Planning for suicide prevention in Thai refugee camps: Using community-based system dynamics modeling

Abstract

Suicide and associated behaviors represent a significant health disparity among refugees and displaced persons. Despite this burden, evidence for prevention programing in these populations is limited. This study aimed to inform the selection and implementation of suicide prevention strategies in refugee camps in Northwestern, Thailand – camps that had experienced recent spikes in suicides and suicide attempts at the time of the study. We leveraged Community Based System Dynamics modeling through a series of four workshops with key local stakeholders and suicide prevention experts, to build a qualitative systems model that accounts for complexities and is aimed at assisting local partners with selecting the most promising strategies for implementation and evaluation. The process expanded local understanding of the causes and consequences of suicide and resulted in selection of priority interventions aimed at reducing suicide in this context. Our research illustrates the application of a novel methodology that aims to account for the complexities of suicide prevention in the context of displacement and helps to optimize local suicide prevention efforts.

Introduction

Suicide deaths and attempt (herein collectively referred to as ‘suicidal behaviors’) are major global health concerns, with profound impacts on individuals, families, and communities (WHO, 2014). The World Health Organization (WHO) estimates that approximately 800,000 people die by suicide each year – resulting in a global mortality rate of around 10.6 deaths per 100,000 people (WHO, 2019) – and there is evidence that for each of these deaths, there may be more than 20 additional suicide attempts (AFSP, n.d.). While suicidal behaviors occur throughout the life course, suicide takes a particularly heavy toll on young people, representing the second leading cause of death among 15- to 29-year-olds worldwide (WHO, 2014). Beyond their social and psychological impacts, suicidal behaviors also result in a substantial economic burden on families and communities due to increased healthcare utilization, lost income, and lost productivity (Shepard, Gurewich, Lwin, Reed, & Silverman, 2016)

This burden may be even more pronounced in humanitarian settings, where rates of suicidal behaviors are thought to be higher due to socioeconomic disadvantage, traumatic exposures, ongoing stressors, mental health problems, and a lack of access to appropriate services (Vijayakumar & Jotheeswaran, 2010). To date, however, there has been limited epidemiologic research on suicidal behaviors among displaced populations. One existing review of suicidal behaviors among refugees found rates ranging from 3.4% to 34%, although the majority of included studies only measured suicidal ideation (Vijayakumar & Jotheeswaran, 2010). More recently, Akinyemi et al. (2015) found a heightened prevalence of suicidal ideation among refugees in Nigeria, with 27.3% of the refugee population reporting suicidal ideation compared to 17.3% of the non-refugee population. In the largest refugee camp in Thailand, a report by the International Organization for Migration (IOM) uncovered a troubling rise in suicide rates from 2015 to 2016, with rates reaching 36.6 deaths per 100,000 people (IOM, 2017). Another study conducted in the same camp found that 9% of maternal deaths were due to suicide, translating to a suicide-related maternal mortality rate of 16.4 per 100,000, much higher than the global average (Fellmeth et al., 2016).

Despite this emerging evidence for the increased burden of suicidal behaviors among displaced populations, few suicide prevention programs have been implemented in these contexts (Haroz, Decker, & Lee, 2018). Indeed, a recent systematic review of suicide prevention and response programs among refugees found only two that specifically targeted populations currently living in displacement (Haroz et al., 2018). Of these, the intervention with the strongest evidence was Contact and Safety Planning (CASP; Vijayakumar et al., 2017), which combines Brief Intervention and Contact (BIC; Fleischmann et al., 2008) with Safety Planning (Stanley & Brown, 2012), and is delivered by community health volunteers. CASP was rigorously implemented and evaluated among Sri Lankan refugees living in India, where it was found to significantly decrease suicide attempts and deaths (Vijayakumar et al., 2017); these results, however, have yet to be replicated in other settings.

