Psychological Distress and Social Support among Con ict Refugees in Urban, Semi-rural and Rural Settlements in Uganda: Burden and Mechanisms of Association

Stephen Lawoko (  s.lawoko@gu.ac.ug ) Gulu University Faculty of Medicine https://orcid.org/0000-0003-4369-7955 Catherine Nakidde Centre for Health and Socio-economic Improvement Eric Lugada Centre for Health and Socio-economic Improvement Maria Ssematiko Centre for Health and Socio-economic Improvement Dunstan P. Ddamulira Agency for Cooperation and Research in Development (ACORD) Andrew Masaba Lutheran World Federation Brian Luswata Directorate of Governance and Regulation, inistry of Health and Community Service: Ministry of Health Community Development Gender Elderly and Children Eric A. Ochen Makerere University College of Humanities and Social Sciences Betty Okot Makerere University College of Humanities and Social Sciences Denis Muhangi Makerere University College of Humanities and Social Sciences Gloria Seruwagi Makerere University College of Humanities and Social Sciences


Introduction
COVID-19 has exasperated the health and wellbeing of population groups worldwide since it was declared a pandemic over a year ago. Beside the fear, anxiety, confusion, and frustrations triggered by the pandemic per se [1], the stringent measures activated by governments to curtail the disease spread have further exacerbated psychological distress. Globally, experts predict that the collective impact of these restrictions on household socio-economics, health and wellbeing, as well as the social infrastructure of communities will extend beyond the lifespan of the pandemic [1][2][3][4][5][6][7]. These trends are further manifest in the increasing burden of intimate partner violence, depression and anxiety [8][9][10][11][12].
Despite the growing body of evidence suggesting deteriorating psychosocial wellbeing at population level, equivocal data from refugee populations is lacking. Currently surpassing 80 million in number globally, con ict refugees are particularly at heightened risk of psychological distress when contrasted with host communities. A myriad of psychosocial problems including Post-Traumatic Stress Disorder (PTSD), depression and anxiety [13][14], and psychological distress triggered by fears of apprehension, risk of deportation, di cult living conditions, poor access to health, social, communication, nancial and legal services [15][16][17][18][19][20][21], have been reported in refugee settings. This is in addition to the stressors imposed by restrictions to contain COVID-19 pandemic at the population level [1][2][3][4][5][6][7]. An assessment of psychological distress status and psychosocial support needs for refugees is therefore particularly imperative in the pandemic era.
Social support can play an important role in preventing, containing or moderating the psychological wellbeing of populations through several mechanisms. While some scholars have emphasized its role as a stress buffer [22][23], others have highlighted its protective function as a coping facilitator [24][25] as well as its role in health promotion [26]. As a stress buffer, social support is envisioned to alleviate the detrimental impact of stressful life events by modifying negative appraisals and promoting problem solving strategies [24,[26][27]. Some scholars argue however that during times of severe or chronic distress, the buffering effect of support may be limited [28][29]. In contrast, as a coping facilitator, social support is envisaged to provide regular directly rewarding experiences such as positive affection, which prevent the development of psychologically distressful outcomes [22,26]. Moreover, a supportive network is hypothesized to promote behaviors bene cial to health such as timely seeking of healthcare and adaptation of healthy lifestyles (e.g. healthy nutritional choices and physical activity) [24,[26][27], thereby reducing the likelihood of psychological distress.
Although population studies have alluded to a breakdown in social support networks during the COVID-19 pandemic, there is a dearth of studies investigating its role in containing psychological distress particularly in refugee populations where the need is augmented. Moreover, refugees live under varying conditions with some residing in urban and others in rural settings, and this has implications on vulnerability to health problems. For instance, it could be assumed that urban refugees may be more prone to social support by virtue of the vast availability of social networks or opportunities for social interaction (e.g. through gainful employment) on the one hand, but may on the other hand be confronted with challenges related with psychological stressors that accompany increased population density and diversity in cities such as unemployment, violence, marginalization and exposure to health-risk behaviours [30][31][32]. Moreover, settlement of refugees in rural/urban areas have in some countries been characterised by cultural homogeneity, and culture is known to play an important role in social integration. In Uganda the context of the current study for example, South Sudanese, Congolese and Somali refugees have been resettled in rural, semi-rural and urban settlements respectively. While on the one hand the collectivist orientation of such settlements may promote social interaction by virtue of shared cultural norms and interest, efforts to integrate into a new society or culture may discourage such intentions [33][34]. Thus, an assessment of the nature of relationship between social support and psychological distress in distinct settlements (i.e. urban vs. rural) is warranted on its own right.
Upon this background, we scrutinized social support and psychological distress, and compared the impact of social support on psychological distress among refugees in urban, semi-rural and rural settlements in Uganda. Speci cally, we addressed the following research questions: Indeed, emerging data in vulnerable groups of refugee women and slum-dwellers have catapulted their risk to stigma, all forms of violence and nancial disadvantage [42][43].While researchers have predicted the economic, psychosocial, physical, and other consequences of COVID-19 on refugees/migrants in Uganda and beyond based on previous epidemics [44][45][46][47], there is a dearth of evidence on the burden of psychological distress, social support and mechanisms linking these phenomena in refugee settings during the COVID-19 pandemic. The current work intends to ll this gap in the evidence using Uganda as a case study. Such data could be useful in the design of interventions to cushion psychosocial problems among refugees through the modi cation of social support agents.

