SAMPLING CONSIDERATIONS: IDENTIFYING APPROPRIATE PARTICIPANTS FOR BEHAVIORAL INTERVENTION RESEARCH
If you aren’t taking a representative sample you won’t get a representative snapshot.
An important aspect in the design of a behavioral intervention study at any stage along the pipeline is the selection of an appropriate sample. Clearly it would be ideal to include an entire population (e.g., all family caregivers of patients with Alzheimer’s disease [AD] or all American adults with high blood pressure) in intervention research. This would enhance the external validity of the study or the extent to which results can be generalized to the population as a whole. However, in most cases, this is not feasible as populations are typically large and geographically diverse.
For example, estimates may vary slightly, but currently about 70 million American adults have high blood pressure (Centers for Disease Control and Prevention [CDC], 2015), and although rates of high blood pressure vary somewhat by geography (e.g., self-reported estimates tend to be higher in some areas of the southern United States as compared to areas in the northwest), there are adults with high blood pressure across the 50 states. Thus, it would not be possible for a researcher conducting a trial to evaluate a behavioral intervention for blood pressure control to include the entire target population, even if the study was multisite. Instead, investigators rely on samples (a subset of a population) to examine proof of concept and to test the feasibility, efficacy, and/or effectiveness of interventions as well as to attempt to generalize the findings and conclusions to an entire population.
Using a sample is advantageous for a number of reasons. Samples involve a smaller number of people and thus are less costly, more time efficient, and require less effort with respect to recruitment and data collection. In addition, it is easier to maintain treatment integrity with smaller numbers of people. Samples can also be selected to reduce heterogeneity. For example, AD is not a unitary disease, but has different symptom presentations at different stages of the disease progression. If an investigator was interested in testing the efficacy of a cognitive training intervention for people with AD, it would generally be more appropriate to evaluate the intervention with people at the mild stages of the disease as those in the later stages of the illness would be unlikely to benefit. Thus, in this case, a sample of the AD population with certain characteristics is more appropriate as opposed to the entire population of people with AD.
In selecting a sample for an intervention study, researchers must be aware of any potential bias in participant selection. Bias can lead to errors in the interpretation of results from an intervention study and limit the ability to generalize the findings to other groups of people. Referring to the blood pressure intervention example described earlier, assume the study involved a nutritional intervention and that the sample was restricted to White males. Obviously, the sample would be biased with respect to gender and ethnicity. The findings could not be generalized to other segments of the hypertensive population such as females and those from other ethnic groups, given differences in the characteristics of these segments of the population (e.g., body size, hormonal differences, cultural food preferences and patterns) that could have an impact on the effectiveness of the intervention.
It is clear that the selection of the sample is an essential consideration in behavioral intervention research. Who should be included in the evaluation of an intervention depends, of course, on the specific research question; the target population of the intervention; study design; and feasibility constraints with respect to budget, time, staff, and participant availability. Two important considerations are the composition of the sample and sample size. The sample should be representative of the target population on characteristics important to the research question and intervention and be of sufficient size to provide adequate power to test the study hypotheses. Unfortunately, oftentimes, researchers focus just on the size of the sample without giving adequate consideration to representativeness and feasibility issues. In this chapter, we discuss the topics of sample composition (who should be included in the evaluation of an intervention) and sample size as well as issues related to feasibility. We also discuss approaches to sampling and which sampling methods are most useful in behavioral intervention research. Our goal is to provide some guidelines to help optimize the selection of samples across the behavioral intervention pipeline.