Hi all! I have a question for you: as an asthma patient, how do you think AI could potentially help meet needs you have that aren't currently being met?
For example, I have been reflecting on how doctors could potentially use AI to more accurately diagnose, treat, and educate patients. For me personally, the big thing would be to somehow use AI to make up-to-date studies and diagnosis criteria widely available to as many GPs, EMTs, emergency care workers, pulmonologists, and respiratory therapists as possible.
For example, many asthmatics follow their asthma action plans and go into the A&E (ER, urgences etc.) in a severe attack but because they aren't wheezing (many asthmatics don't), have normal sats (many studies have shown this is not a good way to assess asthma control), a normalish peak flow (which is possible and especially if a patient is on steroid inhalers + already on pred etc.), and can speak in full sentences, they aren't properly treated and are sent home.
I personally would rather be screened first by a nurse + an AI tool that would ask me a ton of questions and scan my responses against millions of up-to-date papers and textbooks etc. and then have the AI advise the team on treatment than to see a weary generalist who is in a hurry and harkening back to a box they memorized in medical school that says all patients in a bad attack have blue fingers!
I also think AI could help educate us on a lot of other diseases that coexist with asthma like sleep apnea, GERD, rhinitis, and adrenal insufficiency as well as nutrition, recommended vitamins, and so on. Like you could see a physician assistant and do the training(s) via computer then ask them questions and get counseling. (I am also thinking about the doctor shortage and super long waiting times--like a 1+ year to see a specialist!--and how AI could help with that).
Right now, I see my pulmonologist for four minutes at a time and learn everything else from research online and support groups like this one
I totally get privacy concerns around AI and so on but I'm more trying to envision what a patient-friendly tool would look like and how we could build backward from that, if that makes sense.
Thanks so much in advance for any insight you have!
Written by
hilary39
To view profiles and participate in discussions please or .
AI could hopefully give us some constancy on approach to diagnosis.
I was just a bit bored this evening and ticking what my present consultant is doing against the Nice guidelines, and he is not following them.
Each time I go to a different hospital they rediagnose my asthma using a different method. If they all used the same method, no rediagnosis would be needed
The NICE guidelines for asthma are a bit of a dumpster fire anyway though. Not just my opinion, though I have read them for work which is medical related - ALUK weren't happy with a lot of it and neither are some drs, as they appear to prioritise cost savings by trying to 'undiagnose' asthma as much as possible. I probably wouldn't even count as having asthma if those were followed rigidly, and my asthma is confirmed even by my consultant, despite his obsession with FENO and eosinophils.
So I don't blame him for not following those. I feel like an experienced clinician shouldn't have to rigidly stick to guidelines (especially as there is often conflicting advice across them), but I do want to feel like they have a solid reason behind their decisions that is.based on good evidence and not just their hobby horse.
In terms of AI, it could be helpful for providing the knowledge that healthcare professionals need and isn't always easy to access. Perhaps something to analyse/remind human HCPs about the quality of evidence. I find some are obsessed with what they call evidence-based medicine but can be a little selective and don't question the assumptions that the evidence is based on.
Asthma and COPD, along with other areas, have a known problem with the fact that the patients in trials are not representative of the real-world patient population. I wonder if AI could be used to help design trials which are more reflective of real world patients and practice?
Yes I definitely agree and have the same question about how we can get HCPs evidence-based up-to-date guidelines.
I come at this also from my angle as a patient with adrenal insufficiency, too. For example, my current endo insists HC can't cause weight gain since it's a replacement dose and I get so frustrated because just as a layperson I've been able to find several articles that show it's an incredibly common side effect of AI treatment (not to mention I'm in a Facebook group of 7K people devoted to helping people with AI lose weight since it is such a common problem!)
I think the evidence she is thinking of is charts in books or papers she read at some point in the past. I imagine it's hard to keep up with both a giant patient backlog + staying on top of research and I am empathetic to that. But it's frustrating when you can't get basic treatment and education because your doctor isn't up-to-speed. (She also knows nothing about how to treat my hormones which are all v low.)
