Uncontrolled before-and-after studies
An uncontrolled before-and-after study compares an outcome in a period before a change with the same outcome after it, with no comparison group measured over the same time.
Why this level?
The design family and its defining features have not been specified for this hierarchy. NHMRC 2009 sourceA single series measured before and after is Level IV intervention evidence. A distinct historical comparison is III-3, while a controlled before-and-after design with concurrent controls is III-2. The broad pre/post label is insufficient to choose among them.
Where the NHMRC 2009 hierarchy names this design, it places it here:
| Research question | Level | When |
|---|---|---|
| Intervention | III-2 | An observational comparison uses concurrent controls. Statistical adjustment does not change this design level. |
| Intervention | III-3 | The comparison uses historical controls or separate single-arm studies, without concurrent controls. |
| Intervention | IV | A single case series has post-test or pre-test/post-test outcomes, without a comparison group. |
Fit
When to use it, and when not to
Use it when
- The change is being made anyway, and the choice is this study or no evaluation at all.
- You are auditing a change already made, and the data allow nothing stronger.
- You will treat the finding as provisional, a sign that something changed rather than proof of why.
Not when
- You have the outcome at many regular points before and after. Use an interrupted time series instead. Read: Interrupted time series
- A similar site did not make the change. Measure it over the same periods, in a controlled before-and-after study. Read: Controlled before-and-after studies
- The change was started because of an unusually bad period. Results would likely improve on their own. Read: Regression to the mean
Before you start
What you need
- What changed, and exactly whenA clear start date separates the before period from the after period.
- The before and after periodsMake them the same length and cover the same seasons. Monthly values on each side tell you more than two averages.
- How the outcome is countedoutcomeTypeIt must be counted the same way in both periods. A new form or code can look like an effect.
- Why the change was made when it wasIf a bad result prompted it, some improvement would have come anyway.
- An outcome the change should not affectIf it moves too, something broader than the change was going on.
- A missing-data plan that starts from an assumptionDecide why outcome records may be missing before choosing a method. The assumption picks the method, not the share missing.
- Nothing else that affects the outcome changed between the two periods.How to check: This assumption is the whole design, so state it. List staffing, policy, coding and case-mix changes over the same time.
- The outcome was not already drifting up or down.How to check: Plot as many earlier points as the data allow. A trend already under way looks just like an effect.
- The before period was typical, not an unusually bad stretch.How to check: If the change followed a spike, expect a fall anyway. Look at a longer run of earlier data.
- The outcome is measured the same way before and after.How to check: Same definition, same data source and the same people collecting it.
- The patients are alike in both periods.How to check: Compare case mix across the periods. A shift in who is treated can look like an effect.
- Missing values follow the assumption you named.How to check: Records are often less complete in one period, for example before a new form. So missing completely at random (MCAR) is rarely plausible. Missing at random (MAR) means the gaps depend only on recorded factors, such as period, age and ward. Under MAR, multiple imputation can use those factors, but it is not automatically better than analyzing complete records. If bad outcomes go unrecorded more often, even given recorded factors, missing not at random (MNAR) is plausible. Then add a tipping-point analysis, which shifts the imputed values until the conclusion changes.
Plain language
How it works
The outcome is measured during a period before a change and again during a period after it, and the two are compared. No group goes without the change over the same time. So anything else that shifted between the periods looks the same as an effect of the change.
Three additions do most of the work. More time points before and after turn it into an interrupted time series, which can tell a sudden step from a trend already under way. A comparison site turns it into a controlled before-and-after study, analyzed as difference-in-differences. An outcome the change should not affect shows whether something broader was changing.
Worked example
A worked example
- Setting
- One medical ward starts a discharge checklist after a year of unusually high readmissions.
- Question
- Did the checklist reduce readmissions within 30 days?
- Approach
- Readmissions in the 12 months before the checklist are compared with the 12 months after it.
- Result
- Readmissions fell from 15 to 12 in every 100 discharges, a drop of 3 in 100.
- What it means
- The checklist may have helped, but a fall was likely anyway after an unusually bad year. Monthly rates from earlier years would show whether that year was unusual or readmissions were already falling. A ward without the checklist would show whether readmissions fell there too. Neither can rule out another change on this ward at the same time, so the drop is consistent with an effect of the checklist, not proof of one.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
Without a concurrent comparison group, any secular trend, seasonal pattern, regression to the mean, or concurrent change is indistinguishable from the intervention's effect. It is the weakest of the common evaluation designs and should be strengthened wherever possible with multiple pre-intervention time points, a control site, or both.
- Reporting the drop as proof that the change worked.Instead: Say the result is consistent with an effect, and name what else could explain it.
- Changing how the outcome is recorded when the change starts.Instead: Keep the definition, data source and collection the same in both periods.
- Reading a small P value as evidence that the change caused the drop.Instead: A P value can say the drop is unlikely to be chance alone. It cannot rule out a trend or another change.
Inside Mallard
How Mallard uses this
When a design option in Mallard's brief is described as a before-and-after study, the option links to this design.
design.before_after, the id a plan cue records.How to report it
No reporting guideline dedicated to the uncontrolled before-and-after study was found in a PubMed search on 2026-09-16. TREND is the closest fit, written for nonrandomized evaluations of an intervention rather than for this design in particular. Its items on the intervention, the comparison and the threats to validity apply; what it cannot supply is a concurrent control, so state what else changed over the same period. 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
- Segmented Regression in an Interrupted Time Series Study Design (opens in a new tab) · Anesth Analg · doi 10.1213/ANE.0000000000005269
- Overview of clinical study designs (opens in a new tab) · Clin Exp Emerg Med · doi 10.15441/ceem.23.036
- Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls (opens in a new tab) · BMJ · doi 10.1136/bmj.b2393
- 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