Sierra Leone's Former Child Soldiers: A Longitudinal Study of Risk, Protective Factors, and Mental Health

NEW RESEARCH

Sierra Leone’s Former Child Soldiers:A Longitudinal Study of Risk, Protective

Factors, and Mental HealthTheresa S. Betancourt, Sc.D., M.A., Robert T. Brennan, Ed.D., Ed.M.,

Julia Rubin-Smith, M.S.P.H., Garrett M. Fitzmaurice, Sc.D., Stephen E. Gilman, Sc.D.

Objective: To investigate the longitudinal course of internalizing and externalizing problemsand adaptive/prosocial behaviors among Sierra Leonean former child soldiers and whetherpostconflict factors contribute to adverse or resilient mental health outcomes. Method: Maleand female former child soldiers (N � 260, aged 10 to 17 years at baseline) were recruited fromthe roster of an non-governmental organization (NGO)-run Interim Care Center in KonoDistrict and interviewed in 2002, 2004, and 2008. The retention rate was 69%. Linear growthmodels were used to investigate trends related to war and postconflict experiences. Results:The long-term mental health of former child soldiers was associated with war experiences andpostconflict risk factors, which were partly mitigated by postconflict protective factors.Increases in externalizing behavior were associated with killing/injuring others during thewar and postconflict stigma, whereas increased community acceptance was associated withdecreases in externalizing problems (b � �1.09). High baseline levels of internalizing problemswere associated with being raped, whereas increases were associated with younger involve-ment in armed groups and social and economic hardships. Improvements in internalizingproblems were associated with higher levels of community acceptance and increases incommunity acceptance (b � �0.86). Decreases in adaptive/prosocial behaviors were associ-ated with killing/injuring others during the war and postconflict stigma, but partiallymitigated by social support, being in school and increased community acceptance (b �1.93). Conclusions: Psychosocial interventions for former child soldiers may be more effective if they account for postconflict factors in addition to war exposures. Youth withaccumulated risk factors, lack of protective factors, and persistent distress should be identified.Sustainable services to promote community acceptance, reduce stigma, and expand socialsupports and educational access are recommended. J. Am. Acad. Child Adolesc. Psychiatry,2010;49(6):606–615. Key Words: child soldiers, internalizing problems, externalizing prob-lems, prosocial behaviors, longitudinal study

iafymcalehi

m

G lobally, an estimated 300,000 childrenunder the age of 18 years are involvedwith armed forces and armed groups.1,2

Despite the documentation of risks facing childsoldiers because of war-related violence,3 little isknown about what influences long-term mentalhealth trajectories and processes of social reinte-gration. Recent studies on former child soldiersfrom northern Uganda,4 the Democratic Republicof the Congo,5 and Nepal6 have provided insightinto the impact of war experience on reintegra-tion and psychosocial adjustment. Research doc-uments that witnessing, experiencing, and per-
petrating violence, as well as younger age of p
JOURN

606 www.jaacap.org

nvolvement and longer engagement with anrmed group all have negative consequencesor the mental health and social reintegration ofoung people.5-7 Although loss and displace-ent are common adversities confronted by all

onflict-affected children,8 child soldiers facedditional risks such as exceptionally pro-onged and intense exposure to violence.9 Thisxposure can include being forced to kill orarm others1,10 and repeated personal victim-

zation, including sexual violence.10

Prior studies have documented high rates ofental health problems, such as PTSD and de-

ression, among former child soldiers.4,11 How-

AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 49 NUMBER 6 JUNE 2010

ItTtpyFtyaba5wo(ti5b(owaciiwsoo

djlstIaIMHerDdwmbmf

SILmt

SIERRA LEONE’S FORMER CHILD SOLDIERS

ever, wartime exposures alone do not account forthe elevated burden of mental health problems inthese young people,6,7,12,13 raising the question ofhow postconflict factors may contribute to vary-ing degrees of vulnerability to adverse outcomes.One longitudinal study documented that post-conflict experiences such as family support andeconomic opportunity played a role in the mentalhealth of 39 Mozambican males reinterviewed 16years after reintegration.12 More recently, innorthern Uganda, research by Blattman and An-nan7 emphasized that widespread educationaland economic deprivation contribute signifi-cantly to adverse outcomes in former child sol-diers. In Nepal, Kohrt et al. concluded that post-conflict factors such as stigma might contributeto adverse mental health outcomes. Former childsoldiers in his sample showed significantlyhigher symptoms of depression and PTSD com-pared with matched controls even after adjustingfor exposure to traumatic events.6 Similarly, Bet-ancourt et al., in Sierra Leone, observed thatpostconflict experiences of discrimination maysignificantly explain the relationship betweenpast involvement in wounding/killing othersand subsequent increases in hostility. Stigmaalso mediated the relationship between beingraped and increases in depression symptomsover a 2-year follow-up period.13

The present study builds on prior research byinvestigating the role of postconflict risk andprotective factors in the relationship betweenwar experiences and mental health in a cohort ofmale and female former child soldiers followedprospectively over three time points. We hypoth-esized that ongoing risk factors such as stigmaand daily hardships would contribute to poormental health outcomes, whereas protective fac-tors such as being in school, working, or experi-encing social support and community acceptancewould contribute to improved mental healthoutcomes.

