Hostname: page-component-797576ffbb-xmkxb Total loading time: 0 Render date: 2023-12-06T14:26:17.871Z Has data issue: false Feature Flags: { "corePageComponentGetUserInfoFromSharedSession": true, "coreDisableEcommerce": false, "useRatesEcommerce": true } hasContentIssue false

Political violence and mental health in Nepal: prospective study

Published online by Cambridge University Press:  02 January 2018

Brandon A. Kohrt*
Department of Psychiatry and Behavioral Sciences, The George Washington University, Washington DC, USA
Daniel J. Hruschka
School of Human Evolution and Social Change, Arizona State University, Tempe, Arizona, USA
Carol M. Worthman
Department of Anthropology, Emory University, Atlanta, Georgia, USA
Richard D. Kunz
Department of Physical Medicine & Rehabilitation, Virginia Commonwealth University Health Systems, Richmond, Virginia, USA
Jennifer L. Baldwin
Department of Anthropology, University of Illinois-Urbana Champaign, Urbana Champaign, Illinois, USA
Nawaraj Upadhaya
University of Amsterdam, Amsterdam, The Netherlands
Nanda Raj Acharya
Transcultural Psychosocial Organization Nepal, Kathmandu, Nepal
Suraj Koirala
Transcultural Psychosocial Organization Nepal, Kathmandu, Nepal
Suraj B. Thapa
Institute of Psychiatry, University of Oslo, Institute of Psychiatry, Oslo, Norway
Wietse A. Tol
Department of Anthropology, Yale University, New Haven, Connecticut, USA
Mark J. D. Jordans
Department of Research and Development, HealthNet TPO & Centre for Global Mental Health, London School of Hygiene and Tropical Medicine
Navit Robkin
Rollins School of Public Health, Emory University, Atlanta, Georgia, USA
Vidya Dev Sharma
Department of Psychiatry, Institute of Medicine, Tribhuvan University Teaching Hospital, Maharajgunj, Nepal
Mahendra K. Nepal
Department of Psychiatry, Institute of Medicine, Tribhuvan University Teaching Hospital, Maharajgunj, Nepal
Brandon A. Kohrt, Department of Psychiatry and Behavioral Sciences, The George Washington University, 2150 Pennsylvania Avenue, 8th Floor, Washington, DC 20037, USA. Email:
Rights & Permissions [Opens in a new window]



Post-conflict mental health studies in low-income countries have lacked pre-conflict data to evaluate changes in psychiatric morbidity resulting from political violence.


This prospective study compares mental health before and after exposure to direct political violence during the People's War in Nepal.


An adult cohort completed the Beck Depression Inventory and Beck Anxiety Inventory in 2000 prior to conflict violence in their community and in 2007 after the war.


Of the original 316 participants, 298 (94%) participated in the post-conflict assessment. Depression increased from 30.9 to 40.6%. Anxiety increased from 26.2 to 47.7%. Post-conflict post-traumatic stress disorder (PTSD) was 14.1%. Controlling for ageing, the depression increase was not significant. The anxiety increase showed a dose–response association with conflict exposure when controlling for ageing and daily stressors. No demographic group displayed unique vulnerability or resilience to the effects of conflict exposure.


Conflict exposure should be considered in the context of other types of psychiatric risk factors. Conflict exposure predicted increases in anxiety whereas socioeconomic factors and non-conflict stressful life events were the major predictors of depression. Research and interventions in postconflict settings therefore should consider differential trajectories for depression v. anxiety and the importance of addressing chronic social problems ranging from poverty to gender and ethnic/caste discrimination.

Copyright © Royal College of Psychiatrists, 2012 

Low-income countries are home to more than one-sixth of the world's population, but bear a disproportionate burden of the world's violent conflicts. Reference Tol, Barbui, Galappatti, Silove, Betancourt and Souza1 Cross-sectional epidemiological studies have demonstrated the association of torture and political terror with psychiatric morbidity. Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren2 However, a weakness of psychiatric epidemiology studies in low-income countries has been a lack of pre-war estimates of mental illness prevalence. Owing to limitations in studying unpredictable events and the lack of routine psychiatric surveillance in low-resource settings, researchers generally rely upon observations after a violent conflict to estimate how war influences mental health. This has led some mental health experts to conclude that post-conflict psychiatric morbidity is attributable predominantly to war experiences. A review of violent conflict and mental health in low- and middle-income countries exemplifies the problematic conclusions drawn from existing epidemiological studies: ‘The exposure [to war] of large population groups, mostly having no mental health problems prior to the exposure, and the subsequent development, in a significant proportion of the population, of a variety of psychiatric symptoms and disorders represent both a challenge and an opportunity for psychiatrists’ (emphasis added). Reference Murthy and Lakshminarayana3 These authors also suggest that the impact on mental health as a result of war are greater for children, women and elderly people. Reference Murthy and Lakshminarayana3 However, these conclusions are difficult to support without evidence of low psychiatric morbidity prior to political violence. War trauma may not be the sole or dominant determinant of post-conflict mental health. Psychiatric morbidity may have been elevated before the outbreak of war. Populations in low-income countries not only suffer from war trauma but also from other psychiatric risk factors such as poverty, high burdens of infectious disease, high maternal and infant mortality, gender-based violence and limited access to healthcare and education. Reference Desjarlais, Eisenberg, Good and Kleinman4Reference de Jong and de Jong6 Moreover, high prevalence rates among certain demographic groups, such as women and elderly people, may not reflect greater vulnerability to the effects of war. Elevated rates in certain groups could result from factors that operate similarly in non-war settings, such as gender discrimination. Reference Patel, Kirkwood, Pednekar, Pereira, Barros and Fernandes7,Reference Das, Do, Friedman, McKenzie and Scott8 It is difficult to disentangle the influence of war from these chronic societal problems because of the lack of pre-conflict data.

The goal of this study is to use a prospective design to address three questions: (a) is there an increase in mental health problems from the pre- to post-conflict period, (b) if there were an increase, could it be attributed to conflict-related exposures, and (c) are some groups more vulnerable to the effects of conflict on mental health? To answer these questions, the mental health of a cohort in Nepal is compared across two time points: in 2000 before the outbreak of conflict-related violence in their community and in 2007–2008 after the signing of peace accords. This is the first study conducted in a low-income country that examines individual differences in mental health before and after exposure to war.


