Guest blogger this week is Haelo’s own Data Coordinator, Nick John, from the Measurement team, here to discuss the hot topic of the funnel plot.
When working in improvement science one of the most common data visualisations you come across is the funnel plot.
This is a useful way to display data in improvement and always generates a lot of conversation and learning, especially with the people working within the system. However, funnel plots can seem complicated and can confuse people that haven’t seen data presented this way before. Let’s try to answer some of the most common questions.
A funnel plot is a good way to understand variation within a system. For example, if we are working to reduce pressure ulcer rates across Greater Manchester, then it may be useful to know which hospitals have an unusually high rate of pressure ulcers and which hospitals have an unusually low rate of pressure ulcers.
This gives us a smaller, more targeted area to focus our improvement efforts on and a great opportunity to share learning. What is it that this hospital is doing differently to obtain less pressure ulcers and can this be replicated?
In the example above we are looking at the variation of medication omission rates between different wards in a single hospital. If our aim is to reduce the omission rates at this hospital then presenting the data like this can be very useful.
Each dot is a single ward; the further to the right the ward appears, the more patients they have seen over the given time period. The higher the ward appears the more medication omissions the ward has recorded.
The dotted black lines are statistically calculated limits, these are what help us interpret the spread of the wards. Any ward that falls within the funnel is statistically indistinguishable. Wards that fall outside the funnel are statistical outliers.
In this example ward A is above the funnel, so it has an unusually high rate of medication omissions compared to the system as a whole. Ward B is below the funnel, so it has an unusually low rate of omissions.
A funnel plot can be constructed from any measure that is defined with a numerator and denominator and is split by some entity (hospitals, wards, CCGs, specialties, clinicians etc.)
For more from Nick, see his measurement blogs below.