Profile picture
Dr Murphy @DrMurphy11
, 45 tweets, 47 min read Read on Twitter
If you have an interest in #AI, #HealthTech or #PatientSafety - then please read this evidence based thread which tells the story of @DrMurphy11 & an #eHealth #AI #Chatbot.

Read on, or see single tweet summary here👇0/44
is not a 'nerd'; he's a pretty typical NHS consultant with an interest in #PatientSafety.

On 6th Jan 2017 a tweet about an #eHealth #AI #Chatbot in #NHS trials caught his attention, so he thought he'd take a look. 1/44

Evidence here 👇
Dr Murphy downloaded the @babylonhealth App & tried the #Chatbot with a few simple clinical presentations. It quickly became apparent that the #Chatbot had flaws, raising the question if the App had been validated as a triage tool?

As evidenced by👇 2/44
The flaws in the #AI algorithm included time critical Reg Flag presentations such as #ChestPain. Worryingly, Dr Murphy identified a risk that patients suffering #HeartAttacks could be reassured & not advised to seek urgent medical help! 😲

Watch video evidence here 👇3/44
Others on twitter shared Dr Murphy's concerns & questioned the robustness of the evaluation methods used in developing & testing the #AI #Chatbot. A concerned cardiologist suggested the issue should be flagged to the @CareQualityComm.
Evidenced here 👇4/44
Dr Murphy flagged the #AI flaws & #PatientSafety concern to Babylon, who updated the #ChestPain algorithm & offered a chat.

@babylonhealth also provided a link to an unpublished report outlining how the #Chatbot had been tested > arxiv.org/abs/1606.02041…

DM from Babylon 👇 5/44
Still concerned, Dr Murphy submitted his observations to the #CQC. The CQC contacted Dr Murphy & further telephone discussion & email correspondence took place. He felt confident that the regulatory bodies would step-up.

Initial submission to CQC & follow-up email 👇6/44
NB - not being a nerd, Dr Murphy didn't appreciate the role of @MHRAdevices in #eHealth App regulation at this time...

The #CQC indicated that they would involve the #MHRA, & may also play a role in reviewing the App if it was being used to access the Babylon GP services. 7/44
The CQC were sent links to the flawed #ChestPain triage algortithm & a few others including a #Headache triage that could potentially delay the presentation of someone with #Meningitis! 😲

Evidenced here 👇 8/44
All was quiet until Sept 2017, when Dr Murphy became aware that @mgtmccartney had also questioned the evidence base for the App.

This was discussed in a BMJ article > bmj.com/content/358/bm… & on @BBCRadio4.

Listen here (1st 12mins) 👇9/44
bbc.co.uk/programmes/b08…
Until this point Dr Murphy hadn't appreciated that @babylonhealth were claiming the flawed #AI #Chatbot was '100% safe'! 😲

Consequently he re-contacted the CQC & also made @ptsafetyNHS aware of the #PatientSafety concerns.

Email correspondence👇10/44
Dr Murphy downloaded the @babylonhealth #NHS111 App & again identified algorithm flaws that could risk #PatientSafety.

Including, time critical Reg Flag presentations such as 'referred cardiac pain'! 😲

NHS111 App triage 👇11/44
Babylon didn't acknowledge any concerns, & wrote a dismissive response to @mgtmccartney BMJ article suggesting the App had been validated by the NHS & was found to be 'completely safe'!

Babylonhealth response to BMJ👇12/44
bmj.com/content/358/bm…
Babylon did update their website to remove the claim that the #Chatbot had been 'independently' tested, however continued to claim that the #AI #Chatbot was '100% safe'.

Evidenced here 👇13/44
Many other #ClinicalGovernance concerns became apparent at this time; including the fact that the #AI #Chatbot App had bizarre T&Cs & apparently wasn't actually a #MedicalDevice 🤔

Evidenced here 👇14/44
It also became apparent that the regulation of #eHealth Apps was not as robust as you would expect.

Learn about Class I & II #MedicalDevice regs here👇15/44
It was apparent that others had independently tested the @babylonhealth #AI #Chatbot & found that it wasn't as reliable as other #eHealth Apps on the market.

Evidenced here 👇 16/44
Dr Murphy received a response from the @ptsafetyNHS team - stating that the @MHRAdevices compliance team was undertaking a review of the #Chatbot.

Evidenced here 👇 17/44
It gets a bit confusing now as the prior '@babylonhealth App' got re-named '@GPatHand' and was used to access both #NHS GP services & #Babylon private service (as far as I can tell). The #Chatbot appeared much the same.

Evidence of confusion 👇18/44
In Mar 2018 (a year since concerns were 1st raised) fundamental flaws that risked #PatientSafety were still evident in the #AI #Chatbot algorithms.

It raised the question - can a #eHealth #Chatbot be negligent?

Flawed #Headache triage 👇
19/44
Around this time @babylonhealth updated their website with YET ANOTHER #AI #Chatbot called #AskBabylon 😂

Dr Murphy had to put it to the test...

See questions regarding the #AskBabylon #Chatbot here 👇 22/44
#AskBabylon clinical test 1.