Setting and study participants

Nepal is among the world's poorest countries 9 and recently endured the People's War fought between the Communist Party of Nepal (Maoists) and government security forces from 1996 until 2006. Although the People's War officially began in 1996, its effects were differentially felt throughout the country during the early years of the insurgency. Prior to 2000, violence was limited to a few regions of Nepal, and most of the general population considered the Maoists a ‘law and order problem’ rather than a significant military threat. Reference Thapa and Sijapati10 This study was conducted in Jumla, a mountainous district in northwestern Nepal with no conflict mortality prior to 2001. It is important to note that the term ‘pre-conflict’ is employed in this study to refer to the data collection in 2000 because this was prior to any Maoist attacks in the study area of Jumla. Also, ethnographic research in Jumla during 2000 found that Maoist-related concerns did not dominate health narratives among people with anxiety and depression. Reference Kohrt11 It was not until after the Royal Massacre of 2001 when King Gyanendra came to power that the Royal Nepal Army became engaged in the conflict. Of the 14 000 killed in the decade-long war, 92.5% of all deaths occurred after King Gyanedra assumed power in 2001. 12 In the study area of Jumla, there were no war-related casualties until late 2001. Between 2001 and 2006, Jumla suffered the effects of the People's War, including major battles leading to civilian, Maoist and government casualties, destruction of all telecommunication infrastructure, abduction and torture of residents, government blockades, decreased food availability, bombing of schools and destruction of medical infrastructure. 13 (See Tol et al's review for more information on the war history and psychosocial impacts. Reference Tol, Kohrt, Jordans, Thapa, Pettigrew and Upadhaya14 )

The baseline ‘pre-conflict’ data for this study are from a community epidemiology study of depression and anxiety conducted in Jumla in 2000. The baseline study was an evaluation of somatisation, depression and anxiety, Reference Kohrt, Kunz, Baldwin, Koirala, Sharma and Nepal15 as well as caste and gender differences in mental health. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya16Reference Kohrt18 In the baseline study, random sampling of one adult per household was used for recruitment with an nth-household sampling strategy. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya16 After the war ended in late 2006, the research team began tracking original participants for post-conflict interviews, which took place in 2007–2008. In cases of participants who died in the interim of 2000 to 2007, the cause, age and year of death were recorded.

In 2000, all participants were aged 18 years or older. Consent was recorded with a signature for literate participants or a thumbprint for illiterate participants. Participants did not receive compensation. The Department of Psychiatry at Tribhuvan University Teaching Hospital/Institute of Medicine (TUTH/IOM) in Kathmandu provided consultation prior to and during the assessment and gave final approval for the study in 2000. The protocol for the follow-up study in 2007–2008 was approved by the Institutional Review Board of Emory University and the Nepal Health Research Council, with modifications approved by TUTH/IOM. Interviews at both waves of assessment were conducted in participants’ homes with only the interviewer and participant present. Interviews lasted 60–90 min. In 2007–2008, participants with high levels of psychological distress and impaired functioning (58 individuals) were evaluated by the study's principal investigator. A subset of these was referred for psychosocial care (43 individuals).


The 21-item Beck Depression Inventory (BDI) Reference Beck, Ward, Mendelson, Mock and Erbaugh19 and Beck Anxiety Inventory (BAI) Reference Beck, Epstein, Brown and Steer20 were used to assess depression and anxiety symptoms over the prior 2 weeks. Items are scored 0–3 with an instrument range of 0 to 62. Scales were validated for use in Nepal Reference Kohrt, Kunz, Koirala, Sharma and Nepal21,Reference Kohrt, Kunz, Koirala, Sharma and Nepal22 with clinical DSM-IV 23 diagnoses of major depressive disorder or generalised anxiety disorder: area under the curve (AUC) 0.92 (95% CI 0.88–0.96) for the BDI and 0.85 (95% CI 0.79–0.91) for the BAI; internal reliability (Cronbach alpha), BDI α = 0.90 and BAI α = 0.90. Based on the clinical validation of the BDI in Nepal, a score of 20 or higher suggests moderate depression symptoms with the need for mental health intervention (sensitivity 0.73, specificity 0.91). Reference Kohrt, Kunz, Koirala, Sharma and Nepal21 On the BAI, a score of 17 or higher indicates moderate anxiety symptoms with need for intervention (sensitivity 0.77, specificity 0.81). Reference Kohrt, Kunz, Koirala, Sharma and Nepal22 These cut-off scores are intended only to reflect symptom burden at the level requiring intervention; the cut-offs do not indicate diagnoses of major depressive disorder or generalised anxiety disorder. Test–retest reliability Spearman–Brown coefficients for the BDI were 0.84 and for the BAI were 0.88. The validated BDI and BAI were used in the pre-conflict and post-conflict assessments.

The 17-item Post-traumatic stress disorder (PTSD) Checklist-Civilian Version (PCL-C) is a self-report rating scale for assessing PTSD symptoms and severity within the past week. Reference Weathers, Litz, Herman, Huska and Keane24 The English-language measure has good psychometric properties in Western populations Reference Weathers, Litz, Herman, Huska and Keane24 and the validated Nepali version performs similarly, Reference Tol, Komproe, Thapa, Jordans, Sharma and De Jong25 with a cut-off score of 50 or above indicating need for intervention. Reference Thapa and Hauff26 For this study, internal reliability was 0.83. Test–retest reliability was 0.82.

