Transcript

The shadow use of Gen AI in the consultation room

From The BMJ · Episode page · Watch on YouTube

· 38 min · 5,500 words · 53 timestamped sections

This is the full, timestamped transcript of "The shadow use of Gen AI in the consultation room", an episode of The BMJ published December 16, 2025 (38 min, about 5,500 words). It mentions 14 people, works and ideas, including Cameron Abassi, Artificial Intelligence, BMJ, David Navaro, Generative AI and the clinical encounter.

What this episode is about

The discussion centers on the informal and unofficial use of generative AI, termed "shadow AI," in clinical settings, exploring its integration into patient care alongside traditional doctor-patient interactions. A novel concept called "triadic care" is introduced, emphasizing the collaborative roles of patients, clinicians, and AI in enhancing diagnosis and management, while also addressing the challenges and implications of this emerging technology in medicine.

Key takeaways

  • Shadow AI in healthcare is rising—20% of UK GPs use unofficial tools, no training needed!
  • Triadic care: A three-legged stool of patients, clinicians, and AI—because who doesn't love a stable arrangement?
  • Patients now arrive armed with savvy AI insights, challenging the stereotype of clueless web-searchers.
  • AI's empathy isn't human, but its engaging dialogue might just be the next best thing—emotionally speaking.
  • Despite AI’s allure, 70% of patients want doctors in the driver’s seat—because no one likes blind trust.

Who speaks in this episode?

What is discussed, and when

  1. – 5:00: Artificial Intelligence, BMJ, Cameron Abassi, David Navaro
  2. – 20:00: Generative AI and the clinical encounter, Tara Vinan, Naj Lada, University of Helsinki
  3. – 25:00: arthroscopic subacromial decompression
  4. – 35:00: MSF
  5. – 37:49: Cameron Abassi

Notable quotes from "The shadow use of Gen AI in the consultation room"

  • Welcome to the Medicine and Science podcast from the BMJ. I'm Cameron Abassi, editor in chief. In this episode of the podcast, we're taking a deep dive into AI in the consultation room. I think what's more interesting and really.
    about Cameron Abassi · at
  • To explain what that is. I'm joined by David Navaro, a research fellow in generative AI at the Australian Institute of Health Innovation. David, thank you for joining us. I know it's.
    about David Navaro · at
  • Week, Rachel Hinton, our analysis editor, spoke with Charlotte B, who is a researcher, and Marcus Lewis, a GP.
    about Marcus Lewis · at
  • And that collection generative AI and the clinical encounter is free to access on bmj.com.
    about Generative AI and the clinical encounter · at
  • What we found was that the surgery was very effective in the sense that very soon after undergoing surgery all of the patients reported huge improvements in pain scores and in the.
    about arthroscopic subacromial decompression · at
  • I'm Rebecca Kums, head of journalism at the BMJ. We've been reporting all year on the incredible work that MSF are doing in conflict situations around the world. And while that work is essential, we also wanted to highlight other places.
    about MSF · at

Who and what is mentioned in this episode?

People

  • Cameron Abassi — Person · Host · 2 mentions · first at
    Hosts the episode, frames the discussion on AI in the consultation room, and interviews David Navaro while challenging claims around evidence, trust, and hype.
  • David Navaro — Person · Main guest · 1 mention · first at
    Interviewed about 'triadic care' and how generative AI becomes a third party in the clinical encounter, with emphasis on transparency and trust.
  • Marcus Lewis — Person · Referenced expert (audio clip participant) · 1 mention · first at
    Provides a GP perspective on real-world 'shadow AI' use, risks of LLM overconfidence, hallucinations, and the need for clinician expertise to verify outputs.
  • Tara Vinan — Person · Main speaker (research interviewee) · 1 mention · first at
    Orthopedic surgeon-researcher describing a placebo-controlled trial and 10-year follow-up showing arthroscopic subacromial decompression offers no benefit over placebo or exercise therapy.
    I'm Tara Vinan from the University of Helsinki and I was trained as an orthopedic surgeon or a bone blacksmith, but at the moment I'm mostly doing just research and running our residency.
  • Rachel Hinton — Person · Segment interviewer (referenced) · 1 mention · first at
    Introduces and conducts the earlier conversation with Charlotte B and Marcus Lewis about how AI is entering clinical practice.
    You're describing in the clinical encounter will work out. Earlier this week, Rachel Hinton, our analysis editor, spoke with Charlotte B, who is a researcher, and Marcus Lewis, a GP.
  • Naj Lada — Person · Segment narrator/reporter · 1 mention · first at
    Introduces the research segment on arthroscopic subacromial decompression and frames the follow-up discussion and its implications for practice.
    Join us now to talk about why they kept the study going and why their first paper hasn't changed practice. Naj Lada, our UK research editor, reports.

