Profiles of posttraumatic stress disorder and negative world assumptions in treatment-seeking refugees

ABSTRACT

Background: Refugees often suffer from trauma-related psychopathology, specifically posttraumatic stress disorder (PTSD). Negative world assumptions are strongly correlated with the development, course, and severity of PTSD.

Objective: This study aimed to investigate whether there are distinct profiles of PTSD and negative world assumptions (NWA) and examine whether trauma load, torture, and gender differentially predict such symptom profiles.

Method: In a sample of 225 treatment-seeking refugees who had resettled in the Netherlands, latent profile analysis was used to identify subgroups of patients sharing the same profile of PTSD and NWA symptoms. Predictors of profile membership were analyzed via multinomial logistic regression.

Results: A three-profile solution yielded the best model fit: a low PTSD/low NWA profile (23.6%), a high PTSD/high NWA profile (41.8%), and a high PTSD/low NWA profile (34.7%). Participants who reported a higher trauma load, were more likely to be part of the high PTSD/high NWA profile or the high PTSD/low NWA profile in comparison to low PTSD/low NWA profile. Participants who reported having experienced torture were more likely to be part of the high PTSD/high NWA profile in comparison to low PTSD/low NWA profile. Gender did not differentiate between the profiles.

Conclusions: This study reveals that among treatment-seeking refugees resettled in the Netherlands, there are distinct profiles of PTSD and NWA. These profiles indicate that PTSD and NWA are not uniformly experienced among refugees, emphasizing the diversity in their psychological responses to trauma. Among individuals experiencing severe PTSD symptoms, a subgroup was identified of individuals who additionally exhibited negative assumptions about themselves, others, and the world. Recognizing this heterogeneity is crucial in both research and clinical practice, particularly in the context of refugee mental health. Directions for future research are discussed.

HIGHLIGHTS

• Three profiles of PTSD and negative world assumptions were identified in a group of treatment-seeking refugees.

• Directions for future research and the importance of recognizing heterogeneity in psychological responses to traumatic experiences in refugees are discussed.

Antecedentes: Los refugiados suelen sufrir psicopatología relacionada con el trauma, específicamente trastorno de estrés postraumático (TEPT). Los supuestos negativos sobre el mundo están fuertemente correlacionados con el desarrollo, el curso y la gravedad del trastorno de estrés postraumático.

Objetivo: Este estudio tuvo como objetivo investigar si existen perfiles distintos de TEPT y supuestos negativos sobre el mundo (NWA en su sigla en inglés) y examinar si la carga traumática, la tortura y el género predicen de manera diferencial dichos perfiles de síntomas.

Método: En una muestra de 225 refugiados que buscaban tratamiento y que se habían reasentado en los Países Bajos, se utilizó el análisis del perfil latente para identificar subgrupos de pacientes que compartían el mismo perfil de síntomas de TEPT y NWA. Los predictores de pertenencia al perfil se analizaron mediante regresión logística multinomial.

Resultados: Una solución de tres perfiles produjo el mejor ajuste del modelo: un perfil de TEPT bajo/NWA bajo (23.6%), un perfil de TEPT alto/NWA alto (41.8%) y un perfil de TEPT alto/NWA bajo (34.7%). La carga traumática distinguió entre el perfil de TEPT bajo/NWA bajo y los perfiles de TEPT alto/NWA alto y de TEPT alto/NWA bajo, es decir, los participantes que informaron una mayor carga traumática tenían más probabilidades de ser parte del perfil de TEPT alto/NWA alto o del perfil TEPT alto /NWA bajo en comparación con el perfil de TEPT bajo/NWA bajo. La tortura distinguía entre el perfil bajo de TEPT/NWA bajo y el perfil TEPT alto/ NWA alto, es decir, los participantes que informaron haber experimentado tortura tenían más probabilidades de ser parte del perfil de TEPT alto/NWA alto en comparación con el perfil de TEPT bajo/NWA bajo. El género no diferenció entre los perfiles.

Conclusiones: Este estudio revela que en los refugiados reasentados en los Países Bajos que buscan tratamiento, existen perfiles distintos de TEPT y NWA. Estos perfiles indican que el TEPT y el NWA no se experimentan de manera uniforme entre los refugiados, lo que enfatiza la diversidad en sus respuestas psicológicas al trauma. Entre las personas que experimentaban síntomas graves de trastorno de estrés postraumático, se identificó un subgrupo de personas que además mostraban supuestos negativos sobre sí mismos, los demás y el mundo. Reconocer esta heterogeneidad es crucial tanto en la investigación como en la práctica clínica, particularmente en el contexto de la salud mental de los refugiados. Se discuten direcciones para futuras investigaciones.

KEYWORDS:

• Refugees
• posttraumatic stress disorder
• world assumptions
• appraisals
• cognition
• heterogeneity
• torture
• latent profile analysis

Palabras Clave:

• Refugiados
• trastorno de estrés Postraumático
• supuestos sobre el mundo
• evaluaciones
• cognición
• heterogeneidad
• tortura
• análisis de perfil latente

1. Background

Refugees who were forced to flee from war, violence, and persecution often suffer from trauma-related psychopathology. Forms of psychopathology that are common include posttraumatic stress disorder (PTSD), depression, anxiety disorders, and psychosis (e.g. Blackmore et al., 2020; Fazel et al., 2005; Porter & Haslam, 2005; Steel et al., 2009). Fazel et al. (2005) found that refugees who resettled in the Global North are approximately ten times more likely to develop PTSD in comparison to the general population in those countries. A recent systematic review and meta-analysis by Blackmore et al. (2020) found that the increased prevalence of PTSD and depression persists for many years after resettlement. The number of traumatic experiences (i.e. trauma load), the experience of torture, and female gender are among some of the strongest predictors and risk factors for developing PTSD (Porter & Haslam, 2005; Steel et al., 2009).

