March 29, 2026 • 40 mins
Health NZ have had to tell their staff to stop using free AI tools such as Chat GPT for writing clinical notes.
LISTEN ABOVE
See omnystudio.com/listener for privacy information.
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:05):
You're listening to the Weekend Collective podcast from News Talks.
Speaker 2 (00:09):
It be doctor, Do you mean the news? I gotta
bancas loading, you gotta made, you don't need nobody coming
(00:38):
from the start seeing come cute a little bit shame mother,
and welcome back to the Weekend Collective. I'm Tim be Everrgel.
Welcome in, Well, welcome back. This is the Health Hub
(00:59):
and we're going to be having a chat about uh, well,
help New Zealand. We want your call. By the way,
I know one hundred and eighty ten and eighty text
nine two nine two, but Health New Zealand have had
to tell their staff to stop using free AI tell
tools such as chat GPT for writing clinical notes. Look
chat GPT is. I've had some interesting interactions with it,
(01:22):
but I found it generally it's It's also an amazing
tool when you use it for the right thing. But
on its face, that is a kind of troublesome headline,
isn't it, Because you'd hope that, you know, doctors wouldn't
be relying on something that's sort of I imagine there
are AI tools in the background. But we'll dig into this.
But yeah, sounding hearing that they're writing the clinical notes
on chat GPT. Given it can hallucinate on certain things,
(01:44):
you'd be but a little bit nervous about that. But
that being said, of course it could be a great
thing for patients to use to their advantage. I have
used chat GPT because it's a long story and I
don't think I want to overly share it. But I
had a minor procedure of something removed and the scars
(02:04):
that have oppened up my GP well things, you know,
I'm not anyway, I ended up sort of monitoring it,
mostly in getting advice on bandages and all sorts of
things from chat gpt by uploading photos to it, which
was remarkably helpful until a dermatologist basically said, right, it's
now scay, you can just leave it. So yeah, I
(02:24):
actually found it quite useful. And I imagine my guest
is possibly horrified that I've done that, but anyway, he
is a GP dr, Brian Betty, and he's with us
for the Health Hub. Brian, Hello, how are you hey?
Speaker 3 (02:38):
Hello to Nice to be here.
Speaker 2 (02:40):
Yeah, I mean it's an interesting thing, isn't it? The
old AI Because the headline that people are using it
to summarize clinical notes I actually be honest, it didn't
upset me that much because I've used it to sometimes
sum up something I want to say and say, oh,
look just to summarize this for me, and it goes
does a bloody good job.
Speaker 4 (03:01):
Yeah, Look, you're right. I mean, AI is a credibly
useful tool that can be efficient, and in fact we
do use it in the clinics and general practice.
Speaker 3 (03:11):
We use it.
Speaker 4 (03:12):
I think the issue that has arisen here is the
fact that they're using the FREEGTV, not the accredited tools
that are specifically designed for medicine. So now the problem
with using the free chatchy PT, which is what's happened here,
is that we know that about ninety percent of what's
(03:32):
on the Internet has some sort of inaccuracy in it.
So if you're collating information from that source, you quite
like to have in accuracies in terms of what you
document or what you do. Now, the accredited tools, and
there's one called Heidi which has now been run out
in emergency departments and we actually use that in general practice,
(03:53):
is designed for medicine, so you don't run into that problem.
So I think the issue was being very very careful
about the tool you are actually using and that's what's
happened in this case.
Speaker 2 (04:07):
Heidi is that as in the Swiss Alps sort of Heidi.
Speaker 3 (04:11):
It's an unusual name.
Speaker 4 (04:13):
It was developed in Australia and it's called Heidi for
some reason. But it's probably one of the biggest AI
tools that are now used in medicine.
Speaker 3 (04:22):
So just to give you you an example.
Speaker 4 (04:24):
Of productivity, I use it in my clinic and it
probably saves me about two to three minutes of consult
in terms of writing notes. So if I see twenty
twenty five people a day, that's an hour of my
time that's saved that I can be doing other things,
either seeing more patients, or doing paperwork, or contacting specialists
or doing other stuff.
Speaker 2 (04:43):
And I guess it's about I mean, actually, you're right
about the free nature of chat GPT, So I pay
for it and it actually knows, it knows the GISTs
about things that I'm interested in and even my way
of communicating, so it's aware of the sort of language
that I use and everything. It's whereas a free version
of chat GPT that doesn't know you from a bar
AsSalt block a site really.
Speaker 4 (05:05):
And look, that's one of the problems with it, and
you gave the example of looking it up to do
some things.
Speaker 3 (05:10):
One of the problems. I've done research into.
Speaker 4 (05:12):
This into AI and it's benefits in terms of medicine
and patience. Now, what it's found is it's very very
good for things like health literacy or perhaps managing a
problem going forward. But what it is actually very very
poor at is diagnosing problems. So there's two issues you
run into it. Either gives you the wrong information and
(05:34):
it delays a potentially serious problem, or it actually over
eggs it and overdoes it and gives you what we
call hyper anxiety over the potential condition because the questions
are asked out of context, and that becomes a bit
of a problem with use of AI. So the research
is quite interesting into this. So absolute benefits, but there
(05:55):
are some drawbacks with it at this point.
Speaker 2 (05:57):
It's funny how names like Shakespeare said, arose by any
other name would smell as sweet, but there is something
about the name Heidi that makes we think, oh no,
it just is a slightly reassuring, wholesome sort of name.
Let's ask Heidi.
Speaker 3 (06:12):
Yeah. I don't know if that's why they chose it.
Who knows.
Speaker 4 (06:15):
Look there are other tools out there. There's one called Intelltech,
which is given so there's a variety of them, but
they're actually they're actually sort of you know, sort of
okayed by Health New Zealand. So these tools are verified
from privacy point of view, which is very very important
and also in terms of the way they work and
what they actually do. So yeah, but Heidi is I've
(06:37):
got to say an interesting name is is it?
Speaker 2 (06:40):
Also? Because there was something in the news a little
while ago about and it's just the questions around so
a AI. So you go and see a specialist or
a GP and AI is listening to the consultation basically
and taking notes as it goes, which I got to say,
I think sounds absolutely brilliant. There's there's questions around you know,
(07:04):
that information and going into the sort of into the
data sphere and all the private privacy sort of things.
But it is I mean, how do you use that
sort of thing?
Speaker 3 (07:15):
I use it.
Speaker 4 (07:16):
This is exactly what herd he is and that's exactly
what's been trialed and ED departments at this point. So
it records the conversation that will succinctly take out a
note out of that conversation. So the reason it's so
so important that these things are benuch marked by Health
New Zealand and by the experts is for those very
privacy reasons and things that you actually actually mentioned there, Tim,
(07:39):
So it becomes a really really important issue that these
tools are at the absolute privacy standard and that we
can use them with confidence. My other common to be
this that they probably get more accurate notes I've found.
Speaker 3 (07:55):
But look that this technology is with us now, we've.
Speaker 4 (07:59):
Got to learn to use it, otherwise we go back
to pen and paper and using the telephone at the
end of the day.
Speaker 3 (08:05):
And well, that's the thing.
Speaker 2 (08:06):
Medicine and technology goes hand in hand, because you know,
I'd rather have replaced on technology they've got now than
when they you know then it's thirty years ago or
forty years ago. I mean, the technology they've got for
analysis and MRIs and accuracy. But also it does lead
into that question abound robotics.
Speaker 4 (08:27):
And oh look, look, look I totally one hundred percent
agree with you on that I'd rather have have this
type of technology available today in terms of what I
can do in terms of accuracy, and with robotics, the
space we're now moving into potentially in a few years,
if you go in for your knee replacement, essentially your
(08:48):
knee will be positioned on the table. There'll be a
technician there, it'll be a computer program. All the machinery
will work automatically, and potentially yeah that the machine will
do the operation with a surgeon on standby if there
is any problems. This is potentially the space where now
moving into how is it.
Speaker 2 (09:09):
A tough sell for I was going to say, customers patients?
I don't know what the lingo these days is in
the health clients? What what? What are people these days?
Are they patients, customers or clients? It depends when what
year you finished your training.
Speaker 4 (09:24):
Look, look, in many ways, I still use the term patience,
and I'm quite happy to use it. And people can
call me a doctor if they want. That's fine. I'm
quite okay with that. Yeah, yeah, Look, I mean this
type of thing is it's it's interesting where this is
all going now?
Speaker 3 (09:45):
And it is it because?
Speaker 2 (09:48):
Is it because general is the acceptance of this technology?
Is it something that's simply going to happen as a
natural evolution as people and generations get comfortable in the
digital space. So obviously if you were to introduce robotic
to to a population who doesn't really know much about
the Internet and how technology works, they're going to be
(10:10):
very resistant to it. Whereas you know, twenty or thirty
years time, today's twenty year olds, who are fifty or
sixty year olds, you know, they're going to be far
more sanguine about it because simply that's the world they've
got used to.
Speaker 4 (10:22):
Look, look, I think you're absolutely right in that, And look,
I think it's actually in the last five to ten
years we've seen that happen. So there was a lot
of debate about using cell phones for messaging patients when
it first came in a few years ago, and I
know at our clinic there was a big discussion about
how many patients would use it, how many patient would
adopt it rather than ringing the clinic. And what we
(10:44):
sort of failed to understand was that every patient actually
had a cell phone and actually they really liked that
methodology of communication.