Books & papers

Companies & orgs

  • BMJ — Company · Host · 2 mentions · first at
    Publisher and host of the podcast, discussing medical research and AI in healthcare.
    Welcome to the Medicine and Science podcast from the BMJ. I'm Cameron Abassi, editor in chief. In this episode of the podcast, we're taking a deep dive into AI in the consultation room. I.
  • Australian Institute of Health Innovation — Company · Affiliation · 1 mention · first at
    Institution where David Navaro, a research fellow in generative AI, is affiliated.
    To explain what that is. I'm joined by David Navaro, a research fellow in generative AI at the Australian Institute of Health Innovation. David, thank you for joining us. I know it's.
  • University of Helsinki — Company · Affiliation · 1 mention · first at
    Institution where Tara Vinan, an orthopedic surgeon and researcher, is affiliated.
    I'm Tara Vinan from the University of Helsinki and I was trained as an orthopedic surgeon or a bone blacksmith, but at the moment I'm mostly doing just research and running our residency.

Products & tech

Concepts & topics

  • Artificial Intelligence — Concept · Core topic · 2 mentions · first at
    Central theme of the episode, discussing the impact and use of AI in clinical settings.
    Podcast from the BMJ. I'm Cameron Abassi, editor in chief. In this episode of the podcast, we're taking a deep dive into AI in the consultation room. I think what's more…
  • cholera — Concept · Key argument · 4 mentions
    Disease discussed in the context of MSF's work in Nigeria.
    MSF worked closely with the government in Nigeria providing services within a hospital creating specialized units to deal with the cola outbreak. That's a huge logistical challenge when basics.

Sponsors

  • MSF — Company · Sponsor · 2 mentions · first at
    Charity focus of the BMJ's Christmas appeal, highlighting their work in conflict zones and cholera outbreaks.

Full transcript with timestamps

Welcome to the Medicine and Science podcast from the BMJ. I'm Cameron Abassi, editor in chief. In this episode of the podcast, we're taking a deep dive into AI in the consultation room. I think what's more interesting and really not spoken about much is uh what you might call shadow AI use, unofficial AI use.

We look at new research on an ineffective surgical intervention. Of course, a very um bioplausible solution for that is to make room for the tendons or to remove some of the bone.

We finish with the BMJ's Christmas appeal. This year it's MSF. So one of the biggest struggle we have is resources. It's never enough and then you have to get creative sometimes to manage these patients. Firstly, generative AI in the clinic. The BMJ has published a new collection, Generative AI and the clinical encounter. In it, the authors use a relatively new term called triadic care. to explain what that is. I'm joined by David Navaro, a research fellow in generative AI at the Australian Institute of Health Innovation. David, thank you for joining us. I know it's you're on the other side of the world.

It's very different time of day there. Um, tell us what uh triadic care is, David. So very simply put triadic care is the form of care that it's appearing the care itself the diagnosis the management and and the support of patients is not only coming from these doctors in one side the patients in another that's the usual pair that we that we had in medicine for for very long and now with a third element uh in form at all levels and all process by by AI.

Okay. Thank you. Interesting that you say there are those three kind of legs to the stool of triadic care. One is patients, one is clinicians, one is AI. I would as the editor of the BMJ say to you that we did have three legs of the stool already and one of them was evidence. I guess you know evidence here is kind of mingled in with AI. Is it?

Yes. I think evidence is definitely you know it's definitely what AI can provide. I I think um I think also what what it changes is how it is provided. It's it's not only you know it's not only provided as facts it's provided as a as a conversation as an engaging conversation and it can it can even provide some form of support even uh I I wouldn't I wouldn't go all the way to say that

you know it's it's empathic or or or you know or or at any point as a human but uh a lot of people are finding also uh that that level of of of engagement with the eyes and with this dialogue very very entertaining or or very helpful at at some point. So

yeah, I mean just to be clear on that, I mean there are some concerns that that empathy can also be misplaced, but let's come back to those issues in a minute. Um uh let's hear from some more of our authors about how this change that you're describing uh in the clinical encounter will work out. Earlier this week, Rachel Hinton, our analysis editor, spoke with Charlotte B, who is a researcher, and Marcus Lewis, a GP. Let's hear that conversation now.