In the past 30 years, a growing body of research has suggested the important role of basic beliefs about the self, others, and the world, shattered assumptions (Janoff-Bulman, 1989, 1992), negative appraisals in PTSD (Ehlers & Clark, 2000), or posttraumatic cognitions (Foa et al., 1999). Negative appraisals and posttraumatic cognitions in PTSD generally refer to the cognitive changes and dysfunctional cognitions that arise as a result of traumatic experiences. Posttraumatic cognitions are strongly correlated to PTSD symptoms in general (Daie-Gabai et al., 2011; Dunmore et al., 1999; Ehlers et al., 1998; Ehlers & Clark, 2000; McCann et al., 1988; Van Emmerik et al., 2006; Wilker et al., 2017) and in refugees (Schnyder et al., 2015). Posttraumatic cognitions have been found to increase vulnerability for developing PTSD and to predict chronicity and severity of PTSD (Dunmore et al., 1999; Ehlers et al., 1998; Ehlers & Clark, 2000; Ehlers et al., 2000). In addition, studies found that posttraumatic cognitions tend to worsen over time and make recovery exceedingly challenging (Brewin & Holmes, 2003; Dekel et al., 2013; Iverson et al., 2015). In a study on posttraumatic cognitions and PTSD in a Palestinian refugee sample, Van Heemstra et al. (2020) found a significantly higher prevalence of posttraumatic cognitions when compared to a non-refugee reference sample. Additionally, they found that posttraumatic cognitions explained significant variance in PTSD symptoms.

The ‘shattered assumptions’ theory, by Janoff-Bulman (Janoff-Bulman, 1992), elaborated on the role of cognitive appraisals and core beliefs, specifically in relation to traumatization. The theory states that individuals ideally develop positive world assumptions in (early) childhood. These assumptions are fairly resilient to everyday challenges and minor negative experiences. They include beliefs in the benevolence and fairness of the world, the meaningfulness of events, the worthiness and competence of the self, and predictability and controllability of the world and the future. However, when those assumptions are shattered by a traumatic event, this can lead to a crisis in meaning and a re-evaluation of one’s beliefs and values. For example, someone who views the world as benevolent and fair and others as trustworthy, may experience a drastic change in those core beliefs after years of war, indiscriminate bombing of their neighbourhood or having been imprisoned and tortured. Refugees have typically faced prolonged interpersonal trauma, such as war, persecution, murder of loved ones, human rights violations, sexual violence, or torture. Such repeated human-instigated events can be especially threatening to an individual’s beliefs about themselves, others, and the world, and lead to feelings of permanent negative change of one’s personality and perspective of the future (Ehlers et al., 2000). As such, negative world assumptions (NWA) play an important role in refugee mental health.

Although the core symptoms of PTSD are flashbacks or nightmares, hypervigilance, and avoidance, changes in beliefs about the self, others and the world after traumatic experiences are common (e.g. Brewin & Holmes, 2003). Hence negative alterations in cognitions were included in the criteria of PTSD in DSM-5 (American Psychiatric Association [APA], 2013) In the International Classification of Diseases (ICD-11), the classification of Complex PTSD was added, which includes the criterion of negative self-concept, as well as affective dysregulation and disturbed relationships (Karatzias, Cloitre, et al., 2017; Karatzias, Shevlin, et al., 2017). The DSM-5 criteria for PTSD now include the symptoms of ‘persistent and exaggerated negative beliefs or expectations about oneself, others, or the world’, and ‘persistent distorted cognitions about the cause or the consequences of the traumatic event(s) that lead the individual to blame himself/herself or others’. These criteria reflect cognitions that directly link to traumatic experiences, but may not, however, grasp generalized or core assumptions about themselves, others and the world that have been fundamentally changed due to (prolonged) traumatization.

The diagnosis of PTSD can encompass a variety of symptom combinations. Multiple studies have explored heterogeneity in PTSD (e.g. Bonanno & Mancini, 2012; Elklit et al., 2014; Karatzias, Shevlin, et al., 2017; Minihan et al., 2018). Gaining insight into this heterogeneity is vital for understanding and treating mental disorders such as PTSD (Crumlish & O'Rourke, 2010; Palic & Elklit, 2011). However, PTSD symptom heterogeneity in refugees remains relatively unexplored. A study by Minihan et al. (2018) used latent class analysis in a sample of traumatized refugees residing in Australia and identified four distinct PTSD classes: a class characterized specifically by high re-experiencing and avoidance symptoms and classes characterized by severe, moderate, and low PTSD symptoms. Additionally, they found that a higher trauma load predicted membership of the more severe subgroup. A study by Jongedijk et al. (2020) used LPA in a sample of traumatized refugees in the Netherlands to examine subgroups based on PTSD, anxiety and depression symptoms, and somatic complaints. They identified three distinct subgroups reflecting moderate, severe, and highly severe symptomatology. They found that symptom severity of all psychopathology dimensions was distributed equally between the subgroups and concluded that no qualitative differences could be identified in the symptom distribution. Nickerson et al. (2014) explored PTSD and prolonged grief disorder (PGD) symptom patterns in a group of refugees resettled in Australia. They found four distinct subgroups: a combined PTSD/PGD group, a predominantly PTSD group, a predominantly PGD group, and a resilient group. Nonetheless, studies using LPA in samples of traumatized refugees are scarce, and, to date, no study has explored potential subgroups based on the endorsement of PTSD and NWA in refugees.

