Prevalence of post-traumatic stress disorder and depressive symptoms among civilians residing in armed conflict-affected regions: a systematic review and meta-analysis

Data sources and search strategy

The present study aimed to estimate the pooled prevalence of PTSD and depressive symptoms in war or armed conflict-affected regions. For this purpose, two authors (SHA and AZ) independently conducted a thorough literature search across major health databases (ie, MEDLINE(R), Embase Classic+Embase, APA PsycINFO, Ovid Healthstar, Journal@Ovid Full Text, Cochrane, PTSDpubs and CINAHL) from inception until 19 March 2024. Relevant keywords were obtained via the Medical Subject Heading browser tool alongside synonyms; alternate spellings were employed and included “Depression” “Depressive Symptoms” “Emotional Depression” “Major Depressive Disorder” “MDD” “Major Depression” “Acute Post-Traumatic Stress Disorder” “Chronic Post-Traumatic Stress Disorder” “Delayed Onset Post-Traumatic Stress Disorder” “Neuroses, Post-Traumatic” “PTSD” “Post-Traumatic Stress Disorders” “Stress Disorder Post Traumatic” “Posttraumatic Stress Symptoms” “War” “Warzone” “Civil War” “Warfare” “Genocide” “Holocaust” and “Armed Conflict”. The bibliographies of included studies were also thoroughly evaluated to acquire comprehensive search results. Furthermore, relevant published systematic reviews and meta-analyses were shortlisted, and their citations were screened to identify any potentially missed articles. Duplicates were removed manually using the Zotero citation manager.

Selection criteria

We included cross-sectional and cohort studies assessing the prevalence of PTSD and depressive symptoms in the adult, that is, over 18 years, civilian populations witnessing armed conflict and residing in the war-affected region for the duration of exposure. The outcomes could be assessed using both interviews and questionnaires such as the Hopkins Symptom Checklist-25, Clinician-Administrated PTSD Scale, General Health Questionnaire-28, Posttraumatic Stress Disorder Checklist, and so on. Of note, these outcomes were limited in measuring relevant symptoms of PTSD and depression and could only act as a proxy for these diagnoses but not the clinical diagnoses themselves. Studies evaluating prevalence in displaced individuals were only included if the participants were classified as internally displaced population, that is, residing within the same country during the time of conflict. Furthermore, articles were not limited by the scale of armed conflict, including both civil and international wars. Lastly, only articles published in the English language were included.

We did not include original articles evaluating the prevalence among children, military individuals, combatants, refugees, second or third-generation survivors of war, and civilians from non-war-involved countries. Similarly, survivors of terrorist attacks (in an otherwise stable country), non-war trauma exposure and those enrolled in medical interventions like clinical trials were also not included. Overall, the PECOS criteria for the review are the following:

1. Population: adult (18+ years) civilian population residing in the war-affected region for the duration of exposure.

2. Exposure: armed conflict or war, defined as a conflict resulting in at least 1000 war-related deaths in 1 calendar year, in which at least one side is the government of a state.

3. Control: none or baseline.

4. Outcome: prevalence of PTSD and/or depressive symptoms.

5. Study: cross-sectional and cohort studies.

Data extraction and quality assessment

Two reviewers (SHA and AZ) independently screened the title and abstract of articles, and ultimately, full-length articles were assessed against the selection criteria. Following a comprehensive literature search and study selection, two authors (AZ and MZ) independently extracted the following details into two Excel sheets: first author’s name, study year, design, number of participants, type and country of conflict, baseline characteristics, time since the war, epidemiological assessment method employed and relevant prevalence. The quality of included studies was assessed qualitatively using the Joanna Briggs Institute Critical Appraisal Checklist for Prevalence Studies.16 The appraisal tool evaluates each study across the following domains: sample frame, participant sampling, sample size, data collection, study variables, statistical analysis and response rate. It comprises nine questions, where each component is assessed independently and reported as either yes (low risk), no (high risk), unclear or not applicable. Any discrepancies were resolved by discussion with independent reviewers (MYJ and MC).

Outcome measure

The primary outcomes of our meta-analysis were deemed as the pooled prevalence of PTSD and depressive symptoms among the war-affected civilian population.

