7.2 Quasi-experimental Studies
Quasi-experimental methods are the next most robust evaluation method to an RCT. The difference being that in a quasi-experimental design randomisation to condition is not possible. There is usually some existing feature of the groupings or participants and their context that make it impossible to randomly allocate to different conditions, and so decisions must be made in advance as to how to allocate participants to the intervention vs. control groups. However, with this approach confidence in the causal effect of the intervention is lower, as other systematic differences between the two groups could impact the outcomes. Therefore it is important to try and match the two groups as closely as possible on important features that may influence the outcomes. There are statistical analysis techniques that can be used to mitigate some of the design flaw, but generally this approach would be suitable where random allocation is not possible, and preferred to less robust methods.
Example: Education programme
Where the intervention activity is delivered to groups of participants at the same time, who are also clustered in some way, a quasi-experimental design may be the most appropriate. For example, in the education programme, children will be grouped within schools, classes, or key stages for intervention delivery. Within those groups the children are likely to share a number of characteristics, and therefore be similar to one another, but different to groups of children from others schools, classes or key stages (e.g. family make-up, SES, teacher input, developmental stage).
Additionally, it may not be possible to deliver the intervention to come children within a school setting, without the information being shared across peer groups, invalidating any allocation to control groups. This means that using a similarly matched school as a control group would be more appropriate. Again, the children within the matched school would be more similar to each other than the intervention school children, and thus if comparable on other factors, a suitable control condition. Outcomes of interest can be measures before, during, and after the intervention is delivered, as well as at a pre-determined follow-up point to assess retention or maintenance of the intervention impact.