Introduction
Prolonged or complicated bereavement has long been a controversial category in the field of Psychiatry. A turning point in the recognition of the category has been the inclusion of persisting complicated bereavement disorder (PCBD) as a provisional category in DSM5 (American Psychiatric Association, 2013) and the adoption of prolonged grief disorder in the forthcoming ICD-11 (Horowitz et al., Reference Horowitz, Siegel, Holen, Bonanno, Milbrath and Stinson1997; Bonanno and Kaltman, Reference Bonanno and Kaltman2001; Maercker et al., Reference Maercker, Brewin, Bryant, Cloitre, Van Ommeren, Jones, Humayan, Kagee, Llosa, Rousseau, Somasundaram, Souza, Suzuki, Weissbecker, Wessely, First and Reed2013; Bryant, Reference Bryant2014). These developments make it timely to examine the relationship of complicated bereavement and posttraumatic stress disorder (PTSD), disorders that are both triggered by traumatic losses. We investigate the pattern of overlap of the two constellations by applying latent class analysis (LCA) to data from a sample of West Papuan refugees displaced by mass conflict and persecution to neighboring Papua New Guinea (PNG).
Events such as atrocities, extrajudicial executions, and other forms of politically motivated killings commonly expose refugees to traumatic loss (of family and kin) and personal threat (Silove, Reference Silove1999), experiences that are known to trigger both PTSD and complicated bereavement. Further, amongst refugees from collectivist cultures, disruptions to family and kinship groups, eviction from ancestral lands and displacement to foreign and deprived environments, may accentuate symptoms of identity confusion and alienation that are intrinsic to both PTSD and complicated bereavement (Momartin et al., Reference Momartin, Silove, Manicavasagar and Steel2004a; Dorahy et al., Reference Dorahy, Corry, Shannon, Macsherry, Hamilton, Mcrobert, Elder and Hanna2009; Boelen, Reference Boelen2015; Tay et al., Reference Tay, Rees, Chen, Kareth and Silove2015c). We anticipated, therefore, that amongst West Papuan refugees, identity confusion would be prominent amongst symptoms of a combined pattern of complicated bereavement and PTSD.
The ADAPT (Adaptation and Development After Persecution and Trauma) model (Silove, Reference Silove2013) offers a framework for examining the mental health impact of conditions of insecurity, disrupted interpersonal bonds, and undermining of the sense of identity amongst refugees. The model identifies five support systems eroded by mass conflict and displacement, including safety and security; interpersonal bonds and networks; justice; roles and identities; and existential meaning (Silove, Reference Silove2013). Two of these systems (safety and security and bonds and networks) are recognized as central to mental health in the National Institute of Mental Health Research Domain Criteria (RDoC) framework (Clark et al., Reference Clark, Cuthbert, Lewis-Fernandez, Narrow and Reed2017). Consistent with the RDoC, the ADAPT model attempts to bridge the divide between a descriptive or nomothetic approach to identifying mental health problems (represented, e.g. by the operationalized diagnostic criteria used within the DSM system), and an idiographic framework in which personal development and collective experiences shape and give meaning to subsequent psychological reaction patterns. The ADAPT model, therefore, provides a framework for a transdiagnostic approach in which overlapping symptoms of PTSD and complicated bereavement may represent a meaningful response to concurrent exposure to complex ecological and social experiences such as mortal threat, traumatic loss, and disruptions to identities and roles (Rees et al., Reference Rees, Silove, Verdial, Tam, Savio, Fonseca, Thorpe, Liddell, Zwi, Tay, Brooks and Steel2013; Silove et al., Reference Silove, Liddell, Rees, Chey, Nickerson, Tam, Zwi, Brooks, Sila and Steel2014).