Speaker 3 (10:52):
So they don't have to ring the clinic. They don't
have to do that.
Speaker 4 (10:54):
If you've got a result that's normal, or you need
someone to come in, you can just now text them
through these safe texting apps and they can make the appointment,
come in or be reassured the results is okay. So
I think there's already been movement. I think there's going
to be a huge amount of movement around this over
the next five years. I think this is changing quicker
than we fully understand. And look with this whole Heidi thing,
(11:18):
which is a resonant chatch GPP and the rest of it. Yeah,
we're already seeing that change.
Speaker 3 (11:24):
Sort of gallop at us.
Speaker 2 (11:25):
Is not just I mean, we're going to dig into
what AI can be used for, if at all by
the average punter when it comes to just getting a
medical advice, but in terms of treatment choices and suggestions.
What's what's the involvement of AI currently? So you've you've
got a patient there, we go come in with a
(11:49):
range of symptoms, which you know, range of symptoms if
you google it can be anything. Does AI play a role,
Can it play a role? Does it play a role
in helping a doctor suggest treatment options or diagnosis? Because
the superpower of a really good GP i've heard is diagnosis.
Speaker 4 (12:07):
Yeah, so look, you're absolutely right, it is actually diagnosis
because it's looking at the symptoms that a patient has
got in context. It's in terms of what's happening and
taking actually a very good clear history. If you just
put in a symptom like chest pain into AI, you'll
often just focus straight on the heart or cardiac and
this is a good example. The fact of the matter
(12:28):
is there is a myriad of different diagnoses that can
lead to chest pain, from back problems to rib problems,
to lung problems, to the gall bladder, to the stomach,
to acid reflux, to anxiety and the heart. So the
skill of a GP or a doctor is actually to
listen to the history in context and decide whether there's
(12:52):
something serious here that does need immediate follow up, something
that needs investigation, or something that you can actually reassure on.
And we use something called differential diagnosis to do that.
So we run through a list of potential problems in
our mind, we work out where the symptoms fit, and
we start to work out the most likely way this
is going. And that's actually really really important. So something
(13:15):
that AI can't do at the moment is do that
type of type of work, which is why in the
studies are doing at the moment diagnosis is actually problematic
with AI to give a firm answer very very difficult.
Speaker 2 (13:26):
It's funny. I mean, just on the trivial thing. I
was doing talk back and it was I wanted. I
was just doing some research on the latest articles because
Bridget Bardeu had passed away and she was admitted emitted
from the in memoriam at the Oscars and I was
thinking on this and talk about it. And I said,
you know, give us the latest on the you know,
the feedback on Bridget Bardo being admitted, because it wasn't
(13:48):
the story. It was only a few hours old. And
chat GPT goes, Bridget Bardo is still alive, and I said, no,
check again, check. I literally said check again, Bridget Bardot
has passed away and this is the story. And it
came back and it says no. It was a talking
conversation that says I'm going with my original call she's
(14:10):
still alive. And I had to argue with it. I said,
you need to do a deep dive now. And finally
when you were right to challenge me on that, and
it's back and forth going, why did I have it?
It was like a human argument. It wasn't just computer
says no. Computer kept on saying no until I said, naughty.
Speaker 3 (14:25):
Computer, yeah.
Speaker 4 (14:29):
And that's a great example, isn't it of this comment
I made that we know that ninety percent of what
is out there and where AI often collates it from
may actually have misinformation with embedded in it. And that's
part of the problem. Going back to do we use
AI and the clinic, absolutely, but we have very specific
sites that we use that can give us sort of
(14:52):
very the latest updated medical evisidences research and collate that
from verified journals and verified sources. And that's the type
of thing you need to look at. And going back
to patients or what they do with AI, I think
the really really important, important critical thing here is that
they look at the veracity of the site they're looking
(15:13):
at and how sort of valid it is or otherwise.
Speaker 3 (15:16):
And that's what's really really important because the tricky.
Speaker 2 (15:18):
Question I have for you is, and for you who
are listening eight hundred eighty ten to eighty, actually, if
you're out there, have you used AI to check a
symptom or too or actually more maybe more practically to
learn how to dress a cut or a burn, Because Brian,
I can imagine you'd be reticent to say don't use
(15:41):
chat to say you can use chatpt for anything. As
I mentioned before, I used it to actually treat what
was what had become a chronic wound. And I'll be honest,
I wasn't. Every time you turn up to having a
point to check with the nurse, it's fifty bucks. And
I thought, I'm just going to hemorrhage money if I
have to go a couple of times a week for this.
So I shoved a photo into chat GPT and it
(16:02):
says and it described the wound the same accuracy that
my dermatologist did when I eventually saw them. But the
other question would be so that my daughter just had
a blister and she's like, how would I how do
I address this? Daddy? Do I leave it open or something?
And I would think that I could probably ask chat
GPT that, So are there some questions we can ask AI?
(16:23):
And in the same way we might google. It's just
a different way of gurgling. I would have gone to
the Mayo clinic or something online normally. What's your take
on it?
Speaker 4 (16:33):
Look, look, look, what you're describing I think is an
absolute valid use of AI. And as said, going back
to the research into this, what you've just described where
you have a defined problem which is a wound, you know,
what does you do that? Then AI is actually very
good at sort of narrowing that down and sort of
saying what goes on. So I think there's a real
(16:54):
place for this. I think there's a real place for
self management with a lot of these minor conditions that
you may have traditionally gone to see a doctor or
nurse about. However, and I go back to my original
comment where you've got undifferentiated problems or problems with diagnosis.
That's a very very different Kettler fish. And you're right,
(17:15):
you know, things like we can now use take a photo,
we can send them too a verified AI sight for dermatology,
and we'll come back with very very succinct dancewers about
the differential diagnosis and potential management pathways.
Speaker 3 (17:29):
So look, these.
Speaker 4 (17:30):
Things are all becoming part of what we do, and
I think there is validity in terms of a dealing
with these issues. But be sort of patient educational patient awareness,
which I think is a good thing.
Speaker 2 (17:44):
Actually I should I mean not that I want to
overshare too much information. I should, for clarity say that
I still saw a followed up with a dermatologist and
I had a biopsy taken and to me because that
was like, Okay, this is fine, you were looking after it,
but I think you know, I would be worried about
doing that totally in isolation.
Speaker 4 (18:02):
Look at and you've made a really good point there
because often in these situations as well, if it's chronic,
you can do need things like a biopsy to be
done to confirm what visually may look like a like
the issue, because it can be often several different things
that are going on. So yeah, you do have to
be careful about that that type of thing.
Speaker 2 (18:21):
What shouldn't people So okay, when you would say don't
use chat GPT to die, let's not just stop using
that brand because there's claud There's a whole lot of
AI the available to the general punter AI. There is
things that you probably just in the same way you
could google how to dress a wound or a burn
(18:41):
or a scolding. You know what to do for instance,
how to treat it, how to treat a burn I did.
I instantly went to chat GPT because I burnt myself
on a cast iron thing and I started running under
the water and with the other hand, I went, I've
just burnt myself on this How long to run it
under the water for because you know.
Speaker 4 (19:01):
Yeah, look, I mean I think simple questions like that
that are very defined and and you know you've got
to burn, and you know it's minor and it's sitting there.
I think it's a perfectly valid thing to ask the
question and see what it comes back with. I don't
think there's any problems with that, look at it.
Speaker 3 (19:17):
Look.
Speaker 4 (19:17):
Another area that's used quite extensively now is with mental health,
things like anxiety or depression. And again the thing there
is for self education and sort of tools. There's some
very very good, validated sites that actually will take you
through almost a cognitive process with some of those conditions
(19:38):
that traditionally you may have had to pay to see
a psychologist for. And it's about self management and the
patient being able to take their own responsibility for some
of these conditions. So again it comes down to the
veracity of the site and probably the question you're asking, well, we.
Speaker 2 (19:55):
Love your calls one hundred and eighty Have you resorted
to resorted to? That makes it sound like an aspect
of despiration, doesn't it. Have you used AI for some
simple health advice? Give us a call eight hundred eighty
ten eighty text nine two nine two. My guest is
Dr Brian Betty. I've got a few questions to ask him,
but you can get him with yours as well and
(20:17):
your reckons on eight hundred eighty ten eighty because I've
just suddenly remembered something else I used it for, and
I'm gonna ask Brian whether that was a good or
a bad thing. I'm hopefully he's going to say that's
all right. But we'll find out how much I've been
mist diagnosing and mistreating myself using chat GPT. It is
twenty seven past for eight hundred eighty ten eighty backcone. Yes,
(20:41):
we're talking about the role of AI and how any
role it does or currently play in your life when
it comes to keeping an eye on your health or
getting advice or whatever. And I'm going to run a
couple of things plus Dr Brian Betty, who's our guest
taking your callse on eight hundred eighty ten eighty. But
first let's go to Steve Hello.
Speaker 5 (20:58):
Talk about tonight goday, yel I had a well a
nomo que melanoma. Two lots of surgeris you removed from
the league, But there is a typical mail I left
it for que.
Speaker 2 (21:13):
Oh you just dropped out, Steve. I hope we can
get your call back there. Sorry, well, you being a
typical mail you left it for quite a while and
then you dropped out to carry on.
