Controlled before-and-after studies
Why this level?
The design family and its defining features have not been specified for this hierarchy. NHMRC 2009 sourceWith concurrent control sites this corresponds to NHMRC's III-2 row, and with a historical comparison to III-3. The typed vocabulary has no controlled before-and-after fact, so the page is left unclassified rather than guessed. How control sites were chosen, and how many there are, decide how much the comparison can bear.
What question does this help answer?
Did the outcome change more where the intervention was introduced than it changed, over the same time, in places that did not introduce it?
When to use it
Use it when an intervention is introduced in some hospitals, wards or regions and not in others, random allocation is not possible, and the outcome can be measured in the same way in both groups before and after. It is common for evaluating service and policy changes.
When not to use it
Avoid it when no comparison site plausibly shares the intervention sites' trend, or when intervention sites were chosen because their results were unusually bad, which invites regression to the mean. If many repeated measurements exist, an interrupted time series with a control series is stronger, because it can examine trends rather than assume them.
How it works in plain language
The outcome is measured in intervention and control sites during a baseline period and again after the intervention. The analysis compares the change in intervention sites with the change in control sites, which is the logic of difference-in-differences. Unlike an uncontrolled before-and-after study, a change that affects both groups at once, such as a national policy or a seasonal pattern, drops out of the comparison.
What two measurement points cannot show
Choosing control sites
- Choose controls before looking at post-intervention outcomes, and record the criteria used.
- Prefer sites with similar patients, services and baseline outcome levels and, where data exist, similar earlier trends.
- Check that control sites did not adopt the intervention, or a competing initiative, during the study.
- Collect data at the same times and with the same methods in both groups.
- One intervention site against one control site cannot separate the intervention from everything else that differs between two sites. Several sites per group help, and the analysis should reflect variation between sites.
Key assumptions and design requirements
- Without the intervention, the outcome would have changed similarly in both groups. Two measurement periods cannot test this.
- The groups do not differ in concurrent changes, shifts in case mix or data collection.
- The intervention does not spill over into control sites.
- Uncertainty reflects the number of sites, not only the number of patients. With few sites, ordinary cluster-robust standard errors can be unreliable.
Clinical example
Three hospitals introduce a pharmacist-led discharge service and three similar hospitals in the same region do not. Readmissions are measured for six months before and after in all six. A region-wide seasonal change drops out of the comparison, but if the intervention hospitals were already improving faster, two periods cannot reveal it, and monthly data points would. This example is illustrative.
With many time points: a controlled interrupted time series
When the outcome is measured at many regular points before and after the intervention in both groups, the comparison can move from two periods to trends. A controlled interrupted time series fits segmented regression to the intervention and control series together, and estimates the change in level and in slope at the intervention date in the intervention series relative to the control series. A change that affects both series at the same date, such as a national policy, a coding change or a shared seasonal pattern, drops out. Unlike two measurement periods, the data before the intervention show whether the two series were following similar trends. It still assumes the control series would have kept tracking the intervention series had the intervention not happened, and that nothing affecting only one of them happened at the same time. Autocorrelation within each series needs the same attention as in an uncontrolled time series.
What to watch out for
A control group measured at the same two times does not remove differences in how the groups were already changing, and two measurement periods cannot show whether their trends were parallel. With few sites, one unusual site can drive the result, and a control chosen because it looked similar at baseline can still differ in trend, case mix or concurrent initiatives.
How to report it
No reporting guideline dedicated to the controlled before-and-after study was found in a PubMed search on 2026-09-16. TREND is the closest fit, written for nonrandomised evaluations of an intervention. Report how the control sites were chosen, the outcome in both groups in both periods, and the evidence that the two were comparable before the intervention began. 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.
Before using this in your workspace
Record how intervention and control sites were selected, the baseline and follow-up periods, the outcome definition and data source at each site, concurrent initiatives, the number of sites per group, and how site-level clustering will be handled in the analysis.
- Research designs for studies evaluating the effectiveness of change and improvement strategies (opens in a new tab) · Qual Saf Health Care · doi 10.1136/qhc.12.1.47
- Experimental and quasi-experimental designs for evaluating guideline implementation strategies (opens in a new tab) · Fam Pract · doi 10.1093/fampra/17.suppl_1.s11
- Designing Difference in Difference Studies: Best Practices for Public Health Policy Research (opens in a new tab) · Annu Rev Public Health · doi 10.1146/annurev-publhealth-040617-013507
- 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
- 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