Robots, automation, and A.I. tech to replace nurses?

FurBabyMom, MSN, RN · May 8, 2018

I think we could certainly leverage and optimize technological utilization for easier or better work.

That said, some environments will not lend themselves to robots replacing clinicians. Procedure areas seem especially difficult to imagine this takeover due to the insane risks...a patient's life depending on a collection of machines - what happens when there is a total utility failure of either internet or electricity? It's happened before, and it's scary enough with humans.

We have robots at work. We have a "robotic" surgery program which might be better described using the term "robot assisted" because the surgeon has to control the robot and the robot is not utilized for the entire procedure. This seems to be very confusing to people, and maybe rightly so because it's kind of an abstract concept - even when you explain that we use this technology people outside the environment don't generally have understanding of "how" it works. We also have robots that are programmed to transport things - food, linen, trash, etc. They're kinda cool, but sometimes a pain and we've been given instructions for how to override them and turn them into push carts in the event of utility failure or catastrophic event.

brownbook · May 8, 2018

I think everyone knows jobs are changing due to technology.

If I had a crystal ball and could see the future that'd be great.

Educate myself, educate my grand children. Continue to save as much as I can for when jobs are obselete.

Julius Seizure · May 8, 2018

We're really only 15-20 years away, if not less, from a computerized clinician capable of assessment, diagnosis, and treatment that will be monitored by a human counterpart.

When that happens, my new job will be robot nurse monitor. I am down with that. It sounds like it will be easier on my feet.

TriciaJ, RN · May 8, 2018

So the whole thing about "Service Excellence" and scripting: is that so when we get replaced by robots patients won't notice? Now it all makes sense.

bgxyrnf, MSN, RN · May 9, 2018

Robots could certainly play a role... There are many non-value-added activities in my day which could be shouldered by robots... A few which immediately come to mind include:

Fetching water or meals or blankets, for example. Collecting urine specimens from rooms. A mobile pyxis delivering the med to me at the bedside. Fetching other supplies for me at the bedside. Repositioning patients. Tissue retraction for *those* Foley insertions (y'all know the ones I'm talking about). Helping to clean up patients covered in dried feces.

There are also value-added roles that they could easily fill including limited reassessments in some cases under the direction of the nurse. There are certain tasks that would lend themselves to robotics but which would (a) be pretty expensive machines and (b) probably face a tough time being accepted by patients... examples would include IM injections and IV med administration. The problem with med administration - beyond the difficulty getting many patients to embrace it - would be reliability. They certainly could be made to fail safe but there would be plenty of times when they would get stuck and require intervention by the nurse or nurse technologist or whatever it would be to get it to work. Heck, even IV pumps aren't all that dependable and they're pretty damn simple machines.

Ultimately, though, some of the irreplaceable features of nursing is human connection, compassion, empathy, gentleness, reassurance, and therapeutic communication... things which will be exceedingly difficult to design into an AI system anytime soon.

Interestingly, I just read an article which makes the case that what might ultimately be the demise of Tesla Motors is its high degree of automation because robots are pretty inflexible and can't troubleshoot well... and cars are a heck of a lot less complicated than humans, even setting aside human fears and other emotions.

missmollie, ADN, BSN, RN · May 9, 2018

If you know where I could hire a waitress robot, let me know. I'd love to send it in with a tray rather than make the trips back and forth.

SpankedInPittsburgh, DNP, RN · May 9, 2018

Really????

McDonalds can't even get their automated order system to work. Robot cars are running into each other. In the not too distant past Wall Street suffered a major meltdown because of automated trading. Where I work we can't even get the computer systems we have been using for years to run consistently much less make decisions on its own. I'll have returned to the ecosystem by the time Rosie the Robot takes over nursing. Can you imagine a glorified toaster oven doing the critical thinking that nurses do everyday & all the time? "Danger Will Robinson"!!!! Technology and gizmos are cool but still can't replace a human's brain. Help Yes!!! Replace NOOOOOO!!!!

bgxyrnf, MSN, RN · May 9, 2018

Speaking from the perspective of an engineer who has developed and implemented plant-floor automation I can say with some confidence that the primary thing that will keep nurses at the bedside is the shear variety of the tasks which we perform and the huge variation in the "work pieces" on which we perform them.