Although several studies have investigated the relationship between PTSD and complicated bereavement in high-income countries (Prigerson et al., Reference Prigerson, Bierhals, Kasl, Reynolds, Shear, Newsom and Jacobs1996; Horowitz et al., Reference Horowitz, Siegel, Holen, Bonanno, Milbrath and Stinson1997; Momartin et al., Reference Momartin, Silove, Manicavasagar and Steel2004b; Boelen and Van Den Bout, Reference Boelen and Van Den Bout2005; Bonanno et al., Reference Bonanno, Neria, Mancini, Coifman, Litz and Insel2007; Simon et al., Reference Simon, Shear, Thompson, Zalta, Perlman, Reynolds, Frank, Melhem and Silowash2007; Fujisawa et al., Reference Fujisawa, Miyashita, Nakajima, Ito, Kato and Kim2010; Shear et al., Reference Shear, Simon, Wall, Zisook, Neimeyer, Duan, Reynolds, Lebowitz, Sung, Ghesquiere, Gorscak, Clayton, Ito, Nakajima, Konishi, Melhem, Meert, Schiff, O'Connor, First, Sareen, Bolton, Skritskaya, Mancini and Keshaviah2011), there is a dearth of parallel research into the topic conducted amongst refugee populations. Although complicated bereavement was largely distinguishable from PTSD in a study amongst Bosnian refugees, the two constellations overlapped in relation to the domain of intrusive memories (Momartin et al., Reference Momartin, Silove, Manicavasagar and Steel2004b). In a sample of Mandaean refugees in Australia, LCA identified a comorbid complicated bereavement and PTSD class (16%) in addition to pure complicated bereavement (16%) and PTSD (25%) classes, respectively (Nickerson et al., Reference Nickerson, Liddell, Maccallum, Steel, Silove and Bryant2014). Although suggestive of the importance of a combined complicated bereavement and PTSD class amongst refugees, the data did not clarify the psychosocial factors that may be specific to that pattern.
The long history of persecution that West Papuan refugees have experienced has exposed many members of the community to traumatic loss, mortal threat, disruptions to identities, and a sense of belonging (Brundige et al., Reference Brundige, King, Vahali, Vladek, Yuan and Allard2004). Since the invasion and annexation of the territory by Indonesia over 50 years ago, the occupying military has committed widespread atrocities, including torture, disappearances, extra-judicial imprisonment, and mass displacement of communities, in an attempt to suppress the low-grade armed resistance war waged by indigenous groups seeking national independence for the territory (Human Rights Watch, 2014). Successive waves of West Papuan refugees have crossed the border into neighboring PNG, the largest concentration resettling in Kiunga, a remote town near the border with the homeland. The community in Kiunga lives in constant fear of incursions by hostile elements from across the nearby border. In addition, they face multiple deprivations relating to geographical isolation, harsh living conditions including food and water shortages arising from floods and drought, and lack of employment opportunities and services. The absence of mental health services in the region means that West Papuan refugees have not been exposed to international concepts of trauma, PTSD or complicated bereavement.
The aims of our study were to identify a subpopulation of refugees with a combined constellation of complicated bereavement and PTSD and to assess whether that pattern is associated with a specific profile of past and ongoing adversities. Using LCA, we tested the following hypotheses: (1) in this cross-cultural setting, it would be possible to identify subpopulations with relatively independent clusters of PTSD and complicated bereavement, respectively, thereby supporting contemporary international classification systems; (2) in addition, there would be a class manifesting combined symptoms of complicated bereavement and PTSD; (3) the latter class would be characterized by prominent symptoms of identity confusion and alienation; and (4) a distinctive pattern of traumatic events, post-migration living difficulties, and background psychosocial disruptions would characterize the complicated bereavement-PTSD class.
Method
Full details of the methodology are provided in online Supplementary Material 1.
Sample
Between March and September 2016, a full household survey was undertaken of all adult West Papuans living in the nine villages where the community is concentrated in Kiunga. We included persons originating from West Papua or born to at least one West Papuan parent. The analytic sample comprised all those with whom we made contact (486 adults), noting that 18.3% of the community were traveling in other parts of PNG during the entire course of the study.
Ethics
Ethical permission for the study was provided by the University of New South Wales Human Research Ethics Committee and the Medical Research Council of PNG Ethics Committee.
Measures
The Refugee Mental Health Assessment Package is a comprehensive tool assessing psychosocial factors and common mental disorders amongst refugees (Tay et al., Reference Tay, Rees, Chen, Kareth, Mohsin and Silove2015b). Details of the qualitative and psychometric steps taken to develop, adapt and test the mental health measure in a study amongst West Papuan refugees have been provided previously (summarized in online Supplementary File 1).
