9 November 2022: Expert Group on Statistics - C

Chapter 4 of the Expert Group report had been held back to allow a full run at it, due to the complexity involved. The topic was the infection with HCV arising from blood transfusions. There was no actual data source, so the panel had to use a modelling process. They were influenced by the model used in the Penrose Inquiry. Arriving at meaningful figures to complete this task required a considerable amount of work. Basically, the effort was expended to make reasonable assessments of what would have happened without the infection of a group of people with HCV. The difference is the impact which leads to them coming up with a number. Each stage had to come with a set of assumptions. There were unforeseen issues, such as there not being readily available or comparably consistent figures for the total number of blood transfusions carried out. The panel had thought this would be something recorded as a standard matter of public record.

The actual process was described as a “deterministic model” which is basically a spreadsheet exercise. More specifically it is called a “stochastic Monte Carlo analysis”. This is one that runs the spreadsheet many times with reasonable variables changing each time. This produced a recognisable standard deviation curve. “Stochastic” relates to the elements being uncertain, while “Monte Carlo” is the name given to the model because it is as if you were rolling dice many times. It is usual for the spreadsheet model to be run 1,000 times. Because the general population numbers are large enough, it was possible to achieve a pattern. One panel member asked to be indulged to give a more simple description of what happens under the model. Bless him, he made a valiant and earnest attempt, but this writer, who studied statistics for a year at University, was no wiser after his intervention. Maybe think of it like this, one person walking up a flight of stone steps will make no noticeable difference to the steps, but after years of individuals using the stairway, there is a wearing away which shows the commonest route people take, with a degree of variability depending on the depth and width of the dip caused.

It was noted that the mortality rate for transfusion patients was much higher than bleeding disorder infectees. Also, the older someone gets, the greater the chance that HCV will be what actually causes their death. As the session moved on to seeing patterns, there followed a period in the evidence which explored the significantly different baselines and outcomes when comparing Scotland to the rest of the UK. The Scots sure have their own issues which lend themselves to comedic stereotyping.

The panel reported on estimates of up to 400,000 HCV infections in the UK since 1970. They used this as a way of demonstrating how they assessed (or re-assessed) the underlying research and data sets used to come up with particular results, such as the 400,000 figure. There were examples of the panel not being able to support some previous assertions. Clearly, statistics is not a simple process of adding up all the numbers, then dividing the total by the number of numbers. That is too mean an explanation. There are a host of constants, variables, and acceptable ranges which apply. There are some variables which only apply to certain conditions, not all. Then there are the different levels of reliability attached to each data set. One displayed document was so crammed with figures it looked like every screen in the room had pixellated. At the risk of landing squarely back into the realm of gobbledygook, there was described the option to randomly select a single realisation so as to test the veracity of the overall results, … obviously. When one of the experts talked about there being “direct direct evidence” – with the alternative assumption of there therefore being “indirect direct evidence” – then a straw fell from the sky whereupon a whole field of camels suddenly needed lumbar corrective surgery.

One chart demonstrated the statistically significant higher numbers of female HCV infectees for the years including the twentysomethings and the thirtysomethings. This was linked to these being the main childbearing decades, and of course the greatly increased likelihood of a woman needing to receive a blood transfusion. Again, there was the significantly higher mortality rates for transfusion patients, too.