The 64-item Stressful Life Events Rating Scale for Cross Cultural Research (SLERS) was used to assess stressful events over the preceding 12 months prior to the 2000 and 2007 assessments. Reference Li, Ma and Fu27 The SLERS has shown association with poor mental health in other Asian populations. Reference Zheng and Lin28,Reference Kohrt, Hruschka, Kohrt, Panebianco and Tsagaankhuu29 At the baseline pre-conflict assessment in 2000, greater frequency of stressful life events in the preceding 12 months as measured with the SLERS was associated with greater depression, Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya16 anxiety, Reference Kohrt and Worthman17 psychosomatic complaints Reference Kohrt, Kunz, Baldwin, Koirala, Sharma and Nepal15 and general psychological morbidity. Reference Kohrt18

A self-report conflict exposure scale was developed to assess exposures to political violence-related traumatic events during the People's War. The items were chosen based on focus-group discussions, key informant interviews and review of documents about the conflict in Jumla. Mental health professionals who are natives of Jumla reviewed the political violence questionnaire for content and comprehensibility. Additionally, 30 Jumla residents conducted pile sorts and ranking tasks to indicate level of traumatic severity. Factor analysis demonstrated one cultural model for the group of traumatic events, i.e. there were not significant differences in rankings of traumatic events. For the analyses presented here, objective items (such as abduction, witnessed a killing) were retained, and subjective items (such as less access to healthcare, unable to meet daily needs) were removed because of increased risk of appraisal bias from current mental health status. Of the 32 proposed items, the final instrument included 14 items (Cronbach's α = 0.69), which were considered locally severe and putatively objective. For these analyses, the results were dichotomised to ‘0’ no exposure and ‘1’ any exposure. The total number of types of exposure were summed (total: 0 to 14), and individuals were categorized into four groups according to quartiles of total exposures (0–3, 4–5, 6–7, >7 exposures). Only four individuals reported no trauma exposures; they were included in the low-exposure quartile.


We first compared baseline (pre-conflict) characteristics between those who did and did not participate in the follow-up. Chi-squared for categorical outcomes tested differences among three groups: those who participated in the follow-up study, those who died before the follow-up study and those who were lost to follow-up or refused to participate. McNemar chi-squared tests were used to assess crude changes in income, education, stressful life events and depression and anxiety.

Generalised estimating equations (GEEs) were used to model the effect of assessment period (0: pre-conflict, 1: post-conflict), age (divided into 7-year age brackets), gender, caste (Bahun –‘high caste’, Chhetri – ‘high caste’, Dalit – ‘low caste’ and Janajati-Buddhist ethnic minority groups), education (no education v. any education), income (no income v. any income), and stressful life events (median split: <15 events v. ⩾15 events) and conflict exposure (quartiles) on binary outcomes for depression and anxiety. Generalised estimating equations can be used for longitudinal data analysis, Reference Ballinger30 such as this study that includes two data points (pre- and post-conflict) for each individual. We used GEEs to ensure appropriate estimates and inferences from correlated repeated measures within individuals.

We used GEEs to test four models. In the first model, the effect of period of assessment (pre- v. post-conflict) on depression and anxiety was calculated to determine whether rates increased over time. In the second model, gender, age and caste were included to determine whether the effect of period of assessment remained significant after controlling for ageing. Gender and caste were included because of their association with outcomes in the pre-conflict period. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya16,Reference Kohrt and Worthman17 By including current age in 7-year age brackets in the model, we account for the effect of ageing of the sample over the 7 years between assessment periods. In the third model, the level of conflict exposure was entered to determine whether this accounted for the differences in period of assessment. In the final model, stressful life events in the past year and income were entered to determine whether conflict exposure remained significant when controlling for these more proximal stressors.

We used GEEs to determine whether specific demographic groups demonstrated greater change in mental health from the pre- to post-conflict period. Separate models were created for each interaction of interest: exposure (pre- v. post-conflict)×gender, exposure×age and exposure×caste. The sample was dichotomised at 39 years old to compare differential effect by age. For the caste model, Dalit v. all other groups was compared because the Dalit caste displayed the poorest outcomes in pre-conflict analysis. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya16

TABLE 1 Baseline characteristics of all participants (pre-conflict, year 2000) (n =316)

n (%) Test statistic
Total original
participants (n = 316)
participants (n = 298)
Deceased at follow-up
(n = 12)
Lost to follow-up
(n =6)
χ2 P
Gender 5.06 0.08
    Male 183 (57.9) 168 (56.4) 10 (83.3) 5 (83.3)
    Female 133 (42.1) 130 (43.6) 2 (16.7) 1 (16.7)
Age (years) 57.22 <0.001
    18–24 81 (25.6) 79 (26.5) 0 (0) 2 (33.3)
    25–31 67 (21.2) 66 (22.1) 0 (0) 1 (16.7)
    32–38 58 (18.4) 58 (19.5) 0 (0) 0 (0)
    39–45 45 (14.2) 44 (14.8) 0 (0) 1 (16.7)
    46–51 25 (7.9) 22 (7.4) 3 (25.0) 0 (0)
    52+ 40 (12.7) 29 (9.7) 9 (75.0) 2 (33.3)
Caste 14.00 0.03
    Bahun 79 (25.0) 75 (25.2) 2 (16.7) 2 (33.3)
    Chhetri 153 (48.4) 149 (50.0) 2 (16.7) 2 (33.3)
    Dalit 75 (23.7) 65 (21.8) 8 (66.7) 2 (33.3)
    Janajati 9 (2.8) 9 (3.0) 0 (0) 0 (0)
Education 4.22 0.12
    No education 137 (43.4) 125 (41.9) 8 (66.7) 4 (66.7)
    Any education 179 (56.6) 173 (58.1) 4 (33.3) 2 (33.3)
Household income 4.65 0.10
    No income 142 (44.9) 130 (43.6) 9 (75.0) 3 (50.0)
    Any income 174 (55.1) 168 (56.4) 3 (25.0) 3 (50.0)
Stressful life events 0.08 0.96
    0–14 events 201 (63.6) 189 (63.4) 8 (66.7) 4 (66.7)
    ⩾15 events 115 (36.4) 109 (36.6) 4 (33.3) 2 (33.3)
Depression, BDIFootnote a 10.23 0.006
    Below cut-off 213 (67.4) 206 (69.1) 3 (25.0) 4 (66.7)
    Above cut-off 103 (32.6) 92 (30.9) 9 (75.0) 2 (33.3)
Anxiety, BAIFootnote b 6.37 0.04
    Below cut-off 230 (72.8) 220 (73.8) 5 (41.7) 5 (83.3)
    Above cut-off 86 (27.2) 78 (26.2) 7 (58.3) 1 (16.7)

BDI, Beck Depression Inventory; BAI, Beck Anxiety Inventory.

a. Cut-off is 20 or greater.

b. Cut-off is 17 or greater.