In a practical sense, how are you seeing AI coming into the consultation room into into clinical practice? I think one of the best titled pieces that the BMJ's published on this subject uh was we don't need AI to replace clinicians just get our computers to work first. And I think that summarizes the state of officially sanctioned AI uh used by doctors on a day-to-day basis. Slow roll out of uh AI voice transcription softwares otherwise known as AI scribes in primary care. I think what's more interesting and really not spoken about much is what you might call shadow AI use, unofficial AI use, which will be clinicians using commercial tools, uh things like chat, GPT, claude, Gemini.

So I've been leading these surveys on the use of generative AI tools. So we did this first survey in 2024. It was in January and it was a stratified sample of a thousand UK GPS and this survey found 20% were using these tools for any clinical task. Uh nearly three in 10 said documentation but about a quarter said for differential diagnosis and for suggesting treatment options. We redid it this year and it went up to 25%. And um just as Marcus has has said this time was 35% said for documentation um 27% actually said for differential diagnosis and um around a quarter for referrals and treatment options and that was just to sort of reiterate those were UK GPS but what was fascinating was 95% said they had no training. What are you seeing in in the patient community when it comes to to the use of these tools?

We did my group did a scoping review and we found like 75% of all the surveys that look at stakeholders experiences are among clinicians. So we do have very few patient surveys. But what we've got are are pretty interesting in that people do seem to be using these tools to better understand uh medications but also they're using them as a first opinion service and sometimes as a second opinion service

but in the consultation room itself um patients will come in saying that they've done some research uh and although they might not necessarily say I've used AI the stories and the information they arrive with are noticeably more fluent and structured than maybe from a standard web search and they have a deeper level of understanding. I don't think it necessarily challenges the doctor patient dynamic because we're used to patients arriving in the consultation room with ideas about what might be wrong with them and influenced by conversations they had with friends or relatives or other health care professionals or searching the web. So AI is a different way of um coming armed with information.

They they are becoming increasingly open to the use of AI but including with their doctors using AI but what they don't necessarily want um and in significant uh percentage is do not want diagnostic and clinical decisions to be completely outsourced to AI. they want their doctor to have the final judgment and to be in the loop. So I suppose when it comes to trust what we're seeing is increasing openness to triadic care in that sense um a place for AI in the consultation and but with the the trained human professional as the overseer. I mean, Marcus, based on what Charlotte's saying, you know, what what are some of these, you know, potential risks that are coming out with these increasing use, whether it's declared or not within your practice?

How how long have you got? Well, I think if we're accepting that there's a third voice in the consultation, and this voice is immediate, gives very quick answers. The answers are very fluent. They seem wise, confident, and they certainly seem plausible because because of course AIS are trained to give answers that humans approve of. It's it's part of the the way that they're trained. Um, but it's a voice which is not good at saying that it doesn't know. It's not a voice that's good at recommending nothing, dealing with uncertainty.

It's certainly a voice that can make mistakes, whether it's AI hallucinations or things that are more subtle, uh, emitting important details or overemphasizing important details. AI is confidently making recommendations which might be accurate, but maybe not um, considering the possibility of doing nothing or watchful waiting, which is of course a cornerstone of of medical practice. When I do speak about AI use when I'm talking to and and doing teaching for colleagues on this subject, a very basic rule of thumb is apart from very low stakes usage, only use a large language model if you're willing to read the answer and if you've got the expertise to know if there might be errors uh in the answer.

I I obviously completely agree with the points Marcus has made but the the thing I think we should do is also benchmark as to why people will are you turning to these tools as well. Um and all of the things that we can see that can go wrong with AI. Um AI can be people pleasing. It can be have great conviction in its responses um and inverted commas. It can make mistakes and it can be biased. All of those things can happen in care as we know it as well. And this is where I suppose thinking more about triadic care. How do we get the best use of these tools where we can interrogate the responses? Doctors can do it with the patient together. So it's a case of savviness I think with these tools and that's where we need to be offering greater guidance not only to to our in health education for medical students and other health professionals but but patient guidance as well.

