4  Study Design

The ITS study design requires a few components: 1) an intervention, 2) outcome data, and 3) sufficient data.

4.1 The Intervention

The intervention should have a well-defined period, allowing for a distinct separation of pre- and post-intervention periods. If the intervention is gradual, an impact that happens over time, consider modeling it as a slope change within the ITS analysis.

4.2 The Outcome

ITS can handle various data types (counts, continuous, binary), but works best when the outcome is expected to change relatively quickly after the intervention, or after a known lag period. When the timing between the intervention and outcome is more variable or less clear, consider an intermediate or proxy outcome.

4.3 Data Requirements

ITS analysis works best with sequential measurements of the outcome before and after the intervention. While there’s no strict minimum, more time points generally increase statistical power. In a review of almost 200 ITS studies, the median number of time points was 48 (IQR: 30 - 100) with 18 (IQR: 10 - 34) time points used to measure the time before the first intervention@turner2021. Otherwise, simulations can help determine if your study has enough power to detect effects@liu2020. As with any analysis, make sure your data is valid, reliable, and that you understand any changes in collection methods that might coincide with the intervention.