Sampling in Difficult to Access Refugee and Immigrant Communities

Abstract

We evaluated sampling strategies and trust-building activities in a large multiphase epidemiologic study of torture prevalence in populations that were difficult to locate and enroll. Refugee groups under study were Somalis from Somalia and Oromos from Ethiopia who were living in Minneapolis and St. Paul, Minnesota, in 1999-2002. Without a complete sampling frame from which to randomly recruit participants, we employed purposive sampling methods. Through comparative and statistical analyses, we found no apparent differences between our sample and the underlying population and discovered no effects of recruiting methods on study outcomes, suggesting that the sample could be analyzed with confidence. Ethnographic trust and rapport-building activities among investigators, field staff, and immigrant communities made it possible to obtain the sample and gather sensitive data. Maintaining a culture of trust was crucial in recovering from damaging environmental events that threatened data collection.

Trauma and coping in Somali and Oromo refugee youth

“…In addition, generalizability is limited with nonprobability sampling methods. Without the availability of a complete sampling frame, we lacked the ability to assess true representativeness of the sample; however, comparison with available local population data from publicly available records yielded no large discrepancies [33], suggesting an analyzable sample.…”

Consequences of Untreated Posttraumatic Stress Disorder Following War in Former Yugoslavia: Morbidity, Subjective Quality of Life, and Care Costs

“…Instead we used a combination of different recruitment methods, which has been suggested as an appropriate strategy to reduce selection bias when random sampling cannot be achieved (35). A strong systematic selection bias would have particularly affected the prevalence rates of PTSD and the mean score of SQOL in the sample, while the associations between predictive factors and outcomes are likely to be more robust toward sample differences.…”

HADStress Screen for Posttraumatic Stress

“…A near-total 80% population sample was obtained because of the impossibility of identifying all potential participants to conduct a random sample. A detailed sampling description and demographic evaluation of the sample have been published (Spring et al, 2003). The method of sampling Ethiopian refugees depended heavily on targeted cluster sampling (i.e., neighborhoods, demographic clusters).…”