back to article 'Computer algo' blamed for 450k UK women failing to receive breast screening invite

Up to 270 women in the UK could have died of cancer due to a "computer algorithm failure" dating back to 2009 under the NHS' breast screening programme - British health secretary Jeremy Hunt said today. In a Parliamentary statement, Hunt said nearly half a million women between the age of 68 and 71 were not invited to their …

Silver badge
FAIL

Death by Computer

What a time to be alive....... I'm sure some tech will end up taking the fall for this. Ignoring that they were probably under totally unreasonable time constraints and mismanagement.

I bet they won't even be able to produce specification documentation for the project, leave out dreams like comprehensive testing reports.

Sooner or later the IT industry will have its Harrow and Wealdstone rail crash incident and some probably really crap regulation will follow.

16
3
EBG

Re: Death by Computer

More likely the opposite. The drive to " digital railways" could replace tried and tested rail safety systems - fail-safe relay logic - with "IT standard" tech.

5
0
Anonymous Coward

Re: Death by Computer

Someone made a bit of a boob?

3
28

This post has been deleted by its author

Silver badge

Re: The real question is...

It's not as simple as that. Screening is... screening. It gets you sent on for further investigation and tests, not straight on to treatment. Yes, there will still be false positives at the end of that process, but not at the screening false positive rate. That's the kind of thing you take into account when you decide whether to do a screening programme.

25
1

Re: The real question is...

I would have to say that your numbers are rather doubtful given that unless there are additional checks such as biopsies etc. before treatment then there's something very very wrong. But said people would not have been scared and didn't have to go through possibly painful follow up checks.

2
1
Anonymous Coward

Re: The real question is...

If screening shows positive, the next step is ultra-sound. If ultra-sound finds something, the next step is a biopsy. If the biospy finds cancer, then treatment starts.

9
1
Silver badge

Re: The real question is...

I, Misogynist(ic wanker)

Try and understand this concept - false positives just cause worry and stress.

false negatives cause dead teens and women - including my ex partner who was effectively given a death sentence at 14 because the NHS flat out refused to test her for the cancer she actually had.

BUT THERE IS NO FSCKING WAY you can spin this criminally negligent incompetence as any form of positive outcome.

Even had I not lost a partner to cancer and had my own run in with it (a lot less serious thankfully) I'd happily beat your skull in with a claw hammer just for having the gall to write that comment.

You should be so very happy I'm not responsible for your upcoming prostate exam because it'd involve sandpaper and a dab of jalapeño, and that'd be the fun part of the experience!

22
1
WTF?

Re: The real question is...

I fully understand your comments, my wife has had a false NEGATIVE screening in October and repeated false negatives in April, however the surgeon realised that there was something wrong. It turned out that some cancers, even developed cancers do not show up on a screening test. Ultra sound guided biopsies of the right type plus MRI and CTI scans are needed. My wife is now starting her treatment next week.

Please accept my sympathy over your loss.

I am not entirely sure about your proposed prostrate investigation plan for the dumb nut. Surely a suitably heated old fashioned toasting fork should be used for both their examination and possible excision of any offending material.

PS concern was expressed about a possible secondary site in my wife a 'false positive' in wankers nomenclature: a rapidly organised examination showed that the suspect item was not a problem, so treatment can go ahead, in spite of other unconnected existing problems, comorbidities in the jargon.

13
1
Silver badge
Stop

Re: The real question is...

Based on a 2012 study, for every 10,000 women screened 129 will be false positives.

So in half a million "not inviteds" that's 6,450 who weren't unnecessarily treated.

Puts the number of, purely theoretical, deaths into perspective,

I doubt you would consider it a "purely theoretical death" if it was your wife or mother who died because her breast cancer went undetected until it was too late.

7
1
Silver badge

Re: The real question is...

Thanks Richard,

I still really haven't recovered from what happened to her, still so very angry about how badly she was mistreated and abused, and the waste of it all. 10 years this September. I wouldn't trust the NHS as far as Prof. Hawking could throw them.

I hope your wife recovers well, there are some herbal preparations that can help with nausea but you need to check them with the consultant - a good one for bruising is Arnica which is readily available (reduces bruising in general and after surgery). I think she's very lucky to have found a good consultant who's inclined to do a proper job..

Ironically I got told for nine MONTHS I didn't have cancer, paid privately, and the dermatologist took less than 3 minutes to refer me to surgery. I've had abuse from my GP surgery for having the affrontery to get a second opinion and to be in charge of my own medication (for what would probably be called a co-morbidity (certainly would be if I went to Saudi)) to the point I changed surgeries and then after the second BCC showed up they cut it out and promptly lost the biopsy accidenteliberately.