Statistical analysis

We used Comprehensive Meta-Analysis Software V.3.0 (Biostat, Englewood, New Jersey, USA)17 to estimate the pooled prevalence of interested outcomes separately and create forest plots. Due to postulated differences among studies, a random-effects model was employed.18 To evaluate heterogeneity, Q-value, τ² and I² were employed.19 Sensitivity analysis using the leave-one-out method was also performed to see if all the studies contributed roughly equally.20 21 Lastly, to assess the difference between observed effect size (effect size we see in included studies) and true effect size (the effect size we would see in comparable populations) in a random-effects model, a prediction interval was computed to estimate the differences observed at a population level.22

Furthermore, we performed subgroup analyses based on time since the war and the country’s economic status. Studies were classified with respect to data collection, that is, short term (if they collected data within 10 years of war), intermediate (data collection between 11 and 20 years after the war) and long term (data collection following more than 20 years). Similarly, studies were classified according to their economic standing, depending on the World Bank’s classification, using general national income (GNI) per capita. Countries were classified as low income if they had a GNI capita less than $1085, middle income if the GNI capita ranged between $1086 and $13 205, and high income if the capita exceeded $13 205.23 If data were unavailable during years of war, the World Bank’s classification during the data collection period was used.

To identify any potential publication bias, we used Begg’s funnel plot to measure asymmetry, whereas Duval and Tweedie’s trim-and-fit method was employed to estimate the number of trimmed/added studies and their impact on outcomes.24 All analyses were two-sided using p<0.05 to indicate statistical significance.

Literature search

The initial literature search revealed 28 674 articles published until 19 October 2022. However, an updated literature search was performed to identify 38 595 studies published from inception until 19 March 2024. After duplicate removal, 11 564 articles were screened for eligibility. Following screening by title and abstract, 132 studies were assessed full length. After a thorough assessment, 57 articles were deemed eligible for inclusion.25–81 The PRISMA flow chart, shown in figure 1, summarises the literature search; the reasons for exclusion are in the online supplemental appendix 2.

Baseline characteristics and quality assessment

Our review included data from 57 studies, with a total of 64 596 participants meeting our predefined inclusion criteria. Participants from all adult age groups were included, with ages ranging from 18 to over 86 years. The studies evaluated adults from across the globe, as shown in online supplemental material 3, with 85.96% (49 of 57) of studies reporting data from low/middle-income countries (LMICs), while 12.28% (7 of 57) recruited from high-income countries, and one study did not specify the country of conflict. The time to data collection since the war varied from information collected during the war to over 70 years post-war. A total of 17.54% (10 of 57) of studies collected data during an ongoing conflict, while 43.86% (25 of 57) collected data within the first 10 years, that is, referred to as short term. Similarly, 15.79% (9 of 57) and 14.04% (8 of 57) of articles reported prevalence in the intermediate and long term, that is, 11–20 years and beyond 20 years, respectively. Time since the war was not specified in two studies,43 69 while three studies39 40 61 reported outcomes at overlapping intervals. The baseline characteristics and outcomes of individual studies have been described in online supplemental materials 4 and 5, respectively.

Most of the included studies were identified as having a low risk of bias, with the majority opting for appropriate study frames (91.23%), sampling appropriate population (73.68%), well-explained study settings (100%), using valid methods for identification of conditions (98.24%), measured outcomes in a standard, reliable manner (98.24%) and conducted the appropriate statistical analysis (98.24%). However, 19 studies (33.33%) were rated either high risk or unclear for adequate sample size. Similarly, we observed that 22 studies (38.60%) did not report response rates appropriately. The quality assessment results have been summarised in the online supplemental material 6.

Prevalence of PTSD symptoms

In our review, 48 studies reported the prevalence of PTSD symptoms, including data from 40 004 participants. The pooled prevalence was found to be 23.70% (95% CI 19.50% to 28.40%), as shown in figure 2. The prediction interval for the prevalence of PTSD symptoms was estimated as 5.00%–66.00%, as shown in the Online supplemental appendix 7. Excluding the studies by Richards et al 43 and Roberts et al,69 46 studies were included in the subgroup analysis based on time since the war. The highest prevalence was observed during the war, with 33.90% (95% CI 16.30% to 57.50%) screening positive, followed by long, intermediate and short term as 24.60% (95% CI 13.80% to 39.90%), 20.70% (95% CI 12.90% to 31.50%) and 20.20% (95% CI 16.30% to 24.60%), respectively (online supplemental appendix 8).

