We’re planning a study to test the influence of a new guideline on prevalences of different outcomes. Most studies with similar designs seem to compare pre- and post-guideline data, either by comparing just the year before and after a new guideline was implemented or by averaging 2-3 years before and after implementation. Both approaches neglect the possibility of a trend over time, either linear or non-linear. I assume that it would be better to conduct a regression with year (continuous) and the period before and after implementation (binary) as additional predictors. Is anyone aware of a good paper that describes this approach that I could use as a reference?