Certain people are about to find they're in the process of being sued.

To top it all off I get told categorically, by an experienced GP, that BC Carcinoma CANNOT spread - 3 minutes on Google states a 1% metastatic rate - favourites being lung and brain... That GP has at least 25 years experience..

7
4
Anonymous Coward

Re: The real question is...

I wouldn't trust the NHS as far as Prof. Hawking could throw them.

With all due respect, when you're defending the high moral ground (and you were, very well), it's best not to throw it away like that.

And, FWIW, my partner works in PHE on screening programmes. Any expectations that people may have about the omniscience of the NHS should be rapidly dispelled. It's an organisation of people, like any other. It's constrained by resources, but even more so by structures, policies and people, and things go wrong. Frequently. And whilst you may be angry, I can assure you that far more than 270 lives have been lost due to imperfections in screening since 2009. At the local level screening is art, at the national level it is statistics. I really wouldn't use the word "science" for screening, certainly until the accuracy of the techniques available are substantially more accurate than they are.

8
1
Gold badge
FAIL

I'm thinking this is the *mother* of all "off by one" errors.

And in a perfect storm it was the last one, which is the one where data indicates is the most probable one to show a positive result IE need more investigation.

And 9 years to pick this up?

Quality checking not their strong point it would seem.

1
0
Silver badge

Re: The real question is...

If that was the only tantique usque gallus the NHS made I might be prepared to agree with you.

However.

I almost *died* from misprescribed medication and had medication I urgently needed buggered about with by a NHS GP who was nothing to do with my case and caused me and my family 3 years of hassle including contacting consultants behind my back.

I had hip problems at 13, which misdiagnosed (actually flat feet) have caused me a lifetime of spinal pain to the point I'm off work and have more than once been curled up on the floor in agony.

I have had important allergy information suspiciously deleted from my records which could have risked my life..

I've had verbal abuse from GPs and hospital staff - surgeries and treatments denied despite the fact they're mandated..

I've had bloods taken so badly I ended up seconds from intensive care..

That's over and above the cancer situations I've detailed and not including working for the NHS in IT which to be blunt was more terrifying than all that put together - scanners being run by BBC Micros, lockdown passwords set by someone who'd since buggered off to Dubai and forgotten to mention what they were, oh and a patch deployment system that was somewhere to the left of guesswork..

Did I forget to mention being prescribed Fentanyl as I requested (especially on the assumption on my part it was less addictive and safer than morphine - to which I have a family sensitivity (the local NHS effectively murdered my maternal grandmother with it, for her moderate flu, despite it being listed that SHE SHOULD ON NO ACCOUNT TAKE IT, even if Hell was freezing over at that very second)) - no one thought to mention it was up to 25x more addictive than morphine..

Should I go on?

True it might be accidental incompetence or misreading or just a single xx-phobic bigoted GP but there's an old military saying. "once is happenstance, twice is coincidence, three times is enemy action". Last I looked I was on somewhere around number seven or eight..

1
3

Re: The real question is...

May I respectfully suggest that requesting Class A, controlled medications should be left to Doctors rather than patients, particularly if you fail to read the warnings or understand that Fent is an opiate that's used to knock out Great Elephants. It's clearly labelled at every junction how addictive it is, as all opiates are.

Sounds very odd to me that one of the most powerful and addictive controlled drugs were prescribed on request. I take oxycodone regularly and if I walked into any GP and asked for it, morphine or Fent I would be kicked out of the surgery. After the doctors finished laughing. If I'm not in handcuffs.

5
0
Gold badge

Fentanyl?

Like the stuff that Russian special forces used to storm that theatre of hostages held by Chechnyan separatists?

Where a whole bunch of the hostages died of the gas, rather than being shot?

1
1
Anonymous Coward

As sad as this is, it's an easy mistake to make when programming not to account for the end of a date range and not deal in absolutes. I've done it in the past, you learn from it.

4
0
Anonymous Coward

> not to account for the end of a date range and not deal in absolutes.

That's a silly excuse. Those kinds of errors are well known and have been for years. How did this get past coding standards, code reviews, unit testing, integration testing, system testing and user acceptance testing?

I bet £100/day offshore software developers are at the bottom of this. Pay peanuts, get monkeys.

13
0
Anonymous Coward

Agreed, however as in most situations you can't test your own code and employing people to test it costs money.

3
0
Anonymous Coward

Yes you do learn from it. One of the key lessons learnt is that you test boundary conditions. The fact that rudimentary testing didn't spot the error suggests that the design of the algorithm and the testing both matched a specification which had the wrong age limit.

