“Shortened lives”: Jeremy Hunt & breast cancer screening scandal

…earlier this year PHE analysis of trial data from the service found that there was a computer algorithm failure dating back to 2009. The latest estimates I have received from PHE is that, as a result, between 2009 and the start of 2018 an estimated 450,000 women aged between 68 and 71 were not invited to their final breast screening.

A ‘computer algorithm failure’, eh? So not your normal NHS ‘computer error’ or ‘IT systems failure’ but a ‘computer algorithm failure’. That got my attention. I presume Hunt doesn’t know what a computer algorithm failure actually is considering he’s an Oxford Philosophy, Politics, and Economics graduate but I’ll give him the benefit of the doubt. A ‘computer algorithm’ is a set of defined steps of operations that can be undertaken by a computer to solve a problem. In this case the algorithm that defined which women should receive an invitation for a mammogram would be based on their dates of birth, date last checked, etc. A computer program is then written to execute those algorithm instructions then print out the invitation letters. But Hunt didn’t say it was a systems failure, IT glitch or computer error, he said a computer algorithm failure which is odd because an algorithm can only ‘fail’ during testing, i.e. if the steps for solving a problem are incomplete, incorrect or because of data error. For example, the algorithm would be said to have failed if the invitation list contained women who are older than 71 (a date checking error) or don’t have a UK address (another data checking error) or have passed away (another data checking or data timing error) or, in this case, did not include all those women who did meet the criteria. That is, a computer algorithm failure is one normally made when the system is being designed and specified, not when the design is coded or turned into computer instructions or processing thousands or even millions of health records. So Hunt was telling me that this was a PHE design error, i.e. it was attributable to the NHS business design team, i.e. a person or team of people, albeit nearly 10 years’ ago, made a catastrophic design error. But to the Press it’s another NHS IT failure and in a way, I agree it is.

The Daily Telegraph’s Laura Donnelly, health editor & Kate McCann, senior political correspondent wrote: “The apparent IT glitch means that women whose cancer could have been spotted early went on to develop the disease.” Donnelly & McCann compound their error later in their article, “…But a computer error was programmed into the system in 2009, when trials about extending the age range further began.  As a result, women who had reached their 70th birthday were excluded from  the system, meaning that up to 450,000 never received an invitation for their final scan. The blunders were only detected almost a decade later, when PHE attempted  to upgrade its computer systems.”.

What they mean is an incorrectly defined or improperly tested algorithm error was programmed into the invite system in 2009 and the errors were not found until almost a decade later.

The Telegraph wasn’t the only newspaper to misrepresent Hunt’s statement. The Independent’s Health Correspondent Alex Matthews-King wrote: “Nearly half a million women missed out on NHS cancer screening because of an IT error, the government has admitted – adding that hundreds may have died as a result.”.

The Daily Mail’s Tim Sculthorpe, Deputy Political Editor and Martin Robinson, UK Chief Reporter for MailOnline wrote: “The NHS was today engulfed by a breast cancer scandal after it was revealed 450,000 women were denied life-saving scans and up to 270 died early after a ‘colossal’ IT failure lasting almost a decade. The victims aged between 68 and 71 were never sent letters offering them a final routine breast screening because of an IT error lasting from 2009 until this year.”.

The consequences of nearly half a million women not being invited to their final breast screening is disastrous, as Hunt went on to explain,

“…had their lives shortened…”? That’s another new phrase to me, (perchance not to deflect attention from or minimise what has actually happened in any way?), a rather unusual way of saying this number of women may have suffered unnecessary pain, late treatments, life inconvenience, debilitation and early death from (complications due to) breast cancer…

And this error was only found by Public Health England after 9 years of operation of their original invitation system? Really? Did none of these half a million women in all that time highlight to their GP or imaging centre that they hadn’t received their last appointment invitation? I can imagine that if an alert woman did point out that she hadn’t received an invitation her GP might indeed put it down to a system glitch and simply arrange for a ‘special’ mammogram appointment without querying their invitation provider. Maybe. But did half a million women not query their missing appointments, over 9 years, and no-one involved recognised that lots of women were querying this? And this was through a date ‘error’ in what is surely not a complex algorithm? And was the error really found only when PHE was testing a new invitation system? It seemed to me that there was likely more to this story than Hunt was saying, probably because he has rightly set up an independent review to look into it.

