- Research article
- Open Access
Association between distress and displacement settings: a cross-sectional survey among displaced Yazidis in northern Iraq
BMC Public Health volume 21, Article number: 679 (2021)
Globally 70.8 million people have been forcibly displaced from their homes and are at disproportionally high risk for trauma. At the time of this study, there was an estimated 1.6 million internally displaced persons (IDP) in Iraq, more than two-thirds of whom reside in private, urban settings. This study aims to understand the impact of post-displacement accommodation on mental well-being of the Yazidi minority group displaced in Iraq.
Multi-stage stratified sampling was used to randomly select IDPs in camp and out of camp settlements in northern Iraq. Standardized questionnaires evaluated factors including exposure to violence and self-reported distress symptoms (measured by Impact of Event Scale-Revised). A multi-variate linear model assessed the relationship between settlement setting and distress symptoms.
One thousand two hundred fifty-six displaced Yazidi participants were included in the study: 63% in camps and 37% out of camps. After controlling for exposure to violence, social cohesion, unemployment, and access to basic services, IDPs in camps were predicted to have a 19% higher mean distress symptom score compared to those out of camps.
This study provides a framework to investigate post-displacement accommodation as a potential intervention to improve well-being for displaced populations. With a shift towards new models of emergency and long-term housing, it is important to understand the potential and limitations of more decentralized models, and identify effective methods to maintain access to basic services while improving living conditions for both displaced populations and their host communities.
Survey sites and sample selection
Participants for this study were selected among IDPs in out of camp (urban) and camp-based settlements in the Niveneh governorate and in the Erbil and Duhok governorates in the Kurdistan Region of Iraq (KRI). Multi-stage stratified sampling strategies were used for both population groups.. Sample size for the two population groups was determined using the sample size formulation for difference in proportions. We assumed a 90% confidence with a Z distribution, 80% power, and a difference of 20% and a design effect of 2, for a total target of 150 interviews per group. This target was adjusted by 20% for anticipated non-response and rounded for logistical purposes and equal assignment to interviewers, for a target of 200 interviews. We anticipated multiple comparisons within each population groups based on gender and known ethnic diversity. The target sample size for IDP in camps was 1800 interviews. The target sample size for IDPs out of camp was 1200.
For IDPs in camps, 15 out of 31 camps for IDPs were randomly selected proportionate to population size using the latest data from the United Nations High Commissioner for Refugees (UNHCR). Within camps, interviewers were randomly assigned to blocks of similar sizes where they used a geographic sampling method to randomly select participants. Interviews could not be completed in two of the camps because the presence of IS family members posed a security risk for the interviewers. As a result, 1575 interviews were ultimately conducted with IDPs in camps (87.5% of targeted 1800 interviews, within the 20% margin).
A similar multi-stage stratified sampling strategy was used to randomly select IDPs out of camps. Five out of 12 subdistricts were randomly selected proportionately to the out of camp population size based on the latest data from the International Organization for Migration (IOM) (mid-2018) available at the time of the survey. Within the subdistricts, 3 sites (neighborhoods) identified by IOM as hosting IDPs were randomly selected, proportionate to the IDP population size. In one subdistrict, four sites were selected due to the small population at each site. Once at the site, interviewers randomly selected a direction and approached every fifth household. If the household did not host IDPs, they were skipped. Because of the small number of IDPs present at each site, an additional 9 sites (neighborhoods) were randomly sampled, resulting in a final sample size of 1406 IDPs not living in camps (117.2% of the target).
From the two sample of IDPs (in and out of camps), only those that indicated being part of the Yazidi ethnic religious group were included in this study. The rationale for the sub-sample was to increase homogeneity in the socio-cultural background and characteristics that have been shown to be associated with mental health. Among 1575 IDPs residing in camp, 794 were Yazidis (50.4%), and 462 Yazidis were among the sampled 1406 IDPs living out of camps (32.9%).
The study protocol was reviewed and approved by Partners Human Research Committee (PHRC, Protocol #2018P001748)) and a local ad-hoc committee and local authorities in northern Iraq where the data were collected.
Data collection instruments and scale
Respondents were asked questions regarding exposures to violence, social cohesion, access to services, and household assets. Household assets were categorized as medium assets (bed, fridge, and washing machine) and extended assets (mobile phone, internet access, radio, computer, microwave, heater, car, and television) per the UNHCR Vulnerability Assessment .
The 22-item Impact of Event Scale-Revised (IES-R) based on a Likert scale of 0 (not at all) to 4 (in full) was used as a validated instrument to estimate subjective distress and calculate a distress score. Possible scores ranged from 0 to 88, and a score of greater than 33 has been established to be have high diagnostic accuracy for PTSD [31, 32].
Statistical analyses were performed using Stata (v15, SE). Baseline demographics were presented as medians [interquartile ranges (IQR)] for continuous variables or percentages for categorical variables and compared by post-migration settlement location using Wilcoxon rank sum or chi-square test. Given low numbers of non-response, non-response answers were treated as missing values.
Linear regression was used to assess associations between distress scores and settlement location. Exposure to violence variables were converted into binary variables in regression to reflect any history of past exposure. Social cohesion variables were aggregated into a score (ranging from 1 to 29), with 1 indicating the least amount and 29 the highest amount of social cohesion. All variables associated with the outcome of interest with a p-value of < 0.2 in univariable analysis or deemed relevant to post-displacement accommodation were evaluated for inclusion in the final model. Backwards stepwise regression was then performed to derive the final multivariable model which included variables associated with a p-value of < 0.05 as well as variables forced in for known association with distress score.