Similarly, except for Trappler et al 26 (excluded because the country was not specified), 47 studies were included in the subgroup analysis based on the involved country’s economic standing, and a higher prevalence was identified among LMICs as 25.40% (95% CI 20.90% to 30.60%), relative to 11.40% (95% CI 4.90% to 24.10%) in high-income countries (online supplemental appendix 9). The analysis revealed a moderate heterogeneity (Q=75.50, p=0.016, I²=28.48%, τ²=0.83). The sensitivity analysis did not yield any grossly unequal contribution of a single study toward the overall effect size. Begg’s funnel plot was also plotted, as shown in the online supplemental appendix 10, and it depicted a symmetrical spread of studies, with no study being trimmed from either right or left.

Prevalence of depressive symptoms

A total of 37 studies reported the prevalence of depressive symptoms, with data from 51 766 participants included. The pooled prevalence of 25.60% (95% CI 20.70% to 31.10%) was estimated among war-torn civilians, as shown in figure 3. The prediction interval for the prevalence of depressive symptoms was estimated as 5.00%–68.00% (online supplemental appendix 11). Excluding the studies by Richards et al 43 and Roberts et al 69 (excluded because the time from war to data collection was not specified), 35 studies were included in the subgroup analysis based on time since the war. The highest prevalence was observed during the war, with 33.80% (95% CI 19.80% to 51.40%) of adults diagnosed, followed by short, intermediate and long term as 25.30% (95% CI 18.60% to 33.40%), 23.70% (95% CI 17.40% to 31.40%) and 17.50% (95% CI 8.10% to 33.80%), respectively, as depicted in online supplemental appendix 12.

Similarly, a subgroup analysis comprising 37 studies was performed based on the involved country’s economic standing. A higher prevalence was identified among LMICs at 27.00% (95% CI 21.50% to 33.20%), relative to 18.30% (95% CI 9.50% to 32.40%) in high-income countries, as shown in the online supplemental appendix 13. The analysis revealed a moderate heterogeneity (Q=53.43, p=0.020, I²=23.27%, τ²=0.80). The sensitivity analysis revealed a grossly equal contribution of all studies toward pooled effect size. Begg’s funnel plot was also plotted, as shown in the online supplemental appendix 14, and an asymmetry was observed, with seven studies being trimmed from the right side and none from the left. The adjusted prevalence following trim and fill was estimated as 20.57% (95% CI 15.73% to 26.43%).

Main findings

Our analysis of 64 596 adults who had PTSD and depressive symptoms in war-afflicted regions is strong evidence of the correlation between war and mental health illnesses. To our knowledge, this is one of the few studies estimating the symptomatology burden among war-afflicted civilians. Our extant meta-analysis suggests that these disparate symptoms of PTSD and depression (a proxy for clinical diagnoses of PTSD and MDD) point towards a significant unmet need of millions of people affected by man-made disasters.

Consistent with other meta-analyses, our study also found similar pooled estimates of PTSD and depressive symptoms. Hoppen et al reported pervasiveness of 26.51% for PTSD symptoms and 23.31% for depression in adult survivors of the recent wars from 1989 to 2019 globally.13 Similarly, Mesa-Vieira et al found cluster symptoms of PTSD to be predominant (31%) in prevalence, followed by depressive features (25%) and anxiety-related features (14%) among migrants exposed to armed conflict.1 In contrast, our results had a greater pooled prevalence of depression followed by PTSD symptoms among included participants.

We also performed subgroup analysis based on time since the war and the involved country’s economic standing to evaluate its association with the incidence of mental health outcomes. The result of the study by Neria et al that assessed the mental health consequences of the Israel–Gaza 2008–2009 war among young Israeli civilians reported a sharp decrement in the symptomatology of PTSD over time such that the prevalence rate dropped from 20.00% during the war to 3.00% and 2.20% after 2 and 4 months of the ceasefire, respectively.40 It was hypothesised that the overall prevalence of PTSD and depressive disorders may decrease with time, with the highest being reported during the conflict and the lowest in the long term. However, in our analysis, the results for PTSD symptoms were not consistent with the hypothesis, as the long-term prevalence turned out to be higher than the short-term interval. This could be due to the abundant elderly population in this time interval who might have faced greater trauma from other origins over their lifespan. Moreover, the elderly population is more susceptible to adverse age-related health conditions, which may contribute to the existing psychological stress. As reported by Eytan et al, poor health outcomes due to a lack of medical facilities and variations in job requirements following armed conflict may significantly contribute to chronic PTSD symptoms.82 Another explanation could be due to sustained traumatic experiences. Trauma response is a ‘subjective’ response to rapidly changing chronic unpredictable stress and can often result even though the primary insult is over. As such, it could be that post-war circumstances, including sociopolitical upheavals and economic compromises, could act as a perpetual trauma with a gradual surge in PTSD symptoms.

