A systems thinking approach for analysing falls data - a discussion.

Abstract

This workshop will cover a method of analysis for falls/adverse event data called statistical process control charts. This will be an interactive workshop in which we will discuss the challenges to analysing adverse event data and developing clinically meaningful key performance indicators. Workshop participants will be able to create and apply statistical process control charts to adverse event data. There are increasing amounts of adverse event data, but often we do not know what these numbers mean clinically. When looking at adverse events, health services commonly use a specification approach which involves setting numerical targets. For example, if the total number of falls changed for the better compared to the last month (or another arbitrary target) then things are ‘operating okay’. If the total number of falls changed for the worse compared to the target, then you are “in trouble”. In the latter, there’s pressure to come up with an explanation of why the numbers are bad and how you are going to stop it from happening again. This approach is problematic because it does not take into account normal variation. Adverse events, such as falls, are not always predictable and so display both routine variation (noise) and exceptional variation (signals for action). We collect data as a basis for action. However, unless signals are differentiated from noise the actions may be inappropriate, wasting resources and time. Statistical process control charts are a method for interpreting data that separates potential signals from probable noise.