How to interpret the survival funnel plots

How to interpret the survival funnel plots – note that it is not appropriate to produce a funnel plot when there is 100% survival after a procedure across all the centres.

The horizontal axis shows how many of the specific type of operation (procedure) each centre has performed. The vertical axis shows the percentage of patients who survived at each centre for at least 30 days after their operation. The centres are shown using different numbers and there is a key at the bottom. The Table details the number of patients undergoing the specific procedure at the Centres and related the 30 day postprocedural outcomes.

The horizontal grey line is the overall national percentage survival rate for that procedure. There are also two control limits on the graph; a warning limit (Green line, 97.5%) and an alert limit (Red line 99.8%). If a unit's symbol is above the green line then their performance is not statistically different from the national average.

If a unit’s survival rate is at or below the green warning limit, the centre will be alerted to this. The small numbers of operations being performed can mean that a single death will significantly reduce a centre’s percentage survival rates. This means that being at or below the green warning line may not be as significant as it sounds and statistically there is a 1:40 chance that a Centre may fall below this line by chance alone. However, the causes of the deaths will be carefully investigated by the centre, and their performance will be closely monitored in subsequent years.

If a unit’s survival rate is at or below the red alert limit, an investigation into possible reasons and remedial actions will be launched by the appropriate professional and regulatory bodies. Statistically there is a 1:1000 chance that a Centre may fall below this line by chance alone.

If you or your child are due to have an operation at a centre and have any concerns about their results, then you should raise these with your consultant.

Please note: This analysis is based on 2013/16 data exported March 2016. It includes all known deaths reported up until this date. The statutory process for the registration of death across England and Wales does not in all cases guarantee completeness at any specified date. In the case of a death referred to the coroner, in most cases, it cannot be registered until the coroner's enquiries have been completed and a death certificate issued and provided to the Registrar of Births and Deaths. The portal front page provides more information on this process.

The funnel plots that we use to display outcome of congenital procedures are based on a statistical method called the Binomial distribution (http://stattrek.com/probability-distributions/binomial.aspx). This statistical model is based on a series of ‘trials’ (in our case Procedures) and in each trial there is a percentage likelihood of an ‘event’ (in our case a 30-day death). So if the likelihood is 10%, after 10 trials you would expect 1 event and after 20 trials this would jump to 2 events. This sudden ‘jumping’ at low numbers of trials produces a curve which is like a saw-tooth, with sudden changes followed by smooth sections of the curve. For the purposes of displaying the Congenital funnel plots, we use a smoothing technique (‘interpolation’) provided by Professor Sir David Spiegelhalter. This technique takes away the ‘jaggedness’ of the saw-tooth curve, but as with all smoothing techniques, the value at the given point will often be different from the ‘raw’ curve.