METHODStudy Cohort and ProceduresThis prospective longitudinal study was conducted incollaboration with the International Rescue Committee(IRC) and the Post-conflict Reintegration Initiative forDevelopment and Empowerment (PRIDE). Survey in-terviews were conducted at three time points: T1(2002), T2 (2004), and T3 (2008). Subjects were childrenwho had been involved with the Revolutionary United
Front (RUF) and who had then been referred to the a
JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


RC’s Disarmament, Demobilization, and Reintegra-ion (DDR) program in Sierra Leone’s Kono District.14

he IRC’s Interim Care Center (ICC) served five dis-ricts of Sierra Leone; the sample was obtained byooling IRC registries to create a master list of allouth (N � 309) assisted by the ICC from June 2001 toebruary 2002, the most active period of demobiliza-ion. Youth who were between the ages of 10 and 17ears in 2002 and who had contact information avail-ble were approached and invited to participate in theaseline assessment (N � 260). At baseline, no youthnd no caregivers refused consent/assent. At T2,6.5% of the sample (N � 147) had been reinterviewedhen data collection was terminated due to the death

f our collaborating non-governmental organization’sNGO) country director. At T3, we were able to recon-act 68.8 % of the original sample, including manyndividuals whom we did not interview at T2. At T15.3% of the youth (N � 256) lived with at least oneiological parent, whereas this percentage was 53.7%N � 147) at T2 and 34.1% (N � 179) at T3. At all timesf assessment, participants were in a home situationith an identified parent or guardian who was legally

ble to give consent for the child to participate. Nohildren lived in institutionalized settings or situationsn which additional consent from government guard-ans was needed. In all, 47.3% of the original sample

ere assessed at all three waves, whereas 30.8% of theample was assessed at two waves (either at T1 and T2r at T1 and T3), and 21.9% of the sample was assessednly at T1.

Trained Sierra Leonean research assistants con-ucted private face-to-face interviews, first with sub-

ects and then with index caregivers. Because of lowiteracy in the population, all consents/assents andtudy protocols were administered verbally in Krio,he most widely spoken language in Sierra Leone.nterviews lasted 1 to 3 hours. Survey protocols werepproved by internal review at the IRC (for T1) andRB committees at the Boston University School of

edicine/Boston Medical Center (for T2) and thearvard School of Public Health (for T3). Social work-

rs traveled with the research team at all waves toespond to serious emotional or physical health needs.uring T1 and T2, youth who showed signs of imme-iate risk of harm received visits from IRC socialorkers. At T3, 5% of those interviewed were deter-ined to be at immediate risk for harm (mainly

ecause of suicidal ideation) and were referred toental health services via the Community Assistance

or Psychosocial Support (CAPS) program in Kono.

tudy Instrumentsn diverse cultural settings such as postconflict Sierraeone, the assessment of mental health outcomes re-ains a persistent challenge, as constructs must be iden-

ified, framed, and measured in culturally meaningful
nd valid ways.15-17 We used a mix of standard measures
607www.jaacap.org

ssosI.S.flcha

Sps0AotpnbC

iwToflapkahitt

aSdalDdmwoatpfS

BETANCOURT et al.

and locally derived measures developed for use in SierraLeone.18 All measures were selected and adapted in closeconsultation with local staff and community members.Focus groups of youth and adults were used to developadditional questionnaire items and to determine the facevalidity and cultural relevance of standard measures.Similar combined methods for cross-cultural instrumentdevelopment have been used in recent studies of formerchild soldiers.6 All measures were forward- and back-translated to ensure accuracy following a standard pro-tocol.19 Given that several study measures were usedacross a wide developmental range, we examined anumber of factors related to the psychometric propertiesof the measures at each time point. We observed goodinternal consistency and expectable indicators of concur-rent and predictive validity at each time point (i.e., strongcorrelation between conceptually related scales such asthe locally derived scales and scales validated for assess-ing depression and anxiety in adults). Variables expectedto be correlated for conceptual reasons also demon-strated expected correlations (for example, at T3 internal-izing and externalizing symptoms were significantly andinversely associated with community acceptance [r ��0.21 and r � �0.22, p � .01] and positively associatedwith adaptive behaviors [r � 0.44, p � .001]). We alsoexamined the correlation between the three outcomes ofinterest. Across all waves, the correlation between inter-nalizing and externalizing ranged from 0.43 to 0.67. Thecorrelation between adaptive/prosocial behavior andinternalizing ranged from �0.24 to 0.03 and betweenadaptive prosocial behavior and externalizing rangedfrom �0.02 to 0.14.Mental Health Assessment. Mental health was as-sessed across all three waves using the Oxford Mea-sure of Psychosocial Adjustment developed and vali-dated for use among former child soldiers in SierraLeone and northern Uganda.18,20 Six items were ex-cluded from the original 52-item scale because theydid not contribute to the internal consistency of sub-scales. The reduced version of the instrument wasarranged into subscales for externalizing problems(hostility, 12 items, range 0 to 48, T1 to T3 averageCronbach’ � � 0.81), internalizing problems (anxietyand depression, 16 items, range 0 to 64, T1 to T3average Cronbach’s � � 0.79), and a subscale ofadaptive/prosocial behaviors (confidence and proso-cial behaviors, 18 items, range 0 to 72, T1 to T3 averageCronbach’s � � 0.84). Such dimensional approachesassess emotional and behavioral problems along acontinuum rather than solely on diagnostic categories.They are considered to be more sensitive to child devel-opment as well as to cross-cultural differences and, mostimportant to our study, to change over time.16 Within adimensional approach, internalizing problems refer toinward-directed experiences of distress commonly re-lated to symptoms of depression or anxiety. Externaliz-ing problems refer to outward-directed behaviors such as
hostility and aggression. A sum of the total problem s
JOURN

