Distress Due to Displacement and Repatriation of an Entire Island Population in the South Pacific

Publication Date: 2019

Abstract

Displacement due to natural disasters poses challenges for human health, including psychosocial distress, particularly in low- and middle-income countries that lack resources for intervention and recovery. In October 2017, the entire population of Ambae Island, Vanuatu, was evacuated for 4–6 weeks due to an erupting volcano. This study assessed distress symptoms among those displaced, the predictors of distress, and the efficacy of mental health interventions by local health professionals. We surveyed 470 displaced adults following their return home. Questionnaires assessed sociodemographic characteristics, displacement characteristics, dietary patterns, and distress. Prelminary analyses of 99 adults assessed levels and sociodemographic predictors of distress. Mean distress scores did not differ between men or women (p=0.454). Differences in mean scores between non-pregnant (2.1, SD=1.0) and pregnant women (1.2, SD=0.8) approached significance (p=0.063). Sociodemographic variables did not predict distress among men. Among women, distress was correlated with age (r=-0.348, p=0.013) and household size (r=0.384, p=0.008). Approximately 58 percent of participants had high distress scores suggestive of possible post-traumatic stress disorder. Analyses of the full dataset will highlight effects of locally adapted psychosocial interventions on psychosocial distress. Results could help identify those at risk of distress due to displacement and highlight measures to help alleviate distress.

Background and Objectives

Population displacement due to natural disasters is likely to increase as extreme weather events are progressively more common globally (Lee et al. 2014)¹. Even if temporary in nature, displacement poses unique challenges for local infrastructure and the administration of services in affected communities, particularly in low- and middle-income countries (LMICs) that may lack sufficient resources for adequate disaster preparedness and recovery plans (Mirza 2003²). Displacement also poses specific risks to human health, including mental distress and nutritional vulnerability. Distribution of health services during displacement plays an important role in adaptive response in communities and individuals, and lessons learned in LMIC settings can be applied to other vulnerable populations. Further research on the effects of displacement on mental health measures (Siriwardhana 2015³, Patel 2007⁴), and disaster preparedness and outcomes in LMICs (Lee et al. 2014), is needed.

In October 2017, the entire population of one rural island in Vanuatu, a lower-middle income country of 280,000 people in the South Pacific, was evacuated due to high risk from an erupting volcano. More than 12,000 adults and children were moved to rural villages on nearby islands or to the urban capital for a period of 4–6 weeks. The authors have worked in Vanuatu since 2007—including on the island of interest, Ambae—conducting surveys of population health and health behaviors (Dancause et al. 2011⁵, Dancause et al. 2012⁶, Dancause et al. 2010⁷, Dancause et al. 2013⁸, Olszowy et al. 2015⁹, Olszowy et al. 2017¹⁰, Sun et al. 2017¹¹, Weitz et al. 2017¹²). Since 2015, we have incorporated psychosocial health assessments into our surveys. For example, in 2015, with support from the Natural Hazards Center and the Wenner-Gren Foundation for Anthropological Research, we assessed post-traumatic stress disorder (PTSD) symptoms and dietary patterns among 979 women following the Category 5 Cyclone Pam, which hit the archipelago in March 2015. We also worked in collaboration with a team that conducted malariometric surveys of 3,009 people in 12 rural communities (Chan et al. 2017¹³). The following year, we assessed chronic distress, anxiety, depression, and dietary patterns among nearly 1,200 women. Our findings from 2015 show that distress levels following the cyclone were very high, and that high distress among pregnant women in the sample predicted smaller weight among their infants at birth. Controlling for confounding variables, distress explained 8.5 percent of variance in birthweight (p=0.012) and was an even more important predictor than maternal dietary characteristics.

Recognizing the importance of distress following natural hazards, we worked with the Vanuatu Ministry of Health to assess distress among the population displaced from Ambae, and to assess factors that might help to alleviate that distress. In the summer of 2017, a handful of local health professionals participated in mental health training workshops and are now trained to provide basic mental health interventions; this is a modest service, but the first of its kind in Vanuatu. Following the evacuation of Ambae, a small team was sent to provide psychosocial health interventions among displaced families in some villages. This provided a first opportunity to assess the effects of a psychosocial health intervention in the setting. Given the limited mental health infrastructure in Vanuatu (and in many similar LMICs), such basic and small-scale psychosocial health interventions might represent the best efforts possible for many countries. Our assessments could help to highlight what aspects of the intervention worked and for whom, and could guide improvements to the intervention tactics for future applications, as well as the development of similar small-scale, sustainable interventions that could be applied in other LMICs.

The objective of the current project was to assess distress symptoms among people displaced from Ambae, to assess predictors of distress, and to assess the efficacy of mental health interventions delivered by local health professionals.

Methods

This project was approved by the research ethics committee of the Université du Québec à Montreal.

Questionnaires

We modified the questionnaire used in our 2015 study of distress and dietary change due to the cyclone to assess distress and dietary change due to displacement. The questionnaire includes sociodemographic data (e.g., age, parity, education), details of the displacement (e.g., where residents stayed and for how long), distress, and diet (examined using a 24-hour dietary recall and food frequency questionnaire) during and following displacement, and interactions with psychosocial health professionals and other sources of support, such as local chiefs and pastors.