Using latent profile analysis (LPA), we aimed to identify potential subgroups among treatment-seeking refugees residing in the Netherlands based on PTSD symptoms and NWA. LPA is a statistical technique that allows for the exploration of heterogeneity by grouping individuals into homogeneous subgroups based on shared symptom patterns (latent profiles). We hypothesized the existence of profiles characterized by both severity and quality of symptoms. Specifically, we expected to identify profiles that differ in severity of both PTSD and NWA symptoms, as well as different patterns of PTSD and NWA symptoms. Secondly, we aimed to examine whether trauma load, exposure to torture, and gender predict subgroup membership. We hypothesized that trauma load, torture, and female gender differentiate between profiles with higher severity of both PTSD and NWA compared to profiles with less severe symptoms and NWA. Lastly, we performed an exploratory analysis of several types of traumatic experiences as correlates of class membership.

2. Method

2.1. Participants and procedure

Data were collected at ARQ Centrum’45, a Dutch centre that specializes in the treatment of the psychosocial consequences of war, persecution, and violence against refugees, asylum-seekers, and profession-related traumatic events. Participants were trauma-exposed, treatment-seeking refugees referred for specialized psychological and psychiatric treatment. Participants are generally only admitted if they have received previous treatment elsewhere, usually without symptom reduction. As such, participants in this sample commonly experience long-lasting mental health complaints. Data were collected between July 2002 and October 2011. Questionnaires were primarily administered for diagnostic purposes before the start of treatment as part of a routine diagnostic assessment procedure that all refugees who applied for treatment were asked to complete. These data were subsequently archived anonymously for scientific research purposes. Participants were informed that these data were stored anonymously and were given the opportunity to have their data removed from the database. Upon consultation, the Medical Ethics Committee of the University of Leiden stated that no review of the ethical merits of this study or informed consent from patients was needed as the assessment was primarily administered for clinical purposes, and secondarily for research purposes. All patients included in the sample had temporary or permanent refugee status or Dutch nationality, and sufficient language proficiency to complete a diagnostic assessment in Dutch.

2.2. Measures

Harvard Trauma Questionnaire (HTQ). The HTQ is a self-report questionnaire designed to assess traumatic experiences and PTSD symptoms in refugees (Mollica et al., 1992). In the first part of the HTQ, participants are asked which of 20 possible types of traumatic events they experienced, witnessed, or heard of. For each participant, a total score was computed by counting the number of self-experienced traumatic experiences (range 0-20). The second part of the HTQ assesses PTSD according to the DSM-IV (16 items). Symptoms are rated on a 4-point scale ranging from 1 (not at all) to 4 (extremely). Symptom severity with regard to overall PTSD and the symptom dimensions re-experiencing (4 items), avoidance (7 items), and arousal (5 items) were computed by averaging the corresponding item scores. A higher score on the PTSD clusters indicates a higher severity of the symptoms. A mean score of 2.5 on the overall PTSD scale was used as a cut-off score for a probable DSM-IV-based PTSD diagnosis. The HTQ has been shown to have good psychometric properties in studies with refugees (Hollifield et al., 2002). In addition, it has been shown that the PTSD construct as assessed with the HTQ is interpreted in a similar way by refugee and Dutch patients, although refugee patients reported more severe symptoms compared to Dutch patients (Wind et al., 2017).

The number of traumatic experiences was estimated by summing the different types of traumatic events a patient reported having experienced in the first part of the HTQ. For the purpose of sparsity and specificity, we only included those types of traumatic events that were reported as ‘experienced myself’, and we excluded the types of traumatic events that were reported as witnessed or heard of. Whether a patient had been tortured was estimated by combining the items of imprisonment and having experienced torture from the HTQ, again for the purpose of specificity. Only if a patient endorsed both items, it was included as such in the analysis. Our objective for this approach was to prevent false positives and exclusively include individuals who had been subjected to torture as per the parameters defined by the Istanbul Protocol (United Nations. Office of the High Commissioner for Human Rights, 2004). In this study, reliability of the subscales was good, with Cronbach’s alphas ranging from .78 (Avoidance) to .84 (Hyperarousal).

