Before and after study designs

Answers

Statements a and c are true, whereas b is false.

A before and after study design was used to investigate whether exposure to secondhand smoke changed in primary schoolchildren after the implementation of the legislation. The before and after study is experimental by design (a is true). The design involves evaluating the effects of a deliberate intervention—for example, the introduction of clinical guidelines—by comparing the outcomes of study participants investigated before the intervention with those measured afterwards. The participants investigated before and after the intervention would be from the same sites or centres, and may be the same people or different samples. In the study above, the intervention was the introduction of legislation to prohibit smoking in most enclosed public places. The schoolchildren in the sample investigated before the intervention were different from those in the sample investigated afterwards. The two samples of children were selected from the same schools.

A before and after study can be “uncontrolled” and “controlled” in design. A controlled before and after study design is one that measures outcomes in two groups of participants before and after the implementation of an intervention, with one group receiving the intervention and the other—the control group—not receiving the intervention. The control group would have similar characteristics to the intervention group and the children in these groups would be expected to experience similar changes in their environment apart from the intervention. The intervention and control groups would be studied concurrently. The above study used an uncontrolled before and after study design (b is false). A concurrent control group was not introduced—that is, a sample of children from a region where the legislation was not implemented was not included.

Caution is needed when interpreting the results of the above study. Any observed changes in exposure to secondhand smoke were assumed to be due to the intervention. However, without a control group it is difficult to ascribe changes in exposure to secondhand smoke to the legislation. Any observed changes in outcome may have been due to general trends in smoking behaviour or other interventions, such as a public health campaign introduced at the same time as the study. A control group would have permitted the epidemiology of secondhand smoke exposure in schoolchildren in an area where the legislation was not implemented to be recorded. The outcomes for the control group could then have been compared with those for the intervention group, permitting the effects of the legislation to be evaluated.

The uncontrolled before and after study design is a non-randomised one. Non-randomised study designs have been described in a previous question.2 In the above study the children were not randomly allocated to the groups—that is, the group studied before the intervention or the group studied after the intervention. The two groups of children were selected on the basis of belonging to a class in the final year of primary school, with one group being studied before the intervention and the other after. Hence, the samples may not have been comparable in their demography and characteristics. Therefore, any differences between the groups in outcome may not have resulted from the intervention but might have been due to confounding (c is true). However, the researchers did control for any influence that child growth, season, and environment might have had on the outcomes by recruiting children of the same age, from the same schools, at the same time of year. Although statistical analysis would have adjusted for confounding by those variables that were measured, it would not have been possible to record all characteristics that influenced the association between the intervention and outcome measures.