for not networking at all really. Or upgrading equipment. Just a primitive cables to a data station on a cart outside the room would handle it better on every level than trying to figure out which thing isn't beeping when it should. Have it beep if it loses a signal. Better yet it can handle triaging far better when networked. Signal lost and "patient definetly not breathing" are two very different triage levels for one. You want to restore signal as soon as able but it isn't a mobilize a crash team to resuscitate situation. Nearly everything else has abandoned beeps as the sole indicator with good reason.

This makes me think about data backups. It might make sense to only get a notification if a backup failed, but then how do you know the backup system is actually functioning at all?

Regularly restoring from backup?

You can't regularly restore a patient from death unfortunately

How would you tell the difference between a machine that's working and a machine that's been switched off?

Make it play a quiet calming music to the patients taste

I understand where you coming from. However, would the patient know if machine is functioning correctly if it is silent? What happens if it's not plugged in correctly?

I disagree. If the monitor crashes, then its failure state is the same as "everything ok" state.

but what if the alarm isn't working when something is wrong?

so same as train brakes in a way

Trains slow down and use the engine at a slower pace. You don't have a back up heart.

their braking system is passive, no energy means error means brakes engaged; unlike, say, cars where no energy means .. nothing and any accident can occur.

How are cars relevant? I was specifically pointing out trains.

Reply on news.ycombinator.com