Study site and population
We conducted the research at 3 large refugee settlements in different regions of Uganda, hosting over 400,000 refugees: 1. Kisenyi, refugee settlement, an urban refugee setting in the centre of the capital city (Kampala) hosting over 70,000 refugees of mainly Somali origin. The refugees live integrated with their host. 2. Kyaka II refugee Settlement in the South Western part of Uganda, a semi-rural refugee setting hosting multinational refugees from the Democratic Republic of Congo (DRC), Burundi and Rwanda totaling approximately 124,000 refugees. The refugees live partly segregated from their host but with freedom of movement and shared services. The region can be considered as semi-rural, with a blend of rural and urban activities (e.g. farming, industrial activities) 3. Adjumani refugee settlement in North-West Nile Uganda, hosting about 214,000 refugees predominantly of South Sudanese nationality. The refugees live rather segregated from their host but with freedom of movement and shared services. The region is considered as rural, with Agriculture as main activity.

Study design
Cross-sectional survey data on various health and social indicators was gathered from 1014 refugees randomly selected from each of the study sites. For the current study, data on psychological distress, social support, demographic, social and behavioral indicators was of primary interest.

Sampling procedure
Participants were sampled using a two-staged cluster sampling procedure in each settlement. The rst stage involved selecting clusters of zones in the main settlement using systematic random sampling with probability proportional to zone size (PPS). The second stage constituted systematic sampling of households in selected zones. Random numbers procedures were used to choose one adult household member (i.e. 15 years and above) from among all adults in the household to constitute the nal participant. This procedure resulted in 1014 refugees, with the following distribution among the settlements: Adjumani n = 342;Kyaka 354; Kisenyi n = 318.

Ethical Considerations
Thirty Research Assistants (RAs) were trained to collect data using mobile tablets, in a bid to reduce interindividual contact and the risk of COVID-19 spread during interviews. The training oriented RAs on the purpose of the study; ethical considerations; data collection methods and tools; COVID-19 prevention, symptoms, measures and precautions; and standard operating procedures (SOPs) in eldwork in light of COVID-19. The training also involved testing of the data collection tool among a purposively selected refugee sample of n = 30 in each of the 3 settlements, from zones neighboring but not included in the main study. Slight adjustments were made to data collection tools following this exercise.
Informed consent was received from all participants and con dentiality considered. The potential risk and bene ts of the study were explained to all participants and in light of the heightened risk of COVID-19 transmission, we developed Standard Operational Procedures (SOPs) for protection of refugees as well as data collectors, guided by Safety and Security Strategy for COVID-19 of the World Health Organization (WHO) and Uganda Ministry of Health COVID-19 Guidelines.

