The effects of war extend beyond individuals to communities and societies. It has been suggested that group-level determinants of mental health may be important in low-income countries recovering from conflict.1,2 However, the empiric evidence to support this hypothesis is limited. Such evidence would be relevant given the large number of conflicts and their preponderance in low-income countries. Since 1989, 35 countries are recovering from conflicts, and nearly two-thirds of these are low-income.3
The Republic of Liberia is a low-income country on the West African coast recovering from 14 years of civil war that ended in 2003. The conflict is estimated to have claimed 250,000 lives and displaced more than one-third of the nation's 2.5 million inhabitants.4 Fighting initially began in northern Nimba County, where members of the national military engaged a rebel movement. Most of the fighting in Nimba occurred between December 1989 and April 1990, and was characterized by ethnically motivated violence against civilians.5
In this study—informed by published theories about group-level determinants of mental health6–8—we apply multilevel modeling techniques to explore potential associations of village-level experiences of conflict, wealth and education inequality, and ethnic and religious diversity, with the risk of posttraumatic-stress symptoms.
Study Design and Fielding
We employed a 3-stage representative rural cluster sample of households in Nimba County. Data from the Republic of Liberia 2008 National Census were used as the primary sampling frame. In the first stage, we selected 50 census enumeration areas with probability proportional to population size. Next, 30 households were randomly selected from full household listings for each selected enumeration area. Finally, a Kish Table was employed to randomly select a respondent from all eligible individuals in each sampled household.9 Eligible individuals were all those over the age of 18 who resided in the selected household. This method is consistent with a large body of sampling literature, as well as with field implementations and results in a population-representative sample.10,11
We assessed traumatic experiences and posttraumatic-stress symptoms with the Harvard Trauma Questionnaire, which has been used widely in low-resource country settings.12,13 A normalized measure of traumatic experiences was constructed by dividing individual values by the sample standard deviation. The dependent variable of interest, posttraumatic-stress-symptom score, was calculated as the average response to 16 symptom questions. A relative index of household wealth status was constructed using principal-component analysis of household assets.14 For analysis purposes, households were divided into quintiles based on the wealth index.
We were interested in 3-key village-level constructs. (1) Experiences of conflict: A village-level measure of traumatic experiences was constructed as the mean number of individual traumatic experiences reported within a village, normalized by dividing by the sample standard deviation. Similarly, a history-of-displacement measure was constructed as the proportion of persons within a village who reported having been displaced. (2) Wealth and education inequality: Village wealth inequality was determined as the standard deviation of the household wealth index for a given village divided by the standard deviation for the entire sample.15 Village education inequality was determined similarly, using years of education. (3) Ethnic and religious diversity: Village ethnic and religious diversity were determined using fractionalization indices.16 History-of-displacement, inequality, and diversity variables were divided into quartiles, and the results are reported as the highest or lowest quartile versus all others. We also took into account potential village-level confounders, including proximity to a primary road, presence of a health facility, mean asset index, and percent of village attending any school.
Univariate statistics were calculated for individual- and village-level variables. We fit 2 multilevel random-intercept linear-regression models to estimate associations between select covariates and symptom score. All statistical analyses were carried out with Stata software (version 10.0; Stata Corp, College Station, TX).
Of 1464 eligible respondents, 1434 (98%) completed the questionnaire. The average age of respondents was 40 years (SD = 16.3) (Table 1). Fewer than half (46%) of those interviewed were women, and 42% of respondents had never attended school. Respondents reported an average of 17 lifetime traumatic experiences, and more then two-thirds (68%) reported having been displaced during the war. The mean (SD) posttraumatic-stress-symptom score was 1.9 (0.6).
The results of 2 random-intercept models are given in Table 2. The conditional intraclass correlation for the null model was 0.23. In the full model (Model 2), a change in one standard deviation in reported traumatic experiences was associated with a 0.28 unit higher symptom score (95% confidence interval [CI] = 0.25 to 0.31), adjusting for covariates. Persons who were displaced during the war had on average a symptom score 0.09 units higher than persons who were not displaced (CI = 0.03 to 0.15).