It's much easier to blame the computer program than the senior NHS administrators who signed off the spec.

17
0
Bronze badge

So manslaughter is OK

Because preventing it it too costly.

12
0
Anonymous Coward

Emotional damage and stress to all involved.

I know of someone working giving out some of the screening invites. The "computer bug" they found, thankfully in time to send out late, but valid, invites... was as simple as the training saying "type search for cancer screening*" when it should have been "screening for cancer" and the new system missing the old systems search results due to wording ORDER.

They felt horrible for the mistake, but in reality, how were they to know the system was setup with 2 different search terms (old and new) and did not bring up a per word result (screening + cancer) but only brought up a single term result ("cancer screening" exactly).

*not actual term, as I was not working there so don't remember, and to protect identity etc.

6
0
Silver badge

Re: So manslaughter is OK

Would it count as manslaughter? These people died of natural causes, and were going to anyway. Failing to spot that they needed treatment isn't really manslaughter, otherwise every single death of someone untreated would make a doctor liable.

It's definitely a bad thing, but surely it's better to spot/treat/screen most people and miss a few than stop anyone trying to do this in the future for fear of litigation?

12
4
Anonymous Coward

Re: So manslaughter is OK

https://www.cnbc.com/2018/04/11/goldman-asks-is-curing-patients-a-sustainable-business-model.html

Sure it is.

2
0
Silver badge

offshore software developers are at the bottom of this

No, that would be a MANAGEMENT issue. Not a programmer issue.

4
0

Re: Emotional damage and stress to all involved.

"how were they to know the system was setup with 2 different search terms (old and new) and did not bring up a per word result (screening + cancer) but only brought up a single term result ("cancer screening" exactly)."

I can only speculate on how *they* could have known, but *I* would have known because the testing would have included identifying records (which could have been artificial test records, or anonymised real records, or perhaps just plain real records with suitable privacy precautions taken) containing one word, and both words, and no words, together and apart, testing search terms in all the various permutations, and the results compared with the expected output based on the identified test records.

Whether by design or process or procedure or incompetence, what on the face of it sounds like a relatively simple test to consider and execute wasn't done. This is incompetence; spread across which levels (project, programming, testing, management, requirements, etc etc), I couldn't speculate, but this is incompetence.

2
0
Silver badge

Re: So manslaughter is OK

Technically... Possibly.

The important part is "were going to anyway".

If it was a situation where the cancer had already staged and spread *and would have been in that condition* when the screening session was missed then probably it would be impossible to prove that a detection at a late stage and treatment would have resulted in a cure... Therefore you really couldn't argue manslaughter, at best medical negligence.

However. If the cancer would have been detected before staging or very early I or II (for example in a very aggressive cancer like my partners (although that was also to do with her age)) and there is a reasonable chance that a cure could have been effected if it had been caught in time then manslaughter *could* be argued since, as I understand it, manslaughter means you either accidentally caused or *contributed to* by material actions or in this case inaction the death of the individual concerned - although whether it'd succeed is another matter.

The problem is that it would be what lawyers call a "technically interesting" case which to quote the great Sam Vimes (I believe) is lawyerese for "$600 a day plus overtime, and it'll take months".

2
0

Unit test?

'nuff said ...

0
0
Anonymous Coward

>computer program than the senior NHS administrators who signed off the spec

And my experience of the Department of Health is that the specs will change with:

a) Change of minister

b) Change of senior DoH person in charge

c) Because one of the daily newspapers said something rude about something tangentially associated with the project

d) Change of Government

e) Change of 'priorities' - see (a) to (d) above

f) Because they can.

And, of course, none of the technical people involved are given extra time or money to cope with the changes - unless they happen to be working for one of the big outsourcers or are related to one of the civil sevants involved.

1
0
Alert

Software

PHE oversees the service but who actually runs the service? Who developed and maintains the software? Outsourced?

2
0
Silver badge

.."I cannot give all answers today but I can commit [to taking] the necessary steps. I promise will not happen again... [and that we are] united in our resolve to be transparent about what went wrong.....

I think I can guarantee that this is the last we will hear about it, and that the government will fight tooth and nail to avoid any internal findings being published....

5
0
Silver badge
Paris Hilton

Re: Last we would hear of it

I think I can guarantee that this is the last we will hear about it, and that the government will fight tooth and nail to avoid any internal findings being published....

Unless... as the error took place when El Gordo was PM then some PPS might decide that some political mileage could be gained by discrediting Labour. No change there then???

Paris as one would hope that she gets her checkups on time

3
3
Anonymous Coward

Re: Last we would hear of it

>Unless... as the error took place when El Gordo was PM then some PPS might decide that some political mileage could be gained by discrediting Labour

They're implying that - although the 2009 to 2018 range refers to the appointment data analysed not the date the software was implemented initially or updated subsequently.

