Refugees are one of the largest demographics worldwide exposed to severe traumatic experiences. The United Nations Refugee Agency estimated that by mid-2023, there are over 110 million forced displaced people worldwide, 75% of which hosted in low- and middle-income countries (LMICs) (United Nations High Commissioner for Refugees, 2024). In conflict settings, a meta-analysis conducted by the World Health Organization (WHO) estimated that the average prevalence for MDD, anxiety, and PTSD among affected populations was 13% for mild, 4% for moderate, and 5.1% for severe cases (Charlson et al., 2019). However, some groups of refugees and internally displaced persons (IDPs) in LMICs may present significantly higher incidence than others. For instance, a meta-analysis including 10 studies with 5,287 participants in African countries indicated that the pooled prevalence of PTSD was 56.35% for IDPs, and 54.04% for refugees (Andualem et al., 2024). In alignment, the prevalence of PTSD in Liberia was estimated at 48.3% (Galea et al., 2010), compared to 11.8% among Guatemalan refugees in Mexico (Sabin et al., 2003), or 61% among Burmese refugees in Bangladesh (O'Connor and Seager, 2021).

In high-income countries, a comprehensive systematic review including data from 6,743 refugees resettled across seven countries indicate prevalence rates of 9% (99% CI 8–10%) for PTSD and 5% (4–6%) for MDD (Fazel et al., 2005), up to 10 times higher compared to age-matched local populations. Similar to LMICs, studies also show wide ranges of prevalence rates among different cultural groups. For example, among adult Syrian refugees in Norway, prevalence is estimated at 29.7% for PTSD, 30.1% for anxiety symptoms, and 45.2% depression (Nissen et al., 2021). Among African refugees resettled in Italy, one study documented that 79% of a sample met the DSM-5 criteria for PTSD (Barbieri et al., 2019). Interestingly, research examining the relationship between PTSD prevalence at a national level, the frequency of trauma exposure and the overall cultural and socioeconomic vulnerability of countries to adversity found no significant association between PTSD and vulnerability of the country itself (Dückers et al., 2016). Hence, the study found highest PTSD prevalence in developed countries including Canada, the Netherlands and Australia, and lowest in developing nations such as Nigeria, China and Romania (Dückers et al., 2016).

Overall, studies on the prevalence of mental health challenges among refugees demonstrate that ratings can be as elevated across high-income countries and LMICs. Significant discrepancies in results have been attributed to methodological components (e.g., self-report vs., structured interviews, sample size, diagnostic criteria) explaining 12.9% of the variability (Steel et al., 2009). Accounting for these components, meta-analyses estimate global PTSD prevalence rates in refugees at approximately 30% (Steel et al., 2009; Blackmore et al., 2020). Not accounted for, a significant variant in prevalence of PTSD among refugees may also pertain to complex PTSD (CPTSD) (Liddell et al., 2019), postulated to result from prolonged exposure to significant trauma commonly experienced by refugee populations. In addition to PTSD symptomatology, CPTSD is characterized by disturbances in self-organization (DSO), which includes negative self-concept, challenges with relationships, and emotion dysregulation (World Health Organization, 2018). Importantly, among refugee groups, postmigration living difficulties which exacerbates symptom severity have been primarily associated with DSO and CPTSD, rather than PTSD (Hecker et al., 2018; Silove et al., 2018; Tay et al., 2018; Liddell et al., 2019; Schiess-Jokanovic et al., 2021, 2022; Barbieri et al., 2023).

