Rare Populations

Sampling Strategies

Generally, sampling frames for the total population do not contain information identifying members of the rare population; if they do, then the sampling process is trivial. If not, then screening must be used to identify members of the group. With screening, members of the rare population are identified at the beginning of the interview process. However, the costs of screening can often exceed the costs of interviewing, especially if the rare population is only a very small proportion of the sampling frame.

There are several ways in which screening costs can be reduced. Mail questionnaires can be used to identify members of the rare population if the sampling frame has correct address information; for in-person interviews, the use of telephone screening can reduce interviewer costs. If the rare population is geographically clustered, for example, in certain states or urban areas, screening based on clusters becomes less costly.

Besides cost, another potential drawback to the screening approach is response errors of commission and omission (i.e. false positives and false negatives, respectively) during the screening process, especially if many different questions must be correctly answered in order to identify a member of the rare population. Using less stringent criteria during screening to identify members is one approach to reducing response errors, because misclassified members (false positives) can be excluded after the full interview is complete. The main problem here is devising a screening process that avoids erroneously classifying members of the rare population as non-members (false negatives).

Whereas for many rare populations it is impossible to derive a complete sampling frame, one or more incomplete lists may be available. For example, hospital or pharmacy records may identify some, but not all, members of a population with a specific rare disease. A dual frame approach could then be used, in which the partial list is combined with screening of the total population to reduce screening costs. Alternatively, the partial list could be used with a cluster sampling approach to identify areas where members of the rare population are located.

In some situations multiplicity or network sampling can be used to locate and interview members of a rare population. Typically, a member of a household is interviewed and queried as to whether other members of the household or close relatives are members of a special population (although respondents also can be asked about neighbors or members of other social groups to which they belong). Occasionally, the researcher may need only an estimate of the size of the rare population; however, in most cases the researcher wishes to interview members of the rare population. If so, accurate contact information may be difficult to obtain from the respondent. In general, accuracy of reporting depends on the relationship between the respondent and the member of the rare population and the visibility of whatever characteristic(s) defines the rare population. For example, World War II veteran status is generally well known and thus visible to network members, whereas certain illnesses may not be. Researchers must use special weighting when deriving estimates from network samples, because the probability of selection can vary by the size of the network. Costs may be higher than with typical screening methods, especially if addresses are sought.

...