608 www.jaacap.org

core on the Oxford Measure of Psychosocial Adjustmenthowed significant correlations with standard measuresf psychological problems including the nine-item ver-ion of the Child Posttraumatic Stress Disorder Reactionndex (PTSD-RI)21 (T1: r � 0.47, p � .001; T2: r � 0.68, p �0001) and the 25-item Hopkins Symptom Checklist/F-25 (HSCL-25)22 (T1: r � 0.51, p � .001; T2: r � 0.68, p �

0001). These standard measures have not been validatedor use in Sierra Leone, but comparison indicates corre-ation in the expected direction between the core out-ome measure used in this study and standard scales thatave been used to assess the mental health of war-ffected youth in other international settings.4,23

An adapted version of the Inventory of Sociallyupportive Behaviors, consisting of 21 items assessingerceived emotional, instrumental, and informationalupport from others (range 0 to 84, T3 Cronbach’s � �.87), was added at T3.24 The scale of Communitycceptance, administered at all three waves, consistedf six items developed from locally collected qualita-ive data and assessed the manner in which subjectserceived community acceptance. These items wereot phrased to refer specifically to the experience ofeing a child soldier (range 0 to 12, T1 to T3 averageronbach’s � � 0.89).

At T2 (and at T3 for subjects not interviewed at T2),tems from the Child War Trauma Questionnaire25

ere selected to assess individual war experiences.hree types of war experiences were examined in lightf theory and existing research on mental health inormer child soldiers: (1) witnessing general war vio-ence such as massacres or raids on villages; (2) beingvictim of rape; and (3) injuring or killing others.11 Inarticular, stressors such as being raped and injuring/illing others may be considered particularly “toxic”nd may greatly threaten the development and mentalealth of formerly abducted children.9 The survey

nstrument also included two self-reported classifica-ions, i.e., age of abduction and duration of time withhe fighting forces.

The Post-War Adversities Index26 was adapted tossess daily hardships facing youth in postconflictierra Leone. Sixteen items (range 0 to 16) were used toetermine housing and economic insecurity (10 items),s well as to assess interpersonal adversities, such asiving with people who fight or use drugs (six items).emographic and household information such as gen-er, age, and socioeconomic status (SES) (an eight-itemeasure pertaining to family resources and property),as also collected via youth self-report. At each wave

f assessment, subjects were asked whether they werettending school, and at T3 they were also asked ifhey were employed. To allow comparisons betweenredictors with dissimilar scales of measurement, the

ollowing variables were standardized (mean � 0;D � 1): SES, number of violent events witnessed,
tigma score, social support and community accep-


aatmaatmeoe

tibamaticaHmccp

RCTfpa1yftrhvrrndrsvapw

RO


tance; all other variables, including outcomes andpredictors, are in their original metrics.

Stigma and perceived discrimination at T2 and T3were assessed with the Everyday DiscriminationScale27 comprising nine items that capture negativecommunity interactions such as differential treatment,threats, or abuse. Responses were scored on a scale of0 to 2 (0 � never, 1 � sometimes, 2 � always). For eachitem endorsed as “sometimes” or “always,” interview-ers probed the reason for the perceived discrimination.Although eight potential reasons for discriminationwere given (status as a child soldier, religion, gender,age, disability, poverty, level of education, and tribe),only stigma resulting from being a former child soldierwas used in the present analysis (range 0 to 18, T2 toT3 average Cronbach’s � � 0.89). Although we recog-nize that numerous reasons for discrimination are ofpotential importance, we chose to focus on discrimi-nation resulting from being a child soldier (which wasthe most frequently endorsed) for the purposes ofparsimony and to illuminate phenomena specific tochild soldiers in the present analysis.

Statistical AnalysesMultilevel linear growth modeling was used to relatestudy participants’ mental health outcomes over timeto war experiences, as well as to postconflict risk andprotective factors. In these models, trajectories of men-tal health outcomes were represented by two parame-ters: an intercept, which quantifies a young person’sbaseline level of mental health; and slope, whichquantifies changes in mental health over time.28 Coef-ficients for predictors of the slope indicate the relation-ship between time-invariant predictors and changes inmental health over time, wherein a positive coefficientindicates that larger predictor values are associatedwith an increase in the outcome over time. For consis-tency in interpreting coefficients across models, allpredictors were centered on their grand means forboth the intercept and the time slopes. Therefore,regression coefficients are interpreted in terms of de-viations from the sample grand mean.

Analyses also included two time-varying predic-tors, i.e., school participation and community accep-tance, that were measured at all three time points. Anincrease in these variables could relate directly toincreases or decreases in outcomes over time. Each ofthese time-varying predictors has a time-invariantcounterpart (the within-person mean). The time-invariant component for community acceptance isthe average value of community acceptance acrossthe three time points; for school attendance, it is thetotal number of time points at which school atten-dance was reported. In each case, the time-varyingcomponent is then centered on the within-personmean value to represent change only.