World Assumptions Scale (WAS). The WAS (Janoff-Bulman, 1989) is a self-report questionnaire that intends to measure an individual’s assumptive world. It assesses assumptions with regard to the benevolence of the world, the meaningfulness of events, and the worthiness of the self. The WAS consists of 32 items that are rated on a 6-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). The items form eight four-item subscales: Benevolence of the (impersonal) World, Benevolence of People, Randomness, Justice, Controllability, Self-Worth, Self-Controllability, and Luck (see Table S2 in Supplementary Materials for all subscales and corresponding items). A higher endorsement of subscale items reflects a stronger belief in positive world assumptions, with the exception of items in the Randomness subscale, which are negatively worded and are inversely scored. In this study, the WAS was used as an index of NWA. As such, low WAS scores represent high NWA, and high WAS scores represent low NWA. The WAS was found to have adequate psychometric properties for use in clinical and research settings when used with a trauma-exposed sample (Elklit et al., 2007; van Bruggen et al., 2018). In a Turkish sample, the WAS was found to have acceptable internal consistency but poor test-retest reliability (Yılmaz, 2008). The psychometric properties of this (refugee) sample are described in van Bruggen et al. (2018), in which a comparison is made with a Dutch reference group and a group of Dutch healthcare workers. The eight subscales were found to be a consistent model, and nearly all of the factors had adequate reliability, with Cronbach's alpha ranging from .68 (Randomness) to .84 (Luck). In this study, the reliability of the WAS subscales was reasonable, with Cronbach’s alphas ranging from .65 (Self-Worth) to .77 (Self-Controllability).

2.3. Data analyses

First, in Mplus version 8 (Muthén & Muthén, 2017), LPA was conducted to group individuals into homogeneous latent profiles of PTSD symptom clusters and world assumption clusters. To simplify the interpretation of the LPA results, scores on all measures were standardized. The robust maximum likelihood (MLR) estimator was used, which is robust to non-normality of observations. Missing data were accounted for by employing full information maximum likelihood estimation. This means that all available data are used in the analysis. LPA parameter estimates are based on variances in and covariances between the observed indicator variables. For the estimation of the variances and covariances between 79% and 93% of the data (M = 87.14%) were available. Complete data were available for 67% of the participants. Missing data patterns showed that the most frequent missing data pattern is participants with no data on the HTQ (7%). Other missing data patterns were smaller in frequency (≤ 3%) and seemed to be due to random missing items, of which no meaningful pattern could be discerned. Responders and non-responders were compared with regard to age, gender, HTQ symptom cluster scores, or WAS cluster scores. No significant differences appeared between responders and non-responders. To avoid local likelihood maxima, 1000 random sets of starting values in the first step and 100 in the second step of optimization were requested, and 50 initial stage iterations were used. A series of models with increasing numbers of latent subgroups were estimated until no acceptable model fit or substantive meaning was achieved (DiStefano & Kamphaus, 2006; Masyn, 2013).

The optimal number of latent profiles was deduced by fitting the most parsimonious model to the data (one profile), followed by consecutive models with increasing numbers of profiles. The models were then compared by using the Bootstrapped likelihood Ration Test (BLRT), Lo-Mendell-Rubin adjusted likelihood ration test (LMRA), and the Bayesian Information Criterion (BIC). For the BLRT and LMRA, a significant p-value indicates that the estimated model fits the data better than the model with one subgroup less (Nylund et al., 2007). Regarding BLRT, 500 bootstrap samples were requested, with 50 sets of starting values in the first step and 20 in the second step of optimization to avoid local likelihood maxima in each bootstrap sample. A lower value of BIC indicates a better fit of the model to the data (Van de Schoot et al., 2012). To evaluate the classification quality, the entropy statistic was used in combination with the average assignment probabilities. Classification is considered adequate when entropy values are > 0.80 (Celeux & Soromenho, 1996).

Secondly, correlates of subgroup membership were analysed by regressing the latent profiles in the optimal solution on three predictor variables: total number of traumatic events, exposure to torture, and gender. This was done by conducting two multinomial regression models using the three-step procedure in Mplus (Asparouhov & Muthén, 2014). Because data on the predictor variables total number of traumatic events and exposure to torture were available for subsamples of different compositions (total number of traumatic events: n = 159; exposure to torture: n = 198; gender: n = 225) and Mplus handles missing values in the predictor variables with listwise deletion in this context, separate multinomial regression models were estimated for total number of traumatic events and exposure to torture in combination with gender. Finally, we explored the prevalence and distribution of all HTQ events within the subgroups of the optimal profile solution.

3. Results

3.1. Participant characteristics

Data of 225 refugees were used in the study. Demographic variables of the participants in the sample are summarized in Table 1. The sample included participants originating from over 25 different countries (see Table S1 in the Supplementary Material). The majority of the participants were men, predominantly originating from the Middle East and Northern Africa (e.g. Iran, Iraq, Turkey, Afghanistan, Somalia), followed by South-eastern Europe (e.g. Bosnia–Herzegovina or Croatia). Descriptive statistics of the symptom clusters and predictor variables are shown in Table 2. A large proportion of the sample reported clinical levels of PTSD. Participants had experienced approximately ten types of traumatic events on average. A third of the participants reported having been subjected to torture.

Table 1. Sociodemographics.

Characteristics	N	%	M	SD
Gender
Male	157	69.8
Female	68	30.2
Age			42.87	8.75
Region of origin*
Middle East and North Africa	118	52.4
Southeast Europe	72	32.0
Sub-Saharan Africa	11	4.9
Other	16	7.1
Time in the Netherlands (years)			13.3	4.9

* See Supplementary Material Table 1 for full list of countries of origin.

Table 2. Descriptive statistics of symptom clusters and predictor variables.