Among widely utilized measures to screen for PTSD in refugees with acceptable psychometric properties are the Posttraumatic Stress Disorder Checklist (PCL-5) (Pereira-Lima et al., 2019), the Harvard Trauma Questionnaire (HTQ) (Berthold et al., 2019), the International Trauma Questionnaire (ITQ) (Cloitre et al., 2021), and the Impact of Event Scale (IES) (Weiss, 2007). While there is a notable surge in efforts to adapt PTSD screeners for cross-cultural populations in LMICs (Wilson and Tang, 2007), several limitations exist, including few translations, poor accessibility, length of screeners, and the need for trained administers. Among screeners cited, the ITQ presents the best accessibility for refugee populations, with 32 languages freely available for download on an open-source website¹, and has been validated in LMICs among vulnerable populations (e.g., Syrian refugees in Lebanon, Vallières et al., 2018), in addition to being reasonably brief and suitable for self-administration. However, a notable limitation of the ITQ, shared with other measures, is its focus on solely assessing PTSD, lacking transdiagnostic capabilities. This lack has compelled agencies working with refugees to use a combination of measures to document complex outcomes of trauma, resulting in lengthier paper-based efforts that, particularly in LMICs, exhaust limited resources.

The absence of a single instrument to measure varied possible consequences of exposure to trauma led to a collaborative effort by the Global Collaboration on Traumatic Stress (GCTS, Schnyder et al., 2017) to develop the Global Psychotrauma Screen (GPS) (Oe et al., 2020; Olff et al., 2020; Frewen et al., 2021; Rossi et al., 2021). In a short screening format, the GPS broadly assesses several potential consequences of traumatic events, including PTSD, complex PTSD, depression, anxiety, sleep problems, self-injurious behavior, dissociation, substance abuse, and other physical, emotional, or social problems. Furthermore, it documents risk factors for the consequences of trauma, including prior stressful events, history of childhood trauma and mental illnesses, lack of social support, and perceived lack of psychological resilience.

Numerous studies conducted across different countries and populations have demonstrated that the GPS is a valid and reliable instrument (Schnyder et al., 2017; Oe et al., 2020; Olff et al., 2020, 2021; Rossi et al., 2020, 2021; Cao et al., 2021; Grace et al., 2021, 2023; Williamson et al., 2021; Hoffman et al., 2022; Marengo et al., 2022; Havermans et al., 2023; Salimi et al., 2023; Bi et al., 2024; Frewen et al., 2024; Koutsopoulou et al., 2024). Research on the psychometric properties of the GPS present evidence of robust reliability, as well as convergent and concurrent validity with measures of trauma-related symptom domains. For instance, in two general population samples of English-speaking participants aged 18 and over, Frewen and colleagues documented strong internal consistency for the 17 GPS symptom items (α = 0.94) and a cut-off GPS symptom score of 8 for optimal sensitivity (83%) which refers to the screener’s ability to identify the condition, and relative to specificity (71.1%), indicating that majority of patients without the condition were true negatives (Frewen et al., 2021). These resulted in a sensitivity of minimally 80% and area under the ROC curve (AUC) measuring test accuracy = 0.839 for predicting probable PTSD (Frewen et al., 2021), based on scores on the PCL-5 (Blevins et al., 2015). The researchers also identified a single factor via exploratory factor analysis, indicating that one latent construct underlies a substantial part of the outcomes of trauma (Frewen et al., 2021). This corresponds to the high comorbidity between trauma-related disorders.

In a cross-sectional study examining the psychometric properties of the Japanese version of the GPS, including 58 individuals with trauma history, researchers identified excellent internal consistency for the GPS total scores (α = 0.90) (Oe et al., 2020). The GPS total score was highly correlated with the PCL-5, the Patient Health Questionnaire (PHQ-9, Kroenke et al., 1999), a measure for depression, and a self-report GAD measure, the generalized anxiety disorder-7 (GAD-7, Spitzer et al., 2006), (r > 0.79) (Oe et al., 2020). In addition, the GPS subscale scores correlated with corresponding domain scales assessing the several potential consequences of traumatic events, showing good concurrent validity (Oe et al., 2020). In a study validating the GPS in a LMIC, Indonesia, researchers found acceptable AUCs (> 0.70) for all probable diagnoses including PTSD, CPTSD, GAD, and MDD, except for insomnia (Primasari et al., 2024).

