Happy New Year. Especially (for a statistical site) because 2013 is, officially, the International Year of Statistics. More of that later in the week. For now we’ll do a bit of catching up of news that hit over the holidays.
The big story in the UK was Staffordshire Hospital and the Patient Survey. There were hundreds more deaths than expected at the hospital between 2005 and 2009. Some of those deaths were caused by some pretty dreadful failings at the hospital – see the box. To identify such failings earlier Health Secretary Jeremy Hunt has announced a Great Patient Survey.
After a stay in a hospital – any hospital - you will fill in a form, the main thrust which is: would you be happy for a sick relative or loved one to receive the same care that you did in this hospital? Surveys are tricky things, and this one has been attacked as no more than a forum for anecdotal evidence.
Statisticians hate anecdotal evidence. It is unscientific. It is biased. Above all, anecdotal evidence is remarkably unreliable. For a start, you are much more likely to fill in a questionnaire if you had a bad time. So most hospitals in the land, even the excellent ones, will find the majority of the questionnaires slam them. And your bad time can be caused by anything from the trivial to the truly appalling. It can range from ‘The breakfast toast was burnt’ to ‘the surgeon totally messed up my ingrowing toenail and amputated my leg.’ Either one gets the tick in the ‘No Relative of Mine should be Treated Here’ box.
Alternatively, the breakfast could have been perfect, served by a beautifully kindly smiling nurse – which makes you not mention the dirt under the bed. We base hospital policy on evidence that random? No wonder some are sniffy about the proposed questionnaire.
In the other hand, a key moment in statistics was the realization, back in eighteen-ninety-something, that there is wisdom in crowds – even non-expert crowds. People who know nothing about cows can correctly guess the weight of a cow down to the fine ounce – as long as there are enough people. (You just average their answers -see here and here for more.) So people who know nothing about hospitals can correctly spot a duff one – if there are enough people.
That had already happened at Stafford. The people of Stafford who used the hospital knew things were not well. They formed pressure groups - angry ones. No-one took any notice. The staff of the hospital knew things were not well. They, too, were ignored.
What did bring Stafford to light was statistics. A Health Commission investigation in 2008 found 'apparently high mortality rates in patients admitted as emergencies'. Since then a public inquiry, chaired by a QC, sat for 139 days, cost £10m and considered about a million pages of evidence.
It need not have happened; the statistics, and the wisdom of crowds, were there early. But you need someone to take notice of the statistics, and that someone has to have the power and the authority and the will to act upon what the statistics are saying. In stats-speak this is evidence-based decision-making. In real life it is common sense.
Moral? If you are any sort of administrator at all, or in any position of responsibility, look at the stats. They save lives.
Moral two? The best stats in the world are useless if no-one pays them any attention.
Tomorrow: How to live to be 110.