Symptoms of complicated bereavement and PTSD
Details of the qualitative methods used to adapt the symptom measures to the culture are provided in online Supplementary Material 1. These steps included the development and testing of modules to assess complicated bereavement and PTSD symptoms using items consistent with both DSM5 and ICD-11 criteria.
Trauma and stressors
Based on our qualitative inquiries, we identified five items representing traumatic loss, each scored 1, if present. We adapted the Humanitarian Emergency Settings Perceived Needs Scale (Semrau et al., Reference Semrau, Van Ommeren, Blagescu, Griekspoor, Howard, Jordans, Lempp, Marini, Pedersen, Pilotte, Slade and Thornicroft2012) to identify postmigration living difficulties (PMLDs). Based on a four-point ordinal scale (0=not problem at all; 1=a bit of a problem; 2=moderately serious; 3=a serious problem), we generated a mean PMLD score (range: 0.27–4.06). We applied the Adaptive Stress Index (ASI) to assess the five domains of the ADAPT measure. The development and testing of the measure amongst a previous sample of West Papuan refugees in the PNG capital, Port Moresby, are described in online Supplementary Material 1 (Tay et al., Reference Tay, Rees, Chan, Kareth and Silove2015a).
Procedure
Interviews were conducted by a field team in Bahasa Indonesian, English, and Tok Pisin (the pigeon English). The field interviewers were drawn from the West Papuan community and the team was managed by a West Papuan refugee (MK).
The team received 3 weeks’ intensive training from a clinical psychologist followed by 3 months of piloting the interview in the field.
Statistical analysis
We applied LCA to data from the whole sample to identify subpopulations of West Papuans manifesting differential patterns of complicated bereavement and PTSD (Nylund et al., Reference Nylund, Asparouhov and Muthen2007). We applied the Bayesian Information Criterion (BIC), sample size-adjusted Bayesian Information Criterion (SS-BIC), and the Akaike's Information Criterion to assess the fit of serial models (1–4) (McGCutcheon, Reference Mcgcutcheon1987; Collins and Lanza, Reference Collins and Lanza2009). In addition, we applied the Vuong-Lo-Mendell-Rubin (VLMR) and the Lo-Mendell-Rubin (LMR) adjusted likelihood ratio tests to compare successive models (Nylund et al., Reference Nylund, Asparouhov and Muthen2007). We applied the principle of parsimony, the degree of class separation, homogeneity of posterior probabilities within classes, and the interpretability of the classes to assist in deciding on the final model (Collins and Lanza, Reference Collins and Lanza2009). We ranked conditional probabilities according to recommended criteria as high (0.60 or greater), moderate (0.59–0.15) or low (0.14 or lower) (Burstein et al., Reference Burstein, Georgiades, Lamers, Swanson, Cui, He, Avenevoli and Merikangas2012).
We applied multinomial logistic regression to compare the classes with the traumatic event and PMLD counts, and the ASI dimensions. Collinearity required that we tested the three predictors in separate regression models. Analyses were performed in STATA version 13 and Mplus Version 7.
Results
Sociodemographic characteristics
The sample included 486 adults (men 55.9%), the mean age being 35.8 (s.d. = 0.65) years. Half the sample had completed primary school education (46.6%), and 9.5% held post-school degrees or certificates. Over half (56.9%) originated from West Papua, the remainder (43.1%) being offspring of West Papuan parents. The West Papuan born had lived in Kiunga for a mean of 15.6 (s.d. = 0.48) years. The majority (62.6%) of the sample had indeterminate status as displaced persons, 14.6% were PNG citizens, and 2.8% held permissive residency status which conferred the right to remain in PNG with some restrictions. A quarter (n = 115, 23.7%) reported one or more loss (death of a loved one) in the past 12 months and a third (n = 166, 34.2%,) experienced a traumatic event (such as serious injury to self or others, exposure to atrocities, sexual violence), thereby meeting DSM-5 entry criteria for either PCBD or PTSD or both.
Tables 1 and 2 detail exposure to loss and traumatic events as well as PMLDs. The mean score for exposure to PMLDs was 1.82 (s.d. = 0.64).
Table 1. Exposure to conflict-related traumatic events (TEs) amongst West Papuan refugees from Kiunga (n = 486).