However, one of the primary challenges of preventing suicidal behaviors among displaced populations is its multi-causal nature: suicidal behaviors are influenced by interacting factors across the individual, interpersonal, environmental, and cultural levels (Franklin et al., 2017), and the dynamics of displacement contribute added complexity. In order to be successful, it is necessary for prevention approaches to embrace and account for these complicated mechanisms. Systems thinking and its associated models, are grounded in the notion that the behavior of systems is governed by rules and principals that can be discovered and understood (Forrester, 1994; Meadows, 2008; Peters, 2014). At its most basic, systems thinking aims to understand how things are connected to each other within some concept of a whole (Mabry et al., 2008). Our understanding of most phenomenon, whether social, biological, or business related, is based on created mental models (Leischow & Milstein, 2006; Richardson, 2011). Systems modeling aims to make these mental models explicit so that assumptions can be specified and the models can ultimately be calibrated to data and replicated (Barlas, 1996; Hosseinichimeh et al., 2016; Peters, 2014; Schwaninger & Grösser, 2008). There are numerous theories and tools associated with systems thinking that span both qualitative and quantitative methods (Peters, 2014; Sterman, 2006; Wolstenholme & Coyle, 1983).

Systems dynamics models have been suggested as a possible cost-effective approach to help with the identification, selection, and prioritization of strategies aimed at addressing health outcomes and improving program implementation, including those related to suicide (Atkinson, Page, Wells, Milat, & Wilson, 2015; Lyon, Maras, Pate, Igusa, & Vander Stoep, 2016; Page, Atkinson, Heffernan, McDonnell, & Hickie, 2017; Peters, 2014; Zimmerman et al., 2016). For example, Page et al. (2017) created a system dynamics model of suicide to inform policy makers in their decisions to allocate money and resources to suicide prevention efforts in Australia. In suicide prevention, there is a need for preventative approaches that address or consider multiple potential targets. System dynamics models are designed to allow for the consideration of these targets, as well as the relevant emphasis and interactions between targets and strategies.

Our aim was to inform the selection and implementation of prevention strategies for suicidal behaviors in refugee camps in Northwestern, Thailand. This paper describes the use of a participatory approach, Community Based System Dynamics (CBSD) Modeling (Hovmand, 2014), to understand the causes, consequences, and mechanisms of suicide in two refugee camps on the border of Thailand and Myanmar. The current project aimed to build a qualitative model of a system of suicidal behaviors by combining stakeholders’ mental models and making them explicit, while simultaneously building capacity in systems thinking for participants. This type of qualitative model can serve as a building block for future quantitative system dynamics models. Our research illustrates the application of a novel methodology that accounts for the complexities of suicide prevention in the context of displacement and helps to optimize local suicide prevention efforts for maximum impact.

Methods

Setting

Activities for this study were conducted in the towns of Mae Sot and Umphang, both of which lie in close proximity to several of the nine refugee camps located along the western border of Thailand. In addition to serving as a base of operations for many refugee-serving organizations, these towns and the surrounding areas are home to high numbers of irregular migrants without legal status. As of September 2020, approximately 97,000 people live in the camps (UNHCR, n.d.). The camps were established in response to armed conflict in Myanmar and over the years have received refugees fleeing Myanmar in response to continued armed conflict, natural disaster, and political instability in Myanmar. Mae Sot is close to Mae La camp, the largest of the camps, which was established in 1984, and is home to the largest number of residence among all nine camps (37,000 population). Mae La camp has healthcare, education, and community-based service organizations such as women’s groups. Umphang is more isolated and has a camp population of 9,000 people. All camps are, officially, closed meaning that residents are not allowed to leave of their own accord to go to the host community areas, however, these two camps, along with the other seven, do have food rations and other services provided by organizations under the direction of The Border Consortium, an organizing body that oversees these activities in all nine of the camps.

Process and participants

We initiated a series of group model building workshops that brought together local stakeholders from organizations working with displaced populations in Thailand, along with experts on systems modeling, suicide prevention, health systems, humanitarian contexts, and global mental health (Table 1). The purpose of these workshops was to elucidate relevant factors related to suicidal behaviors in this context, collaboratively create a system dynamics model reflecting these key factors, pinpoint underlying feedback mechanisms, and identify locally acceptable intervention strategies over time. Local stakeholders provided their own views about causes, consequences and potential interventions for suicidal behaviors given their understanding of the context and population, while experts provided expertise in what is known about suicide and prevention efforts more broadly. Local stakeholders did not have particular expertise in suicide prevention. To enhance validity and reliability of the process, after each workshop the results of the workshop were summarized and presented back to the stakeholder groups for further comments and suggestions. These summaries were presented in three languages (Karen, Burmese and English) to ensure understanding by local stakeholders and expert participants alike. This longitudinal and iterative process is consistent with existing definitions and approaches for a community based system dynamics approach (Hovamand, 2016).

