Table 1.

Overview of quasi-experimental designs.

Design	Key design elements	Advantages	Disadvantages/threats to validity	Ways to strengthen
Pre-post	•Comparison of outcome of interest before and after intervention. •May or may not include a control group.	•Less cumbersome and simpler to gather data for than other designs (requires data from a minimum of only 2 time points).	•Temporal biases are a key threat to validity; if there are changes in measurement or quality of data over time, this will cause bias. •Control groups, if included, may not be comparable for important covariates. •Concurrent policies challenge validity. •Lags in policy adoption can influence internal validity. •Infectious disease dynamics (e.g., exponential spread over time) can bias results.	•Include comparator groups. •Conduct adjusted statistical analyses. •Specify how time is being addressed in the design and analysis.
Interrupted time series (without control group)	•Data collected at multiple time points before and after an intervention is implemented. •Assess whether there is a level or slope change at the time of intervention (or after a pre-specified lag, if appropriate).	•Each group acts as its own control. •May be only option for studying impacts of large-scale health policies when there are no groups left unexposed to intervention.	•Requires a large number of measurements. •Preferred to have more pre-period data collection. •Relies on the assumption that nothing changed within the study period that would affect the outcome of interest other than the intervention. •Concurrent policies can influence results. •Temporal issues & seasonality are major challenges. •Lag periods must be appropriately conceptualized. •Infectious disease dynamics, such as non-linear functional forms, can bias results.	•Include comparator groups •Ensure adequate number of time points pre- and post-intervention (having sufficient data prior to the intervention will establish existing trends). •Conduct adjusted statistical analyses, with adjustments for time to reduce biases related to seasonal variability. •Adjust for autocorrelations. •Shorten the duration of time periods.
Interrupted time series (with control group)	•Data collected at multiple time points before and after an intervention is implemented in a treatment group and control group. •Most commonly analyzed using a difference-in-differences approach. •Compares the difference in the amount of change in the outcome before and after an intervention is implemented between groups exposed and unexposed to the intervention.	•Controls for observed and unobserved time-invariant variables that differ between groups.	•Requires a large number of measurements. •Preferred to have more pre-period data collection. •Relies on the assumption that nothing changed within the study period that would affect the outcome of interest other than the intervention. •Concurrent policies can influence results. •Temporal issues & seasonality are major challenges. •Lag periods must be appropriately conceptualized. •Inference relies on parallel trends assumption being met.	•Evaluate parallel trends assumption. •Use event-study design that estimates intervention effect at multiple time points before and after implementation (to check for bias and changes over time).