Table 2. Exposure to postmigration living difficulties (PMLDs) amongst West Papuan refugees from Kiunga (n = 486)
Latent class analysis
LCA supported a four-class solution (Table 3). Specifically, model fit indicators improved up to and including the four-class model, but only marginal gains were achieved by increasing the number of classes beyond that point. The decision to adopt the four-class solution was supported further by the VLMR and the LMR adjusted likelihood ratio tests which showed no statistical changes (p = 0.10) when progressing from a four to five class model. Given these findings and the interpretability of the classes, we adopted the four-class model.
Table 3. Goodness-of-fit statistics for latent class 1 to class 5 models
LR χ2, Likelihood Ratio Chi Square; AIC, Akaike Information Criterion; BIC, Bayesian information Criterion; SSABIC, sample size adjusted BIC; VLMR, Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR, Lo-Mendell-Rubin adjusted likelihood ratio tests.
The four classes comprised a complicated bereavement class (11%, n = 55), a posttraumatic bereavement class including items of complicated bereavement and PTSD (10%, n = 50), a PTSD class (11%, n = 54), and a low-symptom class (67%, n = 327). Table 4 shows the posterior probabilities for symptoms of complicated bereavement and PTSD in relation to each class.
Table 4. Conditional probabilities for symptoms of complicated bereavement and posttraumatic stress disorder (PTSD): a four-class solution (n = 486)
In the complicated bereavement class (class 1, comprising 11% of the sample), eight complicated bereavement symptoms exhibited high endorsements by over half of the sample (item probabilities >0.60) and the remaining 10 items received moderate endorsement (probabilities>0.59 and <0.15). Items in the high probability range included core bereavement characteristics of yearning, preoccupations with the deceased, and intense feelings of sorrow and emotional pain. In contrast, all symptoms of PTSD in this class fell into the lower level of moderate or low probability range.
In the combined posttraumatic bereavement class (class 2), comprising 10% of the sample, respondents rated most items of complicated bereavement and PTSD in the high probability range. Specifically, for complicated bereavement, probabilities for 15 out of 18 symptoms exceeded 0.60 (the high range) and the remaining three fell between 0.59 and 0.15 (the moderate range); and for PTSD, 16 of 20 symptoms exceeded probabilities of 0.60 and the remaining four fell between 0.59 and 0.15. For complicated bereavement, high probability items included core features of preoccupations and persistent yearnings but also extended to maladaptive appraisals, bitterness/anger, interpersonal problems and emotional numbness. The high probability items for PTSD in this class included flashbacks, intrusive thoughts, internal and external avoidance, and anhedonia. Symptoms falling into the moderate range included, for complicated bereavement, feelings of confusion, emptiness, and a diminished sense of identity (items that had been added by community members in preceding focus group discussions); and for PTSD, concentration difficulties, insomnia, startle response and hypervigilance.
In the PTSD class (class 3, 11%), seven items fell into the high range (>0.6) including intrusive memories, psychological reactivity, avoidance, anhedonia, detachment, irritability, and posttraumatic amnesia, and the majority (13 out of 20 symptoms) fell into the moderate probability range of >0.59 and <0.15. In this class, virtually all complicated bereavement symptoms fell into the low probability range.
In the low symptom class (class 4), comprising two thirds (67%) of the sample, most symptoms of both complicated bereavement and PTSD yielded low probabilities.
Multinomial logistic regressions
Table 5 presents the findings of the three multinomial logistic regression models. The four-class LCA structure represented the dependent variable and individual predictors were the traumatic loss event count, the PMLD count, and the five ASI domains, in each instance controlling for sociodemographic variables. The comparison class in all analyses was the low symptom reference class.
Table 5. Multinomial logistic regressions examining predictors of latent class memberships (n = 486)
*p < 0.05; **p < 0.01.
No statistically significant effects were evident for socio-demographic factors in any of the three regression analyses. The only significant predictor for the complicated bereavement class (class 1) was the ASI domain of disrupted interpersonal bonds and networks (OR 2.16, 95% CI 1.09–4.28). In contrast, the combined posttraumatic bereavement class (class 2) was associated with higher exposure to traumatic loss events (OR 2.43, 95% CI 1.11–5.34), PMLDs (OR 2.24, 95% CI 1.01–4.6) and greater disruptions in the ASI domains of interpersonal bonds and networks (OR 3.3, 95% CI 1.47–7.38) and roles and identities (OR 2.18, 95% CI 1.11–4.27). The PTSD class reported higher exposure to traumatic loss events (OR 2.17, 95% CI 1–4.77) and was unique in its association with the ASI pillar of safety and security (OR 2.12, 95% CI 1.16–3.89). In common with the posttraumatic bereavement class, the PTSD class reported greater disruptions in the ASI domain of roles and identity (OR 2.21, 95% CI 1.16–4.22).
