Only a badly brought up person would compare mean pre- with mean post-intervention data to estimate an intervention effect in a time series. Such a person would ignore a natural trend (e.g. was the outcome already improving), seasonal effects and auto-correlation. This paper describes two acceptable statistical approaches that mitigate these issues, one of which is favoured on the grounds that is it much simpler to use. This would be useful to those engaged in quality control activities in data-rich environments.
— Richard Lilford, CLAHRC WM Director
- Lagarde M. How to do (or not to do)… Assessing the impact of policy change with routine longitudinal data. Health Policy Plan. 2012. 27:76-83.