I asked this same question in an adrenal insufficiency support group and a woman said it took doctors six months to diagnose her and she became severely ill and almost went into crisis during that time and when she typed her symptoms into a medical chatbot last week the first thing it said after two seconds was that it sounded like Addisons. 🙄 She said she felt seen by the bot in a way she hadn't by a slew of doctors in half a year.
Love the idea of AI helping design trials and conduct more informal surveys of patients to reach a more diverse audience.
I did have a moment reading your post where I read AI as adrenal insufficiency and was thinking well it never helped my asthma 😂😂
It does seem common for many medics to cling to certain things they 'know' from past evidence and not be willing to admit doubt, or to acknowledge evidence that disagrees with their favourite theories. I also have sympathy as there is a lot to keep up with and a lot of guidelines, but I would like to see more acknowledgement from them that what they 'know' might not apply in all cases, or there might be updated evidence. Or minimal evidence in some patient groups. Absence of evidence is not evidence of absence!
I don't know if AI can help with that, it might be more related to how they are trained. But I would like to see more healthcare professionals admit that they don't know everything and how to handle uncertainty. Perhaps if they have regular access to easy summaries it would give them more of a sense of where the evidence is and the existence of gaps?
Very interesting and true. I feel like creating consistency would quickly expose how different every hospital / clinic / region / country is. Like in the UK SMART and FeNO testing are common protocols but I am an American living in France and no one ever talks about SMART and my FeNO has never been tested in either country.
I've never heard of the NICE guidelines, I'll look them up.
It sounds like a great idea, but I'm not sure how many of us who are bad enough to need an emergency trip to hospital are going to be able to answer a 'ton' of questions at that point in time! I'd rather see more effort put into training medical staff to actually LISTEN to patients and to assume that if the action plan states they need to go to A&E then they probably do. I had a very wise GP once who was always open to hearing her patients' thoughts because as she said 'no one else is ever going to know your condition as well as you do'. But sadly, medical professionals like her are a rarity!
However, with medical records being centralised, a system which could flag up key pieces of information might be useful. E.g. medic tells system that this patient has been admitted to A&E with breathing difficulties and the system flags up that this patient's serious asthma attacks are always silent so they're not sent home just because they're not wheezing. Or they enter what seems to be a 'normal' peak flow reading into the system which tells them that this reading is actually worryingly low for this particular patient because their normal reading is so high.
I can also see how AI might have a use in guiding medical staff toward asking the right questions of a patient in emergency care who does not have a diagnosed condition to explain their symptoms. For example, before I was diagnosed with asthma, I was rushed to hospital with breathing difficulties on numerous occasions over the years. Every time, I was asked if I HAD asthma, to which the answer was always 'no' (like many people, including a lot of doctors, I thought you had to wheeze to have asthma, and no one had ever suggested it as a possibility). Not one medic ever thought to follow up the question with 'has anyone ever TESTED you for asthma?' If only they had, I could have been saved years of trouble and the NHS could have been saved the cost of my multiple trips to A&E. As it was, I was usually packed off home with instructions to 'worry less' as it was all put down to anxiety! I will be forever grateful to the student doctor who had the common sense to follow up the first question with the second one - I just wish it hadn't taken so long!
Yes love this--both the idea of better more centralized and easy to access records and guiding staff to ask the right questions.
I keep thinking about how it takes Parkinson's patients an average of five years to get diagnosed (critical time they could otherwise be on meds to slow the disease's progress) and how it's often because the doctors don't ask the right questions.
I am working on an op-ed about how AI could meet patient needs. I come at it from the perspepctive of a patient both with severe asthma as well as adrenal insufficiency caused by taking prednisone so often for my asthma who struggles constantly to get care for both conditions.
My GP also says the same exact thing, that I know my condition(s) better than anyone. We are lucky to have such supportive GPs!
I hear what you are saying and I am so sorry to hear you are juggling asthma with COPD and cancer, too.
