activities. Other strategies for minimizing bias are presented in Chapters 7 and 20, which cover
study designs and causal inference.
Sampling in multiple stages
When conducting large, epidemiologic surveillance studies, it is necessary to do an especially good
job of sampling, because governments use results from these studies on which to base public policy.
As an example, because being obese puts community members at risk for serious health conditions,
government public health agencies have a vested interest in making accurate estimates of the rates of
obesity in their communities.
For this reason, to strive to obtain a representative sample, researchers designing large epidemiologic
surveillance studies use multi-stage sampling. Multi-stage sampling is a general term for using
multiple sampling approaches at different stages as part of a strategy to obtain a representative sample.
Figure 6-1 provides a schematic describing the multi-stage sampling in the U.S. surveillance study
mentioned earlier, NHANES.
© John Wiley & Sons, Inc.
FIGURE 6-1: Example of multi-stage sampling from the National Health and Nutrition. Examination Survey (NHANES).
As shown in Figure 6-1, in NHANES, there are four stages of sampling. In the first stage, primary
sampling units, or PSUs, are randomly selected. The PSUs are made up of counties, or small groups of
counties together. Next, in the second stage, segments — which are a block or group of blocks
containing a cluster of households — are randomly selected from the counties sampled in the first
stage. Next, in the third stage, households are randomly selected from segments. Finally, in stage four,
to select each actual community member who will be offered participation in NHANES, an individual
is randomly selected from each household sampled in the third stage.
That is how a sample of 8,704 individuals participating in NHANES in 2017–2018 was selected to
represent the population of the approximately 325 million people living in the United States at that
time. The good news is that biostatisticians work on teams to develop a multi-stage sampling strategy
— no one is expected to set up something so complicated all by themselves.