Speaker 5 (21:21):
Oh sorry. I lived for a while and my wife
went overseas and come back and said, we've been to
the doctor about the little lump on your league. And
I said, well no, so she probably out the cell
phone the mobile phone and took the photo of it
and seated into huge che GBT or or jogle and
(21:43):
it came back with very seriously, need to go. Never
looked at And while I didn't die post it, you know,
it's pt the bobbies at me and I said, we
made an appointment eventually and I got a biopsy and
it was a melanoma.
Speaker 3 (21:59):
Wow.
Speaker 5 (22:01):
Yeah, it was just you record it on time. Could
have been way more serious.
Speaker 3 (22:08):
Yeah.
Speaker 2 (22:10):
Yeah, hey, Steve, I'm going to put you on hold
there just because your line's jumping in and out. But
actually there are a couple of things that play, aren't
there with Brian his wife who was like that looks serious,
backed up by chat GPT.
Speaker 3 (22:22):
Yep, what one look.
Speaker 4 (22:24):
I think that's a really good example of where AI
can be a real benefit. I mean the thing about
skin lesions obviously, and you've heard about things like mold
map where they do these photographs. It's pattern recognition. So
a photograph or a visual of the lesion itself will
point you in a certain direction that hey, this does
need a biopsy. We need to get a definitive diagnosis
(22:47):
for treatment. I think Steve's example is excellent about what
the potential for CHATTP and what it can actually do
or for AI, and we should stop using chat GTP
or for AI what it can potentially do. And that's
a really really good example of whether it is pattern recognition,
it's a photo. It goes to a bank with a
million zillion photographs says, look, this is potentially a problem.
(23:10):
You should see your doctor and get that looked at,
because the next step is the decision to biopsy, get
a definitive diagnosis.
Speaker 3 (23:18):
Think it's treatment started. So yeah, a really good example
of where I can be used.
Speaker 2 (23:24):
Because the other the other time I used it was
it was at night my one of my daughters had
bad period pain and we had there's a certain I
can't remember the name of the medicine I obviously I
check it in there. There's a type of pain relief
which is is in the proxy and or something that
I can't remember it anyway, but she had taken something else.
(23:44):
It was either ibu, praphene or paracetamol, and and I thought, no,
that's not what we should have you should have been
taking this. And so I actually googled what is compatible?
Which of these is compatible? And I can't remember what
it was, and it says, no, definitely, don't use this
with it, but you can with that. Okay, I use
chat GPT. If I hadn't had it, I probably would
(24:05):
have done Google. Was that? Yeah, I mean, because that
is risky because if I got it wrong, you know,
oh yeah, take it does in paracetamol you can kill yourself.
Speaker 4 (24:14):
Yeah, look, look, look, I think you're right. But again,
it's a singular question about does this work with this
and should I use this together? And I think it's
very easy in those circumstances to get a yes no answer.
I think it's more problematic where it's not a yes
no answer, and there's more variations on what could potentially happen.
(24:34):
So again, go back to Steve's example of a photograph
which was a yes no answer.
Speaker 3 (24:38):
This looked suspicious.
Speaker 4 (24:39):
You should see someone and look, it may have been
okay it turned out it was not, which was a
very very good thing to do, or this should this
product go with this product? At this point it becomes
a binary decision at that point. So I think there
is use in those situations. Absolutely, It's a.
Speaker 2 (24:57):
Tricky one with the other thing that always makes it
feel better because the platform that I use is it
will give you the answer, but it will also have
a little button showing the sources that got it from,
so it might say the Mayo Clinic or whatever the
latest I mean, that's one of that used to be
one of my go tos just for talk back late
at night, you know, latest developments and stuff were Here's
(25:21):
the problem though with Steve had alesion. Would you trust
that the default setting for chat GPT would be if
it was looking if either it would say go to
a doctor, because the other danger it would be if
it went ah, looks all right.
Speaker 4 (25:37):
So that's that I mentioned this at the start that
the two problems with diagnosis are overdoing it and producing
anxiety to say it's something serious when it's not, or
actually delaying diagnosis, and again some of the research shows
that that's potentially what happens because you're right. If it
had gone the other way and said, oh, nothing to
(25:57):
worry about, look just let it see how it goes,
and it turned out to be a melanoma, that's a
delayed diagnosis. So in that situation can be problematic. So
again the site you go to is really really important.
I'd mentioned there are now dermatology sites in New Zealand
Gematology New Zealand that where you can do this type
(26:18):
of thing. You can take a fatal upload and sort
of tell you look, you should see your Wow. Really
there are sites that are starting to do that, and
that's the space we're getting into. So it comes down
to this issue. I think the more we go into
this of using sites that are verified for this type
of thing, and I think that's where we're going to
end up.
Speaker 2 (26:36):
Is that one of the challenges in a way, because
when a sort of dermatologist, she using her experienced eye,
can just go yep, no, that's all right, don't worry
about that. Then she'll get you know, she might go, oh,
let me take a look at that with a light.
Then she'll peer through something, she'll go yeah. But of
course the machines it's hard for humans to get their
head around this. But the AI can literally have examined
(26:58):
millions of images and it's hard to you know, because
we have this intuitive trust of human judgment that it
would make you, you know, if you had a machine
that this went yep, not good, yep, no yep. But
we're getting to that phase where we can do that.
As you say, if you can upload an image to
dermatology website.
Speaker 4 (27:18):
Where we're probably going to get to with it, it
doesn't remove the human factor of seeing in your GP
or seeing a dermatologist. What AI will start to do
with images, whether it be a radiology of a chest,
it might be an X ray of a chest or whatever,
it'll start to screen out the absolute normal ones, the
ones that don't have any problem, and it will start
(27:39):
to say, look, this has potentially got a problem. You
do need to see someone. And we're already seen that.
Probably in radiology, where what the talk is that AI
will you know, it's got ten million chest x rays
to compare. It'll say, no, this chest ray is absolutely normal,
and that'll be accepted as normal. This chest X ray
(28:00):
has got a problem. Therefore that goes to the radiologist
to have.
Speaker 3 (28:03):
A look at it.
Speaker 2 (28:03):
Try fur back, it's tre so.
Speaker 4 (28:05):
Access a triagal, which is essentially what's happening with the
dermatology issue that it's potentially on very good site saying look,
this is a normal lesion on all the indicators, or no,
there is an issue here that you need to see
someone and get this followed up. And I suspect that's
the space we're going to end up on.
Speaker 2 (28:23):
Okay, let's have got a few of the texts. Okay,
this is just on the human medicine side of things,
putting away the AI. What criteria does a GP use
when deciding whether a patient needs to be seen three,
six or twelve monthly? That's a big question.
Speaker 3 (28:40):
It's a huge question.
Speaker 4 (28:42):
Look, the first thing I'd say there is every person
is different, right, everything is individual. No, No, two people
are the same, No, two problems are the same. And
so look, there's a number of things that can come
into it. One is the condition itself. Two is if
you're using medication, that the problems you could run into
(29:04):
with the medication and whether you need to check that. Three,
do you need to tie trait or move the dosages
of something, or check that something isn't developing past where
it is. And four, you know, are we convinced this
is absolutely stable? There's no problems here. Quite happy to
see you in year's time. So look, there's a whole
lot of factors that come into play. There's a whole
(29:24):
lot of things that start to you start to think
about on an individual basis that leads to a decision
about the frequency of visits. And it's actually really, really
hard often to say this should be the frequency that
you're dealing with.
Speaker 2 (29:37):
Yeah, I've got another Oh god, I've got to go
to a break actually, but I've got lots of questions
for you. But by the way, if you would like
to join the conversation, we'd love you to give us
a call. Dr Brian Betty is with us GP and
talking about AI. Do you use it in managing your
health in any respect? And what would you trust it for?
And what wouldn't you trust it for? Oh, eight hundred
(29:59):
and eighty ten eighty it's twenty minutes to five News
Talk there Be with Tim Beveragey with Dr Brian Betty
talking about the use of AI in your lives when
it comes to health. If you texts here Brian, Hi, guys,
my wife's GP uses AI chat box, AI chat box
tool to write up clinical notes we should be using
Heidi voice activated? Does it right in front of us,
(30:19):
him and my wife check and that it's correct and
then entered. He reckoned. It's increases productivity by thirty percent.
He's probably gonna get told they might just be saying
chat GPT as you know genes, so that.
Speaker 4 (30:32):
That will be one of the verified products that he's using.
And I totally agree with the comment. The increase in
productivity is I think quite substantial in terms of what
we've seen.
Speaker 2 (30:43):
What about the other way around, So you have someone
come into your clinic and they've you know, they've got
a complex medical history. Now if you don't, you know,
you see a lot of clients, and especially I guess
if you have you're not seeing that regularly. There's a
lot of information. Is there a role for AI in
giving you a sort of summary, quick summary of where
(31:05):
this person's at because you might have pages of tests
and history for a particular condition. How do you, as
AI have a role in that? How does that work?
Speaker 3 (31:17):
Well? It will do? It doesn't at the moment.
Speaker 4 (31:19):
So it's just about getting the consultation out in a
particular way. We already do that in terms of our
what it called practice management system. So if you're on
a long term medication will be entered, if you've got
a diagnosis from the part or be ented. So I
can take a snapshot of a patient walking through the door.
