Refugee crisis

Approximately 79.5 million individuals worldwide have been forcibly displaced.¹ Of these, 45.7 million have been displaced internally, 29.6 million have been externally displaced and 4.2 million are seeking asylum. Forced displacement typically follows human rights violations, conflict, persecution or events causing serious public disorder, with refugees and asylum seekers experiencing ‘complex traumas’.

Complex trauma

Complex traumas are characteristically interpersonal in nature (i.e. are perpetrated deliberately by other people) and involve repeated prolonged and often multiple traumatic exposures between victim and perpetrator(s). The victim has typically been held in conditions from which they are unable to escape, because of physical, maturational, social, family/environmental or psychological constraints.^{Reference Herman2} Refugees and asylum seekers often report experiencing direct exposure to multiple types of complex traumatic experiences,^{Reference Teodorescu, Heir, Hauff, Wentzel-Larsen and Lien3} including sexual violence, torture, imprisonment, enforced isolation, physical assault, and the murder of friends or family members.^{Reference Liddell, Nickerson, Felmingham, Malhi, Cheung and Den4} In addition to pre-migration traumas experienced in the individuals’ countries of origin, traumatic events can also be experienced during forced migration journeys,^{Reference Beiser5, Reference Phillimore6} for instance, at the hands of human smugglers or traffickers or in refugee camps. This can result in trauma-related mental health disorders.^{Reference Bogic, Njoku and Priebe7}

Complex post-traumatic stress disorder

In response to the observation that complex trauma can lead to mental health sequalae not adequately captured by prevailing diagnostic formulations of post-traumatic stress disorder (PTSD), Herman^{Reference Herman2} in 1992 proposed the term complex PTSD (CPTSD). PTSD typically follows single traumatic events that need not be interpersonally caused, such as natural disasters and accidents.^{Reference Breslau, Davis, Andreski and Peterson8} However, individuals with experiences of complex trauma (particularly in the context of coercive control) may experience symptoms more varied and severe than the intrusion, avoidance and hyper-arousal symptoms classically associated with PTSD.

Enduring personality change after catastrophic events (EPCACE) was included in the ICD-10⁹ to capture the personality changes which can follow catastrophic stressors. Single traumatic events not interpersonal in nature are excluded as possible catastrophic stressors. A diagnosis of EPCACE requires personality changes attributed to trauma to last for a minimum of 2 years and can involve a mistrustful or hostile attitude towards the world, social withdrawal, persistently feeling hopeless or empty, estrangement and continuously feeling on edge.⁹ The criteria have been criticised for being overinclusive, with this diagnosis underutilised in both research and clinical settings.^{Reference Beltran and Silove10}

The DSM-IV¹¹ included an equivalent symptom group referred to as disorder of extreme stress not otherwise specified (DESNOS). Core features of DESNOS include changes in sustaining beliefs, impulse and affect regulation, somatisation, self-perception, relationships with others and dissociation. DESNOS was viewed only as a symptom group within PTSD, as field trials indicated that almost all participants meeting DESNOS criteria also met PTSD criteria.^{Reference Roth, Newman, Pelcovitz, van der Kolk and Mandel12}

The diagnostic category of CPTSD has been incorporated into the ICD-11¹³ as a separate disorder to PTSD. It consists of symptom clusters involving disturbances in self-organisation (DSO) in addition to the classical PTSD clusters of intrusion, avoidance, and hyper-arousal. The DSO symptom clusters include affective dysregulation, disturbances in relational function and negative self-concept. In addition, CPTSD requires endorsing additional symptoms indicating functional impairment. CPTSD reflects the depletion of social, psychological and emotional resources resulting from protracted conditions of trauma,^{Reference Cloitre, Garvert, Brewin, Bryant and Maercker14} leading to greater functional impairment than PTSD.

CPTSD, EPCACE and DESNOS have some overlapping symptoms. All three attempt to capture the more varied mental health sequalae that follow complex trauma. However, the diagnostic conceptualisations are not identical (Fig. 1). Unlike DESNOS, ICD-11 CPTSD is a diagnostic category distinct from PTSD, and unlike EPCACE, CPTSD criteria do not include personality change.

Fig. 1

Overlaps and differences among definitions of enduring personality change after catastrophic events (EPCACE), disorder of extreme stress not otherwise specified (DESNOS), complex post-traumatic stress disorder (PTSD) and ICD-11 PTSD. AD, affective dysregulation; NSC, negative self-concept.