Table 1.

Group model building workshops

Description	Participants	Activities
Workshop 1: Mae Sot, Thailand September 2017	21 stakeholders; 3 JHUa facilitators; 7 Men; 14 Women; 3 stakeholders representing organizations that work with migrant community, 6 that work with internally displaced populations and 12 that work with refugee populations	Introduction: Introduced the CBSD approach.b Nominal group technique (Gallagher, Hares, Spencer, Bradshaw, & Webb, 1993): Participants created a preliminary list of causes and consequences of suicidal behaviors among displaced persons. Intervention options: Participants discussed potential intervention strategies for addressing suicidal behaviors.
Workshop 2: Umphang, Thailand September 2017	8 stakeholders; 3 JHU facilitators; 5 Men; 3 Women; All stakeholders work directly with refugees in the camps	Introduction: Introduced the CBSD approach. Nominal group technique (Gallagher et al., 1993): Participants added to the preliminary list of causes and consequences of suicidal behaviors among displaced persons. Initiating causal loop diagram: Participants developed a causal loop diagram depicting the relationships between key variables connected to suicidal behaviors. Intervention options: Participants discussed potential intervention strategies for addressing suicidal behaviors.
Workshop 3: Baltimore, US October 2017	9 JHU experts 3 JHU facilitators	Model review: Participants reviewed the preliminary system model and discussed missing factors. Intervention options: Participants discussed potential intervention strategies addressing suicidal behaviors.
Workshop 4: Mae Sot, Thailand November 2017	14 stakeholders; 1 JHU facilitator; 7 Men; 7 Women; All stakeholders work directly with refugees in the camps	Model presentation: Participants were presented with a more finalized version of the system model based on input from prior workshops. Intervention selection: Participants reviewed the list of promising intervention strategies, and selected those that they believed would be most appropriate and feasible among the target populations.

^aJohns Hopkins University

^bCommunity Based System Dynamics Modeling

Two initial local workshops took place in September 2017, which were facilitated by three researchers. The first workshop was held in Mae Sot, and included 21 participants representing organizations working with refugee, internally displaced person (IDP), and migrant populations. The second workshop was held in Umphang and included eight participants representing organizations working with refugee populations. Many of the workshop participants were from the displaced and migrant communities in the area (representing Karen and Burman ethnicities). Both of these workshops were structured based on a series of scripts adapted from Scriptapedia, an open source repository containing detailed instructions for carrying out structured participatory modeling exercises (“Scriptapedia,” n.d.). Scripts were designed to introduce systems thinking to local stakeholders, elicit key variables related to suicidal behaviors, explore dynamic relationships between these variables through the creation of a causal loop diagram, and generate potential intervention strategies.

During the first workshop, participants were split into two groups: one group for stakeholders from organizations working with refugee populations, and one group for stakeholders from organizations working with migrant or IDP populations. Participants were first asked to individually list the causes and consequences of suicidal behaviors, which were defined as suicide attempts and deaths among displaced persons. In order to facilitate the inclusion of these factors in the preliminary system model, participants were prompted to formulate these causes and consequences as variables that could increase or decrease over time. Participants were then invited to share the causes and consequences that they felt were the most important to suicidal behaviors, with their ideas documented as initial system model inputs. Finally, participants were asked to brainstorm policies, programs, and other initiatives that could potentially address suicidal behaviors among displaced populations. These could include policies or programs that had been tried in the past, were currently being tried, or could be tried in the future.

The second workshop built on the results from the first, with participants presented with the initial lists of causes and consequences of suicidal behaviors among displaced persons and asked to supplement these lists with key variables that they felt were missing. Facilitators then guided participants in using these variables to construct a preliminary causal loop diagram, which visually depicted the relationships between suicidal behaviors, variables thought to cause these behaviors, and variables thought to result from these behaviors. Throughout this process, participants were instructed to be mindful of the formation of feedback loops within the diagram, with these feedback loops offering potentially relevant points of intervention into the entire system. As with the first workshop, participants were also asked to generate a list of potential interventions targeting suicidal behaviors.