I can look at their past history very clearly, these
are the problems I've had. These the medications are on.
(31:42):
Look at the last note and piece together what's going
on very quickly, so that already happens. What we will
see certainly is AI starting to automatically update what we
call those long term classifications, maybe automatically update the long
term medications based on the consultation. It's starting to move
(32:02):
into that space now, so a lot of that administ
work will be done.
Speaker 2 (32:07):
It's there a danger that means you might be able
to see more clients, but also that's a more intense workload.
How's the balance scurfe for GPS.
Speaker 4 (32:15):
Look, look, I don't know that it gives you any
less time with the patient. Okay, so I'm not an
advocate for that because it still takes time to work
through things.
Speaker 3 (32:26):
Look, it was really interesting.
Speaker 4 (32:28):
We actually did a short study at our clinic on
AI and the use of AI when we introduced it,
and we surveyed about forty patients and said, look, what
did you think about the use of AI and how
did you find it? They said three things, which I
thought was really interesting. Number One, they didn't mind it
at all, the fact it was AI. Number Two, they
felt of it helped the doctor. They were all for
(32:48):
it because they realized under a bit of pressure and
time threasure. But number three, and this was probably the
most important that we hadn't picked up on. They picked
up and they felt that as a doctor I spent
more time with them than I did on the computer
writing notes. So that was really really interesting because the
work of writing notes was being done for me and
(33:08):
I could just concentrate on what the patient was saying
and have a discussion about the issues in front of us. Now,
that was probably for me, the biggest single thing that
actually emerged, because I think that's a bitter quality interaction.
Speaker 2 (33:20):
Yeah, okay, let's take some more calls. Aaron, Hello, thank
you get it.
Speaker 6 (33:24):
Guys, Eda, good listen my whole little story about never
relying on AI too much and I guess, Brian, you're
talking about imaging and things, which is which is a
different kill official memo. But it was sort of a
through January, I really started struggling in terms of breathing
and just getting a deep breath and doing all sorts
of things. And went to AI and I was telling me,
(33:46):
you've probably got RSC or the dishiontory virus, and maybe
we saw my new GPS. We're all he's excellent, he's out,
he's out in the know. Over that period, you sort
of agreed with that, got on antibotics, sort of cleaned
it up with it. But then the janet started getting
worse and worse and worse again. It was just it
was horrendous. And then he'd been very productive. Set all
(34:10):
my bloods like every blood you could possibly imagine, because
I've only been that that sort of fifty four year
old bloke that's just under standard bloods every year. He
did absolutely everything and gave me a call. I went, oh, man,
your heart failure markers are through the roof. And I'm
a very fat person. I don't have any heart failure
or heart disease in my family or anything like that. Long,
(34:32):
long story, short. Yeah, I started seeing a cardiologist that
at ASPOT. He's really really good, got onto medication, but
then went back three weeks later and he looked from
ACG and just went, dude, you're in hospital now, and yeah,
in the cardio version, you know that Bryan stuff, and
(34:53):
you know and and and and and like my my
resting rate heart that was one hundred and seventy beats.
When I got onto ed at Auckland, I didn't feel
like I could feel I could feel palpitations. I could
feel like I like, if I'm training my butt off
at one hundred and eighty beats a minute, I'm nearly
dry reaching. So when I looked up a little monitoring
(35:15):
and see one hundred and seven, I was like, that
can't be right. And they're like, that is very right,
my man.
Speaker 2 (35:21):
And yeah, so got on your doctor.
Speaker 6 (35:27):
Oh for the blood man for the I don't even
realized you could do heart failure markets and blood.
Speaker 2 (35:32):
Actually that is I just learned that myself the other day. Aaron,
But because a sports coach for my daughter's sports team.
He he was doing a yo yo test and he
only had a stent put in six weeks before, and
he had a blood test, and they set up here
into hospital tomorrow, shoved a stent, and away he goes.
I'm not sure if he should have been doing those
yo yos, but anyway, what's your response.
Speaker 4 (35:55):
Yeah, look that that's a really, really good outcome, I've
got to say, And Aaron, and I think, good on
you for going to the doctor. And I think you've
demonstrated there is part of the limitations of using AI,
and it was that comment I made that it can
get it wrong, that it can potentially steer you in
the wrong direction, and it does take an experienced GP
(36:18):
to pick up that, Hey, this doesn't get better as
we expected. There's a whole set of things we need
to do here because there's other possibilities that are going
on here that aren't perhaps obvious to either you as
a patient or to AI.
Speaker 2 (36:31):
Actually, so, is there an enzyme or something that's released
into your blood when you have a cardio sort of incident.
Speaker 3 (36:37):
Yeah, there's a.
Speaker 4 (36:38):
Particular blood test that we can do for heart failure
to sort of demonstrate that that could be the problem
that's going on. It's called a BNP and it's it's
released when the heart's under stress like that, and you
are going into to a backlog of fluid on the lungs.
So it's a very specific test that has done.
Speaker 2 (36:57):
What's what was the surgery You talked about cardiof aversion
or reversion.
Speaker 3 (37:01):
I didn't quit version. Yeah, that's that's essentially.
Speaker 4 (37:05):
The paddles are put on the chest to revert the
heart back into a normal what we call sinus rhythm.
Speaker 3 (37:11):
So I go back to the normal rhythm.
Speaker 4 (37:13):
It sounds like I don't know Aaron's history, but it
sounds like you may have had something called actual fibrillation
or SVT that there's two different things, but it slows
the heart down and gets it back into a normal rhythm.
Speaker 2 (37:23):
I guess would he be under that, wouldn't he?
Speaker 3 (37:26):
Oh?
Speaker 4 (37:26):
No, no, no, you're just giving a light sedation. Often
it's just then zap on the chest and it just
puts electrical current through the through the heart, and it just.
Speaker 2 (37:37):
Not as dramatic as what we see on TV where
you go clear zap boofer.
Speaker 4 (37:43):
Well, there's a little bit of that that goes on.
You don't jump off the beer itself. It's not done
in that way.
Speaker 3 (37:49):
Yes, But.
Speaker 4 (37:53):
Really good example I think of trusting AI exclusively is risky, risky.
Speaker 2 (37:59):
Yeah, okay, hey, we'll take a break. We've back and
just to tick it is nine minutes to five talks.
He'd be news Talks. He'd be with Dr Brian Petty.
We've been talking about the role of AI and managing
our health. We've got a couple of just a couple
of minutes left, Brian. But I've actually got a question
because right now I'm wearing a garment four under two
(38:20):
six five and I've just linked up to my app
which tells me what my heart rate at the moment is,
which is surprisingly relaxed and considering I'm on air right now.
But actually, do these devices can actually? Are they an
aid for keeping an eye on your cardiac health? If
you know you're resting, pulses, whatever it is, and all
of a sudden, it might go, hey, you're I imagine
(38:42):
it would tell me if I suddenly had a dodgy
heartbed I don't know.
Speaker 3 (38:46):
Well, well, look we.
Speaker 4 (38:47):
Are moving into that space with them, and I think
they can be really useful if you're just monitoring your fitness,
for instance, that's fine. Look if you're experiencing palpitations or
irregular heartbeats, and we don't know what does. These type
monitors are really really useful for picking up whether it's
a problem or not. But I I think the issue
with them is if you become too obsessed with them, okay,
(39:09):
because there's always normal variation and heart rates and what
actually goes on and what's normal. If you start to
think is abnormal can be a problem, that can be
very anxiety provoking.
Speaker 3 (39:20):
So I think it's how they used and why they
used a ski question.
Speaker 2 (39:23):
Shall I tell you something that might make you roll
your eyes? So, because I mean, I bought one because
I'm into the exercise, but I did notice I had
quite a slow. My resting pulse when I'm asleep was
quite low, and so I told chat GPT and it
actually it actually just helped me with the exercise thing.
But wasn't it funny?
Speaker 6 (39:44):
Here I go?
Speaker 2 (39:44):
The first thing. I was like, hey, chat GPT, what's
this mean?
Speaker 3 (39:48):
Yeah?
Speaker 2 (39:48):
But there we go.
Speaker 3 (39:50):
It's often an indication that you're very fit, low heart
rate good.
Speaker 2 (39:54):
I won't say what it is on here because that
would be interscree to me, but I did feel reasonably
smug about it. Anyway, Hey, Brian, any closing words with
thirty seconds to go on the AI and you know,
just making sure you get the advice when you need
it from a doctor. Yeah.
Speaker 4 (40:09):
Look look my big thing is that with look look
at can be an aid. But actually with his uncertainty
or clear look see, you'd still see your doctors, see
your nurse.
Speaker 3 (40:20):
Just get advice.
Speaker 4 (40:21):
Yeah, don't don't sit on things, okay, don't rely on
it exclusively.
Speaker 3 (40:25):
It leads to problems.
Speaker 2 (40:27):
Hey, Brian, thank you so much for joining us today.
I always love chatting to you.
Speaker 3 (40:30):
Good stuff a good as gold, Thankston.
Speaker 2 (40:32):
There we go. That is Brian Benny Backscine.
Speaker 1 (40:35):
For more from the Weekend Collective, listen live to News
Talks It'd be weekends from three pm, or follow the
podcast on iHeartRadio
You're listening to the Weekend Collective podcast from News Talks.