A change in one standard deviation in village-level traumatic experiences was associated with a 0.16 unit higher posttraumatic-stress-symptom score (0.06 to 0.26). Persons residing in villages with the greatest history of displacement during the war had on average a symptom score 0.17 units higher than persons in villages that experienced less displacement (0.03 to 0.31). Conversely, persons residing in villages with a more equal household wealth distribution had on average a symptom score 0.13 units lower than persons in more unequal villages (−0.26 to 0.00). The conditional intraclass correlation of the full model was 0.09, and the R-squared value was 0.36.
In a population-representative sample of residents of Nimba County, we found that village-level factors were associated with posttraumatic-stress symptoms in the postconflict setting, over and above the contribution of individual-level factors. Village-level experiences of conflict, including greater levels of traumatic experiences and history of displacement, were associated with increased symptoms. It has been shown that higher group-level exposure to violence is associated with lower social capital,17 the latter being associated with poor mental health.18 These findings suggest that the observed associations may be mediated by measures of social capital. Persons living in villages with more equal household wealth distribution had fewer symptoms. Although this is a novel finding in a low-income country, it is consistent with a study conducted in New York City neighborhoods in the aftermath of the 11 September 2001 terrorist attacks.19 Inequalities may exacerbate the deleterious impact of trauma on social cohesion, the strength of which has been shown to be associated with mental health, particularly in low-resource settings.20 Therefore, the observed association between wealth inequality and posttraumatic-stress symptoms may be mediated by measures of social cohesion. Village ethnic and religious diversity were not associated with symptoms. This is in contrast to recent findings that suggest increased diversity is associated with increased psychopathology, primarily as a result of social discrimination.8 With regard to ethnic diversity, this discrepancy may be explained by the fact that the dominant ethnic groups in Nimba—the Mano and the Gio—are both minorities in Liberia, and may have experienced similar discrimination.21
The individual-level associations documented here are consistent with the available literature and confirm the centrality of individual experiences as determinants of posttraumatic stress symptoms. Namely, individuals with a greater number of traumatic experiences had higher posttraumatic stress symptoms,12,22 as did persons who were displaced during the war.23
While it is difficult to attribute a precise practical effect size to changes in posttraumatic-stress-symptom score, the measures of association between village-level experiences of conflict and symptom score are of the same order of magnitude as those between individual-level experiences of conflict and symptom score. This suggests that these village-level effects are substantial, given the well-established importance of individual-level trauma in determining posttraumatic-stress symptoms.24,25
There were important limitations to this study. First, the multilevel modeling techniques applied here to cross-sectional data could potentially introduce exchangeability issues resulting from differential reporting by included covariates.26,27 In particular, respondents' willingness to report individual-level traumatic experiences may have been associated with the social acceptability of such an acknowledgment, a construct likely captured in the measure of village-level traumatic experiences. In addition, the data are retrospective, with some of the recorded traumatic experiences likely occurring 20 years prior to interviews, and thus recall bias may be an issue. However, we note that the majority of the variance in the reported number of traumatic experiences results from events individuals are unlikely to forget, including torture, imprisonment, and death of a family member (eAppendix, http://links.lww.com/EDE/A392). Finally, no information on individual postdisplacement relocation nor in- and out-migration from sampled villages was collected. Such population movements, if associated with covariates of interest and the outcome, may have introduced bias.
In summary, our findings suggest that characteristics of communities, as well as of individuals, may be important determinants of the high burden of posttraumatic stress in low-income countries such as Liberia that are recovering from conflict. Future longitudinal studies may be helpful in clarifying whether changing community characteristics are temporally associated with changes in mental health, and whether village-level interventions can then be useful to mitigate the psychological consequences of conflict.