The applicability of CPTSD symptomatology has been demonstrated in various trauma samples, including individuals who have suffered childhood physical or sexual abuse,^{Reference Cloitre, Garvert, Brewin, Bryant and Maercker14} former prisoners of war^{Reference Zerach and Solomon15} and institutional abuse survivors.^{Reference Knefel, Garvert, Cloitre and Lueger-Schuster16} Although refugees and asylum seekers are a highly traumatised population, less is known about the prevalence of CPTSD in this population. The increase in the number of forcibly displaced people requiring resettlement in recent years requires physicians and health service providers to be aware of the mental disorders prevalent in this population. Treatment needs for PTSD and DSO symptoms may differ; research indicates that solely trauma-focused treatment can overwhelm those who have suffered complex trauma,^{Reference Herman17} with a phased treatment approach required instead.^{Reference ter Heide, Mooren and Kleber18} An understanding of the current prevalence of CPTSD in this population (and the associated treatment implications) is therefore highly important.

ter Heide et al^{Reference ter Heide, Mooren and Kleber18} have reported findings of an unpublished systematic review of five studies that used validated measures to assess the prevalence of DESNOS among refugees. The review incorporated two non-treatment-seeking and three treatment-seeking refugee populations displaced internally and externally. Relative to a diagnosis of DESNOS, no diagnosis or a PTSD diagnosis was more likely. They concluded that the more severe DESNOS symptoms should not be assumed to occur in a population because of the presence of complex traumatic experiences. However, the authors acknowledged that their conclusions were based on the limited evidence available at the time.^{Reference ter Heide, Mooren and Kleber18} This review also pre-dated the introduction of ICD-11 CPTSD, and there has since been an expansion in the literature examining this disorder among refugee and asylum-seeking populations.

A recent systematic review by Mellor et al^{Reference Mellor, Werner, Moussa, Mohsin, Jayasuriya and Tay19} also reported the prevalence of CPTSD, DESNOS and EPCACE in samples of refugees and displaced individuals, and further examined predictors of CPTSD and the associations between CPTSD and common mental disorders. The results demonstrated a wide range of CPTSD prevalence among studies, with authors positing this to reflect geographic and contextual factors.^{Reference Mellor, Werner, Moussa, Mohsin, Jayasuriya and Tay19} However, not all of the included studies examining CPTSD followed ICD-11 CPTSD diagnostic criteria or used diagnostic tools specific to ICD-11 CPTSD to diagnose participants. Furthermore, not all studies consisted of forcibly displaced samples. There remains the need for further synthesis of data on the prevalence of CPTSD, EPCACE and DESNOS among refugees and asylum seekers in treatment-seeking and population samples. An updated systematic review examining this is therefore timely.

Criteria for study inclusion

Eligible studies were required to include samples of refugees and/or asylum seekers of any ethnicity who were over 18 years of age at the time measures were administered. The use of validated or standardised diagnostic methods specific to the diagnoses of ICD-11 CPTSD, ICD-10 EPCACE or DSM-4 DESNOS to provide the corresponding diagnoses was also required, as was reporting of the prevalence of CPTSD, EPCACE or DESNOS within the refugee and/or asylum-seeking samples. As sampling biases are more likely in smaller samples,^{Reference Fazel, Wheeler and Danesh21} only studies with a sample size greater than 50 were considered eligible. When numerous articles reported the same data from the same sample, the earliest publication was used.

Search strategy

The electronic databases PsychINFO, Embase, Medline, EBSCO and ProQuest PTSDpubs, a database specifically for PTSD and trauma literature, were searched from the time of their inception to 25 March 2020. A final search of these databases was conducted on 8 Jan 2021 to capture any studies published after the first search.

The search terms consisted of terms relating to refugees, asylum seekers, CPTSD, DESNOS and EPCACE. Full search strategies for all databases are provided in Supplementary Appendix 1. The reference lists of studies eligible for inclusion in the review were manually searched by the first author for further relevant studies not captured by the systematic searches. None were found.

Study selection

The identification of studies followed procedures outlined by PRISMA.^{Reference Liberati, Altman, Tetzlaff, Mulrow, Gøtzsche and Ioannidis22} Titles and abstracts were initially screened by U.D.S. for articles deemed potentially relevant. Following this, all authors independently reviewed the full texts of each study to establish whether they were eligible for inclusion. All authors then discussed the final studies to be included.

Data extraction

Data were first extracted on 3 July 2020 and recorded in a tabular form including sample setting; time spent in host country; mean age and age range; measurement tools used to assess CPTSD, EPCACE and DESNOS; measurement tools used to assess additional diagnoses; diagnostic criteria used for diagnosing CPTSD, EPCACE and DESNOS, and additional disorders assessed; and the prevalence of CPTSD, EPCACE and DESNOS, and other diagnoses assessed. Additional study characteristics extracted were gender breakdown, age statistics and duration spent in the host country.