The GPS has also been used and found reliable in a global study examining the mental health of populations among 7,034 participants from 88 countries and 12 UN regions (α = 0.88) (Olff et al., 2021). Moreover, in a large Italian sample, the convergent validity of the GPS was assessed for the GPS symptom total score and the PHQ-9, GAD-7, and the Insomnia Severity Index (ISI, Morin, 1993), all yielding statistically significant correlations (p < 0.001) (Rossi et al., 2021). While the accumulated evidence for the GPS is good with demonstrated cross-cultural cultural capabilities, the accuracy of the screener has not been tested among clinical and refugee populations. As refugee numbers continue to rise worldwide (United Nations High Commissioner for Refugees, 2022), health providers in low-resource settings could significantly benefit from a brief and transdiagnostic screening tool.

This study examined the reliability and clinical validity of the GPS among refugees. We hypothesize that the GPS will present adequate efficiency for detecting a probable diagnosis of PTSD, MDD, and GAD with moderately high sensitivity and specificity (>0.7/0.8).

GPS data was collected between December 2018 and October 2020 as part of a larger clinical trial in Queensland and New South Wales, Australia. The study was approved by the University of Sydney Human Research Ethics Office and was conducted in compliance with ethics committee approval conditions. All participants voluntarily provided oral and written informed consent at baseline. Participants were recruited in collaboration with refugee-led organizations in Queensland, Australia. A protocol detailing the study methodology has been published elsewhere (Pinto et al., 2022).

Inclusion and exclusion criteria were aligned to the aims of the larger clinical trial (see Pinto et al., 2022). Eligibility included trauma-exposed Liberian, Congolese or Sudanese refugees aged 18 and over who migrated to Australia having departed Africa during civil wars, fragile peace periods, or Ebola endemics (1989–2018). Moreover, eligibility for half of the sample (those in the PTSD condition of the original trial) included meeting the diagnostic criteria for clinical or subclinical PTSD, while this was an exclusion criteria alongside the presence of other psychiatric diagnoses for the other half of the sample (those in the control condition of the original trial). All participants were required to have English proficiency. Exclusion criteria included severe medical conditions or genetic disorders, being at acute risk of suicide or homicide, drug or alcohol abuse or dependency in the previous month, current use of psychotropic medication or use within the 2 months preceding the study, and concurrent psychotherapy for PTSD.

All measures included in this present GPS validation analyses were collected during the study baseline phase. The SCID-5 and the CAPS-5 interviews were administered by a trained master’s level clinician, while the BRS and the GPS were completed by participants as a self-report assessment.

Internal consistency of the GPS symptom score and subscales were checked using Cronbach’s alpha, Cronbach’s alpha when item deleted, inter-item and item-total correlations (based on Spearman’s Rho). Cronbach’s alpha >0.75, inter-item correlations between.15 and 0.50 and item-total correlations >0.30 were considered adequate. Clinical validity of the GPS subscales measuring PTSD, dissociation, depression, and GAD was evaluated with respective diagnostic reference standards for PTSD and the dissociative subtype (CAPS-5), and for MDD and GAD (SCID-5). We determined the optimal cut-off with the Youden index², thereby maximizing the sum of the sensitivity and specificity. For the dissociative subtype, we first established the optimal cut-off for PTSD and dissociation, and then combined these two. Thus, scoring above the cut-off for both PTSD and dissociation was scored as a positive screen for the dissociative subtype. We reported corresponding sensitivity, specificity, positive predictive value³, and negative predictive value⁴, including 95% confidence intervals (note that these were corrected for the case–control design), likelihood ratios and receiver operating curves, including the area under the curve. We also tested the convergent validity of the GPS resilience item with the BRS total score with a point-biserial correlation coefficient. Finally, we tested whether risk factor items of the GPS were related to GPS symptom scores via a series of linear regression analyses. Analyses were performed in SPSS (IBM Corp., 2019) and Team RC (2020) using R packages OptimalCutpoints and cutpointsr.