Yeah. Absolutely. There's also research showing AI could potentially start deskskilling professionals. you know, what do you sort of say to this and and see this in your experience related to, you know, the lack of potential training and and preparedness for its use?

But the the question of uh deskkilling, I think, is really interesting. And I suppose just to play devil's advocate here, is it better not to to to have more intimate knowledge of a patient in some circumstances or not? In some circumstances, it may be that the more you understand and you've read through and you've you've written the notes, you intimately understand what's going on with the patient patient. But in other cases, it's possible that not knowing as much might be a good thing cuz it prevents some kind of unwanted stereotyping in relation to their symptoms or their condition or whatever.

Okay, David, I think it was quite interesting hearing that. Um clearly there are some big issues around transparency which I mean I would probably I mean clearly I'm sure we'd agree there isn't enough transparency uh right from which clinician is using AI in which circumstances to help them with I think as was described cognitive task of diag diagnosis and and working out a therapeutic plan uh but also in terms of how an AI or genai um model came up with that particular recommendation. or that particular piece of advice. There's a whole range of things where transparency uh is required but also I think what was clear is that trust is central to this and I think my concern is that there's a high degree of misplaced trust um in the information that's being provided. Um but I just want to before you answer that I just like to pick up on three particular aspects that were discussed.

Number one about patience. uh Charlotte mentioned that there wasn't that much research into patients and one of my concerns around AI is that in our rush to embrace it in the clinical encounter the patient voice is being forgotten. So that's number one. Number two I think there is a bit of a concern when you look at the evidence uh around how how useful generative AI in supporting diagnosis and treatment. I think it's very unclear uh how much benefit we can get in in those areas. Certainly you can see the benefit in back office tasks. Uh there needs to be factchecking and human intervention uh in in those tasks as well. But I'd say the evidence uh that I've seen is much less clear around diagnosis and treatment. And finally, I was quite amused that Marcus used that quote uh from the B BMJ, which I use a lot as well. Um, which is a GP saying, "Don't talk to me about AI. Just get my computer to work." Um, but can you just

address those bigger issues around transparency and trust for us? Sure. I and I think yeah both Marcus and Charlotte make very good points and and yeah I definitely agree that transparency is a big issue and and that's a big part of what we what we trying to uh express and and and convey with triadic care is that these relations this uh AI um assisted diagnosis and AI second opinions we need to bring it to light because I uh on on one side on on on the doctor's side I I think there is um there could be this issue that if you have consulted something with generative AI it can be seen that you know you you are not using the tool that you should be using or there can be issues around trust with the patient patients can be f can also feel uh as it has happened a lot before uh that their opinions or their research are being you know um not considered or or or or just or just perceived as as you know as you have used Dr. Google and

now you have used CHP you know I think we have we have these issues to address in terms of trust also I think it's very important that trust is something that uh you need to build between doctors patients uh and in general uh AI can can play can play a role in both ways can play a role uh decreasing trust uh if you know all now all of a sudden doctor diagnosis might come from an AI and they don't tell me what and you know there there can be this view that uh now uh the the diagnosis is even yeah let's hear where it's coming from uh but we also see that when doctors and patients they have a trustful relation a relation they have been building uh use of AI or or open declaration of the use of these tools the use of uh decision support tools AI diagnosis these AI scribes doesn't really decrease the trust. I I think it's it's a it's a balance to strive. We need to build the trust and AI will be a part of that trust that we

need to build. So, okay. Um, the bit I guess I'm concerned about is that from the paper, your paper, uh, one of the stats, or it could have been the editorial that I came away with was that the public trusts a recommendation from generative AI as much as it trusts a recommendation from a clinician. That's pretty worrying. Yes, I think that's that's part of the issue we have with with current generative AI systems and and and chatbots is the the way they express their answers express generally an an overconfidence and and doesn't really tell you what degree of uncertainty in those answers are without the nuance and I'm and putting that explanation of disease or or clinical reasoning uh into into the uncertainty into into the nuance of medical practice. There can be a lot of issues.

I mean, I guess that's the point of your triadic stool analogy, which is that you got patients, you've got clinicians and AI all working together to make a decision or or come up with a shared decision making process. Um, finally in a on a related point, do you think we're overhyping the use of generative AI in the clinical encounter?

