Abstract
Background
Understanding the burden of common mental health disorders, such as depressive disorder, is the first step in strengthening prevention and treatment in humanitarian emergencies. However, simple random sampling methods may lead to a high risk of coercion in settings characterized by a lack of distinction between researchers and aid organizations, mistrust, privacy concerns, and the overarching power differential between researchers and populations affected by crises. This case analysis describes a sampling approach developed for a survey study of depressive disorder in a Syrian refugee camp in Greece (n = 135).
Discussion
Syrian refugees face an extraordinarily high burden of depressive disorder during the asylum process (43%), necessitating population screening, prevention, and treatment. In order to preserve the informed consent process in this refugee camp setting, the research team developed a two-phase sampling strategy using a map depicting the geographical layout of the housing units within the camp. In the first phase, camp management announced a research study was being undertaken and individuals were invited to volunteer to participate. The participants’ container (housing) numbers were recorded on the map, but were not linked to the survey data. Then, in the second phase, the camp map was used for complementary sampling to reach a sample sufficient for statistical analysis. As a result of the two phases of the sampling exercise, all eligible adults from half the containers in each block were recruited, producing a systematic, age- and sex-representative sample.
Conclusions
Combining sampling procedures in humanitarian emergencies can reduce the risk of coerced consent and bias by allowing participants to approach researchers in the first phase, with a second phase of sampling conducted to recruit a systematic sample. This case analysis illuminates the feasibility of a two-phase sampling approach for drawing a quasi-random, representative sample in a refugee camp setting.
Optimal auxiliary-covariate-based two-phase sampling design for semiparametric efficient estimation of a mean or mean difference, with application to clinical trials