At intake, 100 individuals were screened, among which 20 did not meet the study’s inclusion criteria (n = 7 controls with psychiatric conditions, n = 6 reported substance abuse, n = 4 had poor English fluency, n = 2 were unable to be reached after initial intake, and n = 1 was outside of age range). At baseline, 80 participants were fully assessed, with five not enrolled (n = 4 due to busy schedules, and n = 1 reason undisclosed). Among the final 75 individuals who enrolled, five participants opted out of the study (n = 3 due to busy schedules, n = 1 alcohol abuse disclosure, n = 1 loss of interest), consenting that their data be removed.

Item endorsement of the GPS is shown in Figure 1. Items about self-harm and substance abuse were rarely endorsed (<10%). Notably, two risk factors, childhood trauma and current stressors were highly endorsed, while a history of mental illness and lack of social support were rarely endorsed. GPS symptom scores ranged between 0 and 14 (M = 5.76; SD = 4.55).

Our results indicated that the GPS symptoms score, reflecting the transdiagnostic overall symptom burden, in addition to the GPS subscales that screen for PTSD, dissociation and MDD were reliable, while for GAD and DSO this was not found. GPS validation studies in other populations also generally found that subscales GAD and DSO showed a relatively low internal consistency (e.g., Oe et al., 2020). Both constructs are only assessed with two items, which might be insufficient for a reliable measurement of these constructs. Future studies might assess whether adding more items for these constructs might improve the reliability. However, the main purpose of the GPS is to efficiently screen for probable trauma-related disorders, and to keep the items to a minimum. The present study also documented that substance abuse and self-harm had low inter-item correlations with other GPS symptoms, while self-harm also showed low item-total correlations. Both constructs were rarely endorsed in the current sample compared to previous validation studies and to recent norm data (Oe et al., 2020; Frewen et al., 2021; Rossi et al., 2021; Global Collaboration on Traumatic Stress, 2022). This might be related to the current population (African refugees), since cultural meaning, functions, and manifestation of symptoms such as self-harm might differ cross-culturally (Gargiulo et al., 2021). Moreover, symptoms may differ within cultural groups. It is plausible that the present sample, consisting of 78.5% adults with an active occupation (employed or studying) may be less prone to substance misuse compared to marginalized African refugee youth in Australia (Horyniak et al., 2016; Mwanri and Mude, 2021) which have been prominently documented to abuse alcohol. Likewise, higher ratings among asylum seekers in Australia have been documented among those in off-shore detention centers compared to those in community-based arrangements (Hedrick et al., 2019). Further studies are needed to further investigate the prevalence of self-harm and substance abuse among representative African refugee populations.

We found that the GPS presented adequate efficiency with moderately high sensitivity and specificity for detecting a probable diagnosis of PTSD, dissociation, dissociative subtype and MDD, but we were unable to test this for GAD. The area under the curve for PTSD, dissociation and MDD was above 0.90, reflecting an excellent screener ability of the GPS in this population (Youngstrom, 2014). A clinical cut-off score of 3 and higher on the GPS PTSD subscale was optimal for detecting probable PTSD, and a cut-off score of 1 or higher was optimal on the GPS subscales dissociation and MDD for detecting probable dissociation and MDD, respectively. The sensitivity and specificity for PSTD, dissociation, the dissociative subtype, and MDD documented in this study demonstrate adequate accuracy of the GPS in screening for probable clinical diagnoses assessed with the CAPS-5 and SCID-5. The sensitivity was very high for all these outcomes, indicating that people suffering from the disorder may screen positive on the GPS. For a screening instrument, it is purposeful to be optimally sensitive and thus have rare false negative screenings, as the aim is not to miss potential trauma survivors with these problems. Positive screenings are usually followed-up by a clinical interview, where potential false positives may be detected. Specificity for MDD, and to lesser extend for PTSD, was moderately high, underscoring the importance of follow-up assessments in clinical settings. As noted, validity analyses for GAD were not possible due to the low GAD occurrence in this sample (1.5%) according to the SCID-5. This low endorsement may be explained by the SCID-5 GAD criteria F, “the disturbance is not better explained by another mental disorder” (First et al., 2015b).