I think maybe I was unclear above--I am picturing this as a hybrid of an AI + doctor. I wouldn't want a computer treating me either
Something I've been thinking about a lot is there are 10,000 diseases, 700 symptoms, 20,000 medicines plus 5 million medical papers get published each year.
I'm trying to figure out how we could combine what AI is best at--which is to say accessing all of that information in a millisecond--with what humans are best at which is using their judgment to make decisions.
I personally would want an AI to help a doctor diagnose and treat me but that is a very personal thing and because I have had so many bad situations when it comes to doctors relying on incorrect information.
For example, I also have adrenal insufficiency and have gone into the A&E in a crisis and not gotten proper treatment initially because the doctor mistakenly thought you have to have low blood pressure to be in crisis (it can actually be the opposite and quite high as was the case with me) and in another case, a doctor told me I didn't need to worry because my potassium was low not high (he was mixing up primary and secondary adrenal insufficiency--it's high in the first and low in the second and when both are off it's a sign of crisis).
I've also had issues in the A&E getting proper treatment for an asthma attack so I am just trying to think how AI could meet those gaps, if that makes sense.
I wouldn't want AI in charge, but it could play a role in making those records better and easier to access and use. I find they're often very poorly created with important bits missing, and/or difficult to access, and the human has to wade through a lot of badly recorded data. And then humans are rushed and they create MORE poor records! I know my GP surgery gets very frustrated by this and also that my consultant tends to be seeing only a small amount of the total.
Discharge summaries for example are created by junior doctors with no training in how to create a useful summary, who are rushed for time and have to find pieces of information from various different systems which don't talk to each other well or have different ways of recording information (this is what NHS IT and systems are like). My discharge summaries tend to be misleading and/or missing useful information and context a lot of the time. It wouldn't replace human interpretation, but it might help if they had better access to all the data and a fast way of bringing together all the important facts and putting them into a useful framework.
For example - prompts to record whether heart rate or peak flow or blood tests are pre or post treatment and what treatment was given. Often A&E treatment is just missed off because the inpatient systems are different! The resulting picture is incomplete and unhelpful and can't be interpreted properly by the human medics. And in my experience, it does cause problems repeatedly.
Well I have asthma and a previous wife had asthma long before the day of the computer. Manged just fine.
Before putting ones life/care in the hands of a computer remember computers are programmed by humans, are vulnerable to power cut and even hacking. If AI takes over then God help us.
I've nearly been given the wrong drug or dose at least twice with paper records - I only wasn't because I know what I should be having, which isn't the case for everyone. I'm sure people managed without, but it doesn't mean there isn't room for improvement. Some people managed without modern treatments too - and some didn't and they're not around to talk about it.
As you say, computers are programmed by humans - but humans can also make serious errors even without the help of technology. It doesn't necessarily have to be a choice between a rubbish system, no computers t all, or HAL.
Well GP sharing is not paper records, those days have gone. The health/hospital worker just access you GP records via computer. You are opted into record sharing but you can opt out.
I'm aware of these, but the implementation is still uneven -and you were saying people managed without computers before. I was just pointing out that human error exists even without technology, and 'managing' doesn't mean we can't try to improve things.
Some hospitals do actually still use paper records, or have done very recently/are in the process of switching, and/or have different systems in different parts of the same NHS Trust.
My point was that currently the records are still made by humans before they go into the system, and can have a lot of mistakes and gaps, which isn't necessarily helpful. I don't see AI as something that should take over, but it's possible it could have a role in improving the quality of the record-keeping - not as a replacement for humans, but as an assistance.
A bad record being shared is if anything worse than a bad one not being shared, as it just spreads the mistakes. So it's not enough just to share it across different parts of the NHS unless there's also going to be an improvement in how those records are actually created to start with, and how useful they are to other clinicians viewing them.
I see where you are coming from but with respect I can't agree.