Discussion
LCA yielded a four-class solution comprising three morbid classes and a low symptom class, the latter accounting for over two-thirds of the sample. The three morbid classes, each representing roughly one-tenth of the sample, comprised a complicated bereavement class, a combined posttraumatic bereavement class, and a PTSD class. Notably, the posttraumatic bereavement class was characterized by a distinctive combination of symptoms related to identity confusion and alienation.
Members of the posttraumatic bereavement class reported high levels of adversity reflected in traumatic losses, post-migration living difficulties and disruptions to two of the ADAPT psychosocial support domains of interpersonal bonds and networks and identity and roles. In support of conventional international classification systems, separate classes emerged for complicated bereavement and PTSD, respectively. In keeping with theory, the ASI domain of disruptions to bonds and networks was associated with the complicated bereavement class, whereas erosion of safety and security was related uniquely to the PTSD class. The PTSD class also reported disruptions in roles and identities and exposure to high levels of traumatic loss.
Prior to discussing our findings, we consider the strengths and limitations of the study. We applied a rigorous sampling approach, identifying West Papuans according to the triangulation of census data and a whole-of-household survey. We achieved a high response rate (85.8%) even when absentees from the catchment were accounted for. Although Kiunga is the largest settlement of West Papuans in PNG, restriction of sampling to one location limits the generalizability of our findings. The study is cross-sectional, cautioning against drawing causal inferences from the analyses undertaken. Memory biases could result in either over- or under-reporting of past traumatic events. In addition, exposure to trauma may increase the tendency to report mental health symptoms. There is also a risk of transcultural error in measurement when undertaking studies of this type. To limit the risk, we followed a systematic mixed methods approach to develop and adapt our measures to the culture and context. Our qualitative inquiries suggested that the key mental health constructs of PTSD and complicated bereavement were well recognized by West Papuan refugees even though the population lacked familiarity with international terminology for classifying these reactions. Specifically, West Papuans recognized all the ICD and DSM symptoms of complicated bereavement, applying a specific term (‘Duka Cita’) in Bahasa Indonesian to describe the syndrome in their culture. To avoid excessive complexity, we did not include other diagnostic categories such as depression which could be salient in generating symptoms of identity confusion and alienation. Because of restrictions in the sizes of the respective latent classes, we could only compare each morbid class with the no/low symptom class in our logistic regression analyses. In addition, in applying multinomial regression analysis, collinearity amongst our three predictors required that we assess the impact of traumatic loss events, PMLDs, and the ASI domains separately.
Caveats notwithstanding, our findings offer support for our key hypotheses. Conventional classes of complicated bereavement and PTSD emerged from the LCA. The ASI domain of disrupted bonds and attachments was associated with the complicated bereavement class as would be anticipated. Notably, the PTSD class reported high levels of exposure to traumatic losses, consistent with the criteria of international classification systems (see DSM5). In addition, the PTSD class reported greater erosion of the identity and role domain of the ADAPT model, supporting the importance of identity disturbance as an integral component of the PTSD reaction (Robinaugh and McNally, Reference Robinaugh and Mcnally2011; Dorahy et al., Reference Dorahy, Middleton, Seager, Mcgurrin, Williams and Chambers2015).
The key finding was that 10% of the sample exhibited the combined posttraumatic bereavement pattern. Importantly, the characteristics that were accentuated in this class extended beyond core PTSD and complicated bereavement symptoms to those that fall broadly under the domain of identity confusion and alienation, including maladaptive appraisals (mistrust of others, self-doubt), bitterness/anger, interpersonal problems, emotional numbness, feelings of confusion and emptiness, and a diminished sense of identity. It is notable that most of these symptoms were identified spontaneously by West Papuans participating in the preliminary focus groups.
