Jeremy Hunt plays the numbers
Jeremy Hunt announced last Wednesday that as many as 270 women may have died because an error in a computer algorithm prevented 450,000 women being invited for routine breast cancer screening appointments. Stories about IT glitches will be increasingly common as artificial intelligence enables more and more healthcare to be automated. As things stand, people are still better than computers at detecting early signs of cancer on mammograms, and the neural networks currently being designed to analyse the images are intended for use as an aid rather than a replacement for human decision making. The hope is to engineer systems that combine the different strengths of humans and computers, with outcomes that neither is capable of independently. The sad reality is that we seem to end up with systems that combine an all-too-human capacity for error with a computer’s blunt force, and so wreak havoc at an inhuman scale.
The algorithm at fault here isn’t doing anything clever or subtle; it isn’t detecting who does or doesn’t have cancer. It takes a patient’s age and screening history, and decides whether to invite them for a mammogram, according to a guideline which says that women should be screened every three years between the ages of 50 and 70. It may be hard to see how anyone entrusted with encoding such a simple rule in a computer program could have got it wrong. We don’t yet know what the source of the error was, but it seems to have meant that appointments weren’t issued for some women whose final screening visit should have happened when they were over 68. A woman who is first screened at the age of 50 would normally have her sixth and last screening visit when she is 68. Perhaps the programmers, or those who instructed them, ignored the fact that many women started screening at 51 or 52, or somehow got out of the three-year cycle, and so would have been eligible for an appointment between the ages of 68 and 70.
One reason that such an error was allowed to persist for so long – it seems to have been in place since 2009 – may be that it wasn’t really an error. When the screening interval was extended to cover women up to the age of 70, there wasn’t convincing clinical evidence that this was the right thing to do. Individual programme directors with stretched resources may have felt that a system of triennial appointments for women aged 50 to 68 was a satisfactory approach to implementing the guidance, given that we don’t actually know whether the right age at which to stop screening is 68, 69, 70 or later – or even earlier.
Breast cancer is no less common or less deadly in women over 70. The reason that screening is not routinely offered to older women is that the balance of risks and benefits alters as the competing causes of death increase: the more likely you are to die of something other than breast cancer, the less chance you have to benefit from being investigated for breast cancer, and so the lower the risk of harm you should tolerate from that investigation. It is, however, hard to quantify the risks or the benefits exactly, and so impossible to say exactly when screening should stop. One thing that means is that Jeremy Hunt’s dramatic announcement that 270 women had had their lives cut short was simply another example of his misuse of statistics to attack the NHS.