Following these initial information gathering sessions, a third workshop was held, which included nine participants with expertise in systems approaches, suicide prevention, global mental health, and humanitarian contexts. During this workshop, participants were presented with the locally devised variables, data on suicidal behaviors from the local context based on in-depth interviews and psychological autopsies (IOM, 2017), the preliminary system model, and the initial list of prevention strategies. They were then asked to provide feedback on any variables that they felt were missing from the model based on their expert knowledge, and to provide further ideas regarding potential evidence-based interventions. As the local stakeholders were not experts in suicide prevention, this workshop aimed to add knowledge from the broader suicide prevention field to inform model building efforts.

A final workshop was held in Mae Sot, and consisted of 14 stakeholders from organizations working with refugee populations. During this workshop, participants were presented with the refined system model to comment and further revise as necessary. They were also provided with an expanded list of promising intervention strategies, which included information regarding prior implementation of these strategies as well as overall evidence for their effectiveness in addressing suicidal behaviors taken from the scientific literature. Finally, participants were asked to select several interventions that they believed would be the most appropriate, feasible, and effective among the target population based on their experience working with refugee populations.

Analysis

The analysis process was iterative, with each subsequent workshops allowing for further refinement of the system model. The primary product from the first two local workshops was the collaboratively created causal loop diagram, which presented a preliminary graphical depiction of factors contributing to and resulting from suicidal behaviors among displaced populations in Thailand (Figure 1). Following these workshops, revisions focused on expanding the model based on expert knowledge regarding suicidal behaviors in humanitarian contexts, local data from previous research (IOM, 2017), and further information drawn from notes taken during the local workshops. In addition, the causes and consequences were thematically coded in order to simplify the model through variable aggregation, and variables were grouped at the individual, family, and community levels to facilitate a multi-level approach to modeling. Grouping of factors was done by two authors, EH and SF independently and then compared. Where there were disagreements, final categorization was decided through discussion and consensus between coders. The coded results were then presented back to workshop participants to ensure they were consistent with their interpretations and local beliefs.

Figure 1.

Preliminary causal loop diagram of suicidal behaviors in refugee populations

To transform our qualitative causal loop diagram into a quantitative model that could be used for future quantitative simulations of intervention impact, we used a stock and flow diagram, in which stocks can be conceptualized as reservoirs and flows act between these reservoirs. Stock and flow diagrams have been used in a wide range of public health applications, including epidemiology, implementation science, and studies on community resilience to disasters (Haase, Kabisch, & Rink, 2012; Homer & Hirsch, 2006; Kotir et al., 2016; Links et al., 2019). In suicide prevention, Page et al. 2017 used a stock and flow model to inform selection of policy interventions to prevent suicidal behaviors in Australia. The systems modelers translated the causal loop diagrams into stock and flow diagrams while preserving the variables and relationships generated during the first three workshops. These stock and flow diagrams were subsequently refined through the process described above, resulting in a final model representing causes, consequences and interventions for suicidal behaviors.

Results

Local workshop participants’ opinions regarding the causes and consequences of suicidal behaviors among refugee populations in Thailand, as well as potential intervention strategies, are presented in Table 2. The causes of suicidal behaviors can be grouped into seven general categories: 1) financial insecurities, including issues related to poverty, employment, and debt; 2) psychosocial distress, including feelings of stress, hopelessness, loneliness, embarrassment, shame, self-centeredness, isolation, boredom, rejection, inadequacy, and anger; 3) behavioral problems, including alcohol use, drug use, and gambling; 4) interpersonal conflict, including family and intimate relationship issues, family separation, and lack of social support; 5) traumatic exposures, including sexual and gender-based violence, child maltreatment, and sudden displacement; 6) biological propensity, including chronic medical conditions, cognitive impairments, and family history; and 7) structural and health systems restrictions, including lack of resettlement opportunities, decreased rations, restrictions on movement, and lack of programs. Likewise, the consequences of suicidal behaviors can be grouped into three general categories: 1) psychosocial distress at the individual, family, and community levels; 2) functional impairments for both the individual and family, such as physical disability, decreased productivity, lack of income and employment, and family deprivation; and 3) elevated risk of further suicidal behaviors.

Table 2.