Whenever a minister reads an urgent statement to the House, particularly one that contains new or technical language, one has to assume that someone with some better knowledge of the event and its possible causes wrote the brief and that person just might have written it with some ‘spin’, i.e. to present the story in as favourable light as possible or, as I suspect in this case, to deflect any immediate blame. (Language so used, per advice from an ex-Cabinet Secretary, Sir Richard Armstrong, may give a misleading impression, without actually lying which could later rebound if found out, or be used in a way that is ‘economical with the truth’.) Surely not in this case? Indeed probably so. With potentially hundreds of debilitating early deaths being attributed by the Minister to this error there is good reason for us to concentrate on the words used in the statement because no government wants to be associated with poor administration that has, in this case, led to catastrophe. This is no normal administrative bungle. Because of the horrendous potential impact of this disease on so many unfortunate women it’s a major public administrative disaster. The Aberfan disaster of 1966 killed 116 children and 28 adults. They also ‘had their lives shortened’ but in this case the numbers could be almost twice as high. Someone, not a computer, is to blame for this inadequate specification, design, testing and subsequent sign-off of the algorithm but by implying it is the fault of a complex IT system it is likely no-one will immediately point fingers at those probably to blame (at least, pending the outcome of the inquiry) and this is exactly what seemed to be happening in the days after Hunt’s announcement.

The following few days played out as we might have expected, unfortunately. There was a major hue and cry in the Press for a few days, some papers and TV channels found and interviewed 72 and 73 year-old women who had been diagnosed with breast cancer in the last two years; charities saying they were shocked by the “colossal failure” of the breast cancer screening services and the “shocking incompetence” which allowed the deadly errors to go undetected for so long; it emerged that Public Health England (PHE) which oversees the breast cancer programme had actually been made aware of the potential problems in mid-January but advised health officials that the risk to patients was limited, and Health ministers were told about the fiasco more than six weeks before the announcement – but PHE advised [them] not to make the matter public until it could set up a helpline and a system of checks for those affected; many thousands of anxious women were reported to be badly served by the new helpline because it was staffed by SERCO contractors who had received only one hour’s training, few or none of whom were medically qualified or had any counselling experience; women were told they faced waits of up to a month to hear whether they are affected by the scandal and, of course, mammogram facilities in England were overwhelmed with new appointment requests meaning these ‘missing’ women were put into a remediation scanning queue which quickly grew to six months.

Jeremy Hunt said himself that families would find it “totally devastating” to learn that they had lost or would lose a loved one as a result of “administrative incompetence”. So this is a big deal. For those women affected it could be the biggest issue they have ever faced. At least Jeremy Hunt got in an early apology on behalf of the government (not their fault but the buck stops with them in a nationalised service), Public Health England ​and the NHS: “I apologise wholeheartedly and unreservedly for the suffering caused. But words alone are not enough. We also need to get to the bottom of precisely how many people were affected, why it actually happened and most importantly, how we can prevent it ever happening again.”. In itself this ‘review’ will be helpful but the delay deliberately introduced after the issue was first discovered could be of huge impact to some of the women affected.

Jeremy Hunt announced that he had commissioned an ‘independent review’ (rather than an ‘Independent Inquiry’ – I wonder what the difference is?) to answer

“legitimate questions that need answering: why did the algorithm failure occur in the first place, and how can we guarantee it does not happen again? Why did quality assurance processes not pick up the problem over a decade or more? Were there any warnings, written or otherwise, which should have been heeded earlier? Was the issue escalated to Ministers at the appropriate time? What are the broader patient safety lessons for screening IT systems?