Characteristics	N	M	SD	Range	%
PTSD: re-experiencing	202	3.22	0.72	1–4
PTSD: avoidance	201	2.88	0.63	1–4
PTSD: arousal	200	3.16	0.65	1–4
PTSD: overall	201	3.06	0.57	1–4	73.3% in clinical range
WAS: Benevolence of the world	213	12.10	5.10	4–24
WAS: Benevolence of people	204	13.03	4.62	4–24
WAS: Justice	209	12.23	5.08	4–24
WAS: Self-worth	198	15.78	4.80	4–24
WAS: Randomness	210	16.29	4.95	4–24
WAS: Self-controllability	212	17.42	4.47	4–24
WAS: Luck	210	10.92	5.01	4–24
WAS: Controllability	209	13.04	4.68	4–24
Total number of traumatic event types experienced	159	10.14	5.47	0–20
Torture	83				36.9% of sample

Note. PTSD = posttraumatic stress disorder; WAS = World Assumption Scale.

3.2. Latent Profile Analysis

Fit indices and classification quality for one- to six-profile models are presented in Table 3. All models showed significant p-values of the BLRT. The LMRA yielded a non-significant p-value for all profile solutions. Log-likelihood values increased substantially from one- to two- and then to three-profile solutions before flattening out, indicating diminishing gain in log-likelihood (Figure 1). BIC values decreased from one- to two- and then to three-profile solutions, before increasing for every extra profile (Figure 2). As such, according to the log-likelihood and BIC, the model with three latent profiles is the most parsimonious model with acceptable model fit. This is also in line with LMRA. Entropy remained relatively similar over the various models, with values > .77. In the two-profile solution, there was a clear distinction between a severe profile with high endorsement of almost all symptom clusters and a moderate profile with relatively low scores on all symptom clusters. The three-profile solution, however, showed a clear distinction between three profiles in terms of both severity and quality of the endorsed symptom clusters. The four-profile solution led to a more complex model with profiles that were more complicated to meaningfully define. The three-profile solution was preferred over the four-profile solution because in the four-profile solution, the third and fourth profiles were very similar to the third profile in the three-profile solution (for graphs of the two- and four-profile solutions, please see Supplementary Figures S1, S2). Overall, the three-profile solution appeared to be the most meaningful. Based on the model fit indices, parsimony, classification quality, and theoretical meaning, and interpretability, the three-profile solution was selected as the optimal solution.

Figure 1. Gain in log-likelihood across LPA models with increasing number of profiles.

A graph illustrating the log-likelihood gain across LPA models. It shows log-likelihood values increasing from profile 1 to 6, with a steeper slope up to profile 3, followed by a gradual increase to profile 6.

Figure 2. Gain in BIC across LPA models with increasing number of profiles.

A graph illustrating gain in BIC across LPA models with an increasing number of profiles. It shows BIC values decreasing steeply from profile 1 to profile 3, and then increasing slightly to profile 6.

Table 3. Goodness-of-Fit Indices for Profile Solutions.

	Entropy	BIC	Log-likelihood	BLRT −2LL difference		LMR-A Value	p-value
1 profile	–	6544.423	−3212.634	–	–	–	–
2 profile	0.790	6311.069	−3063.461	298.347	0.0000	293.826	.4109
3 profile	0.772	6211.294	−2981.077	164.767	0.0000	162.271	.0724
4 profile	0.777	6214.856	−2950.361	61.431	0.0000	60.500	.2427
5 profile	0.781	6228.190	−2924.532	51.659	0.0000	50.877	.8670
6 profile	0.793	6246.241	−2901.060	46.943	0.0000	46.231	.2125

Note. Most meaningful model is printed in bold. BIC = Bayesian information criterion; BLRT = Parametric bootstrapped likelihood ration test; −2LL difference = −2 times log-likelihood difference between a N class solution and N-1 class solution; LMRA-LRT = Lo-Mendell-Rubin adjusted likelihood ratio test.

Figure 3 shows the standardized mean symptom scores of PTSD and WAS symptom clusters in the three-profile solution. As the WAS was used as an index of NWA, low WAS scores represent high NWA, and high WAS scores represent low NWA. The first profile (n = 53; 23.6%) was characterized by low scores on PTSD and above-average scores on WAS and was named low PTSD/low NWA. Individuals in this profile reported the lowest PTSD symptoms in comparison to participants in the other profiles, and they reported above average WAS scores, with the exception of Randomness and Self-controllability. The second profile (n = 94; 41.8%) was characterized by high PTSD scores and low WAS scores and was named high PTSD/high NWA. The third profile (n = 78; 34.7%) was characterized by high PTSD scores and above average WAS scores, with the exception of Self-worth, and was named high PTSD/low NWA.

Figure 3. Standardized mean scores of WAS and PTSD symptom cluster endorsement for the three-profile solution. Lower PTSD scores indicate lower symptom cluster endorsement and higher PTSD scores indicate higher PTSD symptom cluster endorsement. Lower WAS cluster scores indicate higher NWA and higher WAS cluster scores indicate lower NWA.

A line graph illustrating standardized mean scores of WAS and PTSD symptom cluster endorsement for the three different profiles in the three-profile solution.

