Staggered interventions with no control groups.

Area of Science:

• Epidemiology
• Econometrics
• Biostatistics

Background:

• Two-way fixed effects models have limitations for evaluating staggered interventions implemented at different times.
• Existing alternative strategies often assume the availability of control groups, which is not always feasible.
• Interrupted time series (ITS) designs are a potential alternative when control groups are absent.

Purpose of the Study:

• To investigate the limitations of standard interrupted time series (ITS) model specifications in the context of staggered interventions.
• To demonstrate that common ITS models can yield biased results when applied to staggered intervention data.
• To propose novel, adapted analytical strategies for ITS analyses with staggered interventions, drawing from difference-in-differences advancements.

Main Methods:

• Review and critique of standard interrupted time series (ITS) model specifications.
• Simulation or empirical analysis to illustrate bias in standard ITS models with staggered interventions.
• Development of modified ITS model specifications inspired by recent econometric literature on staggered difference-in-differences.

Main Results:

• Standard ITS model specifications are shown to produce biased impact evaluations for staggered interventions.
• The extent of bias is dependent on the timing and pattern of intervention adoption across groups.
• Proposed alternative model specifications offer more accurate estimations in the absence of control groups.

Conclusions:

• Standard ITS models are inadequate for analyzing staggered interventions, particularly when control groups are unavailable.
• Adapted ITS analytical strategies are necessary for reliable impact evaluation in such scenarios.
• The findings have significant implications for epidemiological and econometric research involving staggered interventions.