Well, I think with AI and with all that is happening in industry, Silicon Valley, stock markets, etc. there is a certainly there is a degree of hype. I think uh that's that's something that it comes with with most of technologies or many technologies. However, uh abstracting ourselves from uh valuations or the last declaration of of of one of the CEOs of these companies,

I think the the transformative potential of the technology besides besides all that aspect is is is truly great. I think we are we are just starting to grasp what these AIs these advanced forms of of neural networks can can really do and and I think that's that's not hype that's something that will happen uh uh it might not happen next month it it's we won't solve all diseases and and cural cancers as as some of the CEOs are saying with this form of AI but I think they they are transformative on its So we need to explore that. We need to explore that aspect that and and and we need to shape it in a way by by doing the proper research.

David, thank you very much. Let's see where this all leads us. Appreciate you joining us from Australia.

Thank you. And that collection generative AI and the clinical encounter is free to access on bmj.com. Next, arthoscopic subocchromial decompression to reduce shoulder impingement. Surgery for this is still widespread despite us publishing compelling trial data showing it was ineffective. Those trialists have now published a 10-year follow-up of that paper. They join us now to talk about why they kept the study going and why their first paper hasn't changed practice. Naj Lada, our UK research editor, reports.

I'm Tara Vinan from the University of Helsinki and I was um trained as an orthopedic surgeon or a bone blacksmith, but at the moment I'm mostly doing just research and running our residency program.

Give us a kind of highle summary of what you did and what you found. So the patients were very common patient group patients complaining of pain when elevating the arm or or something that was at least formerly considered impinchment of the rotator cuff tendons and the overlying bersa. The diagnosis was created was it 70 or even 100 years ago some at least plus 50 years ago someone came up with the idea that this is is caused by impingement of the tendons and of course a very bioplausible solution for that is to make room for the tendons or to remove some of the bone. So we wanted to test whether this very old theory works in practice. So we we um recruited about 200 patients, randomized them into three groups. One was the standard arthroscopic subacriial decompression. The other group was uh placebo surgery which was diagnostic arthoroscopy. So so the patients basically underwent exactly the same

procedure except we didn't remove any of any any bone. And then we also had a third group, a group that was randomized to get physical therapy and then some homo exercises.

So what did you find? What we found was that the surgery was very effective in the sense that very soon after undergoing surgery all of the patients reported huge improvements in pain scores and in the functional scores. The problem at least from the perspective of being a surgeon was that the the the the improvement was no better in the active or in the ASD group than in the placebo group. So it was basically all placebo. And as we had also had the third group very soon all three groups reported very highly similar comparable recovery comparable outcomes. They were not identical but very similar at two years and then the picture just got kind of kept getting more certain. At 5 years there were no differences and and at that 10 years all the three groups no difference especially between the placebo surgery and the active surgery groups.

Let's talk um I want to make sure we talk about the follow-up period and what was involved in doing a 10-year follow-up of a trial. I mean that seems and you know I think it's 87% of participants are included in this 10-year analysis which is really remarkable. Tell us a little bit about yeah what went into get getting that follow-up but then also what are some of the challenges and things you need to think about when interpreting um those follow-up results? Yeah, I mean 10 years. I don't even want to be shown pictures of what I looked like when we started the trial. Launched the trial in 20 2005 and now we published the trial just before it turns 20 26. So um what went into that? I mean it's just hard labor to keep people involved. It it means treating them well whenever they have complaints, whenever they have concerns. One of the benefits why Finland seems to be kind of one of the better places to do placebo surgery kind

of re research is the people still trust the medical the health care system. They they are nice people. So they trust us and they think that we are we are not using them as guinea pigs. rather we really trying to help and we are trying to find an answer to a question that seems to be relevant and a lot of lot of our patients actually told us that that they understand that this may not benefit them but they want this to to contribute to greater good which is to me like the the essence of of doing research. Mhm. So the earlier trial publications that you've had at two and five years have been influential in informing recommendations not to do um arthoscopic shoulder decompression uh for this indication. This 10-year follow-up would appear to cement that as well. And yet I think it's still it's still happening which I think is why why your this 10-year follow-up is is so interesting. And you know that was

something we talked about when we were considering the paper is that okay the the message is similar but you know it still has it's still relevant because because people are still doing it. Um what's been your experience of you know following these patients up and I guess communicating uh these results with the wider world. What do you think is going on? Why is it why is it still happening when there is this kind of clear clear message? Yeah, I mean this could end up being a very very long story, but just to try to make it short, you you you told that this was our previous publication that was published, the 2-year follow-up was published in 2018 in the BMJ.