Almost a 1/5 of hospital use paper records this is a danger, I agree with that but AI would not help as the paper records would be programmed into an AI ready computer - So the same information would be available to the doctors.
GPs have a system of both paper and computer records and the paper records are help until a patient leaves the practice,
But hospitals would access the online records held by your GP, I know this as a fact as I have seen a doctor in urgent care access my records - he asked did I mind even though I don't think he had to.
The whole point here is AI and would it work - I know GP sharing works and I can't see how a AI computer could help me.
What would help is if all hospitals held patients accurate record in a computer - That would be better than going to trusting an electronic device and all the cost involved. just my opinion but I have seen GP record sharing at work.
I don't think we're talking about the same thing here. I'm talking about how, or whether, AI can help human clinicians to create better records to start with, that are a better reflection of what is actually going on with patients and what happened during an admission or a consultation. And whether the clinical information in that record is helpful for other clinicians who access them. Not about the systems for sharing those records with others.
Again, I don't think AI is going to replace this - because human listening skills are still very important here. But I'm wondering if there's a role for AI in helping the clinician with the information and structure to put down the key facts about their encounter with a patient in a way that's useful to others and that doesn't take them ages.
For example, as others have suggested here, they enter a peak flow of 350 and 'no wheeze on examination during acute attack'. Then they might be prompted to enter, or AI fills in something like 'this patient's best peak flow is 630 and this reading is 55% best, they don't typically wheeze with asthma'. Or if this information isn't available.for a patient, to perhaps fill it in with something like the latest guideline: eg 'No wheeze on examination'* [*GINA guidelines 2023 state wheeze is not always present in acute asthma.]
This does require access to previous shared information of course, but my point is that in this hypothetical scenario, it would be used to help improve the quality of the information provided about that patient. I have no idea whether what I'm suggesting is plausible, but I'm thinking about ways that AI could help improve the quality of clinical information held about patients.
For example, as others have suggested here, they enter a peak flow of 350 and 'no wheeze on examination during acute attack'. Then they might be prompted to enter, or AI fills in something like 'this patient's best peak flow is 630 and this reading is 55% best, they don't typically wheeze with asthma'. Or if this information isn't available.for a patient, to perhaps fill it in with something like the latest guideline: eg 'No wheeze on examination'* [*GINA guidelines 2023 state wheeze is not always present in acute asthma.]
I certainly wouldn't want to be cared for by a computer either - but a computer can search and recall data much faster than a person can, so there would be a definite benefit having a system to do that part of the job for you as a backup.
For example, medical staff do not usually read patient notes thoroughly. This isn't necessarily a criticism, as they are often too rushed to have the time to do it, but it does mean that patients are vulnerable to being given inappropriate treatment if they themselves don't flag up past issues from their medical records. The patient may not always be able to do this themselves, and also, they may not be aware that certain things in their past medical records are relevant.
I knew a man who was landed in hospital by a doctor who prescribed a medication without realising that he was already on another medication which interacted badly with it - the poor patient ended up in hospital for weeks and was left with permanent health issues as a result. A system which would automatically flag up 'this patient is on medication A - medication B is dangerous to take at the same time' would have prevented this happening.
I've also on more than one occasion had a doctor prescribe me something that I have had a severe reaction to in the past - fortunately, each time I've remembered and been able to tell them, but what about people who have poor recall for these things? Also, I once had a GP who altered a medication I was on because he assumed that I was taking it for one condition, whereas I was actually taking it for a different condition - and the new medication would have made me worse! All these issues could be avoided by AI.
I am aware of this. But it still relies on the person accessing the record to read it properly. As far as I am aware, it does not yet automatically flag up potential issues (e.g. a doctor adds a prescribed medication to the patient's record and the system pops up a message that says 'WARNING: Patient cannot take this medication alongside their current medication'. Or A&E enter what they think is a non-worrying peak flow reading and get a message telling them that this is 50% of the patient's normal reading.)
Care plans, outpatient appointments, inpatient stays, discharge letters, personal preferences/decisions and clinical contact"
That is a lot of info from the GP..
