SEARCH

SEARCH BY CITATION

Keywords:

  • Violence;
  • prosocial behaviour;
  • adolescence;
  • parenting;
  • parent—child relationships

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Background

In humanitarian settings, family-level drivers of mental health are insufficiently documented; we examined the strength of caregiver—child associations with two-wave, family-level Afghan data.

Methods

We recruited a gender-balanced sample of 681 caregiver—child dyads (= 1,362 respondents) using stratified random-sampling in government schools in Kabul (364 dyads) and refugee schools in Peshawar (317 dyads). One year after baseline, we re-interviewed 64% of Kabul and 31% of Peshawar cohorts (n = 331 dyads, 662 respondents), retaining fewer Peshawar families due to refugee repatriation. In multivariable analyses adjusted for baseline, we assessed the extent to which caregiver mental health (Self-Report Questionnaire, SRQ-20) was associated with child symptom scores of post-traumatic stress (Child Revised Impact of Events Scale, CRIES), depression (Depression Self-Rating Scale, DSRS), psychiatric difficulties, impact, and prosocial strength (Strength and Difficulties Questionnaire, SDQ).

Results

Caregiver mental health was prospectively associated with all eight measures of child mental health at follow-up, adjusted for baseline. For post-traumatic stress, caregiver mental health had a predictive impact comparable to the child experiencing one or two lifetime trauma events. For depression, caregiver mental health approached the predictive impact of female gender. Thus a one SD change in caregiver SRQ-20 was associated with a 1.04 point change on CRIES and a 0.65 point change in DSRS. For multi-informant SDQ data, caregiver—child associations were strongest for caregiver ratings. For child-rated outcomes, associations were moderated by maternal literacy, a marker of family-level dynamics. Both adults and children identified domestic violence and quality of home life as independent risk and protective factors.

Conclusions

In the context of violence and displacement, efforts to improve child mental health require a thoughtful consideration of the mental health cascade across generations and the cluster of adversities that impact family wellbeing. We identify culturally meaningful leverage points for building family-level resilience, relevant to the prevention and intervention agenda in global mental health.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

The role of caregiver mental health in influencing child and adolescent mental health is receiving increasing attention as a global issue, especially in contexts of war and forced displacement. Caregiver health matters for child health, especially in low-income settings, although the causal pathways responsible for this association are still under debate (LeVine, LeVine, Schnell-Anzola, Rowe, & Dexter, 2012). In global mental health, a ‘call for action’ puts child and adolescent mental health at the top of the international agenda for research, prevention, and treatment (Patel, Flisher, Hetrick, & Mcgorry, 2007; Patel, Flisher, Nikapota, & Malhotra, 2008) especially in humanitarian settings (Tol & van Ommeren, 2012; Tol et al., 2010, 2011b). Nonetheless, very few studies have systematically investigated the associations between caregiver and child mental health in countries affected by chronic poverty, war, or forced displacement (Betancourt, Yudron, Wheaton, & Smith-Fawzi, 2012; Thielman, Ostermann, Whetten, & O'Donnell, 2012).

To date, systematic reviews of literature have highlighted three well-known limitations of the evidence base on child mental health in humanitarian contexts, namely, a paucity of longitudinal data, small and unrepresentative samples, and a narrow focus on individual level rather than family- or community-level data (Betancourt et al., 2013a; Panter-Brick, Goodman, Tol, & Eggerman, 2011; Reed, Fazel, Jones, Panter-Brick, & Stein, 2011). To achieve a truly global perspective on child mental health (Leckman & Leventhal, 2008), we need to better understand changes over the life course and clusters of childhood adversities (Kessler & Al, 2010), as well as the context- and time-specific pathways responsible for risk and resilience in war-affected children (Betancourt, Mcbain, Newnham, & Brennan, 2013b; Panter-Brick & Leckman, 2013; Tol, 2013). A stronger evidence-base is a necessary first step for linking research to humanitarian practice (Tol et al., 2011a).