Each mental health outcome was fit with three
models: the first included war experiences; the second T


dded post-war adversities including daily hardshipsnd stigma due to being a former child soldier; and thehird added post-war protective factors including com-

unity acceptance, social support, school attendance,nd employment. Models also controlled for age, sex,nd SES. In the second and third models, a devianceest measured whether the added variables improved

odel fit over the previous model. The war experi-nces model was compared with a model containingnly demographic covariates. All growth models werestimated with HLM 6.0 Software.29

Missing data were addressed by multiple imputa-ion (MI) at each time point only for those subjectsnterviewed at that time point. This approach reducesias to the extent that values on the observed variablesre informative about the likelihood an item wasissing; increases precision relative to a complete case

nalysis with a smaller sample size; and accounts forhe sampling variability across imputations.30 Ninemputed datasets were generated using the method ofhained equations as implemented in IVEware31 andnalyzed using the multiple imputation feature of theLM software.28 Furthermore, multilevel growthodeling allows inclusion of incomplete cases, so

ases with completely missing time points are in-luded in the models but only with data on the timeoints that are observed for a given subject.28,29

ESULTSharacteristics of Participantsable 1 displays characteristics of the sample of

ormer child soldiers (N � 260). The sample wasredominantly male (88.8%, n � 231), and nearlyll participants were either Christian (52.3%, n �36) or Muslim (47.3%, n � 123). Virtually allouth in the sample reported joining the RUF byorce/abduction (97.7%, n � 254). Youth joinedhe rebels at a mean age of 10.3 years andemained an average of 4.1 years. Participantsad witnessed an average of 6.2 war-relatediolent events. In all, 45% of girls and 5% of boyseported being raped during their time with theebels. More than a quarter of the sample (26.9%,� 70) reported having killed or injured others

uring the war. Comparisons between subjectsetained and those lost to follow-up (results nothown) revealed no significant differences onariables of interest, with the exception of schoolttendance whereby a higher proportion of com-leters versus non-completers (69% vs. 47%)ere in school at baseline (p � .003).

etentionf eligible subjects approached for interviews at
2, one caregiver declined consent for his/her
609www.jaacap.org

ococ.rtsepaaldtcItrgcpptnoii

byp

e poi

BETANCOURT et al.

child to participate, one subject had died, and 19subjects had moved to distant locations thatprecluded follow-up. Midway through T2 datacollection, the death of IRC’s country directorsuspended program activities and halted thestudy, preventing further participant follow-up.As a result, a total of 113 participants (43.5% ofthe original sample) were not recontacted at T2.When the third wave of assessment was con-ducted in 2008 (T3), attempts were made toreinterview all 260 from the original sample.Overall, 81 participants (31.2% of the originalsample) were lost to follow-up at T3; 68 partici-pants could not be found because of insufficientcontact information, four had moved to anothercountry, and nine were deceased.

Growth Models of Children’s Mental HealthProblems and Adaptive/Prosocial BehaviorsThe results of linear growth models for eachmental health outcome over time are shown inTable 2. Deviance tests conducted at each stage—war experiences, postconflict hardships, andpostconflict protective factors—indicated thateach subsequent model fit better than the previ-ous one for all three outcomes.Externalizing Problems. Youth who experiencedstigma because of being a child soldier hadhigher baseline levels of externalizing problems(b � 0.89, p � .016) in models considering onlywar experiences (adjusted for covariates). How-ever, this association was no longer statistically

TABLE 1 Characteristics of the sample (former child soldeach assessment reported as mean (SD) or frequency (%)

Range of Scale

AgeOutcomesExternalizing problems 0–48Internalizing problems 0–64Adaptive/prosocial behaviors 0–72Postconflict hardshipsStigma due to being a child soldier 0–18Postconflict hardships 0–16Postconflict protective factorsSocial support 0–84Community acceptance 0–12In school at time of assessment . . .Working . . .

Note: N � number of participants with complete data available at each tim

significant (b � 0.73, p � .058) upon consideration .

JOURN

610 www.jaacap.org

f postconflict protective factors. Looking athange over time, child soldiers who had injuredr killed others demonstrated significant in-reases in externalizing problems (b � 1.16, p �031). However, after adjusting for postconflictisk factors such as stigma and daily hardships,his effect did not retain statistical significance,uggesting that the association between wartimexperience on externalizing is also explained byostconflict factors. Of the protective factors ex-mined, increased community acceptance wasssociated with decreased externalizing prob-ems (b � �1.09, p � .001), such that a 1–standardeviation (SD) increase in the community accep-

ance score was associated with a 0.10-SD de-rease in the externalizing problems score.nternalizing Problems. In models examining in-ernalizing problems (anxiety/depression), beingaped, longer periods of time with an armedroup, and being subjected to stigma were asso-iated with higher baseline levels of internalizingroblems. After adjusting for all hardship androtective factors, among time-invariant predic-

ors, only being raped remained statistically sig-ificant (b � 4.34, p � .039). Higher average levelsf community acceptance were significantly and

nversely associated with lower baseline levels ofnternalizing problems (b � 1.21, p � .003).

