Controlled interrupted time series
A controlled interrupted time series compares changes in outcome trends after a policy with changes in a comparison series measured over the same periods.
Why this level?
The design family and its defining features have not been specified for this hierarchy. NHMRC 2009 sourceThe comparison series can help separate the policy from concurrent events, but allocation is not random. Different time-varying causes, spillovers or changes in recording can still explain the contrast.
Fit
When to use it, and when not to
Use it when
- A policy has a clear start and repeated outcome measures before and after it.
- A comparison series can help separate the policy from other changes at the same time.
Not when
- You have only one measurement before and after. Use a controlled before-and-after design. Read: Controlled before-and-after studies
- You can randomize enough sites before rollout. Consider a cluster trial. Read: Cluster randomized trial
Before you start
What you need
- Intervention and comparison time seriesUse the same outcome definition, time units and denominators throughout.
- A prespecified impact modelChoose the expected delay, level change, slope change and effect horizon before seeing results.
- A defensible comparisonExplain which outside events it captures and why the intervention cannot change it.
- A plan for absent periodsState why records may be missing and how results could change under other assumptions.
- Without the policy, modeled trend differences would persist.How to check: Check prior trends and concurrent changes; similar prior trends alone do not prove this claim.
- Outcome measurement stays comparable.How to check: Check coding, denominators, case mix and reporting changes in both series.
- The model handles time dependence.How to check: Check seasonality and correlated errors; use a suitable variance model.
- Missing periods follow the stated assumption.How to check: Missing completely at random is unlikely when service disruption creates gaps. Missing at random is plausible if recorded factors explain gaps. Multiple imputation under MAR may help, but it is not automatically better than complete records. If missing not at random is plausible, use a delta or tipping-point sensitivity analysis.
Plain language
How it works
Choose a comparison exposed to relevant outside events but not to the policy. Prespecify the expected timing of change. Fit segmented trends for both series and estimate their differential change at a stated horizon.
Lopez Bernal and colleagues explain why controls can reduce simultaneous-event bias and also introduce bias. Their tutorial addresses time dependence and seasonality. Report uncertainty from the fitted model, not from the simple arithmetic contrast.
Worked example
A worked example
- Setting
- Two ward systems record monthly medication errors before and after a policy.
- Question
- What is the policy contrast at the chosen follow-up month?
- Approach
- Fit both full time series with baseline trends, level changes and slope changes. Compare fitted values with their no-change projections.
- Result
- The intervention series is 18 versus a projection of 30; the comparison is 16 versus 20. The adjusted contrast is -8 errors per 1000 administrations.
- What it means
- The contrast subtracts the comparison change from the intervention change. It is causal only under the design assumptions, not because a model was fitted.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
A comparison series helps only if it captures relevant outside changes without being affected by the intervention or its spillovers.
- Calling two before-and-after snapshots a time series.Instead: Use repeated observations on both sides of the policy date.
- Choosing a control exposed to the same policy.Instead: Check spillovers and intervention coverage before selecting the comparison.
Inside Mallard
How Mallard uses this
Plans do not link to this design yet. Learn search and the Atlas offer it as a reference.
design.controlled_its, the id a plan cue records.How to report it
No reporting guideline dedicated to controlled interrupted time series was found in a PubMed search on 2026-11-07. TREND can guide reporting of nonrandomized public-health evaluations. Report the policy date, series selection, pre/post periods, impact model, effect horizon, uncertainty, seasonality, correlated errors and concurrent changes. A reporting guideline is not a design manual: use it to make the study transparent, not as a substitute for careful design. Check the target journal's own requirements as well, which this page cannot know.
Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls (opens in a new tab) · BMJ · doi 10.1136/bmj.b2393
- The use of controls in interrupted time series studies of public health interventions (opens in a new tab) · Int J Epidemiol · doi 10.1093/ije/dyy135
- Interrupted time series regression for the evaluation of public health interventions: a tutorial (opens in a new tab) · Int J Epidemiol · doi 10.1093/ije/dyw098
- Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: the TREND statement (opens in a new tab) · Am J Public Health · doi 10.2105/ajph.94.3.361