Study Design
Lessons learned from implementing respondent driven sampling to recruit people experiencing homelessness in California: The California Statewide Study of People Experiencing Homelessness Paul Wesson* Ryan D. Assaf Paul Wesson Angelica DeGaetano Meghan D. Morris Margot Kushel
Over 181,400 people experience homelessness (PEH) in California (Ca), accounting for 28% of the US PEH population and half of its unsheltered population. Probability sampling methods, e.g. venue-based sampling (VBS), have advantages over convenience or single-site sampling, but miss “hidden” subgroups. In a statewide representative study of PEH, we conducted multi-stage VBS across 8 CA counties complemented by respondent driven sampling (RDS) to sample “hidden” subgroups of PEH through their social networks: transitional aged youth (TAY), individuals who experienced domestic violence (DV), LGBTQ+ individuals, and day laborers. We adapted RDS to overcome recruitment challenges with options for digital coupons, phone surveys, and mailed or emailed incentives. To assess the approach we describe VBS and RDS participants, recruitment chains, and population homophily estimates (where a value greater than 1 indicates preferential, not random, recruitment). Pre-specified subgroups were more represented in the RDS arm (n=158) than the VBS arm (n=3041) (Table). Recruitment chains were longest for the largest county and for those who opted for a phone interview. We note high levels of homophily for each of the subgroups: TAY= 3.27; LGBTQ+= 1.48; DV= 9.03; day laborers= 1.51. Despite our adaptations, RDS sample size and recruitment chains were smaller than expected. Still, homophily and comparison of sample proportions support RDS for targeted recruitment of hidden subgroups. We offer several recommendations to strengthen RDS for PEH: 1) partner with community liaisons who can champion in the field; 2) conduct interviews at dedicated physical study sites as well as via phone over longer recruitment periods; 3) dedicate trained personnel and resources for the RDS arm. Given our experience and the everyday challenges in the lives of PEH, RDS added value to our overall sampling design by representing voices of those who would have otherwise been missed.