Synthesis and analysis of data

The heterogeneity among the study designs, methods, populations and the reported effect estimates in the eligible studies did not allow for meta-analysis. Consequently, a narrative synthesis was conducted to synthesise findings. The point prevalence of CPTSD, EPCACE and DESNOS among study samples was described and compared. Owing to the expected disparity in the prevalence estimates between treatment-seeking and population samples, these groupings are presented separately. Studies were also grouped by diagnostic system used. Where possible, the prevalence rates of the disorders of interest among refugee and asylum-seeking samples were compared with the prevalence of other diagnoses assessed within the same samples.

Risk of bias assessments

The Joanna Briggs Institute Critical Appraisal Checklist for Prevalence Data²³ was used to assess the methodological quality of each individual study. This considers nine domains: suitability of the sample frame; appropriateness of participant recruitment; sample size sufficiency; adequacy of the setting and participant description; adequacy in reporting the sample; validity of the diagnostic measures; suitability of the statistical analyses; and adequacy of the response rates. The majority view among all authors for each criterion was used.

Treatment-seeking samples

ICD-11 PTSD prevalence was measured in six studies (Fig. 4). Two reported higher ICD-11 PTSD prevalence (38 and 32.9%, respectively) compared with CPTSD (30 and 21.3%, respectively).^{Reference Barbieri, Visco-Comandini, Fegatelli, Schepisi, Russo and Calo24, Reference Hecker, Huber, Maier and Maercker28} Four studies reported lower ICD-11 PTSD prevalence estimates compared with CPTSD.^{Reference Liddell, Nickerson, Felmingham, Malhi, Cheung and Den4, Reference Hyland, Ceannt, Daccache, Abou Daher, Sleiman and Gilmore29, Reference Nickerson, Cloitre, Bryant, Schnyder, Morina and Schick31, Reference Vang, Nielsen, Auning-Hansen and Elklit36}

Fig. 4

Prevalence of the most frequently measured additional diagnoses of DSM-5 post-traumatic stress disorder (PTSD) and ICD-11 PTSD compared with complex PTSD in treatment-seeking and population samples.

One study measured DSM-5 PTSD, reporting a higher prevalence estimate of 79% compared with a CPTSD prevalence of 30%^{Reference Barbieri, Visco-Comandini, Fegatelli, Schepisi, Russo and Calo24}. One study reported prevalence estimates of 82% for DSM-IV PTSD, 71% for major depressive disorder (MDD), 31% for generalised anxiety disorder (GAD), 41% for panic disorder, 49% for dysthymia, 0% for hypomania, 49% for agoraphobia, 49% for social phobia, 44% for obsessive–compulsive disorder, 10% for alcohol abuse disorder, 36% for nicotine dependence and 5% for bulimia, compared with 16% for DESNOS.^{Reference Teodorescu, Heir, Hauff, Wentzel-Larsen and Lien3} Rathke et al reported prevalence estimates of 2.6% for ICD-10 PTSD, 49.5% for ICD-10 PTSD and comorbid EPCACE, 43.9% for ICD-10 PTSD with secondary psychotic features (PTSD-SP) and comorbid EPCACE, 25.8% for ICD-10 and comorbid depression, and 4.2% for ICD-10 PTSD-SP without comorbid depression, relative to 49.5% for EPCACE.^{Reference Rathke, Poulsen, Carlsson and Palic33}

One study^{Reference Mundy, Foss, Poulsen, Hjorthoj and Carlsson30} reported a prevalence estimate of 93.8% for ICD-10 depression and 33.2% for other diagnoses, relative to a lower EPCACE prevalence estimate of 24.2%, in a sample of refugees with existing diagnoses of ICD-10 PTSD.