Looking at change over time, youth who hadecome involved with fighting forces at aounger age showed increases in internalizingroblems (b � �0.35 for each year of age, p �

in Sierra Leone) and distributions of main variables at

T1: N � 260 T2: N � 147 T3: N � 179

5.13 (2.22) 17.41 (2.38) 21.75 (3.16)

9.32 (5.18) 20.24 (6.30) 18.52 (4.44)4.54 (7.64) 35.44 (7.63) 34.83 (6.45)8.27 (7.30) 59.29 (7.46) 57.42 (7.21)

. . . 2.50 (3.63) 1.61 (3.09)

. . . . . . 5.03 (3.54)

. . . . . . 41.32 (13.45)0.50 (2.55) 10.38 (2.41) 9.85 (2.71)165 (63.5%) 111 (75.5%) 86 (48.9%)

. . . . . . 43 (24.0%)

nt that were included in analyses. T1 � 2002, T2 � 2004, T3 � 2008.

iers

1

135

1

02). Thus, being 3 years younger at the time of



TABLE 2 Coefficient Estimates of Linear Growth Models for Mental Health Problems and Adaptive/Prosocial Behavior Among Former Child Soldiers

Externalizing Internalizing Adaptive/Prosocial Behaviors

WarExperiences

PostconflictHardships

ProtectiveFactors

WarExperiences


ProtectiveFactors

WarExperiences


ProtectiveFactors

BaselineIntercept 19.67*** 19.68*** 19.72*** 34.94*** 34.94*** 35.01*** 59.32*** 59.33*** 58.80***Age first involved in fighting �0.21 �0.20 �0.22 0.32 0.34 0.29 �0.05 �0.03 0.05N of years in fighting forces �0.05 �0.05 �0.13 0.56* 0.55* 0.42 �0.20 �0.19 0.06Witness violence 0.20 0.12 0.16 0.22 0.12 0.05 �0.32 �0.37 �0.40Killed/injured others in war 0.66 0.20 0.21 0.26 �0.37 �0.28 1.93 1.53 1.24Victim of rape/sexual assault 3.73 3.48 3.58 4.60* 4.25 4.34* 0.58 0.42 0.83Stigma of being child soldier 0.89* 0.73 1.22** 0.89 0.42 0.91Daily hardship score �0.26 �0.17 �0.26 �0.24 0.59 0.46Social support �0.24 0.29 �0.03Working, not in school �1.33 �0.27 0.95Cumulative school attendance �0.22 �0.10 1.50***Avg. community acceptance �0.60 �1.21** 1.69***Change over timeIntercept �0.29 �0.22 �0.09 0.30 0.47 0.47 �0.89** �0.95** �0.55Age first involved in fighting 0.09 0.11 0.12 �0.35* �0.29* �0.27 �0.12 �0.15 �0.23N of years in fighting forces 0.04 0.07 0.10 �0.33 �0.26 �0.21 �0.06 �0.09 �0.15Witness violence 0.29 0.29 0.22 �0.04 �0.02 0.02 0.18 0.26 0.03Killed/injured others in war 1.16* 0.93 0.90 1.22 0.47 0.34 �2.60*** �1.88* �1.29Victim of rape/sexual assault �0.64 �0.70 �0.78 �0.15 �0.39 �0.38 0.34 0.63 0.13Stigma of being child soldier �0.01 0.03 0.26 0.22 �0.91** �0.61Daily hardship score 0.51 0.40 1.38*** 1.33*** �0.24 0.03Social support 0.38 0.08 0.93**Working, not in school 1.15 �0.22 0.38In school at time of

assessmentIntercept 1.01 0.60 2.69**Change from mean

level of communityacceptance

Intercept �1.09*** �0.86* 1.93***

Note: *p � 0.05, **p � 0.01, ***p � 0.001.

SIERRALEO

NE’S

FORM

ERC

HILD

SOLD

IERS

JOU

RNA

LO

FTH

EA

MERIC

AN

AC

AD

EMY

OF

CH

ILD&

AD

OLESC

ENT

PSYC

HIA

TRY

VOLU

ME

49N

UM

BER6

JUN

E2010

611

ww

w.jaacap.org

iiarimeccwltr

ttwcpldtpphsa

pbfamaoplaMcng

taavocitg

BETANCOURT et al.

abduction translated into an increase betweentime points of 0.10 SD in internalizing problems.Higher levels of daily hardships were also asso-ciated with increased internalizing problemssuch that a 1-unit greater postconflict hardshipsscore was associated with a 1.38-unit increase inthe internalizing problems scale over time (p �.001). Looking to protective factors in the post-conflict environment, community acceptanceagain demonstrated a protective relationshipsuch that each additional SD increase in commu-nity acceptance over time was associated with a0.13-SD decrease in internalizing problems.Adaptive/Prosocial Attitudes and Behaviors. Forpositive developmental outcomes, we observedthat retention in school and higher levels ofcommunity acceptance were associated withhigher baseline levels of adaptive/prosocial be-havior (b � 1.50, p � .001 and b � 1.69, p � .001)(see predictors of “Baseline” slope in Table 2).However, over time, child soldiers who hadinjured or killed others had significant decreasesin prosocial/adaptive behaviors (b � �2.60, p �.001) adjusting for covariates. After adding post-conflict risk factors such as stigma and dailyhardships to the model, the association betweeninjuring/killing others and adaptive/prosocialbehaviors was reduced, but remained significant(b � �1.88, p � .02). Thus, postconflict stigmaresulting from being a child soldier was associ-ated with a decline in adaptive/prosocial behav-iors (b � �0.91, p � .01) over time. However, themagnitude of the association between stigma anddeclines in adaptive/prosocial behavior was nolonger statistically significant upon adding post-conflict protective factors to the model (b ��0.61, p � .054). In addition, after adjusting forpostconflict protective factors, the reduction inprosocial behaviors associated with havingkilled/injured others in war was further reducedto about one half the original effect (b � �1.29,p � .068). Of the protective factors examined,social support (b � 0.93, p � .006) and increasingcommunity acceptance (b � 1.93, p � .001) wereassociated with increased prosocial/adaptive be-haviors over time.