In conflict settings, current literature has questioned the utility of work focused on detailing the prevalence rates of mental health disorders, emphasizing instead the importance of identifying contextually relevant risk and protective factors at multiple levels of the social ecology, namely, individual, family, school, community, and society levels (Reed et al., 2011; Tol, Rees, & Silove, 2013). Thus in Afghanistan, we integrated methodologies from crosscultural psychiatric epidemiology and medical anthropology to conduct the first large-scale, systematic survey of child and adolescent mental health, adverse experiences, and social functioning. Adopting a stratified random-sampling design in three regions of the country, we conducted face-to-face interviews with 11–16 year old children (n = 1,011) and their primary caregivers (n = 1,011) to provide well-grounded evidence on the relationships between violence, suffering, resilience, and mental health outcomes (Panter-Brick, Eggerman, Gonzalez, & Saftar, 2009). One year later, given heightened insecurity and logistic constraints, the follow-up survey was limited to Kabul: we traced the same caregiver—child dyads to identify the prospective impact of life events on child mental health (Panter-Brick et al., 2011). The same two-wave, mixed-method survey with Afghan caregiver—child dyads was conducted in the refugee camps of Peshawar, the main city of North-West Frontier Province (NWFP) in neighboring Pakistan.

In terms of mental health, Afghans experience a high burden of psychiatric difficulties, especially women (Husain, Chaudhry, Afridi, Tomenson, & Creed, 2007; Rahman & Hafeez, 2003; Ventevogel, 2005), linked to exposure to past trauma (Cardozo et al., 2004) and ongoing social and material stressors (Miller & Rasmussen, 2010; Panter-Brick et al., 2011). Our survey showed that by 11–16 years of age, Afghan children experience mental health problems that fall within the expected range of psychiatric difficulties and post-traumatic stress in war-affected populations (Panter-Brick et al., 2009). However, they also show a significant measure of resilience, attending school and working for their families in the hope of socioeconomic advancement (Eggerman & Panter-Brick, 2010). Our qualitative research showed that family ‘unity and harmony’ (Dari: wahdat and ittifaq) was a key cultural value, strongly implicated with family dynamics and mental health in children's narratives of adversity, risk, and resilience. Afghan families face significant challenges everyday in the midst of violent and uncertain environments. In their homeland, they have endured 30 years of armed conflict and a massive disruption to social organization (Dupree, 2004), health, and social services (Trani, Bakhshi, Noor, Lopez, & Mashkoor, 2010; Waldman, Strong, & Wali, 2006). After the fall of the Taliban regime in 2001, reconstruction efforts have massively expanded infrastructure as well as opportunities for school education (Oxfam, 2006). However, significant social and economic frustrations persist, with violent riots and bomb attacks fuelled by a sense of political injustice (Donini, 2007) and widening income inequalities (Goodhand, Dennys, & Mansfield, 2012). In NWFP Pakistan, Afghans have comprised one of the largest single refugee populations in the world, retaining close transnational ties to their nearby homeland. The UNHCR-assisted program of repatriation, initiated for Afghan refugees in 2002 was the largest repatriation program in 30 years (Schmeidl, 2009). By 2006, the Pakistani government accelerated the closure of several refugee camps and by 2007, direct UNHCR material assistance was no longer provided, and refugee schools were closed due to cuts in donor funding. Increasing economic and political precariousness of Afghan refugees in Pakistan led to violent demonstrations in 2005 (following allegations of desecration of the holy Qur'ān in Guantanamo Bay) and 2006 (after the publication of cartoons of the Prophet Mohammad in a Danish newspaper). Both Kabul and Peshawar were thus characterized by episodic civil unrest and persistent insurgent violence, which disrupted social life, interrupted school-based activities, frustrated community relations, and impacted family wellbeing.