In the final analysis, logistic regression was used to determine the effects of pre-conflict depression, anxiety, education and income on post-conflict PTSD, when controlling for age, gender and caste. This was done because PTSD was not assessed in the pre-conflict period. P-values less than 0.05 were considered statistically significant unless otherwise noted. Statistical analyses were performed with SPSS v.16.0 for Windows.


Table 1 presents the baseline characteristics of the sample prior to conflict, categorised by follow-up status. Of the original 316 participants, 298 people were re-interviewed at follow-up (94% of the original participants). Of these 298 follow-up participants, 266 (89.3%) were residing in Jumla and 32 (10.7%) had moved to other Nepal districts. Twelve (3.8%) of the original 316 participants were deceased. Of the remaining six people (1.9%) who did not participate, one refused participation and the other five were lost to follow-up.

There were no significant differences in baseline characteristics between the six people who did not participate and those who did participate in the follow-up study. The three groups (follow-up, deceased and lost to follow-up) only differed in baseline characteristics for age, caste and mental health. Of the 12 participants who died, one was killed in crossfire between Maoists and the Nepal Army. The other 11 died from health problems, most commonly respiratory disease. Participants who died were more likely to be older, Dalit and have had depression (75%) and anxiety (58%) at the baseline pre-conflict assessment. When examining pre- to post-conflict changes in variables among all follow-up participants, the participants were more likely to have any education, any household income and a greater number of recent stressful events during the post-conflict assessment compared with the pre-conflict period (Table 2).

Depression and anxiety prevalence increased from the pre- to post-conflict assessment periods (Table 2). During the pre-conflict period, 92 of the 298 follow-up participants (30.9%) were above the cut-off for depression. At post-conflict assessment, 121 (40.6%) were above the cut-off for depression. For anxiety, 78 of the 298 participants (26.2%) were above the cut-off at baseline. At follow-up, 142 of the 298 participants (47.7%) were above the cut-off. The correlation between BDI measures pre- and post-conflict was 0.37 (P<0.001). For the BAI, the correlation between pre- and post-conflict measures was 0.34 (P<0.001).

TABLE 2 Comparison of characteristics between pre-conflict (year 2000) and post-conflict (year 2007) among follow-up participants (n = 298)

n (%)
Pre-conflict Post-conflict McNemar χ2, P
Education <0.001
    No education 125 (41.9) 92 (30.9)
    Any education 173 (58.1) 206 (69.1)
Household income <0.001
    No income 130 (43.6) 72 (24.2)
    Any income 168 (56.4) 226 (75.8)
Stressful life events <0.001
    0–14 events 189 (63.4) 106 (35.6)
    ⩾15 events 109 (36.6) 192 (64.4)
Depression, BDIFootnote a 0.005
    Below cut-off 206 (69.1) 177 (59.4)
    Above cut-off 92 (30.9) 121 (40.6)
Anxiety, BAIFootnote b <0.001
    Below cut-off 220 (73.8) 156 (52.3)
    Above cut-off 78 (26.2) 142 (47.7)

BDI, Beck Depression Inventory; BAI, Beck Anxiety Inventory.

a. Cut-off is 20 or greater.

b. Cut-off is 17 or greater.

TABLE 3 Generalised estimating equations for effect of assessment period and conflict exposure on depression and anxiety among follow-up participants (n = 298)Footnote a

Model 1 Model 2 Model 3 Model 4
OR (95% CI) P OR (95% CI) P OR (95% CI) P OR (95% CI) P
Depression, BDI Footnote b
Assessment period 0.009 0.22 0.66 0.97
    Pre-conflict (year 2000) 1 (Reference) 1 (Reference) 1 (Reference) 1 (Reference)
    Post-conflict (year 2007) 1.44 (1.09–1.91) 1.27 (0.87–1.84) 0.88 (0.50–1.55) 0.99 (0.54–1.79)
Conflict exposure N/A N/A 0.36 0.56
    0–3 events 1 (Reference) 1 (Reference)
    4–5 events 1.75 (0.89–3.45) 1.49 (0.75–2.95)
    6–7 events 1.42 (0.70–2.91) 1.32 (0.63–2.75)
    8–13 events 1.80 (0.81–4.00) 1.72 (0.76–3.89)
Anxiety, BAI Footnote c
Assessment period <0.001 <0.001 0.54 0.57
    Pre-conflict (year 2000) 1 (Reference) 1.0 (Reference) 1 (Reference) 1 (Reference)
    Post-conflict (year 2007) 2.46 (1.82–3.32) 3.04 (2.03–4.53) 1.22 (0.64–2.31) 1.21 (0.63–2.33)
Conflict exposure N/A N/A <0.001 0.001
    0–3 events 1 (Reference) 1 (Reference)
    4–5 events 2.32 (1.11–4.83) 1.96 (0.92–4.16)
    6–7 events 4.26 (1.93–9.41) 3.17 (1.41–7.23)
    8–13 events 6.64 (2.90–15.19) 5.73 (2.38–13.84)

BDI, Beck Depression Inventory; BAI, Beck Anxiety Inventory; N/A, not applicable.

a. All predictors are measured at pre- and post-conflict, except conflict exposure. Model 1: pre- v. post-conflict assessment period; Model 2: Model 1 plus age, caste, gender and education; Model 3: Model 2 plus conflict exposure; Model 4: Model 3 plus past year stressful life events and household income.

b. Cut-off is 20 or greater.

c. Cut-off is 17 or greater.

The odds of depression and anxiety increased from the pre- to post-conflict period (Table 3, Model 1). When using GEE to account for ageing of the population, there is not a significant effect of assessment period (pre- to post-conflict) on depression whereas the assessment period remains a significant predictor of anxiety (Table 3, Model 2). Figure 1 is a presentation of this effect of ageing for depression and anxiety. The increase in depression from pre- to post-conflict is no greater than what would be expected with ageing of the population alone. In contrast, increases in anxiety were greater than would be expected from ageing.