Speaker 2 (00:09):
It be doctor, Do you mean the news? I gotta
bancas loading, you gotta made, you don't need nobody coming
(00:38):
from the start seeing come cute a little bit shame mother,
and welcome back to the Weekend Collective. I'm Tim be Everrgel.
Welcome in, Well, welcome back. This is the Health Hub
(00:59):
and we're going to be having a chat about uh, well,
help New Zealand. We want your call. By the way,
I know one hundred and eighty ten and eighty text
nine two nine two, but Health New Zealand have had
to tell their staff to stop using free AI tell
tools such as chat GPT for writing clinical notes. Look
chat GPT is. I've had some interesting interactions with it,
(01:22):
but I found it generally it's It's also an amazing
tool when you use it for the right thing. But
on its face, that is a kind of troublesome headline,
isn't it, Because you'd hope that, you know, doctors wouldn't
be relying on something that's sort of I imagine there
are AI tools in the background. But we'll dig into this.
But yeah, sounding hearing that they're writing the clinical notes
on chat GPT. Given it can hallucinate on certain things,
(01:44):
you'd be but a little bit nervous about that. But
that being said, of course it could be a great
thing for patients to use to their advantage. I have
used chat GPT because it's a long story and I
don't think I want to overly share it. But I
had a minor procedure of something removed and the scars
(02:04):
that have oppened up my GP well things, you know,
I'm not anyway, I ended up sort of monitoring it,
mostly in getting advice on bandages and all sorts of
things from chat gpt by uploading photos to it, which
was remarkably helpful until a dermatologist basically said, right, it's
now scay, you can just leave it. So yeah, I
(02:24):
actually found it quite useful. And I imagine my guest
is possibly horrified that I've done that, but anyway, he
is a GP dr, Brian Betty, and he's with us
for the Health Hub. Brian, Hello, how are you hey?
Speaker 3 (02:38):
Hello to Nice to be here.
Speaker 2 (02:40):
Yeah, I mean it's an interesting thing, isn't it? The
old AI Because the headline that people are using it
to summarize clinical notes I actually be honest, it didn't
upset me that much because I've used it to sometimes
sum up something I want to say and say, oh,
look just to summarize this for me, and it goes
does a bloody good job.
Speaker 4 (03:01):
Yeah, Look, you're right. I mean, AI is a credibly
useful tool that can be efficient, and in fact we
do use it in the clinics and general practice.
Speaker 3 (03:11):
We use it.
Speaker 4 (03:12):
I think the issue that has arisen here is the
fact that they're using the FREEGTV, not the accredited tools
that are specifically designed for medicine. So now the problem
with using the free chatchy PT, which is what's happened here,
is that we know that about ninety percent of what's
(03:32):
on the Internet has some sort of inaccuracy in it.
So if you're collating information from that source, you quite
like to have in accuracies in terms of what you
document or what you do. Now, the accredited tools, and
there's one called Heidi which has now been run out
in emergency departments and we actually use that in general practice,
(03:53):
is designed for medicine, so you don't run into that problem.
So I think the issue was being very very careful
about the tool you are actually using and that's what's
happened in this case.
Speaker 2 (04:07):
Heidi is that as in the Swiss Alps sort of Heidi.
Speaker 3 (04:11):
It's an unusual name.
Speaker 4 (04:13):
It was developed in Australia and it's called Heidi for
some reason. But it's probably one of the biggest AI
tools that are now used in medicine.
Speaker 3 (04:22):
So just to give you you an example.
Speaker 4 (04:24):
Of productivity, I use it in my clinic and it
probably saves me about two to three minutes of consult
in terms of writing notes. So if I see twenty
twenty five people a day, that's an hour of my
time that's saved that I can be doing other things,
either seeing more patients, or doing paperwork, or contacting specialists
or doing other stuff.
Speaker 2 (04:43):
And I guess it's about I mean, actually, you're right
about the free nature of chat GPT, So I pay
for it and it actually knows, it knows the GISTs
about things that I'm interested in and even my way
of communicating, so it's aware of the sort of language
that I use and everything. It's whereas a free version
of chat GPT that doesn't know you from a bar
AsSalt block a site really.
Speaker 4 (05:05):
And look, that's one of the problems with it, and
you gave the example of looking it up to do
some things.
Speaker 3 (05:10):
One of the problems. I've done research into.
Speaker 4 (05:12):
This into AI and it's benefits in terms of medicine
and patience. Now, what it's found is it's very very
good for things like health literacy or perhaps managing a
problem going forward. But what it is actually very very
poor at is diagnosing problems. So there's two issues you
run into it. Either gives you the wrong information and
(05:34):
it delays a potentially serious problem, or it actually over
eggs it and overdoes it and gives you what we
call hyper anxiety over the potential condition because the questions
are asked out of context, and that becomes a bit
of a problem with use of AI. So the research
is quite interesting into this. So absolute benefits, but there
(05:55):
are some drawbacks with it at this point.
Speaker 2 (05:57):
It's funny how names like Shakespeare said, arose by any
other name would smell as sweet, but there is something
about the name Heidi that makes we think, oh no,
it just is a slightly reassuring, wholesome sort of name.
Let's ask Heidi.
Speaker 3 (06:12):
Yeah. I don't know if that's why they chose it.
Who knows.
Speaker 4 (06:15):
Look there are other tools out there. There's one called Intelltech,
which is given so there's a variety of them, but
they're actually they're actually sort of you know, sort of
okayed by Health New Zealand. So these tools are verified
from privacy point of view, which is very very important
and also in terms of the way they work and
what they actually do. So yeah, but Heidi is I've
(06:37):
got to say an interesting name is is it?
Speaker 2 (06:40):
Also? Because there was something in the news a little
while ago about and it's just the questions around so
a AI. So you go and see a specialist or
a GP and AI is listening to the consultation basically
and taking notes as it goes, which I got to say,
I think sounds absolutely brilliant. There's there's questions around you know,
(07:04):
that information and going into the sort of into the
data sphere and all the private privacy sort of things.
But it is I mean, how do you use that
sort of thing?
Speaker 3 (07:15):
I use it.
Speaker 4 (07:16):
This is exactly what herd he is and that's exactly
what's been trialed and ED departments at this point. So
it records the conversation that will succinctly take out a
note out of that conversation. So the reason it's so
so important that these things are benuch marked by Health
New Zealand and by the experts is for those very
privacy reasons and things that you actually actually mentioned there, Tim,
(07:39):
So it becomes a really really important issue that these
tools are at the absolute privacy standard and that we
can use them with confidence. My other common to be
this that they probably get more accurate notes I've found.
Speaker 3 (07:55):
But look that this technology is with us now, we've.
Speaker 4 (07:59):
Got to learn to use it, otherwise we go back
to pen and paper and using the telephone at the
end of the day.
Speaker 3 (08:05):
And well, that's the thing.
Speaker 2 (08:06):
Medicine and technology goes hand in hand, because you know,
I'd rather have replaced on technology they've got now than
when they you know then it's thirty years ago or
forty years ago. I mean, the technology they've got for
analysis and MRIs and accuracy. But also it does lead
into that question abound robotics.
Speaker 4 (08:27):
And oh look, look, look I totally one hundred percent
agree with you on that I'd rather have have this
type of technology available today in terms of what I
can do in terms of accuracy, and with robotics, the
space we're now moving into potentially in a few years,
if you go in for your knee replacement, essentially your
(08:48):
knee will be positioned on the table. There'll be a
technician there, it'll be a computer program. All the machinery
will work automatically, and potentially yeah that the machine will
do the operation with a surgeon on standby if there
is any problems. This is potentially the space where now
moving into how is it.
Speaker 2 (09:09):
A tough sell for I was going to say, customers patients?
I don't know what the lingo these days is in
the health clients? What what? What are people these days?
Are they patients, customers or clients? It depends when what
year you finished your training.
Speaker 4 (09:24):
Look, look, in many ways, I still use the term patience,
and I'm quite happy to use it. And people can
call me a doctor if they want. That's fine. I'm
quite okay with that. Yeah, yeah, Look, I mean this
type of thing is it's it's interesting where this is
all going now?
Speaker 3 (09:45):
And it is it because?
Speaker 2 (09:48):
Is it because general is the acceptance of this technology?
Is it something that's simply going to happen as a
natural evolution as people and generations get comfortable in the
digital space. So obviously if you were to introduce robotic
to to a population who doesn't really know much about
the Internet and how technology works, they're going to be
(10:10):
very resistant to it. Whereas you know, twenty or thirty
years time, today's twenty year olds, who are fifty or
sixty year olds, you know, they're going to be far
more sanguine about it because simply that's the world they've
got used to.
Speaker 4 (10:22):
Look, look, I think you're absolutely right in that, And look,
I think it's actually in the last five to ten
years we've seen that happen. So there was a lot
of debate about using cell phones for messaging patients when
it first came in a few years ago, and I
know at our clinic there was a big discussion about
how many patients would use it, how many patient would
adopt it rather than ringing the clinic. And what we
(10:44):
sort of failed to understand was that every patient actually
had a cell phone and actually they really liked that
methodology of communication.
Speaker 3 (10:52):
So they don't have to ring the clinic. They don't
have to do that.