Cardiac #ChestPain triage = fail 😲

Evidenced here 👇 23/44
#AskBabylon clinical test 2.

#DVT #PE triage = fail 😬

Evidenced here 👇 24/44
#AskBabylon clinical test 3.

Ischemic limb = #AI #Chatbot fail

Evidenced here 👇 25/44
#AskBabylon clinical test 4.

This was a 'suspected cancer' presentation with weight loss and fatigue, that the #AskBabylon #Chatbot diagnosed as depression!

Another #AI diagnostic fail...

Evidenced here 👇 26/44
#AskBabylon clinical test 5.

This was a barn door #PE #DVT presentation that you would expect a medical student to get right...

Sadly the #AskBabylon #AI #Chatbot failed, again 😕

Evidenced here 👇 27/44
#AskBabylon clinical test 6.

By this point it was clear the #AskBabylon #Chatbot wasn't very good, so I tried it with a really easy #NoseBleed.

The result, was amusing...

The now infamous #Nosebleed triage👇 28/44
Of concern; the #AskBabylon #Chatbot presented itself as an 'online doctor' with the statement 'I'm here to make sure you get the medical help you need'.

In reality; this was a flawed, unvalidated, non-CE Marked #MedicalDevice!

Evidenced here 👇 29/44
To summarise; between Jan 17 & May 18, Babylon released 4 different #AI #Chatbots;
- Babylon
- Babylon NHS111
- Babylon GPatHand
- AskBabylon

IMO - flaws in the algorithms were readily evident & represented a clinical risk.

My own thoughts on why👇30/44
Remember, Babylon initially promoted their #Chatbot as 'independently' tested & '100% safe'.

@ASA_UK were informed of the misleading claim & raised it with @babylonhealth. By May 18, the claim was removed from the Babylon website.

Evidenced here👇31/44
Babylon have a tendency to look for a #PositiveSpin.

For example, they didn't like a recent @CareQualityComm report & tried to block its publication.

However, their website comments that they 'passed with flying colours'

Evidenced here👇32/44
The evidence thus far, strongly suggests that Babylon marketed flawed, inadequately tested #eHealth #AI #Chatbots.

It's therefore worth considering; how did they meet the #PatientSafety requirements of @NHSDigital?

Possibly because of this?👇33/44
But why would a cutting edge healthcare company promote flawed, inadequately tested #AI #HealthTech as '100% safe'?

IMO;
- The regulations allowed them to
- They thought no one would check
- Corporate culture & conflicts

#ClinicalGovernance👇34/44
But surely #MedicalDevice regulations, ensure that all #HealthTech is safe?
> Nope.

Currently class I devices (most #eHealth Apps) are not assessed or approved by @MHRAdevices.

Companies self certify; without external scrutiny

The Regs don't work👇35/44
New #MedicalDevice regulations come into force in May 2020. It is expected that most #eHealth Apps will require Class IIa 'approval' (not self certification).

You may have noticed, Dr Murphy has become a (reluctant) #eHealth Regs nerd 🤓

New Regs👇36/44
pharmatimes.com/magazine/2018/…
Getting back to Babylon. In June 18 the @FT ran a story about a partnership with @SamsungUK @SamsungHealth which described - 'AI-powered consultations'.

Given my experience with the Babylon #AI #Chatbots, I questioned that description.

The FT 👇37/44
You may recall that #PatientSafety concerns regarding the #Chatbot were first flagged to the @CareQualityComm in Feb 2017, yet more than a year later...

There was still no outcome 😕

April 18 CQC email response👇 38/44
Shortly after the @ft article; @HSJnews ran a story about 'safety concerns' that had been flagged to the CQC, in relation to the Babylon symptom checker #AI #Chatbot.

Yet again, the response from @babylonhealth was dismissive...

See 7 tweet thread👇39/44
Babylon said: “A number of vested interests would like to see us fail, & often resort to anonymous attacks & false allegations to do so. Some claim to be doctors & are deliberately misrepresenting information, which raises serious probity & professional standards issues.”
👇40/44
Dr Murphy wasn't happy with the statement from @babylonhealth, as it suggested;
1. @DrMurphy11 was a bogus doctor with probity issues 🙄

2. The allegations of compliants/concerns were false.

3. The #Chatbot had undergone rigorous testing.

🤨👇41/44
Despite the claims of 'transparency' - #Babylon / #GPatHand have failed to engage in open debate & questions remain unanswered.

In particular; if @HSJnews is correct, it appears the #Chatbot wasn't CE Marked when issues were raised in Feb 17.

? 👇42/44
To summarise this thread.
IMO; the Babylon #AI #Chatbot provides a good example of how NOT to do #HealthTech.
- Fundamental flaws
- Misplaced confidence in system
- No validation
- Confusing T&Cs
- Misleading marketing
- Dismissive of criticism
👇43/44
I hope the thread highlights the critical importance of #ClinicalGovernance & #eHealth #PatientSafety, & will help stimulate wider debate within the #HealthTech sector.

& PLEASE, drop the hype...

Usual T&Cs apply 👇 44/44
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Dr Murphy
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($3.00/month or $30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!