3.3. Multinomial Logistic Regression

Table 4 presents the descriptive statistics of the predictor variables for each of the three profiles. In the first MNR torture and gender were regressed on the latent profile variables. This analysis was based on a subsample of 88% of the total sample (n = 198). In the second MNR total number of traumatic events and gender were regressed on the latent profile variables. This analysis was based on a subsample of 71% of the total sample (n = 159). Results of the two multinomial logistic regression analyses can be found in Table 5. Trauma load differentiated significantly between the low PTSD/low NWA profile and the high PTSD/high NWA profile and between the low PTSD/low NWA and the high PTSD/low NWA profile. Participants who reported a higher number of traumatic event types, were more likely to be part of the high PTSD/high NWA profile or the high PTSD/low NWA profile in comparison to low PTSD/low NWA profile. Trauma load did not differentiate between the two high PTSD profiles. Torture differentiated significantly between the low PTSD/low NWA and high PTSD/high NWA profiles. Participants who reported having experienced torture were more likely to be part of the high PTSD/high NWA profile in comparison to low PTSD/low NWA profile. It did not differentiate between the two high-PTSD profiles. Gender did not significantly differentiate between any of the profiles.

Table 4. Descriptive statistics of the predictor variables within each of the three latent profiles.

	(1) Low PTSD/low NWA		(2) High PTSD/high NWA		(3) High PTSD/Low NWA
	N	%	N	%	N	%
Torture	11	21.6	36	44.4	36	38.7
Gender (male)	32	62.7	55	67.9	70	75.3
	N	M (SD)	N	M (SD)	N	M (SD)
Total number of traumatic events	38	7.13 (5.56)	56	11.43 (5.11)	65	10.78 (5.13)

Note. PTSD = posttraumatic stress disorder; NWA = negative world assumptions.

Table 5. Results of the multinomial regression analyses of the three profiles and the variables total number of traumatic events, torture and gender.

	(1) Low PTSD/low NWA vs. (2) High PTSD/high NWA			(2) High PTSD/high NWA vs. (3) High PTSD/low NWA			(3) High PTSD/low NWA vs. (1) Low PTSD/low NWA
	B	SE	CI	B	SE	CI	B	SE	CI
Total number of traumatic events	0.955*	0.311	0.91–0.96	−0.211	0.250	−0.25 to −0.17	−0.744*	0.298	−0.79 to −0.67
Torture	1.126*	0.490	1.06–1.19	−0.341	0.407	−0.4 to −0.28	−0.785	0.504	−0.81 to −0.67
Gender	−0.094	0.463	−0.16 to −0.06	−0.806	0.472	−0.87 to −0.74	0.900	0.529	0.83–0.97

Note. PTSD = posttraumatic stress disorder; NWA = negative world assumptions; *p < .05; B, log odd; SE = standard error; CI = 95% confidence interval of regression coefficient B.

3.4. Characteristics of the three profiles

The prevalence and distribution of HTQ events are summarized in Table 6 and graphically illustrated in Figure 4. Between the low PTSD/low NWA group and the two high PTSD profiles, relatively large percentual differences (> 15%) can be found. No such differences were found between the two high-PTSD profiles.

Figure 4. Prevalence (%) and distribution of HTQ traumatic event types in the three profiles.

Table 6. Descriptive statistics of HTQ events within the three profiles.

	(1) Low PTSD/low NWA		(2) High PTSD/high NWA		(3) High PTSD/low NWA
	N	%	N	%	N	%
Gender
Male	32	62.7	55	67.9	70	75.3
Female	19	37.3	26	32.1	23	24.7
Lack of food or water	21	46.7	45	61.6	49	59.8
Ill health, no access to medical care	15	33.3	41	56.2	40	48.8
Lack of shelter	14	31.8	41	56.9	42	50.6
Imprisonment	15	34.1	46	63.0	47	57.3
Serious injury	11	24.4	35	49.3	35	42.7
Combat situation	20	45.5	49	68.1	51	62.2
Brain washing	14	31.8	44	61.1	36	43.9
Rape or sexual abuse	7	15.6	14	19.7	10	12.2
Forced isolation	17	37.8	50	70.4	57	68.7
Being close to death	26	59.1	58	71.6	63	75.0
Forced separation from family	23	52.3	55	74.3	56	68.3
Murder of family member/friend	18	40.0	45	61.6	56	70.0
Unnatural death of family member	20	44.5	43	58.9	57	68.7
Murder of stranger	13	28.9	26	35.6	43	51.8
Lost or kidnapped	12	27.3	40	53.3	39	48.1
Torture	18	40.9	44	59.5	45	53.6
Threatened with torture	22	50.0	56	75.7	57	68.7
Threatened to witness torture	12	27.3	45	60.8	44	55.0
Threatened with execution	17	38.6	39	53.4	44	53.7
Frightening life situation	26	59.1	52	78.8	57	73.1
	N	M (SD)	N	M (SD)	N	M (SD)
Total number of traumatic events	38	7.13 (5.56)	56	11.43 (5.11)	65	10.78 (5.13)

4. Discussion

Using LPA, we demonstrated that individuals can be grouped into meaningful clusters based on PTSD symptom presentation and NWA. Our analysis revealed three distinct profiles that differed in nature and severity of PTSD symptoms and NWA: a low PTSD/low NWA profile, a high PTSD/high NWA profile, and a high PTSD/low NWA profile. Overall, the low PTSD/low NWA profile experiences relatively few PTSD symptoms and has generally positive assumptions about themselves, others, and the world. Most noteworthy, they consider themselves to be lucky and worthy. The most interesting differences in NWA, however, appear between the two high-PTSD profiles. One profile exhibits elevated levels of both PTSD symptoms and NWA, while the other profile demonstrates heightened levels of PTSD but maintains relatively positive world assumptions. Notably, certain world assumptions, particularly Controllability an Self-Controllability, are even more positive in this latter profile than in the low PTSD/low NWA profile. All in all, the profiles indicate that PTSD and NWA are not uniformly experienced among refugees, which emphasizes the diversity in their psychological responses to trauma.