Causes, consequences, and potential interventions for suicidal behaviors among refugee populations elicited from local workshops

Causes	Consequences	Potential interventions
Financial insecurities: • Decreased employment opportunities • Increased debt • Increased poverty/income issues Psychosocial distress: • Increased feelings of stress, hopelessness, loneliness, embarrassment, shame, self-centeredness, isolation, boredom, rejection, inadequacy, anger, not being able to complete life plan • Increased mental illness • Increased worries about forced repatriation Behavioral problems: • Increased drug and alcohol use • Increased gambling Interpersonal conflict: • Increased family relationship issues • Increased family separation • Increased interpersonal conflict/fighting • Increased intimate relationship issues • Increased premarital pregnancy • Decreased support systems Traumatic exposures: • Increased child maltreatment • Increased domestic violence • Increased rape/sexual assault • Increased sudden internal displacement • Increased copycat suicide Biological propensity: • Family history/genetics • Increased cognitive impairment/disability • Increased prolonged illness/chronic medical conditions Structural and health systems restrictions: • Decreased capacity to determine those at-risk • Decreased counseling services • Decreased participation in existing programs • Decreased rations • Decreased regulation by CBOs • Decreased third-country resettlement opportunities • Increased commercialism • Increased disease outbreaks • Increased restrictions on movement • Increased young couples not allowed to get married	Psychosocial distress at the individual, family, and community levels: • Increased feelings of fear, being small, frustration, guilt/regret, sadness, shame/embarrassment, stress, worry about being looked down on • Increased social isolation/avoidance of interpersonal interactions • Increased substance use • Increased family relationship issues • Decreased trust • Increased worry/anxiety, trauma, and anger in the family • Increased worry/anxiety in the community • Increased community stigma • Increased copycat suicides • Increased worry about negative response from Thai authorities • Increased organizational embarrassment Functional impairments for the individual and family: • Increased physical disability/health consequences after suicide attempt • Increased medical service needs • Increased avoidance of school • Decreased employment • Decreased productivity • Decreased income for the family • Increased deprivation among family • Increased issues for family members • Increased time spent looking after person Elevated risk of future suicidal behaviors: • Increased risk of another suicide attempt	• De-stigmatization • Engage community leaders/workers • Engage religious organizations • Extra vigilance for possible attempted suicide and train responders • Find means to create hopefulness/justice • Give medical care to person who tried to commit suicide • Health education • Hide means that people use to commit suicide • Home visiting • Increased participation in social support programs • Livelihood/income-generating activities • Mandated training/reporting by all service providers • Peer training • Prevention of drug and alcohol use • Provide basics needs (food/nonfood items) • Psychosocial intervention • Strengthen camp management • Strengthening health information system • Train community health workers • Training on the value of life • Training/learning center for suicide prevention • Training/raising awareness in the family/community

Using these locally derived causes and consequences, we worked with participants from the second workshop to create a preliminary causal loop diagram illustrating the problem of suicidal behaviors in this context (Figure 1). The purpose of this initial paper version of the causal loop diagram was to aid in the identification of key factors that participating organizations may want to target with interventions. While this preliminary system model does not include all of the causes and consequences of suicidal behaviors in this setting, it starts to give shape to some of the dynamic feedback loops that contribute to this issue. For example, decreased social support was identified as a critical factor in increased suicidal behaviors, with ties to many of the consequences of these behaviors including social isolation, stigmatization, and splintered family relationships. This suggests that organizations may want to consider interventions that promote social support as a possible strategy to reducing suicidal behaviors.

With the causal loop diagram as a foundation, the research team created several stock and flow diagrams to capture further systemic complexities related to suicidal behaviors in this context. For clarity, rather than immediately presenting the complete stock and flow diagram, we present it in stages, beginning with half of the diagram that is centered on psychosocial distress (Figure 2). Here, we see a single stock, Suicide Deaths, and a single flow, Suicide Rate, into that stock. In stock and flow diagrams, it is customary to graphically represent the flow rate with a valve. The Suicide Deaths stock represents the subpopulation who have died due to suicide. The remainder of Figure 2 shows the inter-related factors that lead to psychosocial distress. Different color shades are used to indicate factors at the community, family, and individual levels. This portion of the stock-and-flow diagram closely mirrors the causal loop diagram. The most important difference is that the single factor, Increase in Suicide Behaviors, that is in the causal loop diagram is replaced by the following: a factor, Psychosocial Distress, which is connected to the Suicide Rate flow, which leads to the Suicide Deaths stock. Hence, we represent increases in suicide behaviors in slightly more detail as increases in psychosocial distress, which in turn lead to increases in the suicide rate.