Page 7/26
A comprehensive questionnaire covering several areas of relevance to public health and COVID-19 was developed. For the current study, the following variables were of interest:

Dependent variables
The dependent variable for the study was psychological distress, measured using Kessler's Psychological Distress Scale (K10)[48], a previously validated 10-item instrument measuring distress in terms of feelings of nervousness, hopelessness, tiredness, restlessness, dgety, depressed mood, sadness, worthlessness, cheerlessness and loss of effort, during the past 14 days, with a 5-level response ranging from none of the time (score 1) to all of the time (score 5). A composite score for depression is calculated for each participant as the sum of the 10 items. Thus, individual scores for depression scale ranged from 10-50, with higher scores indicative of higher depression risk. Cronbach's alpha testing for internal consistency/reliability of Kessler's Psychological Distress Scale for the current sample was 0.91 indicating very high reliability.

Independent variables
The main independent variable for this study was social support, with the aim to assess its direct impact on psychological distress, together with the assessment of its role in buffering the impact of living in various settlement categories (i.e. urban, rural, semi-rural settlement) on distress. Social support was measured using the Interview Schedule for Social Integration (ISSI) [49], which assesses social support in terms of the Availability and Adequacy of Social Interaction and Social Attachment.
Availability of Social Interaction was assessed using six items inquiring of participants to indicate the number of people: with whom they have common interest, meet and talk to regularly, can speak with openly, can borrow things from and can turn to when in trouble. This was thus a continuous scale ranging between 0-in nity. The participants were in addition requested to rate the Adequacy of these numbers indicating whether they desire more (coded as 1), less (coded as 1) or no change (coded as zero) in the number indicated. Thus, Adequacy of Social Interaction was rated on a total scale ranging between 0-6, with higher scores denoting higher inadequacy (i.e. poor adequacy). Cronbach's alpha testing for internal consistency/reliability of Availability and Adequacy of Social Interaction respectively for the current sample was 0.50 and 0.81 respectively, indicating moderate to high reliability respectively.
Availability of Social Attachment was assessed based on six items inquiring of participants to indicate using a "Yes" (coded as 1) or "No" (coded as 0) response regarding whether there is someone special: from whom they derive support, they feel close to, they share happy moments, they can embrace for comfort, appreciates what they do, and with whom they can share inner thoughts. For social attachment, composite individual scores are calculated as the sum of responses to each item. Thus, scores for social attachment range between 0-6, with higher scores representing higher availability. Adequacy of Social Attachment was assessed by inquiries to participants on whether they desire more (coded 1), less (coded 1) or no change (coded 0) regarding the special persons mentioned above. Thus, scores for this scale ranged between 0-6, with higher scores representing higher inadequacy (i.e. poor adequacy). Cronbach's alpha testing for internal consistency/reliability of Availability and Adequacy of social attachment respectively for the current sample was 0.55 and 0.87 respectively, indicating moderate to high reliability respectively.
Other independent variables included in the study were demographic, social and behavioral characteristics of participants, such as sex, age, marital status, religion, income, employment status, alcohol and substance use and physical activity. As these variables are from previous studies generally known to be associated both with social support and psychological distress, it is prudent to adjust for them in the main analyses in a bid to control for possible confounding.

Statistical Analysis
Cronbach's Alpha coe cients were calculated to assess for reliability of the instruments of interest (i.e. Kessler's Psychological Distress Scale and ISSI)in the current sample. In the assessment and contrasts of the burden of psychological distress in rural, semi-rural and urban refugee populations (research question 1), ANOVA was used. To compare the rates of availability and adequacy of social support among refugees in rural, semi-rural and urban settlements (research question 2), Analysis of Variance (ANOVA) were used. To assess for bivariate associations between Psychological distress and Social support indicators, demographic, social and behavioural independent variables, t-test, ANOVA and Pearsons Correlations tests were run.
In the assessment of the mechanism linking social support to distress (i.e. the direct effects and the stress buffering effects, research question 3), Multivariable Linear Regression (MLR) was used. In the assessment of the direct effect of Social Support on Psychological Distress, MLR including the entire study sample was run with psychological distress as the dependent variable and social support indicators as the main independent variables, but adjusting for independent effects of demographic (including urban/semi-rural/rural residency), social and behavioural factors on the dependent variable.
The direct effect of social support would be con rmed if the social support indicators remained statistically signi cantly associated with distress despite adjustment for possible confounders.
The stress buffering effect was assessed using strati ed MLR with settlement as the main independent variable, but controlling for demographic, social and behavioural factors. Strati cation was by social support availability/adequacy levels categorized under "high" (de ned as lacking in 1 or none of the 6 dimensions studied), "moderate"(i.e. lacking in 2 or 3 of the 6 dimensions studied), "low" (lacking in 4 of the 6 dimensions studied) and "very low" (lacking in 5 or 6 of the 6 dimensions studied). Social support would be considered as a stress buffer if it acted as an effect modi er in the association between psychological distress and urban/semi-rural/rural residency, i.e. if the association between refugee settlement (i.e. rural/urban residency) and psychological distress is stronger where support is weak and weaker where support is strong.
SPSS version 22 was used for all analyses and a statistical signi cance of p < 0.05 was assumed for all tests.