Our findings align with prior research on the heterogeneous nature of PTSD among refugee populations, as reported by Minihan et al. (2018). Furthermore, they underscore the importance of adopting a comprehensive perspective with regard to refugee mental health, as emphasized by Jongedijk et al. (2020). Specifically, our study highlights the significance of critically assessing world assumptions (Janoff-Bulman, 1989, 1992) or negative appraisals (Ehlers et al., 2000; Ehlers & Clark, 2000) in the context of refugee mental health. Consistent with previous research (Porter & Haslam, 2005; Steel et al., 2009), our findings revealed that trauma load and the experience of torture differentiated between profiles characterized by different levels of PTSD symptom severity. However, these factors did not differentiate between the two profiles, which primarily differed in terms of NWA. Moreover, exploring all potentially traumatic event types and their prevalence and distribution within the three profiles did not yield further insights or information (Figure 4). While a difference in prevalence of traumatic event types is visible between the low PTSD/low NWA profile and the other two profiles, only small differences are between the high PTSD/high NWA and the high PTSD/low NWA profiles. Additionally, some events are more prevalent in the high PTSD/high NWA profile, and some are higher in the high PTSD/low NWA profile. Moreover, the pattern of traumatic event types appears relatively similar across all profiles. Therefore, investigating distinct traumatic event types does not appear to elucidate why some individuals experience NWA alongside PTSD while others do not.

Interestingly, we found that gender did not differentiate between any of the profiles. Several studies have shown that women are more likely to develop PTSD than men (e.g. Silove et al., 2017), especially when controlling for the type(s) of traumatic experiences (e.g. gender-based violence). Nonetheless, negative posttraumatic cognitions have been found to be comparable between men and women, especially when controlling for trauma type (Brown et al., 2019; Herta et al., 2017; Reis et al., 2016; Sexton et al., 2018). Possibly the impact of gender on PTSD symptoms differs between samples and may be affected by other variables. Future research is needed to clarify how the interaction between gender and other characteristics of traumatized populations affect PTSD and NWA.

All in all, the analysis did not reveal potential explanations for the difference between the profile with high PTSD severity and low NWA and the profile with high PTSD and high NWA. Gaining a deeper understanding of the contributing factors to the development of both PTSD and changes in beliefs regarding the self, others, and the world is crucial. For example, addressing comorbidity (e.g. Jongedijk et al., 2020), childhood traumatization (Cloitre et al., 2005; McLaughlin & Lambert, 2017), or the influence of daily stress and living difficulties (Miller & Rasmussen, 2017; Minihan et al., 2018) in future research regarding PTSD and NWA might contribute to a more comprehensive and contextual understanding of psychopathology and NWA among refugees. Additionally, while trauma load is frequently used as a measure of severity of traumatization and as a predictor of mental health, it fails to account for patterns and associations between traumatic experiences within individuals. A latent class analysis by Sengoelge et al. (2019) identified subgroups of refugees based on patterns of traumatic experiences, revealing a pattern of interpersonal violence as a key predictor of multiple trauma exposure and a higher severity of mental health problems. This suggests that individuals experiencing one type of violent trauma are more likely to have experienced other types of violent trauma as well, emphasizing the interaction between traumatic experiences, especially within a group known for having experienced an accumulation of potentially traumatic experiences and severe psychopathology in often complex socio-economical contexts.

Moreover, exploring the trajectory of symptom presentations over time is an essential next step in the research of refugee mental health. Longitudinal investigations would allow for the identification of potential causal relationships regarding symptom development and associations among symptoms. For example, Ehlers et al. (2000) found that individuals with chronic PTSD experienced increased levels of perceived negative and permanent change in their personalities and assumptions about life. Such research would also facilitate predictions regarding treatment response and provide valuable insights into the effectiveness of different interventions, enabling the development of individualized treatment approaches tailored to individual needs.

Furthermore, we utilized DSM-IV criteria for PTSD symptomatology, which does not include the DSM-5 addition of ‘negative alterations in cognitions and mood’ and ‘persistent distorted cognitions about the cause or the consequences of the traumatic event(s) that lead the individual to blame himself/herself or others’ (APA, 1994, 2013). Foa et al. (1999) found that the WAS had only low to moderate correlations with measures of psychopathology, and none of the subscales showed substantial correlations with PTSD severity (as measured with DSM-IV). Nonetheless, a significant conceptual distinction between WAS and cognitions as assessed in the DSM-5 lies in the nature of what is measured. The WAS assesses core beliefs developed in an individual’s formative years that can be disrupted due to traumatization. In contrast, the cognitions outlined in the DSM-5 denote negative alterations in cognitions specifically linked to the traumatic event. Further research is needed to examine if similar profiles can be identified using DSM-5 criteria.

