[BeetleB]

If you employ a few hundred nurses, how exactly would you evaluate how well they show empathy?

You can't rely on asking the customer. When they're upset (they often are in these calls), they'll lean towards the negative regardless.

I don't know how well these AIs evaluate, but if they're even a little bit good, it makes sense to use it to screen for outliers, then have a human listen to those outliers and judge.

[kulahan]

"You can't ask people because the experience is so universally terrible they'll just tell you it's terrible" isn't really an argument against surveys, it just means you need more specific questions they'll be fired up to answer

[BeetleB]

Let me suggest the following: Ask the nurses if they want the customer to rate them.

A significant fraction of the calls they answer are patients shouting at them because of:

- Long wait times

- They don't like their doctor

- They don't like the advice they're given (sorry, but we're not going to book you as a high priority appointment if all you can tell me is you have a headache. Sorry, we're not going to prescribe a narcotic for a scraped knee.)

- Several reasons that have nothing to do with the nurse, but the customer will still blame the nurse.

[TheOtherHobbes]

Classic "1 star review because UPS lost my delivery" metric.

I'd guess most people have had a situation where there's a corporate problem, the support person you talk to literally doesn't have the tools or the agency to fix it, but then you're asked to rate their performance on whether or not they solved the issue, with no option to say "Actually they did their best but this isn't their fault."

[fn-mote]

> with no option to say "Actually they did their best but this isn't their fault."

In this case, the lack of such an option is obviously a flaw in the assessment system.

How to fix that? Major political issue, I suppose.

[johnnyanmac]

Ebay solved this decades ago. You can rate the seller and product separately. adding 2-3 ratings won't fix everything of course, but it would be a start.

I suppose it is political because these companies rarely want accurate assessment of their labor to begin with. They want any justification needed to lay anyone off at any time while minimizing legal liability.

[kulahan]

>Major political issue, I suppose

Can’t believe nobody noticed this was a joke

[kulahan]

I’m not sure I understand your point. These things remain true regardless of whether or not you find ways to improve your process. Do we just… not try to improve because there is noise to sift through?

[lostlogin]

> how exactly would you evaluate how well they show empathy?

How would you want yours rated? By someone you have communicated with, or some data centre somewhere?

[BeetleB]

By a knowledgeable/skilled person who listens to the call. (Which the AI solution provides).

I suppose you could do that with the survey as well. It'd be an interesting study to see which is more reliable.

[AlotOfReading]

   By a knowledgeable/skilled person who listens to the call. (Which the AI solution provides).

Can you point me to the information you evidently have about which models Kaiser is using? All I can find is that they're using innovaccer, which can use any of anthropic, openai, and meta models on AWS or azure. Even their published papers don't seem to specify a particular model or capability level, just "AI". For all we know it's a gpt mini or similarly cost-effective model that has the context awareness of a Labrador hearing the word "walk".

[BeetleB]

> For all we know

We don't know, so let's not pre-judge.

[gusgus01]

Or we listen to the experts who are frustrated with the system? They see the effects even if they don't know the AI model causing it.

[lostlogin]

> By a knowledgeable/skilled person who listens to the call. (Which the AI solution provides).

Are you saying that the AI is the same as a knowledgable/skilled person?

[BeetleB]

No - read my original comment.

[lostlogin]

> By a knowledgeable/skilled person who listens to the call. (Which the AI solution provides).

Sorry, I parsed this as claiming ‘the AI solution provides a quality of results the same as a human.’

Are you actually saying that the AI solution should provide a human with the calls it identifies as needing a human review?

[ClumsyPilot]

> You can't rely on asking the customer

It’s not even been five years of AI, and we’ve already arrived at the point where the human is wrong and the AI is right.

Mind you this is in an area where the benchmark is the opinion of the human ! So if the customer is saying you’ve shown enough empathy but AI says you haven’t, then you take opinion of the AI?

Soon we’re going to have a situation where the patient is breathing, but the AI says he’s dead.

[charlieyu1]

I mean the same KPI crap has existed for decades

[malfist]

Actually, you can rely on the customers. They're the only ones that can tell you.

[ak217]

Yeah, this whole discussion cracks me up. I have a number of direct experiences with Kaiser nurses. They repeatedly got into arguments with doctors in front of me, tried to countermand doctors' instructions, ignored their patients, and complained to each other about their patients while they were right there.

Repeated unprofessional behavior with no discernible change after trying to address it. My take is that the Kaiser nursing org has a serious discipline and customer (patient) focus problem.

[drknownuffin]

To one degree or another, this is endemic among nurses. It's part of a broader cultural element: nursing programs have entrenched a culture of nurses vs. doctors. There are literally questions on their licensing exam to the effect of "which of these orders from a doctor should you refuse to enact?" (rather than, say, "which of these orders should you contact the doctor to seek clarification on?" or some other collaborative take). Nurses are taught their job is to protect patients from physicians. Given they don't have the expertise to do that , the general result is more broadly a power struggle in the guise of patient care.

[fn-mote]

I think this is an unfair take.

The necessity of teaching nurses that doctors orders are not sacrosanct comes from the bitter experience of doctors giving orders that are wrong.

Asking for clarification is great, but doctors can be very reluctant to hear. The bottom line is that the nurse must not do certain things and the certification exam is there to make sure they know it.

Think of it in relation to the “anybody can stop the assembly line” part of quality control.