There are compelling reasons why West Papuans would highlight symptoms of identity confusion and alienation. The society is grounded in an agrarian, collectivist culture in which personal identity is strongly bound up with ties to the family, kinship group, and traditional lands. The longstanding campaign of repression pursued by the occupying military has led to major disruptions to this traditional way of life. Mass immigration into West Papua from other parts of Indonesia has added to the sense of threat to the national identity of the indigenous people (Zocca, Reference Zocca1995). In addition, the group participating in our study comprised a population displaced to a remote town across the border where they experienced ongoing deprivations, fear, and a sense of marginalization. Overall, the history of persecution appeared to be mirrored by the pattern of symptoms of grief, fear, and alienation represented by the symptoms of the posttraumatic bereavement class. It also is notable that there is a correspondence between the identified symptoms of identity confusion and alienation within the posttraumatic bereavement class and elements of complex traumatic stress disorder described amongst survivors of human rights violations and abuses in the western literature (Herman, Reference Herman1992; Friedman et al., Reference Friedman, Resick, Bryant and Brewin2011; Ford and Courtois, Reference Ford and Courtois2014; Palic and Elklit, Reference Palic and Elklit2014). Our findings build to efforts to bridge the gap between idiographic and nomothetic models of understanding of traumatic stress by demonstrating the connection between ecological factors, in this instance, politically motivated psychosocial disruptions, trauma and ongoing stressors, and a distinctive pattern of combined PTSD and complicated bereavement observed at the individual level. This approach helps to counter the tendency to regard mental health symptoms as solely serving the function of signifying the presence of pre-defined categories of mental disorder (Akram and Giordano, Reference Akram and Giordano2017; Clark et al., Reference Clark, Cuthbert, Lewis-Fernandez, Narrow and Reed2017).
Our findings have important clinical implications. Although there is some evidence in support of the efficacy of cognitive behavioral interventions in treating complicated bereavement in non-refugee populations (Rosner et al., Reference Rosner, Pfoh, Kotoucova and Hagl2014; Bryant et al., Reference Bryant, Kenny, Joscelyne, Rawson, Maccallum, Cahill, Hopwood and Nickerson2017), no culturally adapted interventions of this type have been systematically trialled amongst refugees (Murray et al., Reference Murray, Cohen, Ellis and Mannarino2008; Bass et al., Reference Bass, Murray, Mohammed, Bunn, Gorman, Ahmed, Murray and Bolton2016). Given that 20% of participants had heightened symptoms of complicated grief (if both relevant classes are counted), there is a pressing need to develop and test appropriate therapies for this reaction pattern amongst refugees. In addition, our findings challenge the prevailing tendency to focus on single diagnoses (usually PTSD) in the refugee field by demonstrating the relevance of comorbid presentations (Silove et al., Reference Silove, Ventevogel and Rees2017) which may require more comprehensive psychological interventions (Tay and Silove, Reference Tay and Silove2016). Specifically, in addition to the focus on issues of safety and traumatic loss, refugees with the posttraumatic bereavement pattern may require special attention to issues of identity confusion and alienation. Failure to focus on the latter issue may account in part for low levels of success reported in some studies treating socially isolated and alienated refugees with complex symptom presentations (Carlsson et al., Reference Carlsson, Mortensen and Kastrup2005; Buhmann et al., Reference Buhmann, Nordentoft, Ekstroem, Carlsson and Mortensen2016). There may be value, therefore, in formulating and trialing a specific approach to psychological intervention that focuses on elements of both complicated bereavement and PTSD for refugees that show the combined pattern.
Conclusions
We identified a combined class comprising complicated bereavement and PTSD amongst refugees, in which symptoms of identity confusion were prominent. That class reported a distinctive profile of exposure to past trauma, postmigration stressors, and psychosocial disruptions. More generally, the study illustrates the value of investigating connections between comorbid patterns of mental health symptoms and distinctive profiles of trauma, postmigration stressors and longer-term psychosocial disruptions that refugees have experienced. Defining these relationships may assist in shaping and refining psychotherapeutic interventions to meet the specific needs of subpopulations of refugees.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291718002027.
Acknowledgements
We thank all our field personnel for their contributions to this project.
Conflict of interest
None.
Funding source
Funding source: National Health and Medical Research Council Program Grant. National Health and Medical Research Council Early Career Fellowship.