To your point - Whoever reads your record should know for example - WARFRIN can't be taken with RIVEROXIBAN or Morphine same time as example barbiturates.
How would AI help? Could be human error programming same as there i suppose someone could read a patients record wrong.
I will say this - Half of time hospitals don't seem to have accessed GP records as they ask patients for details. The same if AI was available, would it always get used, even it was viable. Lots of iffs and buts - I know exactly what I am taking and know what I should not take, Safest way.
As you say though, that is a lot of information! All the information in these shared records is also going to be difficult to read and digest quickly, especially in an acute situation. Things like how the patient presents or whether their peak flow is ok or not might get missed. And computers are more often better at things like complex calculations and consistently performing an operation like 'remind that this is a drug drug interaction with X' or 'did you mean to order this dose which is 10x the standard' than humans, especially when those humans are tired at the end of a long shift.
I'd prefer to have a system which actually checks automatically for the relevant information and reminds clinicians that they shouldn't give X drug with Y drug patient is on, or a patient is allergic to Z drug or has a condition that makes B drug unsuitable. Sure you can have human error with programming, but you can also have checks and balances on that process to reduce it as much as possible. How is it better to rely on standard human error and memory, and the ability to digest a huge amount of complex information from a record without missing anything important, while also trying to stop a patient dying?
There are plenty of things clinicians should know, or may know in a general way but could fail to remember. Or there may be more obscure interactions with lesser used drugs that they're not familiar with.
It's great that you always know what you're taking and what you shouldn't take, but that's not always realistic. What if you were unconscious or otherwise unable to communicate the important points? Being injured or severely ill also tends to affect people's memory and ability to communicate, or maybe they have low health literacy to start with, or they're a child and the parent isn't available or is injured. I have a summary for this type of situation with the important points, but not everyone does or knows how to create one or that they even need it.
I'm definitely a fan of medical alert bracelets for important conditions, and if you can access that person's records then great, though as I said above it's still going to be likely a lot of information to digest all at once in a potentially acute situation. But you can't rely on a patient being able to provide the key information themselves - again that's a lot of potential for human error. If AI could provide a reminder of key points, then perhaps clinicians wouldn't need to read through huge amounts of information to find the important points, or rely on the patient to tell them.
Also, if a patient has yet to be diagnosed, they do not know what condition they have! I know someone who spent a year being dismissed as 'worrying about nothing' before collapsing and being rushed to A&E, to be diagnosed with a cancer that, by then, was no longer treatable. Yes, the doctors she saw over the past year SHOULD have picked up on her condition anyway, and a busy doctor could still decide to ignore the AI system that was prompting further checks, but it might increase the chances of an early diagnosis for someone like this, instead of having them dismissed as suffering from 'anxiety'.
Excellent points in here, thank you so much Lysistrata. This is super helpful as I'm really trying to brainstorm all the ways AI could fill gaps. What you said is exactly what I mean:
How is it better to rely on standard human error and memory, and the ability to digest a huge amount of complex information from a record without missing anything important, while also trying to stop a patient dying?
...What if you were unconscious or otherwise unable to communicate the important points? Being injured or severely ill also tends to affect people's memory and ability to communicate, or maybe they have low health literacy to start with, or they're a child and the parent isn't available or is injured. I have a summary for this type of situation with the important points, but not everyone does or knows how to create one or that they even need it.
As someone who both works in a G.P surgery and is a regular patient in hospital, shared records is not simple. The hospital records and the G.P records operate in different systems, therefore only certain parts of the record can be accessed. For example when I am admitted the hospital can only see my g.p summary such as meds and allergies and not my full records. At the G.P surgery I can only see discharge summaries and treatment when the hospital send them over.
I studied technology in regards to asthma care, and the data shows it massively improves asthma care. Currently there are two asthma guidelines running alongside each other which is confusing for health professionals. If you could enter patient symptoms and meds, technology could then analyse using guidelines and best evidence to suggest a treatment plan and guide Drs. This doesn't replace Drs knowledge but is used as an aid.