Building upon our findings and aligning with previous research connecting PTSD and posttraumatic cognitions in relation to chronicity and recovery (e.g. Dekel et al., 2013; Ehlers et al., 1998; Ehlers et al., 2000; Iverson et al., 2015), several therapeutic recommendations could be considered. The identification of the severe PTSD and high NWA profile suggests the presence of a group that faces prolonged and more chronic psychopathology and that encounters greater challenges towards recovery, especially since NWA represents more generalized core beliefs about the self, others, and the world that have been profoundly disrupted. The diversity in PTSD symptoms and WAS presentations may indicate distinct therapeutic needs or require different therapeutic foci. The primary focus of most trauma-focused treatments is the exposure to and reduction of fear and challenging avoidance behaviour. As such, negative world assumptions may not change with solely these interventions, and residual complaints may remain. Additional interventions targeting dysfunctional negative assumptions associated with the traumatic experience(s) and its meaning with, for example, cognitive therapy (Ehlers et al., 2005), cognitive processing therapy (Schulz et al., 2006), or imagery rescripting (Arntz et al., 2013; Arntz, 2015). Schema-focused therapy has also shown promising results in reducing PTSD symptoms, for example in war veterans (Cockram et al., 2010), and adults with PTSD stemming from traumatic experiences in childhood (Boterhoven de Haan et al., 2019), as it encompasses both imagery rescripting and the exploration and targeting of maladaptive schemas, i.e. dysfunctional core beliefs and cognitions. Lastly, a recently developed intervention targeting trauma-related emotions and cognitions with regard to guilt or shame, Brief Eclectic Psychotherapy for Moral Trauma (BEP-MT), could prove valuable in treating such dysfunctional and rigid feelings and beliefs (de la Rie et al., 2021).

Our findings should be considered in light of several limitations. Given that our sample comprised refugees referred to a highly specialized mental health care facility and therefore included a relatively high number of individuals with elevated trauma exposure, chronic PTSD symptoms, and complex socio-economic contexts, generalization of these findings across other groups of refugees or individuals suffering from PTSD in general should be done with caution. Moreover, the sample exclusively included individuals with a refugee background who demonstrated sufficient proficiency in the Dutch language to fill out the questionnaires used in this study. Consequently, it remains uncertain whether we have inadvertently assessed a subgroup exhibiting a lower symptom burden, considering the relationship between host language acquisition and mental health problems (Kartal et al., 2019). As language is a common barrier for both receiving (mental) healthcare and inclusion in scientific research, it is imperative for future research to prioritize linguistic inclusivity and ensure that participation in studies is not constrained by language barriers (e.g. by translating instruments or by working with interpreters). While the WAS was found to have adequate cross-cultural validity, the conceptualization of PTSD is inherently intertwined with specific historical and cultural developments in Europe and North-America (Jongedijk et al., 2023). Accordingly, instruments and assessments commonly used are also culturally bound and not free of bias. While PTSD appears to be present in different cultural groups across the world (e.g. Hinton & Lewis-Fernandez), cross-cultural variations in PTSD have been found (Heim et al., 2022). Therefore, more research is needed to determine to what extent the co-occurrence of PTSD symptoms and NWA differs between cultural subgroups. Furthermore, whether someone had been subjected to torture was approximated by combining the items of imprisonment and having experienced torture. Although this approach was employed to mitigate the risk of false positive identifications of torture, it should be noted that it may not be a dependable measure for accurately assessing the experience of torture. In addition, it was based on a relatively stringent definition of torture. Future research should include non-state torture (Jones et al., 2018) and genderized torture (Pérez-Sales & Zraly, 2018) to ensure an inclusive and intersectional perspective on torturous experiences. Additionally, we recommend expanding data collection to encompass sociodemographic characteristics, such as duration since migration, employment status or social support, as well as clinical characteristics, such as comorbid psychopathology, because we cannot rule out that the co-occurrence of PTSD symptoms and NWA is moderated by one or more of these variables. Finally, data were collected several years ago, and although there are no specific reasons as to why outcomes with more recently collected data would lead to different results, this possibility can not be dismissed.

In conclusion, our study represents the first examination of subgroups based on PTSD symptoms and NWA in a clinical refugee sample using LPA. Our findings demonstrate the heterogeneity in PTSD symptom presentations among refugees, as evidenced by the identification of three distinct profiles characterized by variations in type and severity of symptoms. Particularly noteworthy is the identification of two subgroups among refugees with severe PTSD complaints: one characterized by low NWA levels, and another exhibiting high NWA levels, indicating an additional burden of negative core beliefs and assumptions about themselves, others and the world. Recognizing this heterogeneity and addressing core beliefs and assumptions is crucial in both research and clinical practice, particularly in the context of refugee mental health.

Ethical standards and patient consent statement

Questionnaires were primarily administered for diagnostic purposes before the start of treatment as part of a routine diagnostic assessment procedure that all refugees who applied for treatment were asked to complete. These data were subsequently archived anonymously for scientific research purposes. Patients were informed that these data were stored anonymously and were given the opportunity to have their data removed from the database. Upon consultation, the Medical Ethics Committee of the University of Leiden stated that no review of the ethical merits of this study or informed consent from patients was needed as the assessment was primarily administered for clinical purposes, and secondarily for research purposes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Open practices and data availability statement

The study reported in this article was not formally preregistered. Data were collected primarily for clinical purposes and are not deposited in a community-recognized repository because participants have not provided informed consent for sharing data outside of the institute. Request to access these datasets should be directed to [email protected].

Additional information

Funding

This study was funded by ARQ Centrum’45 and ARQ Nationaal Psychotrauma Centrum.