Including hitherto unpublished data on Afghan refugees, this paper uses all available longitudinal data for Kabul and Peshawar to address the following question: how strong and consistent are the prospective associations between caregiver and child mental health, relative to other risk and protective factor, for dyads living in chronically poor and violent environments? We ensured that all caregivers interviewed were those adults who assumed primary caregiving responsibility for children at the time of the survey. These were not necessarily biological parents and were not necessarily the same person at baseline and follow-up. In the Afghan context, childcare responsibilities are collective, highly gendered, and readily delegated within extended households to include paternal or maternal relatives and/or co-wives. As in many other Islamic cultures, uncles and aunts share a collective responsibility for their nephews and nieces, a sense of responsibility that extends to decisions about schooling, marriage, employment, and behaviors that have repercussions on ‘family honor.’ Our study thus appraises caregiver—child mental health rather than parent-child mental health, in the Afghan context of a collective distribution of caregiver responsibilities and the culturally expected delegation of care to alternate caregivers when family responsibilities shift in response to conditions in employment, housing, or health. Conceptually, we focused on the children living under the care of adults who were differentiated in terms of overall mental health status and past-year changes in mental health. Both these dimensions of caregiver wellbeing qualitatively matter to Afghan 11–16 year olds, at a juncture in their lives when adults make critical decisions on their behalf, such as allowing them to stay at school in the face of cultural dictates that require young girls to marry and young boys to work to support the social and economic standing of their family. We thus assessed the statistical importance of caregivers being in ‘good’ or ‘poor’ mental health as well as caregivers getting ‘better’ or ‘worse’ over time. We also assessed several dimensional measures of child mental health, including symptoms of post-traumatic stress, depression, psychiatric difficulties, and social functioning.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Study design

This longitudinal study was conducted with international and local ethical approval from the authors' university, the University of Peshawar, the United Nations High Commissioner for Refugees in Peshawar, and the Ministry of Education in Afghanistan. Informed consent was obtained from all participants (school directors, teachers, children, parents or other relatives). We adopted a two-stage random sampling design (Figure 1). At baseline, we recruited gender-balanced samples of 11–16 year olds and primary caregivers (n = 1,362 respondents) using stratified sampling in schools, both in Kabul, Afghanistan (180 boys, 184 girls, 364 primary caregivers) and Peshawar, Pakistan (160 boys, 157 girls, 317 caregivers). One year later, our local team of interviewers sought to re-interview all baseline dyads: we were able to trace 64% and 31% of dyads in Kabul and Peshawar, respectively. We had no refusals for survey participation: attrition occurred because we could not trace children who had left school or had been repatriated in the wake of refugee camp closure.

image

Figure 1. Sampling procedure using stratified random sampling in schools to recruit a gender-balanced sample of 11–16 year old children and primary caregivers at baseline and endline

Download figure to PowerPoint

Our field team consisted of one male and one female bilingual field supervisor, three male and three female interviewers, one professional translator, and one bilingual project manager. The same personnel conducted both baseline and follow-up interviews; three were available to work at both sites, in Kabul and Peshawar. Interviews were conducted in private, independently with children and adults, in Dari or Pashto. The field manager checked data sheets on a daily basis; verbatim statements were translated into English for thematic analysis. All translations and back translations were systematically reviewed for content validity by a group of trilingual fieldworkers and academic faculty with expertise in social work, anthropology, area studies, and clinical psychology. All steps conformed to procedures for preparing instruments in transcultural research (Van Ommeren et al. 1999). Building upon pilot studies conducted in 2003 and 2004, and extensive qualitative work in 2005, we launched the baseline survey in Peshawar in 2005 (May–June) and in Kabul in 2006 (May–July). The follow-up interviews were conducted in Peshawar in 2006 (December–February) and in Kabul in 2007 (October–November).

Mental health (MH) indicators

We developed two-language versions (Dari/Pashto) of several standardized rating scales recommended for crosscultural epidemiological research in school- or community-based samples in conflict settings (Panter-Brick et al., 2009). They were implemented at baseline and follow-up. For caregivers, we used the Self-Report Questionnaire (SRQ-20), a simple and cost-effective screening tool to measure MH (20 items, yes/no answers); in our study, it demonstrated good internal reliability (Cronbach α = .83) and an excellent overlap with the Afghan Symptom Checklist, an instrument developed in Kabul to measure psychological distress with culturally specific terminology (Panter-Brick et al., 2009, 2011). For children, we included (a) the Child Revised Impact of Events Scale (CRIES, 13 items, 4-point scale) for post-traumatic stress, (b) the Birleson Depression Self-Rating Scale (DSRS, 18 items, 3-point scale), and (c) the Strength and Difficulties Questionnaire (SDQ, 25-items) covering behavioral, emotional, and social problems for multi-informant completion. The SDQ allowed explicit comparison of child-rated, parent-rated, and teacher-rated scores about the same child: namely, psychiatric difficulties (Total Difficulties Scores, TDS), interference on domains of social life (Impact Scores), and strength (Prosocial Scores). It provided a genuine dimensional measurement of child MH across the full range of TDS (Goodman & Goodman, 2009). All scales demonstrated acceptable or good internal reliability (CRIES, α = .82; DSRS, α = .69, α = .66 for child-rated TDS, α = .77 for caregiver-rated TDS) and 7-day test-retest reliability (Spearman Brown r = .76, r = .78, r = .57, r = .57 respectively). Being consistent with other ratings, teacher-rated SDQ scores were not reported in this paper.