Speaker 4 (10:54):
If you've got a result that's normal, or you need
someone to come in, you can just now text them
through these safe texting apps and they can make the appointment,
come in or be reassured the results is okay. So
I think there's already been movement. I think there's going
to be a huge amount of movement around this over
the next five years. I think this is changing quicker
than we fully understand. And look with this whole Heidi thing,
(11:18):
which is a resonant chatch GPP and the rest of it. Yeah,
we're already seeing that change.
Speaker 3 (11:24):
Sort of gallop at us.
Speaker 2 (11:25):
Is not just I mean, we're going to dig into
what AI can be used for, if at all by
the average punter when it comes to just getting a
medical advice, but in terms of treatment choices and suggestions.
What's what's the involvement of AI currently? So you've you've
got a patient there, we go come in with a
(11:49):
range of symptoms, which you know, range of symptoms if
you google it can be anything. Does AI play a role,
Can it play a role? Does it play a role
in helping a doctor suggest treatment options or diagnosis? Because
the superpower of a really good GP i've heard is diagnosis.
Speaker 4 (12:07):
Yeah, so look, you're absolutely right, it is actually diagnosis
because it's looking at the symptoms that a patient has
got in context. It's in terms of what's happening and
taking actually a very good clear history. If you just
put in a symptom like chest pain into AI, you'll
often just focus straight on the heart or cardiac and
this is a good example. The fact of the matter
(12:28):
is there is a myriad of different diagnoses that can
lead to chest pain, from back problems to rib problems,
to lung problems, to the gall bladder, to the stomach,
to acid reflux, to anxiety and the heart. So the
skill of a GP or a doctor is actually to
listen to the history in context and decide whether there's
(12:52):
something serious here that does need immediate follow up, something
that needs investigation, or something that you can actually reassure on.
And we use something called differential diagnosis to do that.
So we run through a list of potential problems in
our mind, we work out where the symptoms fit, and
we start to work out the most likely way this
is going. And that's actually really really important. So something
(13:15):
that AI can't do at the moment is do that
type of type of work, which is why in the
studies are doing at the moment diagnosis is actually problematic
with AI to give a firm answer very very difficult.
Speaker 2 (13:26):
It's funny. I mean, just on the trivial thing. I
was doing talk back and it was I wanted. I
was just doing some research on the latest articles because
Bridget Bardeu had passed away and she was admitted emitted
from the in memoriam at the Oscars and I was
thinking on this and talk about it. And I said,
you know, give us the latest on the you know,
the feedback on Bridget Bardo being admitted, because it wasn't
(13:48):
the story. It was only a few hours old. And
chat GPT goes, Bridget Bardo is still alive, and I said, no,
check again, check. I literally said check again, Bridget Bardot
has passed away and this is the story. And it
came back and it says no. It was a talking
conversation that says I'm going with my original call she's
(14:10):
still alive. And I had to argue with it. I said,
you need to do a deep dive now. And finally
when you were right to challenge me on that, and
it's back and forth going, why did I have it?
It was like a human argument. It wasn't just computer
says no. Computer kept on saying no until I said, naughty.
Speaker 3 (14:25):
Computer, yeah.
Speaker 4 (14:29):
And that's a great example, isn't it of this comment
I made that we know that ninety percent of what
is out there and where AI often collates it from
may actually have misinformation with embedded in it. And that's
part of the problem. Going back to do we use
AI and the clinic, absolutely, but we have very specific
sites that we use that can give us sort of
(14:52):
very the latest updated medical evisidences research and collate that
from verified journals and verified sources. And that's the type
of thing you need to look at. And going back
to patients or what they do with AI, I think
the really really important, important critical thing here is that
they look at the veracity of the site they're looking
(15:13):
at and how sort of valid it is or otherwise.
Speaker 3 (15:16):
And that's what's really really important because the tricky.
Speaker 2 (15:18):
Question I have for you is, and for you who
are listening eight hundred eighty ten to eighty, actually, if
you're out there, have you used AI to check a
symptom or too or actually more maybe more practically to
learn how to dress a cut or a burn, Because Brian,
I can imagine you'd be reticent to say don't use
(15:41):
chat to say you can use chatpt for anything. As
I mentioned before, I used it to actually treat what
was what had become a chronic wound. And I'll be honest,
I wasn't. Every time you turn up to having a
point to check with the nurse, it's fifty bucks. And
I thought, I'm just going to hemorrhage money if I
have to go a couple of times a week for this.
So I shoved a photo into chat GPT and it
(16:02):
says and it described the wound the same accuracy that
my dermatologist did when I eventually saw them. But the
other question would be so that my daughter just had
a blister and she's like, how would I how do
I address this? Daddy? Do I leave it open or something?
And I would think that I could probably ask chat
GPT that, So are there some questions we can ask AI?
(16:23):
And in the same way we might google. It's just
a different way of gurgling. I would have gone to
the Mayo clinic or something online normally. What's your take
on it?
Speaker 4 (16:33):
Look, look, look, what you're describing I think is an
absolute valid use of AI. And as said, going back
to the research into this, what you've just described where
you have a defined problem which is a wound, you know,
what does you do that? Then AI is actually very
good at sort of narrowing that down and sort of
saying what goes on. So I think there's a real
(16:54):
place for this. I think there's a real place for
self management with a lot of these minor conditions that
you may have traditionally gone to see a doctor or
nurse about. However, and I go back to my original
comment where you've got undifferentiated problems or problems with diagnosis.
That's a very very different Kettler fish. And you're right,
(17:15):
you know, things like we can now use take a photo,
we can send them too a verified AI sight for dermatology,
and we'll come back with very very succinct dancewers about
the differential diagnosis and potential management pathways.
Speaker 3 (17:29):
So look, these.
Speaker 4 (17:30):
Things are all becoming part of what we do, and
I think there is validity in terms of a dealing
with these issues. But be sort of patient educational patient awareness,
which I think is a good thing.
Speaker 2 (17:44):
Actually I should I mean not that I want to
overshare too much information. I should, for clarity say that
I still saw a followed up with a dermatologist and
I had a biopsy taken and to me because that
was like, Okay, this is fine, you were looking after it,
but I think you know, I would be worried about
doing that totally in isolation.
Speaker 4 (18:02):
Look at and you've made a really good point there
because often in these situations as well, if it's chronic,
you can do need things like a biopsy to be
done to confirm what visually may look like a like
the issue, because it can be often several different things
that are going on. So yeah, you do have to
be careful about that that type of thing.
Speaker 2 (18:21):
What shouldn't people So okay, when you would say don't
use chat GPT to die, let's not just stop using
that brand because there's claud There's a whole lot of
AI the available to the general punter AI. There is
things that you probably just in the same way you
could google how to dress a wound or a burn
(18:41):
or a scolding. You know what to do for instance,
how to treat it, how to treat a burn I did.
I instantly went to chat GPT because I burnt myself
on a cast iron thing and I started running under
the water and with the other hand, I went, I've
just burnt myself on this How long to run it
under the water for because you know.
Speaker 4 (19:01):
Yeah, look, I mean I think simple questions like that
that are very defined and and you know you've got
to burn, and you know it's minor and it's sitting there.
I think it's a perfectly valid thing to ask the
question and see what it comes back with. I don't
think there's any problems with that, look at it.
Speaker 3 (19:17):
Look.
Speaker 4 (19:17):
Another area that's used quite extensively now is with mental health,
things like anxiety or depression. And again the thing there
is for self education and sort of tools. There's some
very very good, validated sites that actually will take you
through almost a cognitive process with some of those conditions
(19:38):
that traditionally you may have had to pay to see
a psychologist for. And it's about self management and the
patient being able to take their own responsibility for some
of these conditions. So again it comes down to the
veracity of the site and probably the question you're asking, well, we.
Speaker 2 (19:55):
Love your calls one hundred and eighty Have you resorted
to resorted to? That makes it sound like an aspect
of despiration, doesn't it. Have you used AI for some
simple health advice? Give us a call eight hundred eighty
ten eighty text nine two nine two. My guest is
Dr Brian Betty. I've got a few questions to ask him,
but you can get him with yours as well and
(20:17):
your reckons on eight hundred eighty ten eighty because I've
just suddenly remembered something else I used it for, and
I'm gonna ask Brian whether that was a good or
a bad thing. I'm hopefully he's going to say that's
all right. But we'll find out how much I've been
mist diagnosing and mistreating myself using chat GPT. It is
twenty seven past for eight hundred eighty ten eighty backcone. Yes,
(20:41):
we're talking about the role of AI and how any
role it does or currently play in your life when
it comes to keeping an eye on your health or
getting advice or whatever. And I'm going to run a
couple of things plus Dr Brian Betty, who's our guest
taking your callse on eight hundred eighty ten eighty. But
first let's go to Steve Hello.
Speaker 5 (20:58):
Talk about tonight goday, yel I had a well a
nomo que melanoma. Two lots of surgeris you removed from
the league, But there is a typical mail I left
it for que.
Speaker 2 (21:13):
Oh you just dropped out, Steve. I hope we can
get your call back there. Sorry, well, you being a
typical mail you left it for quite a while and
then you dropped out to carry on.
Speaker 5 (21:21):
Oh sorry. I lived for a while and my wife
went overseas and come back and said, we've been to
the doctor about the little lump on your league. And
I said, well no, so she probably out the cell
phone the mobile phone and took the photo of it
and seated into huge che GBT or or jogle and
(21:43):
it came back with very seriously, need to go. Never
looked at And while I didn't die post it, you know,
it's pt the bobbies at me and I said, we
made an appointment eventually and I got a biopsy and
it was a melanoma.