I have found and seen with my own eyes - Hospitals share GP records. I have seen on my online patient records Lancashire Health have several times shared my medication, my mental health status not once but several times. When your record is shared, you can see that under 'Shared GP Records" - I don't know about the other way round, I don't think GPs can access hospital records.
I have Bladder Cancer as well as Heart and Lung Problems and before I was diagnosed the cancer I rang 111 as Blood in urine - it cleared but came back - I rang 111 again and I got a ring from urgent care doctor asking me to go to the hospital with a sample, he thought it was an infection and before he prescribed, he checked what medication I was taking on his PC - I was then given a script for antibiotics.
Which I took but bleeding started again, when my GP got to know about this she referred me to Urology under I think a 28 day rule.
My sole point has been and nothing else, I doubt AI could have helped my situation at all, All computers have to be programmed by humans - that to me makes AI pointless and costly.
They only share very basic information such as medication and diagnosis, Not extensive history. It is extremely frustrating that I have to continually explain my asthma and when it started as the hospital cannot see my previous g.p consultations.
I appreciate for your situation AI couldn't improve things, however there are a hundred different scenarios that can help. Yes technology is programmed, however the evidence is clear and the national review of asthma deaths as well as asthma and lung uk state that technology is the way forward.
I am not against technology at all. Look how X Ray and CT Scans ect changed the face of medicine.
AI is another matter - In fact the defense industry and governments are more than interested in AI - Mistakes would be unaccountable.
Asthma I have COPD and Bronchial asthma and it can be distressing - I think being the problem is so huge, there should be dedicated data bases for things like asthma heart problems and diabetes.
It would take humans to to set up a database and humans to access it. To me that is the safest way.
It’s even worse than hospital and GP operating different systems: when I go to my diabetic clinic, the nurse doing the initial checks can see my GP summary and latest test results. The consultant, who really needs the results, can’t. There is apparently no way for the nurse to transfer the results from her computer to the different system the doctors use - she can’t even print them off so I can take them in with me. I either have to get the surgery to print them before I go, or log into the NHS app on my phone and let the consultant scroll through the results herself. Occasionally I can persuade the surgery to email them to me, then I email them to the hospital before the appointment, but for some reason my surgery (tho otherwise very good) doesn’t like using email. I’m with Lysistrata on the usefulness of flags on a system, but basics like getting the computers within a clinic to communicate surely have to come first!
I agree - I've experienced similar as a patient where A&E systems especially cannot be connected to other systems in the same trust and it makes it look as though I barely needed to come because nothing from A&E is available to my consultant, the person writing the discharge summary, or sometimes even to my GP!
I've heard from friends working in the NHS that it's very frustrating from their point of view too. My GP surgery also may not get information from the hospital I go to for similar reasons of systems not talking. They have asked me for physical copies for them to scan in. Mind you, discharge summaries are so often a bit rubbish that this may be a blessing in disguise for me...
It's so interesting what you say about studies showing technology improves asthma care. What you said here is exactly what I'm picturing
If you could enter patient symptoms and meds, technology could then analyse using guidelines and best evidence to suggest a treatment plan and guide Drs. This doesn't replace Drs knowledge but is used as an aid.
I think another area where AI could be useful would be in looking at your general health rather than specifics, as in you might visit a GP several times over a year or 2 for different symptoms, each time that visit is treated in isolation because GPs are 2 busy, but there must be cases where a link is missed and treatment delayed. So in the case of people on this forum how many are pure asthma, asthma/copd or just copd? I mention this as that question just occurred to a nurse practitioner on my last visit with an ongoing exacerbation.
I know what you mean - especially for things like autoimmune conditions which can produce totally different symptoms at different times, and because you rarely present with all of them at once, it can be misdiagnosed or not diagnosed at all. No GP is going to read your entire medical history every time you come in, and the patient themselves isn't going to realise that all these different issues are linked. But an AI system could spot that the different symptoms you've presented with in the past are all linked to --------------condition and could pop up a note to the doctor to suggest you get tested for that condition.