Trauma, risk and protective experiences

We developed a Traumatic Events Checklist adapted from the Harvard Trauma Questionnaire previously used in Afghanistan (Cardozo et al., 2004) to assess lifetime trauma (20 yes/no events), and we qualitatively identified the nature of traumatic experiences. We assessed past-year stressors and protective factors with a separate checklist. Stressors (15 items) included health, family events, debt, domestic, and community conflict, whereas protective factors (12 items) related to improved health, neighborhood relationships, family circumstances, living conditions, school or work conditions, and perceived neighborhood trust. Such items were identified to be culturally relevant risk and protective factors from content analyses of our 1,011 caregiver and 1,011 child interviews at baseline (Eggerman & Panter-Brick, 2010). We randomly selected item-starting points at each interview, and used show-cards to illustrate ratings on current status (bad/so-so/good) and intervening-year changes (worse/same/better).

Family socioeconomic and demographic data

We collected data on household composition, displacement history, parent education and occupation, and children's school attendance, performance, and work activity after school. We used several measures of socioeconomic status: number of household material items, number of wage earners, and the caregiver's evaluation of household economic vulnerability on a 4-point scale, namely, food insecure (very poor), unable to buy items such as clothing (poor), able to afford most commodities (average), or able to cover their needs (better off).

Statistical analyses

Analyses were conducted in three steps. First, we removed 4 cases (0.3% of 1,362 interviews) with incomplete information for MH, described the cohort using standard descriptive statistics, and tested for site and gender differences. To assess attrition bias, we compared participants retained versus lost-to-follow-up on all sociodemographic and MH data. Second, we evaluated the predictors of MH at T1 using multiple regressions. Third, we identified the predictors of MH at T2 for children participating in surveys at both time-points (n = 331). We examined the age distribution of child MH symptom scores; cohort effects did not simply mirror age-related changes, as established in previous work from age-related changes evidenced in cross-sectional data (n = 1,011 Afghans, 11–16 year old) (Panter-Brick et al., 2011). Given the nature of delegated childcare responsibilities in Afghan families, we ran sensitivity analyses restricted to dyads whose caregivers were the same person at baseline and follow-up (n = 201), and also compared findings among dyads in which caregivers were biological parents or other close relatives. All analyses were performed accounting for the probability of selecting boys and girls within schools and potential clustering by size of schools (using the STATA Version 12.1 survey command for a complex multistage design, STATA Corporation, College Station, TX).

We used T2 child MH as the dependent variable, adjusting for T1 child MH score, child age, and household wealth. We produced separate multivariate regression models for eight dimensional outcomes of interest: (i) child-rated CRIES, (ii) child-rated DSRS; (iii-v) child-rated SDQ total difficulties, impact, and prosocial scores, and (vi-viii) caregiver-rated SDQ total difficulties, impact, and prosocial scores. Caregiver MH (SRQ-20) was the independent variable of interest. We used a pair of indices to characterize the situation of children who lived under the care of adults with different overall MH status (mean SRQ-20) and under the care of adults whose MH improved or worsened over time (T2-T1 differences in SRQ-20 scores). Following Oldham (Oldham, 1962), we used caregiver mean SRQ-20 scores rather than SRQ-20 score at baseline to avoid spurious correlation and bias in the regression coefficients (our pair of indices, mean and difference SRQ-20 scores was uncorrelated in our dataset, r = −.008, p = .880). We tabulated readily interpretable regression coefficients, wherein positive coefficients indicated that increases in caregiver MH scores were associated with increases in child MH scores over time. We showed standardized regression coefficients for caregiver MH, such that adjusted regression coefficients were independent across rows of tabulated data: the size of coefficients indicated the relative strength of associations between child MH outcomes and predictor variables.