1.Summerfield D. Conflict and health: War and mental health: a brief overview. BMJ
2.Silove D, Ekblad S, Mollica R. The rights of the severely mentally ill in post-conflict societies. Lancet
3.Human Security Report Project. miniAtlas of Human Security. Vancouver, BC Canada: Human Security Report Project; 2008.
4.National Transitional Government of Liberia. Joint Needs Assessment. Monrovia, Liberia: National Transitional Government of Liberia; 2004.
5.Bourke Gerald. “Thousands Flee Massacres by Liberian Army: President Doe's Campaign Against Rebels in the Bush has Turned Into Genocide Against a Rival Tribe, Gerald Bourke Writes From Sanniquellie.” The Independent. January 29, 1990. Available at: http://www.Lexis-Nexis.com/Universe
. General News Topics. Accessed February 10, 2009.
6.Kawachi I, Subramanian SV. Measuring and modeling the social and geographic context of trauma: a multilevel modeling approach. J Trauma Stress
7.Patel V, Kleinman A. Poverty and common mental disorders in developing countries. Bull World Health Organ
8.Williams DR, Neighbors HW, Jackson JS. Racial/ethnic discrimination and health: findings from community studies. Am J Public Health
. 2008;98(suppl 9):S29–S37.
9.Kish L. A procedure for objective respondent selection within the household. J Am Stat Assoc
10.Demographic and Health Surveys:Sampling Manual
. Calverton, MD: Macro International Inc; 1996.
11.United Nations Statistical Division. Household Sample Surveys in Developing and Transition Countries
. New York: United Nations Publication; 2005.
12.Scholte WF, Olff M, Ventevogel P, et al. Mental health symptoms following war and repression in eastern Afghanistan. JAMA
13.Roberts B, Damundu EY, Lomoro O, Sondorp E. Post-conflict mental health needs: a cross-sectional survey of trauma, depression and associated factors in Juba, Southern Sudan. BMC Psychiatry
14.Filmer D, Pritchett LH. Estimating wealth effects without expenditure data—or tears: an application to educational enrollments in states of India. Demography
15.McKenzie D. Measuring inequality with asset indicators. J Popul Econ
16.Vigdor JL. Interpreting ethnic fragmentation effects. Econ Lett
17.Colletta NJ, Cullen ML. ViolentConflictand theTransformation of Social Capital: Lessonsfrom Cambodia, Rwanda, Guatemala, and Somalia
. Washington, DC: World Bank; 2000.
18.De Silva MJ, McKenzie K, Harpham T, Huttly SR. Social capital and mental illness: a systematic review. J Epidemiol Community Health
19.Ahern J, Galea S. Social context and depression after a disaster: The role of income inequality. J Epidemiol Community Health
20.Fone D, Dunstan F, Lloyd K, Williams G, Watkins J, Palmer S. Does social cohesion modify the association between area income deprivation and mental health? A multilevel analysis. Int J Epidemiol
21.Ellis S. The Mask of Anarchy: The Destruction of Liberia and the Religious Roots of an African Civil War
. London, United Kingdom: Hurst & Co Ltd; 2001.
22.Roberts B, Ocaka KF, Browne J, Oyok T, Sondorp E. Factors associated with post-traumatic stress disorder and depression amongst internally displaced persons in northern Uganda. BMC Psychiatry
23.Thomas SL, Thomas SD. Displacement and health. Br Med Bull
24.Mollica RF, McInnes K, Poole C, Tor S. Dose-effect relationships of trauma to symptoms of depression and post-traumatic stress disorder among Cambodian survivors of mass violence. Br J Psychiatry
25.de Jong JT, Komproe IH, Van Ommeren M, et al. Lifetime events and posttraumatic stress disorder in 4 postconflict settings. JAMA
26.Duncan C, Jones K, Moon G. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med
27.Oakes JM. The (mis)estimation of neighborhood effects: causal inference for a practicable social epidemiology. Soc Sci Med