Strategies for obtaining probability samples of homeless youth

Background: In order to generalize to the homeless youth population a probability sample is essential, yet few studies have attempted to obtain probability samples of homeless youth from a diverse array of settings in terms of types of sites and/or geographic areas. Methods: For a study on the relationship between the social context and risky behaviors in homeless youth, we designed and obtained a probability sample of 419 homeless youth from 15 shelters, 7 drop-in centers and 19 street hangouts located in four regions of Los Angeles County with the highest concentrations of homeless youth. Results: We describe a two-stages time-location sampling methodology, as well as key fielding strategies, that were adopted to obtain a large and representative probability sample of homeless youth. We also discuss how the study results would have differed if we had sampled the homeless youth from only a specific type of sites (e.g. only from shelters or only from street hangouts) or from only a narrower geographic area within Los Angeles County (e.g. only from Hollywood). Conclusions: Homeless youth that can be found in shelters differ considerably from youth that do not use shelters both in terms of demographic characteristics and rates of risky behaviors. Similarly, we found differences in the characteristics and behaviors of homeless youth across geographic regions. Our findings demonstrate how limiting the type of sites or geographic areas from which youth are sampled can considerably bias study results.