Population samples

ICD-11 PTSD was measured in three studies, with the prevalence ranging between 6 and 21.3%. Two studies reported greater ICD-11 PTSD prevalence estimates (20.9 and 13%) relative to CPTSD (4.9 and 3%),^{Reference Frost, Hyland, McCarthy, Halpin, Shevlin and Murphy26, Reference Tay, Rees, Chen, Kareth and Silove35} with the remaining study reporting a lower PTSD prevalence estimate (21.3%) compared with CPTSD (50.9%).^{Reference Grossman, Hoffman, Shrira, Kedar, Ben-Ezra and Dinnayi27} DSM-5 PTSD was assessed in two studies, with prevalence estimates of 12^{Reference Tay, Rees, Chen, Kareth and Silove35} and 12.1%^{Reference Silove, Rees, Mohsin, Tam, Kareth and Tay34} reported; DSM-5 PTSD prevalence estimates were higher than CPTSD prevalence in both studies. One study^{Reference Tay, Rees, Chen, Kareth and Silove35} reported prevalence estimates of 13% for DSM-IV PTSD and 13% for ICD-10 PTSD, whereas the CPTSD prevalence was 3%. Another study^{Reference Silove, Rees, Mohsin, Tam, Kareth and Tay34} reported prevalence estimates of 56.2% for MDD, 26.5% for GAD, 8% for panic disorder, 9.9% for persistent complex bereavement, 9.9% for separation anxiety disorder and 5.6% for intermittent explosive disorder, compared with 9.3% for CPTSD.

Methodological quality

Risk of bias assessments for all studies are presented in Table 2. Two studies used samples drawn through random selection.^{Reference de Jong, Komproe, Spinazzola, van der Kolk and Van Ommeren25, Reference Frost, Hyland, McCarthy, Halpin, Shevlin and Murphy26} Two studies attempted to sample the entire population of eligible refugees.^{Reference Silove, Rees, Mohsin, Tam, Kareth and Tay34, Reference Tay, Rees, Chen, Kareth and Silove35} One study used a consecutively drawn sample.^{Reference Vang, Nielsen, Auning-Hansen and Elklit36} The sampling methods in the remaining ten studies were unclear or non-random.

Table 2

Methodological quality of included studies assessed using the Joanna Briggs Institute Critical Appraisal Checklist for Prevalence Studies²³

Y, Yes; N, No; U, Unclear.

The samples used in two studies were drawn from census data of refugees living in selected settlements.^{Reference Silove, Rees, Mohsin, Tam, Kareth and Tay34, Reference Tay, Rees, Chen, Kareth and Silove35} One study^{Reference de Jong, Komproe, Spinazzola, van der Kolk and Van Ommeren25} drew its sample from a list of registered displaced individuals, one study^{Reference Frost, Hyland, McCarthy, Halpin, Shevlin and Murphy26} used a subsample selected from a randomly drawn and nationally representative sample of US non-institutionalised adults, and one study^{Reference Grossman, Hoffman, Shrira, Kedar, Ben-Ezra and Dinnayi27} drew its sample from post-liberation camps. The remaining ten studies drew samples from treatment centres.^{Reference Teodorescu, Heir, Hauff, Wentzel-Larsen and Lien3, Reference Liddell, Nickerson, Felmingham, Malhi, Cheung and Den4, Reference Barbieri, Visco-Comandini, Fegatelli, Schepisi, Russo and Calo24, Reference Hecker, Huber, Maier and Maercker28–Reference Rathke, Poulsen, Carlsson and Palic33, Reference Vang, Nielsen, Auning-Hansen and Elklit36}

As they were not stated in any of the included studies, the required sample sizes for prevalence estimates for treatment-seeking and population samples were calculated using the following formula^{Reference Naing, Winn and Rusli37}: $n = {{Z^2P( {1-P} ) } \over {d^2}}$, where n is the sample size, Z is the level of confidence statistic, P is the anticipated prevalence and d is the precision. Expected CPTSD prevalence estimates of 31.4% in treatment-seeking samples and 4.9% in population samples were assumed by calculating the median values of the reported prevalence estimates within these two sample types. Median values were used owing to the variability in prevalence estimates among studies, so as to reduce sensitivity to outliers. In addition, 95% confidence intervals and 5% precision were used in sample size calculations for both treatment-seeking and population samples. Required sample sizes for treatment-seeking and population samples were 331 and 72, respectively. Thus, sample sizes were insufficient in all ten studies with treatment-seeking samples but were sufficient for studies with population samples.

Two studies developed measures consisting of items from several scales^{Reference Frost, Hyland, McCarthy, Halpin, Shevlin and Murphy26, Reference Nickerson, Cloitre, Bryant, Schnyder, Morina and Schick31} but did not validate them. One used items representing DSO clusters from a PTSD scale but reported good internal consistency (Cronbach's α = 0.81).^{Reference Barbieri, Visco-Comandini, Fegatelli, Schepisi, Russo and Calo24} Another used unvalidated precursors to the International Trauma Questionnaire but reported good and high internal consistency for both within the sample (Cronbach's α = 0.75 and 0.92).^{Reference Liddell, Nickerson, Felmingham, Malhi, Cheung and Den4}