The year after the VMJ rapid recommendation group published a very I don't know whether it was influential but at least a very clear guideline recommending against this surgery virtually in any and all imaginable patients with with this kind of complaint. But as as it seems to happen in every medical discipline, the specialists know better. And I mean this isn't this isn't unique to orthopedic surgery. This is this is very common across pretty much the the entire range of medical disciplines that the specialists always know better. what they do not know better and I can say for certainty that they do not know better the principles of evidence-based med medicine or or how research evidence should be assessed but they just know better uh and and this is what kind of happened. I mean a very influential, a very credible guideline organization sums up all the evidence using the the best available

techniques and methods and ends up con concluding that that his surgery is is futile. And then we have an avalanche of papers written in all kinds of orthopedic journals and and letters submitted to even the the BMJ criticizing the the paper and the methods and and whatever. So I mean I guess this is the problem with with medicine that that people tend to believe what they want to believe and just but I but I also want to kind of emphasize that the the I'm I'm not sure whether I should be proud of the Finnish surgeons because we were probably the the ones doing the most of these surgeries. So the our incidences were the highest before the fact trial was published and we actually had at least one prior randomized trial comparing the surgery to nonoperative treatment but our rates have gone down by 90%.

But I I guess we are exception to the rule because as as you kind of pointed out I think there are still like nearly 30,000 of these procedures performed each year in the UK even though a very similar study to ours placebo surgery control trial came out from the UK in 2018. So the same year as our fimbec trial. So evidence doesn't seem to move surgeons.

Yeah. I mean and as you say is it is a a story we see repeated um through many disciplines of medicine where particularly where you know a treatment is established before it's been studied in a trial and it has this kind of plausible mechanism. So you think well of course it it should work. It it does work. So what do you think is needed for things to change? If you read why the BMJ rapid recommendation and my disclaimer is or Gordon Guy is my huge huge huge mentor and hero. Uh but anyway, how the the guideline panels are assembled, you know, having methodologists and and and content experts and patients and having balance panels rather than all these specialty societies having panels with experts and with with with loads of not just financial but other conflicts of interest. So I think that that would be I would really appreciate the difficulty of of of saying that medical reversals will never affect uh occur unless we as

a medical discipline as a community accept the fact that people with any kind of intellectual or financial conflicts will not be allowed to make decisions on the viability of of the treatments that they are given. And that article, arthoscopic subacchromial decompression versus placebo surgery for subocchromial pain syndrome is available on bmj.com. Every year at this time, the BMJ launches its Christmas appeal and this year we've chosen medicine s from tier as our charity. It's not the first time we've raised money for MSF, but given the ongoing conflict in Gaza, Sudan, and Yemen, we felt it was right to help them again.

I'm Rebecca Kums, head of journalism at the BMJ. We've been reporting all year on the incredible work that MSF are doing in conflict situations around the world. And while that work is essential, we also wanted to highlight other places they are active. I had a chance to speak to one of MSF doctors working in Nigeria in a colera treatment center. MSF worked closely with the government in Nigeria providing services within a hospital creating specialized units to deal with the cola outbreak. That's a huge logistical challenge when basics like clean water and power are lacking.

They also support volunteer doctors and nurses who help run these wards. My name is Dr. Shafalad. I think I've been practicing medicine for over seven years now. I'm the emergency medical activities manager. My day never really ends per se. Most times my day even starts before the sun, you know, rises. Sometimes I I can sp you don't sleep deep because you know there's something waiting for you. So by 5 you're already preparing the day. So here during the cola outbreak there's a WhatsApp we create where you post the cases that came through the night. You highlight the worst cases that came through the night. So my day starts by going through that. if there's any mortality, we have to check for that mortality and you know continue following up in the morning. So I read those messages and I pray to God there's no death and then by 8:00 I'm doing a walk around. So the walk around is first I want to see uh what is the situation

in the world. I want to be sure that we're giving the best care and we're using the protocol properly and ensuring that the patients are getting the best. So I do those walk around. I do a multiple walk around as the day goes. So most times my day starts really early.