In addition to caregiver MH, our final regression models included three individual-level variables measured at baseline (child sex; age; lifetime trauma exposure), one area-level variable (Kabul vs. Peshawar), and five family-level variables measured at follow-up (past-year domestic violence, major family conflict, serious family illness, better family unity, and better family life at home). We included variables with demonstrable statistical significance in previous analytical steps, and demonstrated that results were similar whether past-year life events were based on child-only, caregiver-only, and any child/caregiver data. We checked for other potentially important variables, such as father and mother literacy, and for potentially important interactions, such as site*gender, site*caregiver MH, gender*caregiver MH, and literacy*caregiver MH. We ran sensitivity analyses comparing results by parent literacy (with and without primary school education) and by site (with and without the Peshawar cohort that suffered high attrition).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Sample characteristics at baseline

The Kabul and Peshawar samples (n = 1,362 respondents) had similar demographic composition. In Kabul, caregiver—child dyads (n = 364) included 180 boys and 184 girls, with 161 mothers, 85 fathers, 37 close female and 81 close male relatives as primary caregivers. In Peshawar, caregiver—child dyads (n = 317) included 160 boys and 157 girls, cared for by 103 mothers, 80 fathers, 56 close female and 78 close male relatives. Combining both sites, children averaged 13.1 (SD 1.59) years of age and 4.99 (SD 2.22) years of formal education. Most lived in very cramped conditions, in households comprising a median of 10 family members. Half the households (57%) were reliant on a single wage-earner (57% in Kabul, 56% in Peshawar). Families had been displaced many times: three-quarters (n = 502, 74%) were forced to move home for conflict or economic reasons, once (31%), twice (16%), or three or more times (52%) since the birth of the child.

As expected, child MH outcomes were moderately intercorrelated (n = 331, data not shown). Caregiver SRQ-20 correlated with all child outcomes (n = 331 DSRS, r = .18, p = .001; child-rated TDS, r = .34, p < .001; child-rated Impact scores, r = .18, p = .001), except for CRIES (r = .07, p = .197).

Site and gender differences

We found site differences with respect to child age, child education, parent literacy, and socioeconomic status. In Peshawar relative to Kabul, children were one year younger at baseline (p < .001), lagging in primary school education by 1.5 years (p < .001), with a higher proportion of nonliterate fathers (p = .054) and nonliterate mothers (p = .046). The proportion of food-insecure households was the same (43%) at both sites, but more families considered themselves better off (in the sense of ‘having enough money to cover their needs’) in refugee camps (29% in Peshawar vs. 17% in Kabul, p < .001). We found no gender differences for sociodemographic variables, except that more Kabuli boys than girls were working after school (data not shown).

Follow-up sample

We recontacted 64% of the Kabul sample (n = 234 dyads), but only 31% of the Peshawar sample (n = 98 dyads) due to refugee repatriation that occurred in the interval between the two assessment periods. The follow-up samples were demographically similar at each site. In Kabul, we traced 115 boys and 119 girls, along with 43 fathers, 101 mothers, and 90 other close relatives; in 39% of cases, illness or work responsibilities meant that close relatives other than baseline respondents had assumed primary childcare responsibilities at follow-up. In Peshawar, we traced 55 boys and 42 girls, 32 fathers, 27 mothers, and 38 other close relatives; also in 39% of cases, out-migration, illness, or work meant that caregivers were not the same close relative at both time-points. Among the 331 families in our cohort, within-household caregiver substitution occurred in 130 cases: these were mostly gender-specific switches (88 cases, involving fathers, uncles, brothers, and grandfathers, or alternatively, involving mothers, sisters, sisters-in-law, grandmothers, and mothers' co-wives), with fewer switches occurring across gender (18 cases involving the biological parents and 24 cases involving other relatives). For all results of interest, sensitivity analyses demonstrated the same findings for the cohort including all caregiver—child dyads (n = 331) and the cohort restricted to dyads with the same caregiver at both time-points (n = 201). Moreover, the same findings pertained for data excluding the Peshawar cohort that suffered significant attrition.