Results
Demographic, social and behavioural characteristics of Participants 57% earning less than 50,000 Uganda Shillings per week) and were uneducated (40%).

Participants ratings of Availability and Adequacy of Social Interaction by Settlement
On average, refugees rated their availability of social interaction rather high, ranging between 23-28 people across the settlements (table 2)

Discussion
Few studies in refugee settings have scrutinised the individual level social resources available to refugees and how these may manifest on their psychological wellbeing. Accordingly, we assessed the burden of psychological distress and scrutinized the mechanisms linking social support to distress among refugees resident in urban, semi-rural and rural settlements in Uganda.

Burden of Psychological Distress in Urban, Semi-rural and Rural Refugee Settlements
Levels of psychological distress were on average moderate among refugees, but with notable variations across settlements, i.e. signi cantly higher distress levels were observed among refugees resident in semi-rural and rural settings when contrasted with peers resident in urban settlements. While such data, to the best of our knowledge is previously lacking in refugee cohorts, population studies have generated contradictory results in this respect. Some scholars have supported the notion of heightened psychological distress in urban areas due to stressors related with urbanization such as unemployment, violence, marginalization, discrimination and increased exposure to health-risk behaviours, while other researchers have envisioned residents in urban settings to have an upper edge in health as a result of better access to health and social services in urban than in rural settings [30][31][32]. Our ndings could be a re ection of the latter circumstances.

Availability and Adequacy of Social Interaction and Social Attachment in Urban, Semi-rural and Rural Refugee Settlements
Refugees rated their availability of social interaction (which encompassed access to resources required for regular social activities such as conversation and meeting people) and availability of social attachment (which covered access to social relations of emotional relevance such as embracing for comfort and sharing inner thoughts) as high, but rated adequacy of such resources (i.e. whether such resources were perceived su cient) as low. Moreover, the results from the regressions analyses found adequacy of social interaction/attachment rather than availability of social interaction/attachment to correlate positively with psychological distress (i.e. with increasing adequacy, levels of distress reduced). Corroborating previous research in general population studies [50][51], these ndings contribute new data to the refugee literature demonstrating that the effectiveness of social support in moderation of psychological distress among refugees is anchored on the perceived meaning and value of the support network (here measured by adequacy of social interaction and attachment), rather than the support network's morphological aspects (e.g. density and size of the network).
Availability of social interaction and attachment was highest among refugees residing in urban settlements when contrasted with peers residing in rural and semi-rural settlements. A plausible explanation for this result could stem from the fact that urban refugees lived in integrated rather than separate settlements with the host communities, and this may have provided opportunities for social networking with host community members, to a higher extent than in rural settlements where the residential areas are distinctively for refugees. On the other hand, adequacy of social interaction and attachment was highest among refugees resident in semi-rural and rural settlements when contrasted with peers resident in urban settlements. These results are more di cult to reconcile. While the ndings linking availability of support networks to urban settlement are largely congruent with previous works in general populations [30][31][32], the ndings linking adequacy of support to rural settings appear to contradict those works. Social support is a multi-dimensional concept and discrepancies in results across studies could stem from variations in its conceptualization between the studies. For instance, some studies have measured social support in terms of the form of support received (e.g. emotional support, instrumental support, cognitive guidance, informative and appraisal support) [52][53], while others have focussed on the morphology of the network (e.g. size) [50][51], or value of relationships within the network (e.g. relationship reciprocity) [54]. Thus, cross-study comparisons should be done with caution.
Mechanisms Linking Social Support to Psychological Distress in Urban, Semi-rural and Rural Refugee Settlements The regressions analyses ruled out availability of social interaction and availability of social attachment respectively as independent predictors of Psychological Distress, but con rmed adequacy of social interactions and adequacy of social attachment respectively as predictors of such distress, i.e.
Psychological Distress levels reduced as adequacy of social interaction and adequacy of social attachment respectively increased. Furthermore, our study generated evidence in support of the stress buffering function of social support [22][23], which postulates that the deleterious effects of psychosocial stressors on health can be reduced or eliminated with the existence of a strong support network. Speci cally, we hypothesized and demonstrated that social support can moderate the impact of stressors due to rural/urban settlement on psychological distress in refugee populations.