It is relatively easy to make a robot to do one or two things the same way every single time on the same size, shape, and material of work piece. It becomes substantially more difficult to design in the flexibility to accommodate a wide range of operational parameters (that is, do this same thing on every patient from a neonate to a 600 lb person... and be intelligent enough and properly 'sensored' to know when this patient falls outside of your range).

It becomes even more difficult when you try to have the same machine perform multiple functions... select the proper size bp cuff (we typically have five at each patient location), pick the right extremity (let's see, multiple lymph nodes removed from L upper, old dialysis shunt on right, possible DVT in R calf... OK, pick the left calf), place the cuff (from very big to very small), know that it's in the right place to give and accurate measurement, recognize that the patient is twitching which may yield bad data, encourage them to hold still, recognize that the patient is getting ticked because their HTN is causing the cuff to retry a couple of times and it hurts because it's up to 230 on the systolic, reassure the patient that it's (a) vital and (b) almost done, then decide whether that high or low number is reliable... maybe reposition the cuff and/or the patient... maybe pick a new limb... and then you're not able to get a pressure... is it because there's an air leak? Hm, where... the cuff or the tubing... or is the pump actually defective... troubleshoot that and do it all again... or is it that the patient is really so hypotensive that you just can't get a good reading... does that low number correlate with their mental status? Well, check their radial pulses... strong and equal? What about their pedal pulses? What about the cap refill? OK, that number isn't consistent with the clinical presentation... I'll grab a doppler and use that... The expert nurse rips through that whole situation in a couple of minutes... the robot is spinning around in circles with the lights flashing and smoke trailing out of its joints. And taking a BP is about as simple a task as there is... until you really start to think it through.

To the layman, automation seems simple; to the engineer, it's terribly complex.

And then who's going to invest in the development of these robots? The development costs will be extremely high if you're serious about actually replacing a person and demanding six-sigma reliability. And then the actual cost of the machines... easily several hundred thousand dollars for something fairly simple and much, much more if you want to be mobile, intelligent, adaptable, and reliable... like millions per unit.

And robots don't move around terribly well. They can follow set paths pretty simply but they really struggle to locate themselves and their actuators (the business end) without set references to work off of. I'm not saying it can't be done but what seems pretty simple from the outside is a remarkable feat of engineering... and then you want it to be reliable and not hurt people... any people... ever...

Edited to add: I failed to mention another impediment... layout... none of the facilities at which I have worked or currently work are laid in a way that would facilitate automation. Hell, we can barely get some of the beds out of the rooms without ripping things off the walls, knocking things over, and taking chips out of the door jambs and that's with three people trying to navigate. In many cases we can't even get a properly sized bed into the room nor can we even fit a suitable lift... and then we end up with so many patients whose body habiti (habituses?) that they don't fit the design range of the equipment...

I'll wager that I understand electromechanical automation better than anybody else reading this thread and... I'm not worried... at all.

It'll be decades before we see anything beyond the most rudimentary bits of automation (like passing trays or delivering meds to the nurse at the bedside) outside of a handful of specialty units or facilities. I'm much more concerned about the depressing effect on wages and working conditions being driven by the political climate over the last 20 years.

**NICU Guy, BSN, RN** · May 10, 2018

30 yrs ago, I did a term paper on the automation of the air traffic control system (first degree was aviation). Airplanes have auto pilot, communicate with the air traffic control radar so the aircraft knows where the other planes are in their area, and planes can land themselves. It could be easily automated. The problem is that when you completely automate the system and there are no problems for an extended period of time, humans become complacent. When the system fails, humans are slow to react effectively and people will die. There is no way to program every possible issue that comes up and the system may not recognize that there is a problem.

The same with nursing, there are a million issues that could possibly come up and it would be extremely difficult for a robot to recognize every one of the potential one in a million series of events.

DannyBoy8, RN · May 10, 2018

Pffft facilities won't pay for an extra nurse much less a bot that will surely cost a pretty penny. Didn't the Google car run someone over?
There will always be vacancies on the night shift on floors like Ortho/Renal.

Uber car ran over someone who stepped in front of it. Uber driverless car not at fault, just like a human driver.

