• Healthcare quality improvement
• Continuous quality improvement
• Patient safety

We believe that the use of the term ‘interrupted time series study’ to describe two reports from this year’s February issue of BMJ Quality & Safety is potentially misleading.1 ,2 Putting aside the merits of the studies themselves, we would like to clarify what the interrupted time series design entails to promote a productive conversation of its particular strengths and shortcomings.

The interrupted time series design has been identified as “one of the most effective and powerful of all quasi-experimental designs”.3 It is defined by collecting data on “multiple instances over time before and after an intervention (interruption) is introduced to detect whether the intervention has an effect significantly greater than the underlying secular trend”.4 The Standards for Quality Improvement Reporting Excellence guidelines point out that the interrupted time series technique entails taking secular trends into account when estimating intervention effects.5 This is the key feature that distinguishes interrupted time series studies from other evaluation methods, where observations are usually aggregated into pre-intervention and post-intervention periods, thus ignoring underlying trends during the baseline period as well as ‘shifts’ in trend once the intervention is introduced.

The authors seem not to have used time series data in their analysis. From our reading of the articles, they have based their effect size estimation on a comparison of the mean change from the pre-period to the post-period, for the treatment and control groups. In our understanding, the conventional characterisation of this method would be a ‘difference-in-differences approach’, or a ‘controlled before-after study’. We do not dispute the validity of this design. Rather, we want to draw attention to the fact that different quasi-experimental designs are subject to differing threats to their internal validity. As such, taxonomical precision is crucial to ensure that the merits of each study are appropriately considered. To this end, we would advocate for a formal, shared lexicon of terms in program evaluation studies that aligns with previous literature.3