Implications and Recommendations
The burden of psychological distress and social support in refugee settings drum for the need of surveillance of these phenomena as a basis for informed action. The need to revitalize, strengthen or restructure existing social networks cannot be overemphasised, with observance of COVID-19 prevention measures notwithstanding. Humanitarian actors will need to invest more in mental health and psychosocial support services (MHPSS), tailoring these interventions to the unique needs of different refugee categories. The leveraging of professionals across disciplines (e.g. social workers, psychologists, community health-workers) to be more actively involved in the MHPSS response to distress is apparent, particularly in rural/semi-rural refugee settings, during this COVID-19 era where refugees' access to the natural individual support networks (e.g. relatives and friends) have been temporarily constrained by government directives. A multi-sectoral team including main actors in humanitarian settings, should be set up to address how such measures can be implemented without necessarily exposing refugees to increased risk of COVID-19 alongside other challenges to implementation.
Moreover, the ndings con rming the direct and buffering hypotheses of social support have important implications for interventions to manage distress in refugee settings. Firstly, the direct pathway linking speci c aspects of social support (i.e. inadequacy) to psychological distress provides new evidence to support the design of focused (as opposed to generic) interventions, aimed at strengthening social support among refugees. In addition, the study has demonstrated the potential utility of the stress buffer framework of social support in the identi cation of irreversible stressor in refugees' lives (e.g. settlement in urban/rural settings) that could lead to psychological distress. Moreover, social support could, within such frameworks, be applied to predict the thresholds at which such stressors may lead to severe psychological distress, information of importance in early detection and prevention strategies.

Study Limitations
While the study carried the strength of a large representative sample of refugees in Uganda, utilized validated instrumentation and applied conventional statistical methodology to respond to the research questions, some of its weaknesses deserve acknowledgement. First, the cross-sectional design does not permit inference of causality, which is implied in some of the hypotheses tested. For instances, we can only rmly ascertain associations between social support and distress. Secondly, the homogenous cultural composition of refugees in each of the studied settlements (i.e. predominance of Somalis in the urban settlement vs. South Sudanese and Congolese in the rural and semi-rural settlements respectively) makes it di cult to disentangle associations between distress and urban/rural settlement on the one hand from associations between distress and country of origin on the other hand. Thus, the observed differences in distress between settlements could be masking differences in distress due to nationality. Due to the presence of collinearity in our data with regard to these two variables (i.e. settlement and nationality), we could not address this question by way of analysis. Future research focussing on the separation of these two effects is warranted using the appropriate study design.

Conclusion
There is a settlement inequality (i.e. rural vs. urban) in psychological distress and social support among con ict refugees in Uganda. Both the direct and the buffering effects of adequacy of social interaction and attachment can be forwarded to explain the rural/urban discrepancy in psychological distress, with   Adequacy of Social Attachment** 0.10 1 The analysis was adjusted for settlement, sex, religion, occupation, age, education level and physical activity. Due to high collinearity between nationality and settlement, nationality was excluded from the analysis