The next step was to evaluate whether conflict-related traumas can account for observed changes in anxiety. Participants experienced a range of exposures to political violence (Table 4). When conflict exposure was entered into the GEE, the effect of pre- v. post-conflict assessment period was no longer significant for anxiety (Table 3, Model 3), suggesting that conflict exposure accounts for the difference in anxiety levels between the two assessments. Moreover, this showed a dose–response effect of number of types of conflict events on anxiety. In the final GEE model, conflict exposure remained significant in a dose–response relationship for anxiety even when controlling for stressful life events in the past year and household income (Table 3, Model 4). Table 5 presents the final model including the effects of gender, age, caste, education, income and stressful life events. For depression, these factors are significant. For anxiety, gender, age, caste and stressful life events are significant in addition to conflict exposure. For both depression and anxiety, Dalit caste had the greatest odds of poor mental health compared with other ethnic/caste groups.

We evaluated whether any specific demographic group showed greater change in mental health outcomes between the pre- and post-conflict periods. The absolute increases were greater for anxiety v. depression for all groups (Fig. 2). Using GEE accounting for the two observations per person, interaction effects were examined in separate models for gender×exposure, age×exposure, and Dalit×exposure. In each of the separate models, interactions were not significant (P<0.05). Women, elderly people and marginalised caste groups did not show a greater increase in psychiatric morbidity compared with other demographic groups.

FIG. 1 (a) Depression (Beck Depression Inventory (BDI)) and (b) anxiety (Beck Anxiety Inventory (BAI)) by age at time of pre- v. post-conflict assessment.

There are no participants in the 18- to 24-year-old age group at the post-conflict assessment because they aged out of this cohort. Error bars represent standard error of the proportion above the cut-off. Cut-off for BDI is 20 or greater, and cut-off for BAI is 17 or greater. The ‘*’ refers to pre- v. post-conflict differences P<0.05 for the specific age group.

Post-traumatic stress disorder was assessed only during the post-conflict period. Based on the validated cut-off score of 50 or greater, 42 participants (14.1%) met criteria for PTSD. In a logistic regression including gender, age, caste, conflict exposure and pre-conflict measures of education, income, depression and anxiety, the only significant predictors were conflict exposure (P<0.05) and pre-conflict anxiety, with the latter showing a protective relationship for PTSD risk (odds ratio (OR) = 0.33, 95% CI 0.11–0.97).

TABLE 4 Conflict exposure among follow-up participants (n =298)

Exposure type n (%)
House searched by armed groupFootnote a 256 (85.9)
Witnessing someone killed by armed group 234 (78.5)
Forced to feed and shelter armed group 220 (73.8)
Witnessing someone beaten by armed group 213 (71.5)
Witnessing bomb explosion 191 (64.1)
Threatened by armed group for political involvement 121 (40.6)
Family member tortured by armed group 75 (25.2)
Forced into political involvement by armed group 64 (21.5)
Family member abducted by armed group 37 (12.4)
Property damaged in battle 35 (11.7)
Displaced due to conflict 34 (11.4)
Sexual violence perpetrated by armed group 26 (8.7)
Domestic violence perpetrated by family member in armed group 23 (7.7)
Family member killed by armed group 14 (4.7)
Total conflict exposure (quartiles)
    0–3 exposures 75 (25.2)
    4–5 exposures 85 (28.5)
    6–7 exposures 92 (30.9)
    8–13 exposures 46 (15.4)

a. ‘Armed group’ can refer to either government forces (Nepal Army and Armed Police Force) or Maoist forces (People's Liberation Army).


Main findings and their signficance

The goal of this study was to examine the effects of exposure to war on mental health using a prospective cohort design comparing prevalence of depression and anxiety prior to the outbreak of widespread violence v. after peace accords were signed. This is the first study to assess the same group of individuals before and after a violent conflict in a low-income country. Crude prevalence rates increased. Anxiety rose from 26.2 to 47.7%. Participants had nearly three times greater risk for anxiety during the post-conflict period compared with the pre-conflict period, even when controlling for ageing of the cohort. The number of conflict exposures had a dose–response effect on anxiety. When controlling for pre- to post-conflict changes in recent stressful life events, conflict exposure continued to predict anxiety in a dose–response manner.

Depression rose from 30.9 to 40.6%. In contrast to anxiety, the increase in depression was attributable to ageing of the population but was not associated with war trauma. Regarding reputed vulnerable groups, we found that women, elderly people and those from marginalised caste groups were not more susceptible to increases in either depression or anxiety because of the conflict. These groups had a high symptom burden prior to conflict exposure. In summary, exposure to political violence led to an increase in anxiety but not depression in a rural community in Nepal, and the effects of political violence on psychiatric morbidity were felt equally across demographic groups.

These findings challenge some generalisations about the effects of war on mental health. Counter to the assumption that societies have low psychiatric morbidity prior to war, Reference Murthy and Lakshminarayana3 this study suggests that (a) post-conflict societies may have had high rates of mental health problems prior to conflict exposure, especially in marginalised communities such as this study site, (b) exposure to conflict may play a dominant role in anxiety prevalence but may not be the primary risk factor for other forms of psychiatric morbidity such as mood disorders, which may be influenced more by ongoing non-war risk factors such as poverty and ethnic/caste-based discrimination, and (c) putative risk groups may not show a greater rise in psychiatric problems as a result of conflict exposure but instead may have chronically elevated risks observable even during peacetime.

FIG. 2 (a) Depression (Beck Depression Inventory (BDI)) and (b) anxiety (Beck Anxiety Inventory (BAI)) by demographic group pre- and post-conflict.

Error bars represent standard error of the proportion above the cut-off. Cut-off for BDI is 20 or greater, and cut-off for BAI is 17 or greater. The ‘*’ refers to pre- v. post-conflict differences P<0.05 for the specific demographic group.

