You will be very familiar with the quality output from NHS111 sent to GPs from the standard version, where a human reads a list of questions from a computer to a patient, and based on their answers tells the patient what the computer tells them to do. Data shows the disposition is 8% to ambulance with a further 7% advised to attend A&E.
That process has now been fully automated into NHS 111 Online, and the number advised to attend A&E has reached 24.7%. Yes, almost one in four patients seeking help “when it’s not an emergency” is told that after all it is an emergency.
Over many years we’ve run audits, n>200k, where GPs labelled just 0.4% or 1 in 250 consultations as emergency. So we can estimate the proportion of false positives in the NHS 111 online dispositions at around 98.4%.
As you would expect, I have tested the system myself, in the interests of science. During a break in the weather we managed a cycle club run this morning and after 53 miles I am shot to pieces, so I tried “aching legs”. Answering just 6 questions, all truthfully and all negatively, except that the aching was all over, sure enough I got the result in big red letters: “Phone 999 now for an ambulance”
Folks, if you read the press release they are actually proud of this. I put the figures to Simon Stevens at a Reform meeting last week. Earlier he had spoken of the benefits of machine learning in assessing images, an application which makes sense as computers are good at comparing similar things and looking for differences. When it came to 111 he assured the room that results would improve when the algorithms had more cases from which to learn. They’ve had a million, how many do they want?
247,000 innocent patients have been advised by this abominable software to call an ambulance or attend A&E. Tariffs vary but if we take an average of £200 (ambulances much more) that’s around £50m, almost all of it overtreatment. We are told that ambulances and hospitals are under pressure and want to reduce demand… is anyone in Skipton House paying attention?
Suppose these million patients had been diverted to a qualified human to assess their needs and respond appropriately, that £50m could have been invested in front-line services so beloved of politicians.
My only hope is that most patients had the good sense to ignore the advice. But what have we come to when the NHS is spending money to make people afraid?
We get asked to bid for contracts with CCGs to provide online consultations through the intriguingly named “Dynamic Purchasing System”. Most of the specifications require an algorithm to triage patients automatically to an appropriate disposition. We explain why we won’t do that because it would be both wasteful and dangerous. We don’t win any DPS contracts.
It doesn’t have to be like this. Public money could be spent on evidence based interventions. Ploughing our own evidence based furrow can be lonely but we’re going to keep on doing what works. As far as we know at 20,000 per week we’re already the largest provider of online consultations, mostly from practices self-funding because they make such a difference to their workload and profitability. They absolutely love it.
Cheer up, believe that better is possible, right now, and smile with the staff at Witley and Milford practice.
Harry Longman
PS Barry Sullman showed me the numbers from his CCG this week, Balaam St winter A&E attendances fell by 24% in his first year with askmyGP. That’s the power of human intelligence.
Shudder at NHS 111 digital’s first million
Read this NHS Digital press release and shudder: “NHS 111 online hits one million triages mark”
You will be very familiar with the quality output from NHS111 sent to GPs from the standard version, where a human reads a list of questions from a computer to a patient, and based on their answers tells the patient what the computer tells them to do. Data shows the disposition is 8% to ambulance with a further 7% advised to attend A&E.
That process has now been fully automated into NHS 111 Online, and the number advised to attend A&E has reached 24.7%. Yes, almost one in four patients seeking help “when it’s not an emergency” is told that after all it is an emergency.
Over many years we’ve run audits, n>200k, where GPs labelled just 0.4% or 1 in 250 consultations as emergency. So we can estimate the proportion of false positives in the NHS 111 online dispositions at around 98.4%.
As you would expect, I have tested the system myself, in the interests of science. During a break in the weather we managed a cycle club run this morning and after 53 miles I am shot to pieces, so I tried “aching legs”. Answering just 6 questions, all truthfully and all negatively, except that the aching was all over, sure enough I got the result in big red letters: “Phone 999 now for an ambulance”
Folks, if you read the press release they are actually proud of this. I put the figures to Simon Stevens at a Reform meeting last week. Earlier he had spoken of the benefits of machine learning in assessing images, an application which makes sense as computers are good at comparing similar things and looking for differences. When it came to 111 he assured the room that results would improve when the algorithms had more cases from which to learn. They’ve had a million, how many do they want?
247,000 innocent patients have been advised by this abominable software to call an ambulance or attend A&E. Tariffs vary but if we take an average of £200 (ambulances much more) that’s around £50m, almost all of it overtreatment. We are told that ambulances and hospitals are under pressure and want to reduce demand… is anyone in Skipton House paying attention?
Suppose these million patients had been diverted to a qualified human to assess their needs and respond appropriately, that £50m could have been invested in front-line services so beloved of politicians.
My only hope is that most patients had the good sense to ignore the advice. But what have we come to when the NHS is spending money to make people afraid?
We get asked to bid for contracts with CCGs to provide online consultations through the intriguingly named “Dynamic Purchasing System”. Most of the specifications require an algorithm to triage patients automatically to an appropriate disposition. We explain why we won’t do that because it would be both wasteful and dangerous. We don’t win any DPS contracts.
It doesn’t have to be like this. Public money could be spent on evidence based interventions. Ploughing our own evidence based furrow can be lonely but we’re going to keep on doing what works. As far as we know at 20,000 per week we’re already the largest provider of online consultations, mostly from practices self-funding because they make such a difference to their workload and profitability. They absolutely love it.
Cheer up, believe that better is possible, right now, and smile with the staff at Witley and Milford practice.
Harry Longman
PS Barry Sullman showed me the numbers from his CCG this week, Balaam St winter A&E attendances fell by 24% in his first year with askmyGP. That’s the power of human intelligence.
What you can read next
The Flaw of Averages
Dear Mrs May
Making it easier for patients (shock!)