A test of nonrespondent bias in a family-based study: a research note.

The increasing use of family samples from "normal" populations (as opposed to clinical samples) in substance abuse research raises concern about the validity of responses from participating families. Nonparticipating families may have greater numbers of substance users who wish to conceal problems, when compared to families which cooperate with researchers. Unobtrusive analyses of institutional records of a health maintenance organization comparing families agreeing to participate in a substance use study (N = 508) with those families unwilling to participate (N = 538) reveal no significant differences in the use of chemical dependency services between the two groups. Surveys of the non-participating families indicate that the major reasons for noncooperation were the lack of time and inconvenience in getting family members together. These findings suggest the feasibility of sampling families from normal populations.