The relevance of these findings is inherently aligned with the advantages of screening over diagnostic tests. Trauma survivors may suffer from a wide range of disorders and problems, as well as comorbid presentations, that would require a large test battery to detect. This benefits both the individual and the healthcare system, for instance “placing fewer demands on the healthcare system and being more accessible as well as less invasive, less dangerous, less expensive, less time-consuming, and less physically and psychologically discomforting for clients” (Trevethan, 2017). These advantages become particularly valuable for trauma-impacted populations in low-resource settings. Compared to existing screening measures, the GPS presents the unique transdiagnostic quality in a simple and brief format, suitable for both research and clinical settings.

We also examined associations between the GPS symptom scores and risk and protective factors. We found that current stressors and childhood trauma were related to higher GPS symptom scores, while lack of resilience, social support, and history of mental illness were not significantly associated with the GPS symptom scores. Lack of social support, and a history of mental illness, were both only endorsed by a few people, much less compared to previous GPS validation studies in other samples (Oe et al., 2020; Frewen et al., 2021; Rossi et al., 2021). While speculative, we suggest that the low mental illness history endorsement may be explained by a lack of awareness rather than a lack of illness, given the low accessibility to mental health professionals in the participants’ home countries. As addressed in the pilot test of the Japanese version of the GPS, cross-cultural differences may exist in response to traumatic experiences, including underreporting and stigma (Oe et al., 2020). In support, studies with samples of Sudanese refugees in Australia have indicated stigma (Savic et al., 2016) and low mental health literacy levels (May et al., 2014) to be prevalent, but the same is not known for Liberian and Congolese populations.

Lastly, we examined the convergent validity of the GPS resilience item concerning the BRS total score and found non-significant results. Challenges with the BRS scale have been previously reported in another study with refugees, indicating potential inadequacy in the BRS index and contextual nuances when assessing resilience cross-culturally (Tay et al., 2021). Alternatively, one item measuring psychological resilience might not be sufficient for such a complex construct. Future studies might correlate the GPS resilience item with other measurements of resilience such as the Resilience Evaluation Scale (van der Meer et al., 2018).

This study presented some limitations. First, since the data was collected as part of a larger treatment trial, the enrolment criteria involved excluding participants with substance abuse. Moreover, while the study sample presents a reasonably balanced gender ratio (M = 31, F = 39), of those who fulfilled the PTSD criteria, 65.7% of the participants were female. This gender imbalance has been documented among displaced African women (Idemudia et al., 2014) and is consistent with the previous evidence of women being at greater risk for developing posttraumatic stress outcomes, including PTSD (Olff et al., 2007; Olff, 2017; Ramikie and Ressler, 2018). Although the year range of trauma exposure in the inclusion criteria was extensive to accommodate refugees exposed to war and conflict across 3 countries at different times, it is possible that age at onset and number of years in resettlement may have impacted severity levels, alongside post-migration difficulties most associated with DSO and CPTSD. The validity indices presented for the SCID-5, CAPS-5, and BRS were obtained from western populations and not representative of refugees and Sub-Saharan Africans. While a strength of the study is the use of golden standard clinical interviews to assess diagnoses of all axis I psychiatric conditions, few or no endorsements in relevant disorders, such as GAD and sleep disorders, limited the present analyses.

Future studies with larger samples are recommended to establish the usability of the GPS as an efficient screening tool for trauma-related disorders. In particular, studies are needed to investigate the validity of the GPS in screening anxiety, self-organization, and resilience among refugees. Moreover, while participants’ English language skills were satisfactory for the present study, new studies validating translated versions of the screen in languages widely spoken by underrepresented populations across conflict-impacted regions would be valuable. Considering the present results are based on English-speaking African refugees resettled in a developed country, the generalizability of the present results to global refugee populations, including culturally diverse groups, requires caution and further research. While gender disparities may be challenging to address in future studies, methods should be designed to ensure gender-sensitive analyses (Olff, 2019). Lastly, considering the difficulties faced with the BRS instrument, future studies should cautiously examine the cultural appropriateness of resilience (and other) measures for refugee populations, particularly in the absence of normative data.