Our study also suggests different trajectories and risk factors for mood v. anxiety disorders. Studies in Algeria and Afghanistan also have demonstrated trauma exposure associated with increased risk of anxiety but not depression. Reference de Jong, Komproe and Van Ommeren31,Reference Lopes Cardozo, Bilukha, Crawford, Shaikh, Wolfe and Gerber32 In a study of 2000 survivors of 9/11, there were different profiles for depression including one group characterised by chronic depression with risk factors including high stressors, low social support, a lifetime history of depression and non-terror chronic traumas; Reference Nandi, Tracy, Beard, Vlahov and Galea33 this depression subgroup parallels the profile of depression in our Nepal sample. Furthermore, there is increasing evidence for differing trajectories and risk factors for mood and anxiety disorders in non-conflict affected populations. Reference Penninx, Nolen, Lamers, Zitman, Smit and Spinhoven34,Reference Nandi, Beard and Galea35 Therefore, whereas others have argued that the distinction between anxiety and depression in low-income settings is not clinically relevant, Reference Patel36 the differences in trajectory suggest some divergence in public health prevention strategies for mood and anxiety disorders.

TABLE 5 Generalised estimating equations for predictors of depression and anxiety among follow-up participants (n =298)Footnote a

Depression, Beck Depression InventoryFootnote b Anxiety, Beck Anxiety InventoryFootnote c
OR (95% CI) P OR (95% CI) P
Gender <0.001 <0.001
    Male 1 (Reference) 1 (Reference)
    Female 3.02 (1.85–4.92) 4.55 (2.67–7.74)
Age, years <0.001 <0.001
    25–31 1 (Reference) 1 (Reference)
    32–38 2.05 (0.90–4.64) 0.44 (0.21–0.92)
    39–45 3.56 (1.57–8.06) 1.12 (0.51–2.47)
    46–51 7.60 (2.94–19.68) 1.86 (0.80–4.33)
    52–80 10.91 (4.58–25.99) 2.81 (1.28–6.20)
Caste 0.04 <0.001
    Chhetri 1 (Reference) 1 (Reference)
    Dalit 1.69 (1.01–2.83) 5.80 (3.44–9.78)
    Bahun 1.06 (0.65–1.75) 2.51 (1.55–4.06)
    Janajati 0.10 (0.01–0.92) 0.11 (0.01–1.19)
Education 0.03 0.26
    Any education 1 (Reference) 1 (Reference)
    No education 1.72 (1.07–2.76) 1.33 (0.81–2.20)
Household cash income <0.001 0.44
    Any income 1 (Reference) 1 (Reference)
    No income 2.49 (1.65–3.76) 1.19 (0.77–1.85)
Stressful life events in past year 0.005 <0.001
    0–14 events 1 (Reference) 1 (Reference)
    ⩾15 events 1.86 (1.20–2.88) 2.55 (1.60–4.05)
Conflict exposure 0.56 0.001
    0–3 events 1 (Reference) 1 (Reference)
    4–5 events 1.49 (0.75–2.95) 1.96 (0.92–4.16)
    6–7 events 1.32 (0.63–2.75) 3.17 (1.41–7.23)
    8–13 events 1.72 (0.76–3.89) 5.73 (2.38–13.84)
Assessment period 0.97 0.57
    Pre-conflict (year 2000) 1 (Reference) 1 (Reference)
    Post-conflict (year 2007) 0.99 (0.54–1.79) 1.21 (0.63–2.33)

a. All predictors are measured at pre- and post-conflict, except conflict exposure.

b. Cut-off is 20 or greater.

c. Cut-off is 17 or greater.

Strengths and limitations

The strengths of this study include the use of validated instruments, the ability to employ a prospective sample, and a follow-up rate of 98% of living original participants. Regarding limitations, it is difficult to specify what qualifies as ‘pre-conflict’ mental health. As with most settings of political violence, the historical date a war begins, the date violence reaches a community and the time when a community suffers the psychological sequelae of political violence all may differ. Even though direct violence did not occur in Jumla before this study was conducted, anticipatory concerns cannot be excluded. Anxiety rates may have been lower in this community prior to the war's officially recognised start date in 1996. There is likely a gradient ranging from anticipating conflict to actually experiencing violence. The increase in anxiety that occurred after direct violence reached the community in 2001 and the dose–response pattern suggest that direct exposure to violence has an added effect upon anxiety beyond possible pre-violence anticipatory anxiety. Ceiling effects also may have influenced the findings in that high levels of depression prior to the conflict may have precluded observing a significant increase at the post-conflict assessment. A major limitation is that PTSD was not assessed in 2000, so the change in prevalence attributable to the conflict could not be quantified.


These findings have implications for global mental healthcare. Whereas high levels of mental health problems alone demand attention and intervention, prevention and palliative treatment will differ depending on the nature of war-related factors v. other risk factors. Reference de Jong37Reference Betancourt, Agnew-Blais, Gilman, Williams and Ellis39 More detailed knowledge of the impact of political violence v. other chronic social problems can lead to better-informed interventions in post-conflict settings with scarce resources. The identification of mental health problems in these settings preceding conflict demonstrates the need for investment in mental healthcare infrastructure and other psychosocial services in impoverished communities. Reference Patel and Thornicroft40 Social marginalisation can be as damaging to mental health as war trauma: being from a low caste Dalit group has the same effect size (OR = 5.80) as having experienced conflict exposure at the highest quartile (OR = 5.73) for anxiety. Therefore, although trauma-specific interventions may be helpful for anxiety, social interventions to reduce and prevent depression and anxiety in this setting should focus on education, poverty reduction, minimising stressful life events, improving healthcare and reducing other risk factors associated with ageing, low caste and female gender.

Mounting evidence suggests that the impact on mental health of war is not inevitable, but rather intimately related to the social, economic and cultural conditions that precede and follow violent conflict. Reference Pedersen5,Reference Miller and Rasmussen38,Reference Betancourt, Agnew-Blais, Gilman, Williams and Ellis39,Reference Porter and Haslam41 These suppositions are in no way intended to diminish the emphasis placed upon the suffering caused by political violence. Rather, the goal is to increase attention to the equally damaging forces of chronic injustice in the form of ongoing financial, social and health-related threats that erode everyday mental health. Moreover, pre-conflict marginalisation may increase vulnerability to exploitation by militant groups. Maoists used the promise of eradicating social inequities to recruit adults and children into their military. Reference Tol, Kohrt, Jordans, Thapa, Pettigrew and Upadhaya14,Reference Kohrt, Tol, Pettigrew, Karki, Singer and Hodge42 This leads to a final important observation of this study: psychiatric morbidity did not decrease for any demographic group. This challenges suggestions, such as those proffered by militant revolutionary movements, that violent uprising is a form of psychological emancipation. Ultimately, addressing risk factors for poor mental health such as poverty, lack of education, inadequate healthcare and gender- and ethnic/caste-based discrimination, in addition to trauma healing, may not only ameliorate mental health problems, but also help to reduce vulnerability to exploitation and involvement in political violence.