1
0
Silver badge

Re: Last we would hear of it

Steve Davies 3: "as the error took place when El Gordo was PM then some PPS might decide that some political mileage could be gained by discrediting Labour. No change there then???"

You moron. You utter fucking moron. If they'd hushed it up, you'd be saying "evil Tories hushed it up". When they announce the problem, you're spouting shit that is "evil Tories, just making it public to crap on Saint Gordon the Fiscally Incontinent".

And in any event , do you really think that this is about discrediting a government that is already the most discredited in the past fifty years? FFS, that government are responsible for the Iraq war, and probably well over 100,000 deaths, as well as running up a public sector deficit to guarantee the need for austerity (not that we've actually seen any austerity).

5
3

Check check and check again.... imagine if you were a supermarket manager and nothing went through a checkout/till system for 10 years... oh its a coding fault

3
1
Silver badge
FAIL

Lies, damned lies etc

The reports are very clear "computer modelling suggested that between 135 and 270 women may have had their lives shortened."

NB key words: 'modelling', 'suggest', 'may', 'between'

BBC News front page headline "Breast scan error 'shortened 270 lives' "

Epic fail by BBC headline writers. More what you'd expect from the Sun

19
2
Anonymous Coward

This is why ...

SWMBO and I have committed to putting everything into a shared (Outlook as it happens) calendar.

(With a local Lightening database for backup).

SWMBO has just (Monday) had a replacement IUD - the last was fitted in 2013, and we set the appointment to chase up for an appointment then. It pinged 6 weeks ago, and we got the appointment letter a week later.

Currently in there are appointments up to 2024.

The great thing about using calendar entries is you can attach documents, forward, rearrange, and filter by category. It's also great for historic data when you need to make benefits claims.

30 years ago I would have evangelised about this, but it's like rolling a stone up hill, so as long as me and mine are sorted (awaits downvotes) that's enough.

Before too many people squeal, it's worth noting that my local Dr and Hospital never report less than 10% DNA, and of the women invited for screening, 29% did not attend.

9
0

For the love of god where is the testing ..

This should be simple

1. create a set of birthdays since 1938

2. run date selection routine

3. check output matches known result.

Of course they say its an "Algo" issue because that makes it sound clever and all modern like, when it is a simple coding error which has been known about since dot.

I hope the contractor who coded this pays NHS compensation

9
0
Anonymous Coward

Re: For the love of god where is the testing ..

Although you don't know whether it's an error in coding, or an error in analysis.

1
0
Silver badge

'Computer algo' blamed

No, people write computer programs, people provide the training data and managers sign off and release it. Other managers deploy it.

See also Irish Cervical Cancer screening. Most of the experts told the managers the outsourcing to USA wouldn't work (they were detecting much less in trial), partly because USA screening was designed for annual tests and Irish tests are / were every three years. Most of the experts then resigned.

It's the managers that are killing people. Manslaughter by deliberate negligence. Not computer algorithms.

12
0
Silver badge

Re: 'Computer algo' blamed

No, people write computer programs, people provide the training data and managers sign off and release it. Other managers deploy it.

Thank you for that. It irritates me no end when some catastrophe is blamed on a "computer bug." No, it's the fault of a person who screwed up while writing the program, or the people supervising the project, not a magically appearing "bug."

Worse is when blame is placed on an algorithm. Every choice made by an algorithm is decided by some person, usually for a specific reason. When you get screwed over it's because someone decided you should be, not because HAL decided not to open the pod Bay doors.

All of this leads the general public to think that they're powerless against "computers, " and that problems are caused solely by technology, not by the people who implement it.

Politicians know this, corporations know this, and tech giants especially know this.

9
0
Silver badge

You've missed the best part. If you request removal from the cervical smear list on account of hysterectomy or other reasons..

THEY AUTOMATICALLY REMOVE YOU FROM THE BREAST SCREENING LIST.

So for helping them out they repay you by putting your life at risk, especially since many women who've had full hysterectomy have had it because of cancer like my former partner.

I'd also like to see the list of the women involved because I'd bet there's a surprising number of minorities on there.. blame the racism on the computer..

I've always wondered if my partners race (half Thai) had anything to do with the execrable care she received from the No Hope Service. My personal experience in the last 3 years backs that up.

0
3
Silver badge
Unhappy

Jemma, I have read both your comments and feel very, very saddened. I used to call it the National Homicide Service after my mum was denied treatment (in my case a question of a GP who had reached his annual budget for referrals, and managed to delay the referral of my mum by three months), your No Hope Service is much, much better, imho.