Speaker 3 (21:59):
Wow.
Speaker 5 (22:01):
Yeah, it was just you record it on time. Could
have been way more serious.
Speaker 3 (22:08):
Yeah.
Speaker 2 (22:10):
Yeah, hey, Steve, I'm going to put you on hold
there just because your line's jumping in and out. But
actually there are a couple of things that play, aren't
there with Brian his wife who was like that looks serious,
backed up by chat GPT.
Speaker 3 (22:22):
Yep, what one look.
Speaker 4 (22:24):
I think that's a really good example of where AI
can be a real benefit. I mean the thing about
skin lesions obviously, and you've heard about things like mold
map where they do these photographs. It's pattern recognition. So
a photograph or a visual of the lesion itself will
point you in a certain direction that hey, this does
need a biopsy. We need to get a definitive diagnosis
(22:47):
for treatment. I think Steve's example is excellent about what
the potential for CHATTP and what it can actually do
or for AI, and we should stop using chat GTP
or for AI what it can potentially do. And that's
a really really good example of whether it is pattern recognition,
it's a photo. It goes to a bank with a
million zillion photographs says, look, this is potentially a problem.
(23:10):
You should see your doctor and get that looked at,
because the next step is the decision to biopsy, get
a definitive diagnosis.
Speaker 3 (23:18):
Think it's treatment started. So yeah, a really good example
of where I can be used.
Speaker 2 (23:24):
Because the other the other time I used it was
it was at night my one of my daughters had
bad period pain and we had there's a certain I
can't remember the name of the medicine I obviously I
check it in there. There's a type of pain relief
which is is in the proxy and or something that
I can't remember it anyway, but she had taken something else.
(23:44):
It was either ibu, praphene or paracetamol, and and I thought, no,
that's not what we should have you should have been
taking this. And so I actually googled what is compatible?
Which of these is compatible? And I can't remember what
it was, and it says, no, definitely, don't use this
with it, but you can with that. Okay, I use
chat GPT. If I hadn't had it, I probably would
(24:05):
have done Google. Was that? Yeah, I mean, because that
is risky because if I got it wrong, you know,
oh yeah, take it does in paracetamol you can kill yourself.
Speaker 4 (24:14):
Yeah, look, look, look, I think you're right. But again,
it's a singular question about does this work with this
and should I use this together? And I think it's
very easy in those circumstances to get a yes no answer.
I think it's more problematic where it's not a yes
no answer, and there's more variations on what could potentially happen.
(24:34):
So again, go back to Steve's example of a photograph
which was a yes no answer.
Speaker 3 (24:38):
This looked suspicious.
Speaker 4 (24:39):
You should see someone and look, it may have been
okay it turned out it was not, which was a
very very good thing to do, or this should this
product go with this product? At this point it becomes
a binary decision at that point. So I think there
is use in those situations. Absolutely, It's a.
Speaker 2 (24:57):
Tricky one with the other thing that always makes it
feel better because the platform that I use is it
will give you the answer, but it will also have
a little button showing the sources that got it from,
so it might say the Mayo Clinic or whatever the
latest I mean, that's one of that used to be
one of my go tos just for talk back late
at night, you know, latest developments and stuff were Here's
(25:21):
the problem though with Steve had alesion. Would you trust
that the default setting for chat GPT would be if
it was looking if either it would say go to
a doctor, because the other danger it would be if
it went ah, looks all right.
Speaker 4 (25:37):
So that's that I mentioned this at the start that
the two problems with diagnosis are overdoing it and producing
anxiety to say it's something serious when it's not, or
actually delaying diagnosis, and again some of the research shows
that that's potentially what happens because you're right. If it
had gone the other way and said, oh, nothing to
(25:57):
worry about, look just let it see how it goes,
and it turned out to be a melanoma, that's a
delayed diagnosis. So in that situation can be problematic. So
again the site you go to is really really important.
I'd mentioned there are now dermatology sites in New Zealand
Gematology New Zealand that where you can do this type
(26:18):
of thing. You can take a fatal upload and sort
of tell you look, you should see your Wow. Really
there are sites that are starting to do that, and
that's the space we're getting into. So it comes down
to this issue. I think the more we go into
this of using sites that are verified for this type
of thing, and I think that's where we're going to
end up.
Speaker 2 (26:36):
Is that one of the challenges in a way, because
when a sort of dermatologist, she using her experienced eye,
can just go yep, no, that's all right, don't worry
about that. Then she'll get you know, she might go, oh,
let me take a look at that with a light.
Then she'll peer through something, she'll go yeah. But of
course the machines it's hard for humans to get their
head around this. But the AI can literally have examined
(26:58):
millions of images and it's hard to you know, because
we have this intuitive trust of human judgment that it
would make you, you know, if you had a machine
that this went yep, not good, yep, no yep. But
we're getting to that phase where we can do that.
As you say, if you can upload an image to
dermatology website.
Speaker 4 (27:18):
Where we're probably going to get to with it, it
doesn't remove the human factor of seeing in your GP
or seeing a dermatologist. What AI will start to do
with images, whether it be a radiology of a chest,
it might be an X ray of a chest or whatever,
it'll start to screen out the absolute normal ones, the
ones that don't have any problem, and it will start
(27:39):
to say, look, this has potentially got a problem. You
do need to see someone. And we're already seen that.
Probably in radiology, where what the talk is that AI
will you know, it's got ten million chest x rays
to compare. It'll say, no, this chest ray is absolutely normal,
and that'll be accepted as normal. This chest X ray
(28:00):
has got a problem. Therefore that goes to the radiologist
to have.
Speaker 3 (28:03):
A look at it.
Speaker 2 (28:03):
Try fur back, it's tre so.
Speaker 4 (28:05):
Access a triagal, which is essentially what's happening with the
dermatology issue that it's potentially on very good site saying look,
this is a normal lesion on all the indicators, or no,
there is an issue here that you need to see
someone and get this followed up. And I suspect that's
the space we're going to end up on.
Speaker 2 (28:23):
Okay, let's have got a few of the texts. Okay,
this is just on the human medicine side of things,
putting away the AI. What criteria does a GP use
when deciding whether a patient needs to be seen three,
six or twelve monthly? That's a big question.
Speaker 3 (28:40):
It's a huge question.
Speaker 4 (28:42):
Look, the first thing I'd say there is every person
is different, right, everything is individual. No, No, two people
are the same, No, two problems are the same. And
so look, there's a number of things that can come
into it. One is the condition itself. Two is if
you're using medication, that the problems you could run into
(29:04):
with the medication and whether you need to check that. Three,
do you need to tie trait or move the dosages
of something, or check that something isn't developing past where
it is. And four, you know, are we convinced this
is absolutely stable? There's no problems here. Quite happy to
see you in year's time. So look, there's a whole
lot of factors that come into play. There's a whole
(29:24):
lot of things that start to you start to think
about on an individual basis that leads to a decision
about the frequency of visits. And it's actually really, really
hard often to say this should be the frequency that
you're dealing with.
Speaker 2 (29:37):
Yeah, I've got another Oh god, I've got to go
to a break actually, but I've got lots of questions
for you. But by the way, if you would like
to join the conversation, we'd love you to give us
a call. Dr Brian Betty is with us GP and
talking about AI. Do you use it in managing your
health in any respect? And what would you trust it for?
And what wouldn't you trust it for? Oh, eight hundred
(29:59):
and eighty ten eighty it's twenty minutes to five News
Talk there Be with Tim Beveragey with Dr Brian Betty
talking about the use of AI in your lives when
it comes to health. If you texts here Brian, Hi, guys,
my wife's GP uses AI chat box, AI chat box
tool to write up clinical notes we should be using
Heidi voice activated? Does it right in front of us,
(30:19):
him and my wife check and that it's correct and
then entered. He reckoned. It's increases productivity by thirty percent.
He's probably gonna get told they might just be saying
chat GPT as you know genes, so that.
Speaker 4 (30:32):
That will be one of the verified products that he's using.
And I totally agree with the comment. The increase in
productivity is I think quite substantial in terms of what
we've seen.
Speaker 2 (30:43):
What about the other way around, So you have someone
come into your clinic and they've you know, they've got
a complex medical history. Now if you don't, you know,
you see a lot of clients, and especially I guess
if you have you're not seeing that regularly. There's a
lot of information. Is there a role for AI in
giving you a sort of summary, quick summary of where
(31:05):
this person's at because you might have pages of tests
and history for a particular condition. How do you, as
AI have a role in that? How does that work?
Speaker 3 (31:17):
Well? It will do? It doesn't at the moment.
Speaker 4 (31:19):
So it's just about getting the consultation out in a
particular way. We already do that in terms of our
what it called practice management system. So if you're on
a long term medication will be entered, if you've got
a diagnosis from the part or be ented. So I
can take a snapshot of a patient walking through the door.
I can look at their past history very clearly, these
are the problems I've had. These the medications are on.
(31:42):
Look at the last note and piece together what's going
on very quickly, so that already happens. What we will
see certainly is AI starting to automatically update what we
call those long term classifications, maybe automatically update the long
term medications based on the consultation. It's starting to move
(32:02):
into that space now, so a lot of that administ
work will be done.