Funding was provided through a National Institute of Mental Health – National Research Service Award (NIMH-NRSA F31 MH075584), a Wenner-Gren Dissertation Fieldwork Grant, a Fulbright Fellowship and the Graduate School of Arts and Sciences of Emory University.


Renu Shrestha, Krishna Maya Neupane, Mukunda Chaulagain, Indra Rai, Diya Khatri, Raj Budthapa, Lok Bahadur Rawal, Khem Singh Kathaya, Chandra Devkota, and Sabitri Devkota conducted field research in Jumla. Thanks to Ganesh Rokaya and his family for hosting the research team and to 4S Nepal and the Jumla Hospital staff for providing logistical assistance. Thanks to Peter Brown, Mel Konner, and Ian Harper for reading earlier manuscript drafts. Special thanks to Christina Chan.


Paper presented in part as the John P. Spiegel Fellowship Lecture for the Society for the Study of Psychiatry and Culture 2011 Annual Meeting, Seattle, Washington, USA.

See editorial, pp. 255–257, this issue.

Declaration of interest



1 Tol, WA, Barbui, C, Galappatti, A, Silove, D, Betancourt, TS, Souza, R, et al. Mental health and psychosocial support in humanitarian settings: linking practice and research. Lancet 2011; 378: 1581–91.Google Scholar
2 Steel, Z, Chey, T, Silove, D, Marnane, C, Bryant, RA, van Ommeren, M. Association of torture and other potentially traumatic events with mental health outcomes among populations exposed to mass conflict and displacement: a systematic review and meta-analysis. JAMA 2009; 302: 537–49.Google Scholar
3 Murthy, RS, Lakshminarayana, R. Mental health consequences of war: a brief review of research findings. World Psychiatry 2006; 5: 2532.Google Scholar
4 Desjarlais, RR, Eisenberg, L, Good, B, Kleinman, AM. World Mental Health: Problems and Priorities in Low-Income Countries. Oxford University Press, 1995.Google Scholar
5 Pedersen, D. Political violence, ethnic conflict, and contemporary wars: broad implications for health and social well-being. Soc Sci Med 2002; 55: 175–90.Google Scholar
6 de Jong, JTVM. Public mental health, traumatic stress, and human rights violations in low-income countries. In Trauma, War, and Violence: Public Mental Health in Socio-cultural Context (ed de Jong, JTVM): 191. Kluwer Academic/Plenum Publishers, 2002.Google Scholar
7 Patel, V, Kirkwood, BR, Pednekar, S, Pereira, B, Barros, P, Fernandes, J, et al. Gender disadvantage and reproductive health risk factors for common mental disorders in women. Arch Gen Psychiatry 2006; 63: 404–13.Google Scholar
8 Das, J, Do, Q-T, Friedman, J, McKenzie, D, Scott, K. Mental health and poverty in developing countries: revisiting the relationship. Soc Sci Med 2007; 65: 467–80.Google Scholar
9 World Bank. Nepal at a Glance. World Bank, 2007.Google Scholar
10 Thapa, D, Sijapati, B. A Kingdom under Siege: Nepal's Maoist Insurgency, 1996 to 2003 (updated edn). The Printhouse, 2004.Google Scholar
11 Kohrt, BA. Political Violence and Mental Health in Nepal: War in Context, Structural Violence, and the Erasure of History (PhD Dissertation). Anthropology Department, Emory University, 2009.Google Scholar
12 Informal Sector Service Centre (INSEC). Fact Sheet: Number of Children Killed by State and Maoists. INSEC, 2008.Google Scholar
13 Watchlist. Caught in the Middle: Mounting Violations Against Children in Nepal's Armed Conflict. Watchlist on Children and Armed Conflict, 2005.Google Scholar
14 Tol, WA, Kohrt, BA, Jordans, MJD, Thapa, SB, Pettigrew, J, Upadhaya, N, et al. Political violence and mental health: a multi-disciplinary review of the literature on Nepal. Soc Sci Med 2010; 70: 3544.Google Scholar
15 Kohrt, BA, Kunz, RD, Baldwin, JL, Koirala, NR, Sharma, VD, Nepal, MK. ‘Somatization’ and ‘comorbidity’: a study of Jhum-Jhum and depression in rural Nepal. Ethos 2005; 33: 125–47.Google Scholar
16 Kohrt, BA, Speckman, RA, Kunz, RD, Baldwin, JL, Upadhaya, N, Acharya, NR, et al. Culture in psychiatric epidemiology: using ethnography and multiple mediator models to assess the relationship of caste with depression and anxiety in Nepal. Ann Hum Bio 2009; 36: 261–80.Google Scholar
17 Kohrt, BA, Worthman, CM. Gender and anxiety in Nepal: the role of social support, stressful life events, and structural violence. CNS Neurosci Ther 2009; 15: 237–48.Google Scholar
18 Kohrt, BA. Vulnerable social groups in post-conflict settings: a mixed-methods policy analysis and epidemiology study of caste and psychological morbidity in Nepal. Interv 2009; 7: 239–64.Google Scholar
19 Beck, AT, Ward, CH, Mendelson, M, Mock, J, Erbaugh, J. An inventory for measuring depression. Arch Gen Psychiatry 1961; 4: 561–71.Google Scholar
20 Beck, AT, Epstein, N, Brown, G, Steer, RA. An inventory for measuring clinical anxiety: psychometric properties. J Consult Clin Psychol 1988; 56: 893–7.Google Scholar
21 Kohrt, BA, Kunz, RD, Koirala, NR, Sharma, VD, Nepal, MK. Validation of a Nepali version of the Beck Depression Inventory. Nepal J Psychiatry 2002; 2: 123–30.Google Scholar
22 Kohrt, BA, Kunz, RD, Koirala, NR, Sharma, VD, Nepal, MK. Validation of the Nepali Version of the Beck Anxiety Inventory. J Instit Med 2003; 25: 14.Google Scholar
23 American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorder (4th edn) (DSM-IV). APA, 1994.Google Scholar
24 Weathers, FW, Litz, BT, Herman, DS, Huska, JA, Keane, TM. The PTSD Checklist (PCL): Reliability, Validity, and Diagnostic Utility. Paper presented at: 9th Annual Meeting of International Society for Traumatic Stress Studies, October 1993 ( Scholar
25 Tol, WA, Komproe, IH, Thapa, SB, Jordans, MJD, Sharma, B, De Jong, JTVM. Disability associated with psychiatric symptoms among torture survivors in rural Nepal. J Nerv Ment Dis 2007; 195: 463–9.Google Scholar
26 Thapa, SB, Hauff, E. Psychological distress among displaced persons during an armed conflict in Nepal. Soc Psychiatry Psychiatr Epidemiol 2005; 40: 672–9.Google Scholar
27 Li, C, Ma, A, Fu, L. Investigation of stressful life events and mental health in railway workers. Chin Ment Health J 1994; 8: 4952.Google Scholar
28 Zheng, Y-P, Lin, K-M. A nationwide study of stressful life events in Mainland China. Psychosom Med 1994; 56: 296305.Google Scholar
29 Kohrt, BA, Hruschka, DJ, Kohrt, HE, Panebianco, NL, Tsagaankhuu, G. Distribution of distress in post-socialist Mongolia: a cultural epidemiology of yadargaa. Soc Sci Med 2004; 58: 471–85.Google Scholar
30 Ballinger, GA. Using generalized estimating equations for longitudinal data analysis. Organiz Res Method 2004; 7: 127–50.Google Scholar
31 de Jong, JTVM, Komproe, IH, Van Ommeren, M. Common mental disorders in postconflict settings. Lancet 2003; 361: 2128–30.Google Scholar
32 Lopes Cardozo, BL, Bilukha, OO, Crawford, CA, Shaikh, I, Wolfe, MI, Gerber, ML, et al. Mental health, social functioning, and disability in postwar Afghanistan. JAMA 2004; 292: 575–84.Google Scholar
33 Nandi, A, Tracy, M, Beard, JR, Vlahov, D, Galea, S. Patterns and predictors of trajectories of depression after an urban disaster. Ann Epidemiol 2009; 19: 761–70.Google Scholar
34 Penninx, BWJH, Nolen, WA, Lamers, F, Zitman, FG, Smit, JH, Spinhoven, P, et al. Two-year course of depressive and anxiety disorders: results from the Netherlands Study of Depression and Anxiety (NESDA). J Affect Disord 2011; 133: 7685.Google Scholar
35 Nandi, A, Beard, JR, Galea, S. Epidemiologic heterogeneity of common mood and anxiety disorders over the lifecourse in the general population: a systematic review. BMC Psychiatry 2009; 9: 31.Google Scholar
36 Patel, V. Cultural factors and international epidemiology. Br Med Bull 2001; 57: 3345.Google Scholar
37 de Jong, JTVM. Trauma, War, and Violence: Public Mental Health in Socio-Cultural Context. Kluwer Academic/Plenum Publishers, 2002.Google Scholar
38 Miller, KE, Rasmussen, A. War exposure, daily stressors, and mental health in conflict and post-conflict settings: bridging the divide between trauma-focused and psychosocial frameworks. Soc Sci Med 2010; 70: 716.Google Scholar
39 Betancourt, TS, Agnew-Blais, J, Gilman, SE, Williams, DR, Ellis, BH. Past horrors, present struggles: the role of stigma in the association between war experiences and psychosocial adjustment among former child soldiers in Sierra Leone. Soc Sci Med 2010; 70: 1726.Google Scholar
40 Patel, V, Thornicroft, G. Packages of care for mental, neurological, and substance use disorders in low- and middle-income countries. PLoS Med 2009; 6: e1000160.Google Scholar
41 Porter, M, Haslam, N. Predisplacement and postdisplacement factors associated with mental health of refugees and internally displaced persons: a meta-analysis. JAMA 2005; 294: 602–12.Google Scholar
42 Kohrt, BA, Tol, WA, Pettigrew, J, Karki, R. Children and revolution: the mental health and psychosocial wellbeing of child soldiers in Nepal's Maoist army. In The War Machine and Global Health (eds Singer, M, Hodge, GD): 89116. Altamira Press, 2010.Google Scholar
Figure 0