Crikey it all started at age 14, makes me think of my daughters. Not knowing all the story, I assume you were probably not much older when you went through that with her... all my deepest sympathies!!! I do hope you are well surrounded by family and friends etc!

The removal from certain lists entails removal from other, non-related, lists??? This is of course unacceptable and needs to be fixed! I could not find anything on that, though (found nothing after a quick few g00gle's and am a bit busy right now, will try again later), I would really like to make sure this gets fixed, has it been reported [on] ? This listing issue seems to be one list to rule them all, want out of one, get out of all ...

As for racism, this seems to be a listing blunder. I'm not implying people making decisions in the UK's health system are never bigots, racists, homophobics, or whatever. But in this case it is much more likely a listing blunder, which, as your story implies, can kill, so we must make sure it gets fixed!

All the best,

Hans 1

PS: Your particularly fitting treatment (imho) for the heartless comment@rd up there (last part of your previous reply) managed to squeeze a smile out of me!

2
1
Silver badge

The situation was nasty - involving hpv and familial abuse and that was by no means the worst of it. My rule of thumb is if you've a body part and it can get cancer - you are NEVER too young to get cancer in it. That's the excuse they gave her - you can't have cancer you're too young.

If your girls are offered the vaccine I'd suggest they take it. Personally I'd make it mandatory. It took my partner almost a decade to die, some of it stuck in a hospital bed to the point if I showed her pictures of my parents garden or just pictures around the area she'd be in floods of tears.

Unexplained weight loss, tiredness, back pain, bloating or fainting are signs and if doctors dismiss you get a second, third, fourth opinion..

In teens you have very very little time to act because the sky high hormone levels overdrive the tumour - so what might take 10-15 years in an adult woman takes 10-15 months in a teenage girl.

I've had another idea for said commentard involving testicles - pinky mice and my friendly but always hungry Western Hognose snakes. They're generally adorable little snakes that pretend to be rattlesnakes - but don't get between them and food, they're aggressive feeders and they chew a venom into the food which causes swelling in humans... Cue testicles that could have been used at Trafalgar in the 6 pounders (about apple sized) and (from my experience of finger and hand bites) swell up to the hardness of wood... Not that I'm vicious or anything...

I'm not going to comment on this any more because it's starting to stress me. But just be aware that the symptoms of cancer in teenagers can be diffuse and difficult to pin down, and the No Hope Service are not necessarily your friend in this situation.

1
0
Silver badge

"At this stage it is unclear whether any delay in diagnosis would have resulted in any avoidable harm or death and that is one of the reasons [for] establishing an independent review."

The the ambulance chasers are already circling this one and rubbing their hands together.

5
0
Anonymous Coward

Maybe not a bad thing ....

since it'll tie them up for years, keeping them from other vulnerable people.

IANAL but even I can see this is a stretch ... especially in a population that was already headed towards the grave anyway.

Even if you can prove a person who died from breast cancer (and that'll be fun, as there are myriad ways to die from cancer that don't say "cancer" on the death certificate) you are still left having to show that had they received a call for screening AND they had responded (remember that 29% non attendance) AND the screening had picked up their cancer (remember false negatives) AND that result had triggered a course of treatment (bearing in mind what was available at the time) AND that course of treatment had been successful (because all treatments have a failure rate) THEN you might just be in with a chance.

To be honest I'm more enraged about the casual admission that some poor innocent was blown to pieces by a trigger happy RAF ... pity his family.

2
0
Silver badge
Flame

Final appointment?

Why are there final appointments?

Which politicians decided that people after 70 do not get cancer?

2
0
Anonymous Coward

Re: Why are there final appointments?

Maybe the chances of catching anything beyond that age that can be meaningfully treated drop below an acceptable level ?

Sorry to break it to you, but cancer or no cancer, the chances of you dying at 70 are greater than at 60 - all things being equal.

0
0
Silver badge

Re: Why are there final appointments?

As I understand it, the main reason is that there's no point in treating early slow-growing breast cancer above a certain age.

Many of the treatments would kill you, while a slow growing cancer wouldn't actually affect you in any way before you die of something age-related.

More aggressive cancers are an entirely different thing, however those aren't generally found this way - or at least, not early enough to help.

Check your breasts and your balls. Seriously, do that often.

3
0

Noticing 45K missing invites a year does not require careful code analysis

I'm guessing that would translate into at least 25K fewer visits per year, which an Accountant should have noticed even if eagle-eyed code reviewers didn't spot anything.

5
0

Page:

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Forums

Biting the hand that feeds IT © 1998–2018