The use of network sampling for locating the seriously ill

Network sampling, based on the experience of conducting the large-scale 1986 Robert Wood Johnson Foundation National Access to Health Care Survey, was shown to be a viable alternative to more conventional procedures for oversampling persons with a low incidence and prevalence of health conditions. The network sampling approach was a more economical and methodologically less obtrusive means of increasing sample size of persons with desired characteristics than conventional procedures. The approach requires estimating the incidence, prevalence, or both of the required characteristics in the population. As the proportion of persons in the population with the characteristic increases, network sampling may lose its attractiveness. However, this has to be judged both in economic terms and in terms of the methodologic consequences of using other procedures to increase the size of subgroups of interest. As discussed, additional methodologic work is required on the consequences of network sampling for study group representativeness and sampling bias. However, network sampling, in the authors' opinion, is sufficiently promising to warrant methodologic explorations to answer these and other relevant questions about its use.