Could you tell me a little bit about the realities of dealing with color patients?

Dealing with cola. First of all, we all know cola usually most times occurs in a resource limited setting. So one of the biggest struggle we have is resources. It's never enough. And then you have to get creative sometimes to manage these patients because and manage the staff because you come to a color unit, it's vomiting and diarrhea

consistently. You need IV fluids, you need this. We're always almost running out of um you know supplies. As of now I'm speaking to you, we've run out of antibiotics asycin is finished

really. very very essential resources you're running out of. Yeah, we're running out of acettomy. Our ringers is going down and before our stock arrives, we have to get creative. You have to start estimating weights and pushing for things. So a lot of the times it's me the patient and me the hospital management and it's uh it's a lot of things because this center the CTC is located inside the center and the center is also under equipped.

So does the MSF team take over a ward or a series of wards then in order to provide the care and how many patients on average are you dealing with? Okay. So when we arrived there was a a CTC created but on arrival the CTC had nothing. It had no circuit. It had no organization. It had nothing. So when we came we had to create the circuit create the IPC protocol create the water was a problem. Water border is a problem. We don't have water. So we depend on water trucking. So we have to truck in water. We had to find a triage. We created a temporary tent where we triage and stabilize patients on a good day and 24 hours we can admit over 30 patients with colera over 30. I have requested for with the logistics to help us fabricate you know color beds cuz our color beds currently are not enough. So I'm really proud of my team because with the little have made a lot of difference

and and and can you explain to me what MSF logistics team is and what they do? Okay. So the logistics team here provide us with almost every other thing we need stationaries waste disposal dust bins. When we're running out of water, we call them to say, "Look, we running out of water, we're running out of hand washing soaps or we need a new tripod stand for drips."

So, yeah, we we work something I'm requesting. Yes. So, yeah, we work closely together.

What difference do donations to MSF make in concrete turns on the ground to people like you doing the work you do? I think um for that if I'm going to talk to if a donor was in front of me one on one I would probably tell them they are providing that the as little as you the money they €1 no amount is even small because that amount someone donates is buying ringas it's buying OS it's buying chlorine chlorination tablets it's providing clean water that donation is saving one life at the same time saving a generation.

That donation is giving hope. That donation is equipping the nurses. That donation is equipping the doctors. That donation is going beyond medicine. Your donation feeds somebody. You give someone a hot plate of food. They probably and I can say from working and being from this part of the country a hot plate of food is a luxury at this point. Water is a luxury. These are basic necessities of life but it becomes a luxury in a place like this. And when you donate, you're lighting up a soul. You're giving hope for the future.

You're making a big difference. And if you want to donate to that Christmas appeal, we've added links in the podcast notes. That's it for this episode of the Medicine and Science podcast. Next episode is the last one of this year and we'll be taking you through the Christmas BMJ, but no spoilers. Until then, I'm Cameron Abassy. Thanks for listening.

Questions about this episode

Who speaks in "The shadow use of Gen AI in the consultation room"?

Cameron Abassi (Host), David Navaro (Guest), Charlotte B (Guest), Marcus Lewis (Guest), Tara Vinan (Guest), Rebecca Kums (Guest), Dr. Shafalad (Guest).

What is "The shadow use of Gen AI in the consultation room" about?

The discussion centers on the informal and unofficial use of generative AI, termed "shadow AI," in clinical settings, exploring its integration into patient care alongside traditional doctor-patient interactions. A novel concept called "triadic care" is introduced, emphasizing the collaborative roles of patients, clinicians, and AI in enhancing diagnosis and management, while also addressing the…

What is mentioned in this The BMJ episode?

14 people, works and ideas are mentioned, including Cameron Abassi, Artificial Intelligence, BMJ, David Navaro, Generative AI and the clinical encounter, Marcus Lewis. Each one on this page links to its own page with every episode that mentions it.

How long is "The shadow use of Gen AI in the consultation room"?

The episode runs 38 min; the transcript is about 5,500 words in 53 timestamped sections.

More from The BMJ

About this transcript: the episode's spoken words as captioned on YouTube, split into timestamped sections. Reclip.Me locates every person, work and idea in the text, keeps the time it is first named, and quotes the words around it verbatim. Last updated January 24, 2026.