The results of the subgroup analysis based on the country’s economic status revealed the dominance of both PTSD and depressive symptoms among LMICs, in contrast to high-income countries. Similar to our results, the study by Hoppen et al revealed higher PTSD and depressive symptoms among war-affected LMICs.13 This is due to numerous challenges causing mental distress in LMICs that include poor infrastructure, unemployment, financial instability and food insecurity, which are further aggravated consequently by these conflicts. Moreover, the lack of healthcare facilities and trained professionals compromises satisfactory disease management. In a community-based survey in Rwanda,25 Bolton et al observed a lack of globally approved management strategies for depression, with medications and one-on-one therapy considered extremely costly and budget-unfriendly. Moreover, the efficacy of Western-style psychotherapy and regimens has not yet been proven in non-Western or LMIC populations.25

Our analysis revealed some heterogeneity across both PTSD and depressive symptoms. Other than disparity across the economic standing of involved countries and time since the war, the use of different diagnostic tools, varying trauma severity experienced by participants, differences across sample size (ranging from 27 to 12 841), study designs (cross-section, cohort) and varying target populations (ranging from university students to older adults) could have potentially contributed to the observed heterogeneity.

Several factors may predispose specific populations to develop mental illnesses despite similar trauma exposures. Luitel et al stated that the female gender, older age, inaccessibility of clothes and more perceived negative impact of the conflict in the community were strongly correlated with features of anxiety, depression and PTSD. Furthermore, the study found a significant relationship between disability, measured by the WHO Disability Assessment, and mental disorders.53 Another study reported that decreasing PTSD symptoms over time is associated with multiple elements, including younger or older age, male gender, lower education, social support, fewer traumatic experiences during the follow-up period, employment status and absence of psychiatric comorbidity, particularly with depression and other anxiety disorders and minor initial PTSD symptoms.83

In contrast to the civilian population, a higher prevalence of mental health disorders may be observed among combatants and refugees because of greater trauma exposure. While most studies report a higher prevalence among militant groups,84 Lim et al reported features of depression and anxiety significantly more prevalent among militant populations as compared with civilians, whereas no significant difference was observed for PTSD symptoms.6 This may be due to repetitive traumatic events experienced during training or different time zones, leading to gradual desensitisation and, ultimately, a mere emotional response.6

Implications

The findings of this meta-analysis underscore the critical need for comprehensive mental health interventions targeting civilians in war-affected regions. The incidence of mental health disorders among war-afflicted populations can be alleviated by routine screening to diagnose at-risk individuals and timely administration of culturally appropriate psychiatric interventions, including psychosocial support, counselling sessions, group therapy and pharmacotherapy. As the conflict leads to detrimental effects at the community level, destroying the families and social networks, community-based interventions must be promoted where family members, neighbours and caregivers can provide natural support to the affected individuals instead of external professionals such as counsellors.85 If offered through a formal health and educational system, psychological interventions can attain greater coverage and sustainability.86 In addition, governments and international non-government organisations must collaborate to ensure adequate supply provision and improve living conditions, reducing physical and psychological stresses. Similarly, global forums should be urged to highlight the unendurable toll that conflicts take on affected populations and advocate against such violent acts for a better and peaceful world. Lastly, further research is needed to address gaps in understanding, including longitudinal assessments and the effectiveness of interventions tailored to the unique needs of war survivors.

Limitations

Our study carried certain limitations. First, the epidemiological assessment methods for PTSD and depressive symptoms varied across included studies, with some employing questionnaires for data collection while others opting for mini-interviews. Similarly, no standardisation was observed across the questionnaires, tools and diagnostic cut-offs. Although all participants experienced armed conflict firsthand, the trauma levels varied across different studies. In the study of Branas et al,54 17.44% of the participants had firsthand experience of civil war, while Rugema et al 64 reported that 35.4% of women and 37.5% of men experienced traumatic episodes during the genocide period in Rwanda. Similarly, the varying severity and duration of the wars were not considered and might have impacted the outcomes. All of these factors could have contributed to the observed heterogeneity as well. Lastly, due to a lack of phenomenological consensus across the definitions of war, terrorism and armed conflict, a few studies might have been missed.