TABLE 1 Baseline characteristics of all participants (pre-conflict, year 2000) (n =316)

Figure 1

TABLE 2 Comparison of characteristics between pre-conflict (year 2000) and post-conflict (year 2007) among follow-up participants (n = 298)

Figure 2

TABLE 3 Generalised estimating equations for effect of assessment period and conflict exposure on depression and anxiety among follow-up participants (n = 298)a

Figure 3

FIG. 1 (a) Depression (Beck Depression Inventory (BDI)) and (b) anxiety (Beck Anxiety Inventory (BAI)) by age at time of pre- v. post-conflict assessment.There are no participants in the 18- to 24-year-old age group at the post-conflict assessment because they aged out of this cohort. Error bars represent standard error of the proportion above the cut-off. Cut-off for BDI is 20 or greater, and cut-off for BAI is 17 or greater. The ‘*’ refers to pre- v. post-conflict differences P<0.05 for the specific age group.

Figure 4

TABLE 4 Conflict exposure among follow-up participants (n =298)

Figure 5

FIG. 2 (a) Depression (Beck Depression Inventory (BDI)) and (b) anxiety (Beck Anxiety Inventory (BAI)) by demographic group pre- and post-conflict.Error bars represent standard error of the proportion above the cut-off. Cut-off for BDI is 20 or greater, and cut-off for BAI is 17 or greater. The ‘*’ refers to pre- v. post-conflict differences P<0.05 for the specific demographic group.

Figure 6

TABLE 5 Generalised estimating equations for predictors of depression and anxiety among follow-up participants (n =298)a

Submit a response


No eLetters have been published for this article.