DISCUSSIONFormer child soldiers’ acute war experiences havelong-term consequences, but the nature and extentof these consequences are influenced by postcon-
flict risk and protective factors. In our sample, T
JOURN

612 www.jaacap.org

njuring/killing others was associated with increas-ng externalizing problems and decreasing levels ofdaptive/prosocial behaviors, whereas beingaped predicted higher baseline levels of internal-zing problems (anxiety/depression). However,

ost of the associations between these war experi-nces and longitudinal outcomes were not statisti-ally significant in models that accounted for post-onflict factors. The relationships between acutear trauma, internalizing and externalizing prob-

ems, and positive psychosocial adjustment wereherefore significantly shaped by postconflict expe-iences.

Our results also suggest that different forms ofrauma have differing degrees and types of long-erm mental health impact. General witnessing of

ar violence, for example, was not associated withhanges in mental health, whereas being raped wasredictive of higher levels of internalizing prob-

ems but not of increased externalizing problems orecreased prosocial behavior. Our findings—par-

icularly the associations we observed betweenerpetration of violence, increases in externalizingroblems, and decreases in adaptive/prosocial be-aviors—are consistent with relationships ob-erved in other studies of former child soldiers3,5

nd adult war veterans.32

Our findings point to a number of importantostconflict factors that influence long-term well-eing and that may represent appropriate targetsor intervention. Youth who remained in schoolnd those who experienced higher levels of com-unity acceptance had higher baseline levels of

daptive/prosocial behaviors adjusting for allther factors. In addition, decreases in adaptive/rosocial behaviors were associated with higher

evels of stigma, whereas increases were associ-ted with social support and school attendance.ost promisingly, in the presence of increasing

ommunity acceptance, youth demonstrated sig-ificant improvements in all outcomes investi-ated.

Taken together, these findings highlight the po-ential of postconflict intervention programs thatddress community dynamics such as acceptancend stigma and that bolster strengths at the indi-idual level through social support. In Sierra Le-ne, thoughtful attention was given to promotingommunity acceptance of former child soldiersmmediately after the war.14 Community sensitiza-ion campaigns assisted the initial phases of reinte-ration but did not continue over the long term.33

he nationwide sensitization efforts were particu-



co

aitoswFmwolAeavba

lhipscsiebacbesytsibfivhriafee


larly strong before the creation of the Sierra LeoneTruth & Reconciliation Commission, which wasmeant to coordinate mass campaigns but limited infunding. Given the long-term positive effects ofimproved community acceptance on former childsoldiers’ mental health, interventions aimed at bol-stering and maintaining this protective processmay strengthen and sustain the impact of initialefforts.

Our findings underscore aspects of community-based intervention features—including efforts toreduce stigma, increase community acceptanceand increase school attendance—that are crucialto the mental well-being of war-affected youth.However, despite the fact that our findings sup-port broad, community-based approaches, we donot disregard the possible beneficial effects oftargeting interventions at individual child sol-diers. Without specific comparison groups (i.e., asample of non–war-exposed children with be-havioral problems), there is not sufficient evi-dence to rule out war-related factors as ulti-mately etiological and influential in subsequentpostconflict adjustment.

Study limitations include moderate retentionrates, given the challenging field conditions andearly termination of the T2 data collection; use ofself-reports; and a lack of information on pre-warlevels of mental health problems in Sierra Leone,which limit our ability to interpret the magnitudeof the mental health problems observed. Further-more, although our measure of mental health out-comes was developed and validated for use inSierra Leone, clinical cut points are unavailable forthis context. In addition, we cannot make strongclaims about causality in these data, as war-relatedexperiences were reported retrospectively. Ourfindings are also limited in their general applicationto the country as a whole, as our sample comprisesyouth who originated from five of 14 districts inSierra Leone (Kono, Bo, Kenema, Pujehun, andMoyamba) demobilized over a specified time pe-riod; this design was necessitated in part by thesignificant ethical and logistical difficulties in-volved in obtaining a country-level representativesample of former child soldiers. However, we haveno reason to believe that the children we inter-viewed are very different from the children eitherin their own districts or in other districts that wereserved via the DDR process. Finally, without acomparison group of self-reintegrated youth or
youth never associated with armed groups, we c


annot broadly generalize our findings to commentn other groups of war-affected youth.

Study strengths include the use of dimensionalssessments capturing a broad range of mental healthndicators in a culturally sensitive manner as opposedo a narrow focus on trauma-related disorders. More-ver, by including multiple types of outcome mea-ures, including positive/adaptive outcomes, weere able to identify different patterns of sensitivity.

or example, all outcomes were associated with com-unity acceptance, but rape was more associatedith internalizing problems, while killing/injuring

thers was more associated with externalizing prob-ems and deficits in prosocial/adaptive behaviors.lthough stigma influenced both internalizing and

xternalizing problems, internalizing problems werelso influenced by daily hardships and age of in-olvement with armed groups. Adaptive/prosocialehavior, meanwhile, was most sensitive to schoolccess and social support.