Distress was evaluated using a Bislama adaptation of the Impact of Events Scale - Revised (IES-R) (Weiss and Marmar 1997¹⁴). This 22-item scale describes symptoms from three categories relevant to PTSD: intrusive thoughts, hyperarousal, and avoidance. Participants report the extent to which each behavior describes how they felt over the preceding seven days. Responses are on a five-point Likert scale, from “Not at all” (0) to “Extremely” (4). Items were written to reflect symptoms relative to the displacement. We translated questions and responses into Bislama and reviewed them with native speakers. Some vocabulary was not directly translatable, leading to repetition in some items. We retained three questions from each category of PTSD symptoms. In the current preliminary analyses, scores ranged from 0–4, and the mean of the response values for all nine items was used for the analyses.

Sample

We distributed questionnaires at local hospitals and clinics, churches and community groups, and local markets. The goal of the convenience sample was to reach as wide a demographic as possible to allow for broad characterization of distress at the population level. Data were collected in 13 villages representing four main regions (North, South, East, and West) of the island. Questionnaires were completed by 470 individuals.

Analyses

Data entry and analyses are underway. In preliminary analyses, we assessed sociodemographic predictors of distress among 99 adults from six villages who represented all four regions of the island. We tested whether mean scores differed by sex and pregnancy status using one-way analysis of variance (ANOVA) and correlations between sociodemographic variables (i.e., household size, number of children, years of education, and age) and distress. We also calculated the percentage of participants with high scores suggestive of possible PTSD.

Results

Mean distress scores were similar among regions (North=1.8, standard deviation (SD)=1.1; South=2.0, SD=1.1; East=2.0, SD=0.9; West=1.7, SD=1.0) (p=0.787) and did not differ between men (1.8, SD=1.1) or women (2.0, SD=1.0) (p=0.454). Differences in mean scores between non-pregnant (2.1, SD=1.0) and pregnant women (1.2, SD=0.8) approached significance (p=0.063; based on a significance level of 0.05).

Correlations among distress and sociodemographic variables differed by sex. Among men, distress was not correlated with age (r=-0.121, p=0.409), household size (r=0.123, p=0.403), number of children (r=0.041, p=0.809), or years of education (r=0.020, p=0.893). However, among women, distress was correlated with age (r=-0.348, p=0.013) and household size (r=0.384, p=0.008), suggesting higher scores indicative of PTSD among younger women and women from larger households.

Based on cutoff points used in our previous studies in Vanuatu, 58 percent of participants had high distress scores suggestive of possible PTSD, with similar prevalence among men (55 percent) and women (60 percent) and across regions (North=44 percent, South=58 percent, East=63 percent, West=65 percent).

Applications of this Research

The major applications of this research will be evident following analyses of the full dataset, and in particular, differences in distress scores among people who participated in mental health interventions and those who did not. However, our preliminary analyses allow for some initial comparisons to our results obtained after the 2015 cyclone in Vanuatu.

For example, our 2015 analyses, which were focused specifically on women, showed no correlations between distress and age (r=-0.038, p=0.257) or number of children (r=0.056, p=0.091); however, they highlighted correlations between distress and years of education (r=-0.136, p<0.001), which persisted in multivariate analyses that controlled for confounding variables. This suggests greater distress among women with less education. Similar patterns have been observed in other LMICs (Telles, Singh, and Joshi 2009¹⁵). In contrast, the current preliminary results showed that education was not a predictor of distress among men or women.

Second, prevalence of high symptoms suggestive of possible PTSD was lower than that after Cyclone Pam, but was still remarkably high. Participants expressed concern about their livestock and gardens while displaced, which was likely a major source of distress for both men and women. One year after the original assessments following Cyclone Pam, prevalence of high symptoms was less than half that at three months following the disaster, dropping from 70.5 percent to 33.7 percent. Recovery following displacement will be based on a variety of factors including not only interventions, but also the state of households, gardens, and livestock. Follow-up surveys are planned to assess the changes in distress over time.

Differences in distress predictors and prevalence between the two studies might be due, in part, to the severity of the events and the timeframe of the studies. The displacement resulted in damage to gardens (which were overtaken by weeds, destroyed by animals, and might suffer long-term damage due to volcanic ash) and often the loss of livestock. These hardships affected nearly everyone in the population. In contrast, the cyclone resulted in the complete destruction of some homes but spared others, resulting in widely different distress levels within villages. Distress symptoms following the displacement were lower than those after the cyclone, but potentially more widespread.

Heavy ashfall from the Manaro volcano resulted in damage to crops and contamination of water supplies across Ambae. ©Amanda B. Roome

Furthermore, assessments following the displacement were conducted almost immediately after the population had been repatriated to their island, whereas those after the cyclone were conducted 3–4 months after the disaster, thus capturing longer-term symptoms. Sociodemographic factors, such as age and education, likely affect the recovery profile through related factors, such as health literacy and access to services. Differences in distress based on sociodemographic characteristics might become apparent in the year following repatriation.

Our future analyses will highlight the role of interventions on patterns of distress following the displacement. Vanuatu, like most LMICs, lacks the infrastructure, budget, and personnel to conduct detailed, long-term, individual interventions with people suffering from distress. Simple, locally-administered mental health interventions must be prioritized if they are to be sustainable. If effective, such interventions applied in Vanuatu could help to guide similar projects in other LMICs, and could also be useful additions to existing interventions in the United States and other industrialized countries.

References