Figure 2.

Initial form for the stock and flow diagram of suicide behaviors in refugee populations

While psychosocial distress was not the sole cause of suicidal behaviors, after review during workshop 4, it was agreed upon that many of the other causes of suicidal behaviors were indirect causes of these by way of increasing psychosocial distress. While we could have used Figure 2 as the final form of our stock and flow model, the suicide prevention experts found that this model could not readily accommodate most of the potential suicide-prevention interventions which more specifically target distinct points in the process towards suicide attempt and death that are all aggregated into a single flow in Figure 2. Hence, we decided to expand this flow to include more stocks to represent the multiple stages of suicidal behaviors and recovery, with the flows between these stocks. Figure 3 shows the resulting expansion of this part of the model, with stocks in the upper portion of the figure starting from Non-Suicidal Thought, progressing with flows to Ideation and Suicide Attempts stocks before ending at Suicide Deaths. Here, as in compartment models, each stock is a separate subpopulation with a given state, such as the state of suicide ideation. The flows represent transition from one state to another, such as a person who is initially in the ideation state transitioning to a person who attempts suicide.

Figure 3.

Expansion of the portion of the model for suicide behaviors and recovery

While the upper portion of Figure 3 shows the stocks representing the stages toward suicide, the lower portion shows survival and recovery pathways. We begin at the right side of the figure, where those who attempt suicide would flow to the Suicide Deaths stock if they were successful or to the Attempt Survivors stock if they survived. Individuals in this stock can transition to the Non-Suicidal Behaviors stock at which point they move to the Non-Suicidal Behaviors stock if they are recovering or return to the Suicide Ideation stock if they did not recover. In addition, some in the Non-Suicidal Behaviors stock can move to the Participation in Social Programs stock, provided that such programs exist in the community. Assuming that these programs are effective, all of the participating individuals transition back to the Non-Suicidal Behaviors stock.

We are now ready to combine the model components shown in Figures 2 and 3 into a single model, which is shown in Figure 4. Here, increases in the Psychosocial Distress factor impacts every flow in the model as indicated by the arrows between this factor and the valves. Most of the arrows are solid, indicating positive impact (e.g., increases in the flows toward the Ideation stock). Some of the arrows are dashed, indicating negative impacts (e.g., decreases in the flow from Ideation to Non-Suicidal Behaviors). In the opposite direction from the stocks to the factors on the left side of the figure, increases in the Attempt Survivors stock results in increases in the Chronic Illness and Unemployment factors, while increases in the Participation in Social Programs stock has an attenuation effect on Psychosocial Distress.

Figure 4.

Final form for the stock and flow diagram of suicide behaviors in refugee populations

Inputs from local stakeholders, as well as discussions with suicide prevention experts, yielded a number of potential interventions for further consideration. These included: restricting access to means, targeting individuals with previous attempts, gatekeeper training, suicide surveillance and case management, safety planning; male engagement strategies, interpersonal therapy, providing clinics with atropine to treat pesticide poisoning, community awareness activities such as public education campaigns, culture-based approaches to increasing community engagement, targeted support for individuals who have experienced sexual- and gender-based violence, peer-to-peer mentorship, and religious leader engagement. From this list of potential interventions, local stakeholders, using the model to identify key points in the suicide risk process, settled on four strategies that they felt were likely to be appropriate, feasible, and effective in this setting: restricting access to means, targeting individuals with previous attempts, gatekeeper training, and community awareness activities. Stakeholders also hypothesized where in the model these strategies would have an effect (Figure 4).

The relationships of each of these four intervention strategies on the flows are shown in Figure IV. Starting from the left, Community Awareness Activities acts on the population who are transitioning from Non-Suicidal Thought to Ideation; hence this intervention would reduce the flow to the Ideation stock. Next, Gatekeeper Training seeks to identify those who are close to attempting suicide so that they can be treated. This results in a diversion of some of the flow to the Suicide Attempts stock towards the flow to the Non-Suicidal Behaviors stock. Restricting Access to Means acts to reduce the flow from Suicide Attempts to Suicide Deaths. Finally, Targeting Individuals with Previous Attempts results in higher rate of recovery among those who attempted suicide. This is indicated by increases in the flow from the Attempt Survivors stock to the Non-Suicidal Behaviors stock.