Attrition bias

We examined baseline characteristics for follow-up (FU) versus lost-to-follow-up (LTFU) cohorts (Table 1aa). Cohorts were similar in terms of gender balance, age distribution, kin relation to child, child labor participation after school, and orphan status. They were also similar with respect to experiences of forced displacement (76% in FU and 71% in LTFU, p = .134) and expectations of having to move in the next year (31% in FU and 37% in LTFU, p = .142). Fewer food-insecure households were retained to FU (combined sample, p = .057, but nonsignificant differences per site).

Table 1a. Sociodemographic characteristics at baseline, for cohorts ‘lost versus retained’ to follow-up (n = 681 caregiver—child dyads)
 CohortLost to follow-up n = 350Retained for follow-up n = 331p-value
  1. a

    Significant differences in Kabul.

  2. b

    Significant differences in Peshawar.

  3. c

    Significant differences only in the combined sample.

ChildSex, n (%)
Male170 (49%)170 (51%).467
Female180 (51%)161 (49%)
Site, n (%)
Kabul130 (37%)234 (71%)<.001a,b
Peshawar220 (63%)97 (29%)
Age, years, mean (SD)13.02 (1.64)13.22 (1.55).096c
Education, years in school, mean (SD)4.42 (2.25)5.57 (2.08)<.001a,b
Working after school, n (%)
Yes62 (18%)65 (20%).520
No288 (82%)266 (80%)
CaregiverKin relation to child, n (%)
Father74 (21%)91 (27%).122
Mother137 (39%)127 (38%)
Other male relative83 (24%)76 (23%)
Other female relative56 (16%)37 (11%)
Age, years, mean (SD)36.22 (12.8)36.04 (12.05).850
HouseholdFather, n (%)
Living117 (90%)215 (92%).544
Died13 (10%)19 (8%)
Mother, n (%)
Living345 (99%)326 (98%).929
Died5 (1%)5 (2%)
Socioeconomic position, n (%)
Food insecure165 (47%)131 (39%).057c
Poor45 (13%)55 (17%)
Average57 (16%)74 (22%)
Better off83 (24%)71 (21%)
Father's education, n (%)
Not literate185 (55%)126 (39%)<.001a,b
Literate151 (45%)198 (61%)
Mother's education, n (%)
Not literate281 (82%)233 (71%).002b
Literate63 (18%)93 (29%)
Ever been forced to move home, n (%)
Yes248 (71%)254 (76%).134
No101 (29%)76 (24%)
Likely to move in the next year, n (%)
Yes103 (37%)83 (31%).142
No173 (63%)182 (69%)

There were notable differences for education. The FU cohort had children with one extra year of schooling at baseline (5.57 vs. 4.42 years, p < .001), and included more literate fathers (p < .001) and literate mothers (p < .001). Site-specific data (not tabulated) showed differences in children's years of schooling in Kabul (p = .001) and Peshawar (p = .014), in father literacy in Kabul (p = .066) and Peshawar (p = .005), and in mother literacy in Peshawar (p = .003) but not Kabul (p = .823).

There was little evidence of a MH-related attrition bias (Table 1bb). A difference in SDQ Impact Scores was noted, but only for child-rated data and not caregiver-rated data. Differences in trauma exposure (p < .001) were driven by the Kabul sample, in which significantly more adolescents with zero lifetime trauma were retained to FU (Kabul, p = .002; Peshawar, p = .203). We noted that caregiver—child MH associations at baseline were similar for FU and LTFU, except for child-rated DSRS and TDS where caregiver—child associations were stronger in LTFU (DSRS: FU r = .21 and LTFU r = .28, p = .053; TDS: FU r = .32 and LTFU r = .38, p = .035; data not tabulated).

Table 1b. Baseline mental health characteristics for child and caregiver cohorts lost versus retained to follow-up (n = 681 caregiver—child dyads)
 CohortLost to follow-up n = 350Retained for follow-up n = 331p-value
  1. a

    Significant differences in Kabul.