Speaker 2 (32:07):
It's there a danger that means you might be able
to see more clients, but also that's a more intense workload.
How's the balance scurfe for GPS.
Speaker 4 (32:15):
Look, look, I don't know that it gives you any
less time with the patient. Okay, so I'm not an
advocate for that because it still takes time to work
through things.
Speaker 3 (32:26):
Look, it was really interesting.
Speaker 4 (32:28):
We actually did a short study at our clinic on
AI and the use of AI when we introduced it,
and we surveyed about forty patients and said, look, what
did you think about the use of AI and how
did you find it? They said three things, which I
thought was really interesting. Number One, they didn't mind it
at all, the fact it was AI. Number Two, they
felt of it helped the doctor. They were all for
(32:48):
it because they realized under a bit of pressure and
time threasure. But number three, and this was probably the
most important that we hadn't picked up on. They picked
up and they felt that as a doctor I spent
more time with them than I did on the computer
writing notes. So that was really really interesting because the
work of writing notes was being done for me and
(33:08):
I could just concentrate on what the patient was saying
and have a discussion about the issues in front of us. Now,
that was probably for me, the biggest single thing that
actually emerged, because I think that's a bitter quality interaction.
Speaker 2 (33:20):
Yeah, okay, let's take some more calls. Aaron, Hello, thank
you get it.
Speaker 6 (33:24):
Guys, Eda, good listen my whole little story about never
relying on AI too much and I guess, Brian, you're
talking about imaging and things, which is which is a
different kill official memo. But it was sort of a
through January, I really started struggling in terms of breathing
and just getting a deep breath and doing all sorts
of things. And went to AI and I was telling me,
(33:46):
you've probably got RSC or the dishiontory virus, and maybe
we saw my new GPS. We're all he's excellent, he's out,
he's out in the know. Over that period, you sort
of agreed with that, got on antibotics, sort of cleaned
it up with it. But then the janet started getting
worse and worse and worse again. It was just it
was horrendous. And then he'd been very productive. Set all
(34:10):
my bloods like every blood you could possibly imagine, because
I've only been that that sort of fifty four year
old bloke that's just under standard bloods every year. He
did absolutely everything and gave me a call. I went, oh, man,
your heart failure markers are through the roof. And I'm
a very fat person. I don't have any heart failure
or heart disease in my family or anything like that. Long,
(34:32):
long story, short. Yeah, I started seeing a cardiologist that
at ASPOT. He's really really good, got onto medication, but
then went back three weeks later and he looked from
ACG and just went, dude, you're in hospital now, and yeah,
in the cardio version, you know that Bryan stuff, and
(34:53):
you know and and and and and like my my
resting rate heart that was one hundred and seventy beats.
When I got onto ed at Auckland, I didn't feel
like I could feel I could feel palpitations. I could
feel like I like, if I'm training my butt off
at one hundred and eighty beats a minute, I'm nearly
dry reaching. So when I looked up a little monitoring
(35:15):
and see one hundred and seven, I was like, that
can't be right. And they're like, that is very right,
my man.
Speaker 2 (35:21):
And yeah, so got on your doctor.
Speaker 6 (35:27):
Oh for the blood man for the I don't even
realized you could do heart failure markets and blood.
Speaker 2 (35:32):
Actually that is I just learned that myself the other day. Aaron,
But because a sports coach for my daughter's sports team.
He he was doing a yo yo test and he
only had a stent put in six weeks before, and
he had a blood test, and they set up here
into hospital tomorrow, shoved a stent, and away he goes.
I'm not sure if he should have been doing those
yo yos, but anyway, what's your response.
Speaker 4 (35:55):
Yeah, look that that's a really, really good outcome, I've
got to say, And Aaron, and I think, good on
you for going to the doctor. And I think you've
demonstrated there is part of the limitations of using AI,
and it was that comment I made that it can
get it wrong, that it can potentially steer you in
the wrong direction, and it does take an experienced GP
(36:18):
to pick up that, Hey, this doesn't get better as
we expected. There's a whole set of things we need
to do here because there's other possibilities that are going
on here that aren't perhaps obvious to either you as
a patient or to AI.
Speaker 2 (36:31):
Actually, so, is there an enzyme or something that's released
into your blood when you have a cardio sort of incident.
Speaker 3 (36:37):
Yeah, there's a.
Speaker 4 (36:38):
Particular blood test that we can do for heart failure
to sort of demonstrate that that could be the problem
that's going on. It's called a BNP and it's it's
released when the heart's under stress like that, and you
are going into to a backlog of fluid on the lungs.
So it's a very specific test that has done.
Speaker 2 (36:57):
What's what was the surgery You talked about cardiof aversion
or reversion.
Speaker 3 (37:01):
I didn't quit version. Yeah, that's that's essentially.
Speaker 4 (37:05):
The paddles are put on the chest to revert the
heart back into a normal what we call sinus rhythm.
Speaker 3 (37:11):
So I go back to the normal rhythm.
Speaker 4 (37:13):
It sounds like I don't know Aaron's history, but it
sounds like you may have had something called actual fibrillation
or SVT that there's two different things, but it slows
the heart down and gets it back into a normal rhythm.
Speaker 2 (37:23):
I guess would he be under that, wouldn't he?
Speaker 3 (37:26):
Oh?
Speaker 4 (37:26):
No, no, no, you're just giving a light sedation. Often
it's just then zap on the chest and it just
puts electrical current through the through the heart, and it just.
Speaker 2 (37:37):
Not as dramatic as what we see on TV where
you go clear zap boofer.
Speaker 4 (37:43):
Well, there's a little bit of that that goes on.
You don't jump off the beer itself. It's not done
in that way.
Speaker 3 (37:49):
Yes, But.
Speaker 4 (37:53):
Really good example I think of trusting AI exclusively is risky, risky.
Speaker 2 (37:59):
Yeah, okay, hey, we'll take a break. We've back and
just to tick it is nine minutes to five talks.
He'd be news Talks. He'd be with Dr Brian Petty.
We've been talking about the role of AI and managing
our health. We've got a couple of just a couple
of minutes left, Brian. But I've actually got a question
because right now I'm wearing a garment four under two
(38:20):
six five and I've just linked up to my app
which tells me what my heart rate at the moment is,
which is surprisingly relaxed and considering I'm on air right now.
But actually, do these devices can actually? Are they an
aid for keeping an eye on your cardiac health? If
you know you're resting, pulses, whatever it is, and all
of a sudden, it might go, hey, you're I imagine
(38:42):
it would tell me if I suddenly had a dodgy
heartbed I don't know.
Speaker 3 (38:46):
Well, well, look we.
Speaker 4 (38:47):
Are moving into that space with them, and I think
they can be really useful if you're just monitoring your fitness,
for instance, that's fine. Look if you're experiencing palpitations or
irregular heartbeats, and we don't know what does. These type
monitors are really really useful for picking up whether it's
a problem or not. But I I think the issue
with them is if you become too obsessed with them, okay,
(39:09):
because there's always normal variation and heart rates and what
actually goes on and what's normal. If you start to
think is abnormal can be a problem, that can be
very anxiety provoking.
Speaker 3 (39:20):
So I think it's how they used and why they
used a ski question.
Speaker 2 (39:23):
Shall I tell you something that might make you roll
your eyes? So, because I mean, I bought one because
I'm into the exercise, but I did notice I had
quite a slow. My resting pulse when I'm asleep was
quite low, and so I told chat GPT and it
actually it actually just helped me with the exercise thing.
But wasn't it funny?
Speaker 6 (39:44):
Here I go?
Speaker 2 (39:44):
The first thing. I was like, hey, chat GPT, what's
this mean?
Speaker 3 (39:48):
Yeah?
Speaker 2 (39:48):
But there we go.
Speaker 3 (39:50):
It's often an indication that you're very fit, low heart
rate good.
Speaker 2 (39:54):
I won't say what it is on here because that
would be interscree to me, but I did feel reasonably
smug about it. Anyway, Hey, Brian, any closing words with
thirty seconds to go on the AI and you know,
just making sure you get the advice when you need
it from a doctor. Yeah.
Speaker 4 (40:09):
Look look my big thing is that with look look
at can be an aid. But actually with his uncertainty
or clear look see, you'd still see your doctors, see
your nurse.
Speaker 3 (40:20):
Just get advice.
Speaker 4 (40:21):
Yeah, don't don't sit on things, okay, don't rely on
it exclusively.
Speaker 3 (40:25):
It leads to problems.
Speaker 2 (40:27):
Hey, Brian, thank you so much for joining us today.
I always love chatting to you.
Speaker 3 (40:30):
Good stuff a good as gold, Thankston.
Speaker 2 (40:32):
There we go. That is Brian Benny Backscine.
Speaker 1 (40:35):
For more from the Weekend Collective, listen live to News
Talks It'd be weekends from three pm, or follow the
podcast on iHeartRadio
Popular Podcasts
Hey Jonas!
Hey Jonas! Hosted by Kevin, Joe, and Nick Jonas. It’s the Jonas Brothers you know... musicians, actors, and well, yes, brothers. Now, they’re sharing another side of themselves in the playful, intimate, and irreverent way only they can. Spend time with the Jonas Brothers here and stay a little bit longer for deep conversations like never before.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com