I am sure they will find a role for AI but humans need to start by always checking records of Medication. Allergies, Test Results and maybe feed that into an AI capable computer but would have to solve a huge problem names - To type a name into a computer would take Name, D.O.B and NHS Number so the AI PC can locate the correct patient - Would this help?
That should be feasible. Currently, I have to give my name, address and DOB whenever I call or visit the surgery or hospital, so entering Name, DOB and NHS number should not be any more difficult than that.
re: AI -- if things go wrong, who would be responsible?
I think a network sharable between hospitals and A&Es showing that patient X already has been diagnosed with asthma, showing recent tests, allergies, what worked, what did not work etc -- would have a more immediate impact. Every time I move I have to bring 200 pages worth of documents showing that I have asthma. Every hospital wants to do their own tests etc. Like many on this forum, I have high lung function for my height, I do not weeze, and NO does not show any abnormalities.
I guess it would depend on what 'went wrong' and how AI was being used. Obviously, if AI were being used independently with no oversight or input from medical professionals, then the capacity for things to go wrong would be huge. But if it were being used more as a tool to 'nudge' a medic to ask specific questions or to flag up a potential issue that might have been overlooked, then I can't see things going any more wrong than if the medic was just dealing with things on their own.
There are a lot of comments talking about how we don't need this because medical records are already available etc, etc, etc. But that ignores the number of people who get sent home with no treatment at all or with treatment that ends up being harmful because the person treating them didn't spot something vital in their records. An AI system that made decisions about patient care independently would be very very scary. But an AI system that flags up important facts that the medic has potentially overlooked could be very useful.
Thanks so much for your thoughtful replies. I am working on an op-ed on this topic and I really wanted to hear patients' perspectives so this is super helpful.
I am going to summarize the big things I took away from this thread and feel free to reply with anything else I should add.
Things AI could help with:
1) Records
-helping all the people in the loop access the same records (GPs, nurses, consultants, doctors)
-helping doctors quickly scan lengthy patient notes for relevant details
-helping point out a patient's past presentation of symptoms (e.g. patient often has a normal peak flow because they typically test at 120% of predicted FEV, patient didn't wheeze during last severe attack)
-helping write better discharge summaries and flag information doctors should include (e.g. heart rate, peak flow, blood tests)
2) Diagnosis
-providing doctors with a summary of guidelines and best evidence (updated constantly) to check against their own more intuitive diagnosis
-creating more consistency across clinics, hospitals, GP offices etc. i.e. what kind of guidelines should they use? (e.g. NICE)
3) Treatment
-flagging potential drug interactions
-keeping doctors up to date on new recommendations and guidelines on medicines (e.g. there is a huge movement to make biologics accessible earlier so people have less of a risk of developing long-term complications like adrenal insufficiency from prednisone; if something like this changed, this could be an easy way to notify all doctors instantly of guideline changes [particularly important with something like biologics which are so hard to access])
4) Involving more patients in studies, trials, and discussion
-AI could help doctors survey their patients much more frequently to keep a pulse on their needs
-AI could help create more diverse trials and studies that reflect real-world patients by helping doctors get best practices for community engagement (e.g. I just attended the European Respiratory Society Congress and was in a session talking about this exact thing and one group shared how they've been working with churches, mosques, and non-governmental agencies to recruit patients from as wide a possible background as possible in terms of ethnicity, religion, age, socioeconomic background and so on)
Things people are anxious about when it comes to AI:
1) What if the AI is wrong?
2) What if doctors become too reliant on AI?
3) Since humans make machines, how will we prevent human error and bias in AI?
4) What if the AI is hacked?
5) What if there is a power failure and doctors can't access AI tools (though presumably this would equally threaten a lot of life-saving equipment in any hospital!)?
Hi Hilary, this is really interesting! Do you have a journal or other publication in mind to submit to?
