Interrupted Time Series (ITS) Analysis
Also known as: ITS analysis, segmented regression of time series, Kesintili Zaman Serisi (ITS) Analizi
Interrupted Time Series analysis is a quasi-experimental design that estimates the effect of a single, well-dated intervention by comparing the trajectory of an outcome before and after it occurs. Formalised as segmented regression by Wagner and colleagues (2002) and popularised as a public-health evaluation tutorial by Bernal, Cummins and Gasparrini (2017), it separates the intervention's impact into a change in level and a change in slope.
Key highlights
- Strong quasi-experimental design for evaluating a single, well-dated intervention without a randomised control group.
- Separates the intervention effect into an interpretable level change and slope change.
- Widely used and validated in policy evaluation and health-services research.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
ITS is appropriate when you have a continuous or count outcome measured repeatedly over time, the intervention date is known precisely, and there are enough pre-intervention observations (at least about 24 points) to estimate a stable baseline trend. It is most reliable when autocorrelation is modelled, and when seasonality and underlying trends are accounted for so they do not confound the intervention estimate. Adding a comparable control series (controlled ITS) strengthens it by removing external trends. It is less suitable when the pre-intervention window is short or when concurrent events coincide with the intervention.
Strengths & limitations
- Strong quasi-experimental design for evaluating a single, well-dated intervention without a randomised control group.
- Separates the intervention effect into an interpretable level change and slope change.
- Widely used and validated in policy evaluation and health-services research.
- Needs a sufficiently long pre-intervention series (about 24 observations) to estimate trend and seasonality reliably.
- Autocorrelation between observations must be controlled, otherwise standard errors are seriously misleading.
- Seasonality and underlying trends can confound the intervention estimate if not modelled.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How is the intervention effect read off an ITS model?
Two coefficients capture it: the level-change term (β₂) is the immediate jump in the outcome right after the intervention, and the slope-change term (β₃) is how much the trend tilts afterwards relative to the pre-intervention slope. Together they describe the intervention's short-term and longer-term impact against the projected baseline.
Why does autocorrelation matter so much in ITS?
Observations close in time tend to be correlated, which violates the independence assumption of ordinary regression. If left uncorrected, the standard errors are understated and the intervention effect looks more significant than it is. HAC (Newey-West) standard errors or an ARIMA error structure correct for this.
How many pre-intervention observations do I need?
A common guideline is at least about 24 observations so the baseline trend and any seasonality can be estimated reliably. With a shorter pre-intervention window the counterfactual projection becomes unstable and the effect estimate untrustworthy.
What is a controlled ITS?
A controlled ITS adds a comparable series that did not receive the intervention. By comparing the change in the treated series to the control series, external trends and shocks affecting both can be netted out, strengthening the causal interpretation.
Sources
- 1.Bernal, J. L., Cummins, S., & Gasparrini, A. (2017). Interrupted time series regression for the evaluation of public health interventions: a tutorial. International Journal of Epidemiology, 46(1), 348-355.
- 2.Wagner, A. K., Soumerai, S. B., Zhang, F., & Ross-Degnan, D. (2002). Segmented regression analysis of interrupted time series studies in medication use research. Journal of Clinical Pharmacy and Therapeutics, 27(4), 299-309.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Interrupted Time Series. ScholarGate. https://scholargate.app/causal-inference/interrupted-time-series