Discussion

This project aimed to inform the selection and implementation of prevention strategies for suicidal behaviors in refugee camps in Northwestern, Thailand by local stakeholders. Using a participatory community-based system dynamics process, we were able to generate a wide range of causes, consequences, and potential interventions to prevent suicidal behaviors among refugees from Myanmar living in camps in Thailand. This community-based process helped expand understanding of the complexities of suicidal behaviors in this context and engage local stakeholders around possible solutions through group model building activities and discussion.

Instead of narrowly focusing on only one causal pathway or intervention, participants were encouraged to consider multiple overlapping causes and consequences of suicidal behaviors and could ultimately use that knowledge to inform a more comprehensive suicide prevention strategy. This conceptualization of suicidal behaviors as a complex problem is consistent with the literature showing multiple etiological factors that combine to produce risk (e.g. Franklin et al. 2017). In fact, the pathways to risk for suicidal behaviors identified during this process (Figure 4) are consistent with pathways found in other non-refugee populations (Franklin et al. 2017; Page et al. 2017) indicating preliminary evidence that these same pathways may be very relevant for the current context.

In terms of prevention strategies, the consideration of multiple and overlapping etiologic pathways identified through the group model building process, yielded local consideration of a wider array of intervention approaches. For example, awareness raising was commonly mentioned among stakeholders prior to completing the modeling workshops. However, over the course of the process, it became clear to stakeholders through their interactions with others in the groups and the group model building process, that the problem was more complex and that there were other points of interventions that may be appropriate for different levels or risk and/or may lead to more substantial impact. Awareness raising targeted the flow from non-suicidal thought to suicide ideation, which may be helpful in a universal prevention approach, but would not completely stop suicides given that some people would need more intensive or targeted interventions (Figure 4). While school-based awareness programs have been shown to be effective in reducing ideation and attempts, the evidence for broader public awareness campaigns has shown no reductions in actual suicide (Zalsman et al., 2016). The process of combining participants mental models of suicide and the resulting in a shared understanding about all the possible causes and consequences of suicide and potential solutions, is a key outcome of our approach.

The stock and flow diagram in Figure 4 can provide insights into the system of suicidal behaviors even before it is translated into a quantitative model. For instance, there is a feedback loop that can be traced from the Attempt Survivors stock to the Chronic Illness factor, resulting in higher Psychosocial Distress. This represents the idea that if one survives an attempt, this might result in some sort of chronic disability resulting from an injury, that will in turn increase one’s psychosocial distress which may subsequently raise someone’s risk for subsequent suicidal behaviors. An intervention that disrupts this feedback loop, for example a psychosocial or coping intervention targeted at suicide attempt survivors, might ultimately help drive down suicide risk in this population. A tighter feedback loop can be seen with flows from Suicide Attempts to Attempt Survivors, proceeding to Non-Suicidal Behaviors and Suicide Ideation before returning to Suicide Attempts indicating that some people who survive suicide attempts may recover completely, while others experience relapses of suicide risk. This is consistent with literature that indicates that the most potent risk factor for an attempt, is having attempted previously, yet despite this being the strongest known risk factor, most people who attempt have no history of this behavior (Franklin et al. 2017). An important property of these feedback loops is that they are reinforcing, meaning that increases in any factor or stock within the loop results in a cascade of causal impacts that cycles back to create further increases in the original factor or stock.

Nearly all of the proposed interventions shown in Figure IV could attenuate this trend and result in reduction in the cyclical flows. For instance, the Awareness Raising and Gate-keeper Training interventions will restrict the flows towards more severe suicide behaviors in the upper portion of the feedback loops, whereas the Caring Contacts will increase the flows towards recovery pathways in the bottom portion. The Restriction to Means intervention is critical because it is the only intervention that will reduce the flow from Suicide Attempts to Suicide Deaths. Its impact on the feedback loops, however, is indirect and limited, and hence may be insufficient to counter the reinforcing spiraling effects of these loops. Our stock and flow diagrams allow prevention planners to explicitly understand where in the system interventions target, the expected impacts on direct and indirect causes, and how multiple interventions may work together to more effectively reduce suicide in these settings (SPRC, n.d.).