Overall, this prospective study contributes to theiterature on the processes that influence mentalealth in former child soldiers over time. By exam-

ning postconflict variables, we identified broaderrocesses that work to shape the long-term psycho-ocial adjustment and well-being of children asso-iated with armed forces and armed groups; inhort, we found that mental health in this group isnfluenced by much more than past war experi-nces. Although our general finding may applyroadly, the specific processes that influence risknd resilience are likely to be shaped by culture andontext.34 Although stigma and discrimination maye seen as universally detrimental human experi-nces, the social meaning of being a former childoldier may differ, for example, in a context whereouth involvement in conflict is focused on joiningogether to fight a common enemy compared withituations in which forced abduction plays a signif-cant role. Similarly, although the experience ofeing raped may be considered universally harm-ul, the implications of this experience certainlynteract with gender and culture to influence ad-erse outcomes.10 In Sierra Leone, other researchas observed that, although both boys and girlseported rape, stigma against girls more commonlyncluded insinuations of sexual impropriety thatre particularly damaging to the social prospects ofemales.13,35 In this manner, the nature of both warxperiences and postconflict experiences must bexamined carefully across settings.

Based on our findings, we suggest that post-
onflict adversities and resources must be given
613www.jaacap.org

BETANCOURT et al.

closer attention in both research and servicedelivery. Such understanding of the impact ofmultiple war-related and postconflict factors isimportant for identifying appropriate interven-tion targets. The results of this study emphasizethe importance of investing in longer-term (i.e.,beyond the immediate postconflict period) ap-proaches to monitoring and supporting the well-being of war-affected youth. &

Accepted March 18, 2010.

Drs. Betancourt and Brennan and Ms. Rubin-Smith are with theFrançois-Xavier Bagnoud Center for Health and Human Rights at theHarvard School of Public Health (HSPH); Dr. Betancourt is also withthe Department of Global Health and Population at HSPH; Dr.Fitzmaurice is with the Laboratory for Psychiatric Biostatistics at McLeanHospital of Harvard Medical School; Dr. Gilman is with the Depart-ment of Society, Human Development and Health and the Departmentof Epidemiology at HSPH.This study was funded by the United States Institute of Peace,USAID/DCOF, Grant #1K01MH077246-01A2 from the NationalInstitute of Mental Health, the International Rescue Committee, and the
François-Xavier Bagnoud Center for Health and Human Rights.
among former child soldiers in Sierra Leone. Soc Sci Med.2010;70:17-26.

1

1

1

1

1

1

2

2

2

2

2

2

2

JOURN

614 www.jaacap.org

The authors thank Sidney Atwood, Kathleen McGaffigan, and LauraKhan for their assistance in data management and analysis. Mr.Atwood is with the Division of Global Health Equity at Brigham andWomen’s Hospital; Ms. McGaffigan is with the Harvard School ofPublic Health; Ms. Khan is a student at HSPH. The authors also thankin particular Lloyd Feinberg and John Williamson for their invaluablesupport, and Catherine Weisner for her assistance on Wave 3follow-up. The authors also thank Marie de la Soudiere for her help inconceptualizing and completing the 2002 and 2004 study phases,and to Moses Zombo and the Sierra Leone research team, whosededication and energy brought this study to fruition. Finally, the authorsthank the families and amazing young people who shared their liveswith us.

Disclosure: Drs. Betancourt, Brennan, Fitzmaurice, Gilman, and Ms.Rubin-Smith report no biomedical financial interests or potential con-flicts of interest.

Correspondence to Dr. Theresa S. Betancourt, Research Programon Children and Global Adversity/François-Xavier BagnoudCenter for Health and Human Rights, Harvard School of PublicHealth, 651 Huntington Avenue, 7th Floor, Boston, MA 02115.E-mail: [email protected]

0890-8567/$36.00/©2010 American Academy of Child andAdolescent Psychiatry

DOI: 10.1016/j.jaac.2010.03.008

REFERENCES1. Coalition to Stop the Use of Child Soldiers. Child soldiers: Global

report 2008. London 2008.2. The hidden health trauma of child soldiers. Lancet. 2004;363:831.3. Betancourt TS, Borisova II, Brennan RB, et al. Sierra Leone’s

former child soldiers: a follow-up study of psychosocial adjust-ment and community reintegration. Child Development (inpress).

4. Derluyn I, Broekaert E, Schuyten G, De Temmerman E. Post-traumatic stress in former Ugandan child soldiers. Lancet. 2004;363:861-863.

5. Bayer CP, Klasen F, Adam H. Association of trauma and PTSDsymptoms with openness to reconciliation and feelings of re-venge among former Ugandan and Congolese child soldiers.JAMA. 2007;298:555-559.

6. Kohrt BA, Jordans MJD, Tol WA, et al. Comparison of mentalhealth between former child soldiers and children never con-scripted by armed groups in Nepal. JAMA. 2008;300:691-702.

7. Blattman C, Annan J. The consequences of child soldiering. RevEcon Stat. (in press).

8. Lustig SL, Kia-Keating M, Knight WG, et al. Review of child andadolescent refugee mental health. J Am Acad Child AdolescPsychiatry. 2004;43:24-36.