[ak217]

I agree (with the obvious qualification that there are excellent nurses out there who do a great job and don't let this stuff get in the way of helping their patients, some Kaiser nurses included). But I also see a marked difference in behavior and outcomes in other hospitals I've been to. Yes, there are still some unprofessional nurses in those networks as well, but judging by the outcomes, the hospitals don't let them do damage.

[lostlogin]

> you can rely on the customers.

Patient or customer? I even struggle with that, but I guess that’s what people are in a privatised healthcare system.

[munk-a]

Most of the folks here on HN are dealing with customer feedback in systems automation in one form or another - it's pretty unavoidable in this age of LLM trendiness. The customers of healthcare (in both private and publicly funded systems) are the patients. So while the term might not be super natural it's an understandable one to use.

[datsci_est_2015]

Patients are not customers, or at least I don’t want to live in a world where patients are considered customers. Customers and vendors are usually more of a symmetric relationship: price transparency, alternatives, lack of urgency. These are all characteristics of transactions that healthcare often lacks.

[markdown]

If the hospital is owned by private equity, the owners definitely think of the patients as customers. Doctors and nurses shouldn't, but the owners do.

[Quekid5]

Lack of empathy tends to trickle down, alas.

[fn-mote]

The customer has more power than a patient.

I definitely want to be a customer.

[deejaaymac]

Customer, user, patient...it's all the same.

[groby_b]

But they don't want to, because figuring that out is your job as a supervisor.

If you outsource that work to customers/patients, you'll end up with the car dealership model, where the sales rep begs you to give a 10 on every single question including on the interior design so they don't get fired.

That's the part most of this discussion misses. Supervisors exist for a reason. Congrats on your flat org structure, you fucked up an important feedback channel.

[BeetleB]

> where the sales rep begs you to give a 10 on every single question including on the interior design so they don't get fired.

Oh yes, and the nurses did employ strategies like that pre-LLM (don't know if they still do). They had to be very strategic about it (you can't just say "Rate me a 10.")

[ImPostingOnHN]

Not asking the customer because you're afraid they'll tell you they're upset is a good indicator that you should do it more, and fix the issues.

You can ask the customer enough times that unreasonable customers or surveys are averaged out.

A good question might be "why are you upset?"

[weard_beard]

Isn’t it a running joke at this point that if you do what customers ask instead of focusing on the highest quality of service you get worse outcomes and the customer is still unhappy?

[gusgus01]

I thought that was more about correctly interpreting their asks, and not just actioning without consideration. If all your customers are saying your chair could use a cushion, it doesn't mean add a cushion, it means the chair is uncomfortable and you need to investigate why.

[ImPostingOnHN]

If all your customers are unhappy, then you probably aren't providing the high quality of service you think you're providing. After all, they're the judge, not you.

[kxrm]

I am quite heavily in AI, and I would say I am pro AI. However this use-case for AI is putting AI in the wrong position. AI should be in service to all humans. An administrator building out a middle management KPI based on AI is a misapplication of AI.

Hospital systems are incentivized to avoid the real problems with healthcare. People want timeliness and they want quality care which hospital systems are not incentivized towards in the US. The incentives are profit, which given budgets means corners cut.

Triaging is an opaque system to the patient. It's an important process to doll out finite resources but it also very frustrating to be told, "soon" when you've been waiting 15 hours to see someone. Frankly, if I were King for a day, the first thing I would do is break up the monolithic hospital systems and build out more urgent care.

I would also try to find a way to facilitate transferring less critical patients from ERs to urgent care centers. Right now a hospital won't take the risk, especially if you are sitting in waiting room because beds are full. You can't easily punt a patient because them leaving would be against medical advice.

[wisty]

There's a great academic book called "Stupid Rules". Excessive rules (or KPIs - often rules in disguise) exist because we don't like authority.

Get the doctor to assess the nurse. Or the head nurses if you don't trust doctors. The nurses have managers, and if none of the doctors or head nurses can be trusted with a simple matter like assessing whether nurses are doing their jobs then you got bigger issues.

Oh no, the boss might play favourites if it's not an objective measure! Oh the injustice /s

But stupid rules or KPI also allow favourites. You can use an officious 30 point checklist and play favourites while ticking boxes. You can even rig "objective" data by controlling other factors (e.g. giving someone difficult customers do deal with).

Yeah, data driven would be nice, if you have good data. But data driven is a power tool. You don't measure SLOC or reward token use in software because of perverse incentives.

[xp84]

When reading the headline I was thinking we were talking about evaluating things like whether a nurse asked the right questions of the patient from a best practices point of view (say you have <insert condition> and the best practices for that are to ask the patient about pain level and which side it's coming from and check in with them every X hours).

But evaluating tone and empathy? Great, now every nurse is gonna be wasting their time and energy making sure to recite the best canned, optimized text-adventure incantations for the KPI every time they enter the room instead of using their brains to see what the patient actually needs.

"Hello Mr. Smith our patients are our top priority at Kaiser and your nursing staff here at Kaiser Raccoon City are here to make sure you are cared for, comfortable, and safe. If you have any concerns or are feeling anxiety be sure to press the nurse button and we will be happy to assist you, we appreciate the trust you place in us and are eager to celebrate your recovery with you." < nurse now realizes Mr. Smith has been choking and losing consciousness while she was reciting that spiel >

[BeetleB]

> Get the doctor to assess the nurse.

Definitely don't do this. I know doctors. I know nurses. Plenty of doctors view nurses as their slaves.

And besides, doctors aren't qualified. These are different roles.