The implications of the process described in this paper for suicide prevention services for refugees is multifold. First, despite the disproportionate burden of suicidal behaviors among refugee and displaced populations (Vijayakumar & Jotheeswaran, 2010), there are limited prevention programs designed specifically for these groups and even fewer with robust evidence (Haroz et al. 2020). This process allowed a shared understanding of the pathways to suicide in these communities, and as such, a shared understanding of factors that could be targeted through multiple intervention approaches. Second, we know from literature outside of refugee populations that implementing multi-tiered and cost-effective approaches to prevent suicidal behaviors is critical, as no single intervention is sufficient for reducing suicide in at the community level (SPRC, n.d.). Our systems model also illustrated the need for interventions to consider the broader socio-ecological factors that contribute to suicide risk in this population. A public-health and multi-tiered approach to suicide prevention, whereby universal, selective and indicative prevention interventions are implemented and evaluated based on local data, is warranted to comprehensively address suicide risk with displaced populations. And third, while our qualitative model produced valuable insights, future work should consider quantitative approaches that balance intervention effectiveness with costs and implementation considerations, to help local decision makers in prioritizing prevention efforts.

Systems thinking and associated tools and models have the potential to be highly relevant to planning and implementing mental health programs (Trani, Ballard, Bakhshi, & Hovmand, 2016; Zimmerman et al., 2016). These models have been used in suicide prevention policy efforts (Page et al., 2017), but have never been applied to suicide prevention efforts with refugee and displaced populations. Our model generates initial insights into potentially relevant suicide prevention strategies in the context of being displaced from Myanmar into camp settings in Thailand. This mirrors the current state of the field suggesting “soft systems modelling techniques”, or qualitative approaches, are likely most useful right now, with a need to improve quantitative modeling to more fully unlock their potential to improve public health (Carey et al., 2015).

Limitations

Our time allocated to conducting the local workshops was limited, leaving insufficient time to elicit the relative strength of contributing factors and develop all possible feedback loops. Information on relative strength of associations and other feedback loops would perhaps contribute more insight. While the process includes a broad range of stakeholders and was iterative in nature, there is inherent subjectivity in our model that may introduce biases and limitations. For example, it is possible that some of the causes, consequences and potential solutions were combined into categories that may not reflect all possible groupings, such as family history of suicide, which was grouped as a biological cause, but may reflect more of a spiritual cause and thus may warrant different intervention approaches. We tried to mitigate these occurrences through review of groupings with stakeholders but could not completely eliminate this type of bias. The range of possible interventions was generated across stakeholder groups, but the basis for selecting which interventions were of most interest was based on evidence taken from published literature for each potential intervention. This evidence was not generated in a similar context, making the generalizability of the effects of these interventions to the study setting unclear. It is also important to acknowledge that the workshops did not involve local stakeholders with expertise in suicide, thus the causes, consequences and potential interventions reflected their expertise in the local context rather than a comprehensive insight of suicidal behaviors. Data from other studies, such as the psychological autopsies and in-depth interviews performed by IOM (2017) which reflect the voices of refugees directly affected by suicide, were considered later during our model building process and provide a more valid understanding of suicide in the local context. Finally, causal pathways are rarely well-defined in mental and behavioral health, thus it is challenging to capture a causal model of suicide in this context.

Conclusion

Implementation of a participatory community-based system dynamics modeling approach facilitated a process to involve local stakeholders in breaking down the components of the complex problem of suicidal behaviors and prevention in this setting. Stakeholders contributed to a more in-depth understanding of the factors that contribute to these behaviors and were, with input from an expert team, able to identify evidence-informed interventions likely to be able to be successfully implemented for refugees. The findings underscore the importance of collaboration between researchers and local stakeholders to understand specific contexts of suicidal behaviors in order to identify promising interventions that can be adapted and implemented. Furthermore, the findings show that existing prevention interventions shown to be effective among non-displaced populations could be implemented to address contributing factors to suicidal behaviors and, thus, further studies should be conducted that involve implementation and evaluation of these interventions among displaced and refugee populations.

Public Health Significance Statement:

The causes of suicidal behaviors are multi-factoral and complicated, making prevention challenging particularly in health disparity populations. This paper describes a stakeholder involved process to map out the complexities of suicide risk among displaced persons from Myanmar who are living in Thailand in order to inform local suicide prevention approaches.

Summary for Asian American psychology:

This study advances our methodological approaches to understanding suicide prevention research for Asian American communities.