Pre Post Design

A pre post design is a quasi experimental study design that measures an outcome before and after an intervention in a single group of participants, without including a concurrent control group. Also referred to as a one group pretest posttest design, it is one of the simplest and most widely used evaluation frameworks in applied research.

Pre post designs are popular in educational research, quality improvement initiatives, public health interventions, and pilot studies because they are practical, inexpensive, and easy to implement. They require no randomization and no control group, making them feasible in settings where experimental designs are impractical or unethical. However, their simplicity comes at a significant cost to internal validity. The absence of a control group means that observed changes cannot be confidently attributed to the intervention, as numerous alternative explanations may account for the pre to post difference. Researchers, reviewers, and policymakers must understand these limitations to avoid drawing causal conclusions from pre post data alone.

The basic structure of a pre post design involves three steps: measuring the outcome of interest at baseline (the "pre" measurement), delivering the intervention, and then measuring the same outcome again (the "post" measurement). The difference between pre and post scores, often tested with a paired t test or its nonparametric equivalent (the Wilcoxon signed rank test), is the observed change, a first estimate of the intervention's effect that the threats described next may distort.

The central problem with reading that change as the intervention's effect is that several threats to internal validity cannot be ruled out without a control group. Maturation refers to natural changes that occur over time regardless of the intervention (for example, children's reading ability improves as they age). History refers to external events that occur between the pre and post measurements and could influence the outcome independently. Testing effects arise when the act of taking the pretest itself influences post test performance through practice or sensitization. Regression to the mean occurs when participants are selected because of extreme baseline scores and would have moved closer to the average on remeasurement regardless of intervention.

Researchers can partially mitigate these threats by adding design features. Including a nonequivalent control group, even one that is not randomly assigned, substantially strengthens the design by providing a baseline for comparison. Using multiple pre measurements or time series data can help distinguish intervention effects from underlying trends. Extending the post measurement period with multiple follow up assessments can reveal whether observed changes are sustained. Once a comparison group is added, statistical techniques such as analysis of covariance (ANCOVA) adjusting for baseline scores can improve precision, though they do not eliminate confounding from unmeasured variables.

Despite its limitations, the pre post design remains a valuable starting point when resources are limited or when the goal is to generate preliminary evidence of feasibility and potential efficacy before investing in a larger, controlled trial.

Use a pre post design for pilot or feasibility studies, quality improvement projects, or educational evaluations where a control group is not feasible. It is appropriate when the primary goal is to generate preliminary evidence or assess whether an intervention is worth testing in a more rigorous design. Do not rely on a pre post design to establish causation or to make definitive claims about intervention effectiveness. When possible, add a comparison group, use an interrupted time series design, or employ a stepped wedge design to strengthen causal inference.

A common mistake is interpreting a significant pre to post change as evidence that the intervention caused the improvement, ignoring plausible alternative explanations such as regression to the mean or concurrent environmental changes. Researchers often fail to consider the magnitude of the change in context, reporting statistical significance without discussing whether the change is clinically or practically meaningful. Another error is selecting participants based on extreme baseline scores (such as the lowest performing students), which maximizes the apparent effect due to regression to the mean. Finally, failing to measure potential confounders or document concurrent events between measurements makes it impossible to assess alternative explanations.

A hospital implements a new electronic medication reconciliation system and measures medication error rates for six months before and six months after implementation, with about 10,000 different admissions in each period. The error rate declines from 8.4 per 1,000 admissions (84 errors) to 5.1 per 1,000 admissions (51 errors); because the admissions are different patients, the two periods are compared as independent groups (p < 0.01). However, during the same period, the hospital also hired additional clinical pharmacists and a national patient safety campaign was launched. Without a control hospital for comparison, the researchers cannot determine how much of the improvement is attributable to the electronic system versus these other changes.