Back when I was a computer programmer (I’m now retired), AI was regarded as a toolkit of bits and pieces to do boring but necessary tasks - like routing a print-out to an appropriate printer in an organisation with 2,000 of them.
I went on a training course to write full-blown AI systems and my first was a diagnostic system for lung diseases. These were acknowledged to be (on average) slightly superior to doctors but the input was entirely text input, which rendered them all but useless.
The current ‘revolution’ revolves around the ability of the systems to directly read X-rays, blood-tests, patient history etc, eliminating the need for someone to examine the input and tell the computer about it.
An excellent, informative interview from Rory with Tom Whicher, brilliantly written as ever. Do I sense that the NHS is improving in recording useful pockets of digital data? Perhaps now we should ensure that they can seamlessly and safely interface? County Durham & Darlington NHS Trust's Organ Donation Committee is seeking to open a digital link between the donation register, the NHS App and the Electronic Patient Record. What are our chances of pulling that one off? Wish me luck!
This is really interesting - especially much of the Quality Improvement work in health and social care focuses on getting people to record and understand the data themselves so they can pinpoint the areas where changes can have the most impact. Most of this is done on excel or even transcribed in manually. I wonder how AI will impact on this, will it allow a focus on the human side of change? Or will the use of AI to make judgements on areas for improvement actually have the opposite effect, of disengagement from the change process because I can see the top down model of the existing system absolutely buying in to this!!!
For diagnostic related stuff (e.g. ML on imaging), a key issue seems to be the ROC (receiver operating characteristic) curve. ie how well the combination of technology and user fit the usecase. It can be very good. Or not.
For all anything that relates to operational processes and the corresponding data, the NHS seems to me to be a basket case: process and data semantic variations are messy, largely invalidating any insights; and for some reason, the whole thing seems to have the worst organisational memory of any enterprise - e.g. compare, say, A+E performance vs 2010, and corresponding clinician spends: back then, some effort was put in to improving processes; and it worked. Now, less so.
Personally, I find the RACI for any interaction with the NHS totally opaque and nearly always am surprises the I'm supposed to be setting up appointments based on events that happened months ago and which I, as an individual, don't have the resources to track.
Meanwhile, presumably, many missed appointments are contributing the 9k people who die each year and take so long to find that post-mortems cannot identify the cause of death.
Working as a clinical psychologist in the NHS some decades ago, I once received a referral from a consultant enclosing his letter to the patient’s GP referring to the patient as being “as mad as a box of frogs”. Thankfully this was before the days when copies of clinic letters were sent routinely to the patient as well as the GP but I still felt compelled to admonish the consultant for his use of a derogatory and clinically meaningless term to describe someone with a mental health problem.
Back when I was a computer programmer (I’m now retired), AI was regarded as a toolkit of bits and pieces to do boring but necessary tasks - like routing a print-out to an appropriate printer in an organisation with 2,000 of them.
I went on a training course to write full-blown AI systems and my first was a diagnostic system for lung diseases. These were acknowledged to be (on average) slightly superior to doctors but the input was entirely text input, which rendered them all but useless.
The current ‘revolution’ revolves around the ability of the systems to directly read X-rays, blood-tests, patient history etc, eliminating the need for someone to examine the input and tell the computer about it.
So I am very excited about the possibilities.
An excellent, informative interview from Rory with Tom Whicher, brilliantly written as ever. Do I sense that the NHS is improving in recording useful pockets of digital data? Perhaps now we should ensure that they can seamlessly and safely interface? County Durham & Darlington NHS Trust's Organ Donation Committee is seeking to open a digital link between the donation register, the NHS App and the Electronic Patient Record. What are our chances of pulling that one off? Wish me luck!
This is really interesting - especially much of the Quality Improvement work in health and social care focuses on getting people to record and understand the data themselves so they can pinpoint the areas where changes can have the most impact. Most of this is done on excel or even transcribed in manually. I wonder how AI will impact on this, will it allow a focus on the human side of change? Or will the use of AI to make judgements on areas for improvement actually have the opposite effect, of disengagement from the change process because I can see the top down model of the existing system absolutely buying in to this!!!
For diagnostic related stuff (e.g. ML on imaging), a key issue seems to be the ROC (receiver operating characteristic) curve. ie how well the combination of technology and user fit the usecase. It can be very good. Or not.
For all anything that relates to operational processes and the corresponding data, the NHS seems to me to be a basket case: process and data semantic variations are messy, largely invalidating any insights; and for some reason, the whole thing seems to have the worst organisational memory of any enterprise - e.g. compare, say, A+E performance vs 2010, and corresponding clinician spends: back then, some effort was put in to improving processes; and it worked. Now, less so.
Personally, I find the RACI for any interaction with the NHS totally opaque and nearly always am surprises the I'm supposed to be setting up appointments based on events that happened months ago and which I, as an individual, don't have the resources to track.
Meanwhile, presumably, many missed appointments are contributing the 9k people who die each year and take so long to find that post-mortems cannot identify the cause of death.
Did a consultant ever pen a letter starting "This grumpy old sod"? I'll bet a few have been tempted!
Working as a clinical psychologist in the NHS some decades ago, I once received a referral from a consultant enclosing his letter to the patient’s GP referring to the patient as being “as mad as a box of frogs”. Thankfully this was before the days when copies of clinic letters were sent routinely to the patient as well as the GP but I still felt compelled to admonish the consultant for his use of a derogatory and clinically meaningless term to describe someone with a mental health problem.
As ever, very informative. Thank you very much, all your work is very much appreciated.
Very interesting article/interview. Looking forward to the next one!
Nhs finance model needs to be replaced urgently.