One thing that occurred to me as I was reading the list was about this:
-helping point out a patient's past presentation of symptoms (e.g. patient often has a normal peak flow because they typically test at 120% of predicted FEV, patient didn't wheeze during last severe attack)
-helping write better discharge summaries and flag information doctors should include (e.g. heart rate, peak flow, blood tests)
I think part of where AI could be helpful here may also be in prompts to provide context when clinicians create records. Not just to provide context for what is normal for individual patients (though this is really important I think) - but also for the background and context of when a test or reading was done.
I get so much crap and often worse treatment or misunderstandings because someone will do a test like arterial blood gas post-treatment when I'm feeling better, or while an oxygen-driven neb is running - but not record any of that background alongside the results! Or they won't record pre and post treatment peak flows properly to show reversibility, or heart rate pre and post neb to show the fact that heart rate can actually come DOWN post neb because the body is working less hard to breathe. Which can support the idea that you can and should give nebs to a patient during an asthma attack whose heart rate is high because asthma itself can cause that.
My consultant likes to cruise through my notes from admissions and seems to get some weird ideas that I know are incorrect from this kind of failure to provide context - and other consultants have been the same.
I think this context and timing is really important in asthma, and I would bet it is also for other conditions (an example that comes to mind is context for blood sugar readings in diabetes - has the patient eaten recently, and what did they eat? Or in AI - what time of day was the cortisol test taken, and did the patient withhold steroids?)
Totally agree. I might have more questions for you as I go as I really appreciate your thoughtful feedback and insights (as always) and because I want to be sure I am capturing nuances like what you're saying here.
I am thinking of writing the op-ed to a large newspaper in the US as 1) a way to point out gaps in the system and do some advocacy around severe asthma and adrenal insufficiency 2) start a conversation about how AI could fill some of those gaps.
I'll keep you posted as I go if that's ok as I might bounce a few more ideas off you.
Please do! You're probably more than aware of this, but there are some journals which would probably be interested too, especially from a patient perspective. Very happy to hear from you and discuss.
in my opinion, AI could be very advantageous in patient management . Before that can happen, diagnostic equipment needs updating.
I lived on the continent and for almost 10 years had pulmonary function tests carried out by a technician and state of the art technology.
My results would be clearly printed out to include previous data, charts and summaries. The computer would generate potential diagnosis such as …results indicate both obstructive and restrictive lung disease or FeNo 50% increase on previous result…
Last year, I paid privately to have my lung function tests done as I wasn’t being assessed by GP or Nurse.
To my astonishment, I was the first patient to use the new machine in Glasgow. The same machine I had been using for almost 10 years.
The nurse was reading instructions and was not sure how to read the data. The consultant was also confused, it appeared to be a bit high tech for him.
My diagnosis was asthma stable but you are too fat which is why the machine suggests obstructive and restrictive disease …nonsense!!!
The machine calculated my results based upon the data fed in by the nurse such as sex, height and weight and the results of each test.
I complained to my GP and brought a copy of my results to compare. This demonstrated that my condition was no worse as I am known to have poor diffusion at the Alveoli.
My GP response was that he was an old man not used to this technology. Fine, but he should have allowed someone tech savvy to interpret the data.
In my case, I was stable however, I dread to think what the outcome of an unstable silent asthmatic might have been.
Therefore, typical U.K., Lauch a new piece of technology that no one has the training for. The nurse was only there to feed in data but also no clue how to interpret it!
I paid almost £1000 for the pleasure.
The infrastructure should be built before we start using a half baked idea … my opinion.
Content on HealthUnlocked does not replace the relationship between you and doctors or other healthcare professionals nor the advice you receive from them.
Never delay seeking advice or dialling emergency services because of something that you have read on HealthUnlocked.
How reliable is a spirometer in accessing asthma severity or a flare up?
Severe asthma and support in the community & Oxygen
tiger milk mushroom
Peoples experience of biologics
Could this be Asthma?