9. National Scientific Council on the Developing Child. ExcessiveStress Disrupts the Architecture of the Developing Brain: Work-ing Paper #3. Cambridge, MA: Center on the Developing Child atHarvard University; 2005.

10. Johnson K, Asher J, Rosborough S, et al. Association of combatantstatus and sexual violence with health and mental health out-comes in postconflict Liberia. JAMA. 2008;300:676-690.

11. Betancourt TS, Borisova I, Rubin-Smith JE, Gingerich T, WilliamsT, Agnew-Blais J. Psychosocial adjustment and social reintegra-tion of children associated with armed forces and armed groups:the state of the field and future directions. Austin, TX: PsychologyBeyond Borders; 2008.

12. Boothby N, Crawford J, Halperin J. Mozambique child soldier lifeoutcome study: lessons learned in rehabilitation and reintegrationefforts. Global Public Health. 2006;1:87-107.

13. Betancourt TS, Agnew-Blais J, Gilman SE, Williams DR, Ellis BH.Past horrors, present struggles: the role of stigma in the associa-tion between war experiences and psychosocial adjustment

4. Williamson J. The disarmament, demobilization and reintegrationof child soldiers: SOcial and psychological transformation inSierra Leone. Interv Int J Ment Health Psychosoc Work CounselAreas Armed Conflict. 2006;4:185-205.

5. Canino G, Alegria M. Psychiatric diagnosis—is it universal orrelative to culture? J Child Psychol Psychiatry Allied Disc 2008;49:237-250.

6. Achenbach TM, Rescorla LA. Multicultural Understanding ofChild and Adolescent Psychopathology: Implications for MentalHealth Assessment. New York, NY: Guilford Press; 2006.

7. Betancourt TS, Speelman L, Onyango G, Bolton P. A qualitativestudy of psychosocial problems of war-affected youth in northernUganda. J Transcult Psychiatry. 2009;46:238-256.

8. MacMullin C, Loughry M. Investigating psychosocial adjustmentof former child soldiers in Sierra Leone and Uganda. J RefugeeStud. 2004;17:460-472.

9. Van Ommeren M, Sharma B, Thapa S, et al. Preparing instrumentsfor transcultural research: use of the translation monitoring formwith Nepali-speaking Bhutanese refugees. Transcult Psychiatry.1999;36:285.

0. MacMullin C, Loughry M. Proposal to Assess Psychosocial Ad-justment of Former Child Soldiers in Sierra Leone; Phase 1 FinalReport: Construction of Research Instrument. New York, NY:International Rescue Committee; 2000.

1. Pynoos RS, Rodriguez N, Steinberg AS, Stauber M, Frederick C.UCLA PTSD Index for DSM-IV child version. Los Angeles, CA:UCLA Psychiatry Trauma Service; 1998.

2. Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L. TheHopkins Symptom Checklist (HSCL): a self-report symptominventory. Behav Sci 1974;19:1-15.

3. Layne CM, Olsen JA, Baker A, Legerski JP, Isakson B, Pašalic A.Unpacking trauma exposure risk factors and differential path-ways of influence: predicting post-war mental distress in Bosnianadolescents (in press).

4. Barrera M, Ainlay SL. The structure of social support: a concep-tual and empirical analysis. J Community Psychol. 1983;11:133-143.

5. Macksoud MS. Assessing war trauma in children: a case study ofLebanese children. J Refugee Stud. 1992;5:1-15.

6. Layne CM, Stuvland R, Saltzman W, Djapo N, Pynoos RS.
Adolescent Post War Adversities Scale (unpublished instrument).1999.


mailto:[email protected]

3

3

3

3


27. Williams DR, Yu Y, Jackson JS, Anderson NB. Racial differencesin physical & mental health: socioeconomic status, stress, anddiscrimination. J Health Psychol. 1997;2:335-351.

28. Singer JD, Willett JB. Applied Longitudinal Data Analysis: Mod-eling Change and Event Occurrence. New York, NY: OxfordUniversity Press; 2004.

29. Raudenbush SW, Bryk AS, Cheong YF, Congdon RT. HLM 6:Hierarchical linear and nonlinear modeling. 2nd ed. Chicago, IL:Scientific Software International; 2005.

30. Rubin DB. Multiple imputation for nonresponse in surveys. NewYork, NY: Wiley; 1987.

31. Raghunathan TE, Lepkowski JM, Van Hoewyk J, SolenbergerP. A multivariate technique for multiply imputing missing



values using a sequence of regression models. Surv Methodol.2001;27:85-95.

2. Hofmann SG, Litz BT, Weathers FW. Social anxiety, depression,and PTSD in Vietnam veterans. J Anxiety Disord. 2003;17:573-582.

3. Zaeh S, Betancourt T. Mental Health and Psychosocial Servicesfor Youth in Post-Conflict Sierra Leone [case study] (in progress).

4. Betancourt TS, Khan KT. The mental health of children affectedby armed conflict: protective processes and pathways to resil-ience. Int Rev Psychiatry. 2008;20:317-328.

5. Denov M, Gervais C. Negotiating (in)security: agency, resistance,and resourcefulness among girls formerly associated with Sierra
Leone’s Revolutionary United Front. Signs. J Women Cult Soc.2007;32:885-910.
615www.jaacap.org

Documents

Sierra Leone's Former Child Soldiers: A Longitudinal Study of Risk, Protective Factors, and Mental Health