DannyBoy8, RN · May 10, 2018

Speaking from the perspective of an engineer who has developed and implemented plant-floor automation I can say with some confidence that the primary thing that will keep nurses at the bedside is the shear variety of the tasks which we perform and the huge variation in the "work pieces" on which we perform them.
It is relatively easy to make a robot to do one or two things the same way every single time on the same size, shape, and material of work piece. It becomes substantially more difficult to design in the flexibility to accommodate a wide range of operational parameters (that is, do this same thing on every patient from a neonate to a 600 lb person... and be intelligent enough and properly 'sensored' to know when this patient falls outside of your range).
It becomes even more difficult when you try to have the same machine perform multiple functions... select the proper size bp cuff (we typically have five at each patient location), pick the right extremity (let's see, multiple lymph nodes removed from L upper, old dialysis shunt on right, possible DVT in R calf... OK, pick the left calf), place the cuff (from very big to very small), know that it's in the right place to give and accurate measurement, recognize that the patient is twitching which may yield bad data, encourage them to hold still, recognize that the patient is getting ticked because their HTN is causing the cuff to retry a couple of times and it hurts because it's up to 230 on the systolic, reassure the patient that it's (a) vital and (b) almost done, then decide whether that high or low number is reliable... maybe reposition the cuff and/or the patient... maybe pick a new limb... and then you're not able to get a pressure... is it because there's an air leak? Hm, where... the cuff or the tubing... or is the pump actually defective... troubleshoot that and do it all again... or is it that the patient is really so hypotensive that you just can't get a good reading... does that low number correlate with their mental status? Well, check their radial pulses... strong and equal? What about their pedal pulses? What about the cap refill? OK, that number isn't consistent with the clinical presentation... I'll grab a doppler and use that... The expert nurse rips through that whole situation in a couple of minutes... the robot is spinning around in circles with the lights flashing and smoke trailing out of its joints. And taking a BP is about as simple a task as there is... until you really start to think it through.
To the layman, automation seems simple; to the engineer, it's terribly complex.
And then who's going to invest in the development of these robots? The development costs will be extremely high if you're serious about actually replacing a person and demanding six-sigma reliability. And then the actual cost of the machines... easily several hundred thousand dollars for something fairly simple and much, much more if you want to be mobile, intelligent, adaptable, and reliable... like millions per unit.
And robots don't move around terribly well. They can follow set paths pretty simply but they really struggle to locate themselves and their actuators (the business end) without set references to work off of. I'm not saying it can't be done but what seems pretty simple from the outside is a remarkable feat of engineering... and then you want it to be reliable and not hurt people... any people... ever...
Edited to add: I failed to mention another impediment... layout... none of the facilities at which I have worked or currently work are laid in a way that would facilitate automation. Hell, we can barely get some of the beds out of the rooms without ripping things off the walls, knocking things over, and taking chips out of the door jambs and that's with three people trying to navigate. In many cases we can't even get a properly sized bed into the room nor can we even fit a suitable lift... and then we end up with so many patients whose body habiti (habituses?) that they don't fit the design range of the equipment...
I'll wager that I understand electromechanical automation better than anybody else reading this thread and... I'm not worried... at all.
It'll be decades before we see anything beyond the most rudimentary bits of automation (like passing trays or delivering meds to the nurse at the bedside) outside of a handful of specialty units or facilities. I'm much more concerned about the depressing effect on wages and working conditions being driven by the political climate over the last 20 years.

Maybe with your experience you can speak to the other ways 'automated tech' can manifest itself. Automation isn't necessarily 'robotic'. i.e. Machine learning algorithm for predicting preterm labor - this exists, not yet reliable, but getting there.

As for the investment side of this equation. Most tertiary medical systems have data science centers with billions in total funding. Our medical colleagues are running this show and developing what will probably be an entirely new 'specialty'. Nurses with an interest in healthcare, innovation, and technology need to get in the game. To date, Duke nursing is the only entity to have responded to the trend.

Sour Lemon · May 10, 2018

Uber car ran over someone who stepped in front of it. Uber driverless car not at fault, just like a human driver.

Yes, try telling that to the family. I believe at least one of them has already received a settlement. Another telsa crashed into an 18 wheeler:

"In a statement, Tesla said it appeared the Model S car was unable to recognize 'the white side of the tractor trailer against a brightly lit sky' that had driven across the car's path."

Another failed to detect a concrete median and that driver was killed, too.