  2. b

    Significant differences in Peshawar.

ChildLifetime trauma, n (%)
0 event51 (15%)106 (32%)<.001a
1–2 events169 (48%)130 (39%)
≥3 events130 (37%)95 (29%)
Mental health, self-rated
SDQ total difficulties score, mean (SD)10.49 (5.20)10.44 (5.11).898
SDQ impact score, mean (SD)1.21 (2.05)1.60 (2.09).014b
SDQ prosocial score, mean (SD)8.98 (1.38)8.98 (1.24).992
CRIES, mean (SD)8.74 (10.35)7.95 (10.05).323
DSRS, mean (SD)9.90 (4.96)9.51 (4.63).288
Mental health, caregiver-rated
SDQ total difficulties score, mean (SD)12.05 (5.9)11.36 (5.66).125
SDQ impact score, mean (SD)1.15 (1.92)1.37 (1.99).142
SDQ prosocial score, mean (SD)8.34 (1.70)8.51 (1.64).186
CaregiverMental health, self-rated
SRQ20, mean (SD)8.49 (5.05)7.88 (4.62).105

Changes from baseline to follow-up

We showed one-year changes for selected outcomes in Figure 2. CRIES trajectories were flat (post-traumatic stress symptoms did not alleviate for Kabul, and showed high standard errors for Peshawar). Depression scores (DSRS) changed for Kabul boys (−29%, p < .001), Kabul girls (−22%, p < .001), and Peshawar girls (−21%, p = .020), but not Peshawar boys (−16%, p = .124). SDQ difficulties scores improved in Kabul (TDS, −32% for boys, p < .001, and −24% for girls, p < .001), but remained unchanged in Peshawar. As data were consistent, prosocial scores, impact scores, and caregiver-rated SDQ scores were not illustrated. We established that cohort-level changes exceeded the age-related changes observed in cross-sectional data, as reported elsewhere (Panter-Brick et al., 2011).

image

Figure 2. p-values of trends: ***p ≤ .01, **.01 < p ≤ .05, *.05 < p ≤ .10, ns, not significant. Data for the whole cohort (as illustrated) are similar to data for the cohort restricted to the same caregivers at both time-points

Download figure to PowerPoint

Caregiver MH showed pronounced gender differences (SRQ-20 scores for women are double those of men, p < .001); the same results pertained where caregivers were the same at both time-points. Cohort-level changes were significant (−24% for Kabul men, p = .005; −43% for Peshawar men, p = .002; −17% for Kabul women, p = .001; and −23% for Peshawar women, p = .057). One-year trends with the same caregivers at both time-points were steeper for males (−45% for Kabul, p < .001; −58% for Peshawar, p < .001) but not females (−9% for Kabul, p = .083; −24% for Peshawar, p = .08).

Regression models

Caregiver MH status was consistently and significantly associated with all eight child MH outcomes at follow-up, adjusted for baseline. As shown by the sign of adjusted regression coefficients in the first row of Table  2, children whose caregivers had higher mean SRQ-20 scores showed higher CRIES, DSRS, TDS, and Impact scores, and correspondingly lower Prosocial scores at follow-up. As shown by coefficients tabulated in the second variable row, associations between differences in caregiver SRQ-20 over time and child MH outcomes were not so consistent.

Table  2. Predictors of child mental health (n = 331 caregiver—child dyads)
 Self-rated child outcomes at follow-upCaregiver-rated child outcomes at follow-up
CRIESDSRSTDSImpactProsocialTDSImpactProsocial
  1. As shown by coefficients tabulated in the second variable row, associations for child outcomes at follow-up were adjusted for baseline scores, child age, and socioeconomic status. Adjusted regression coefficients, with 95% confidence intervals shown in brackets. CRIES: post-traumatic stress symptoms; DSRS: depression scores; Strength and Difficulties Questionnaire data for TDS, Total Difficulties Scores, Impact scores, and Prosocial scores. Significance level shown as: *.05 < p ≤ .10; **.01 < p ≤ .05; ***p ≤ .01.

Caregiver mental health
Overall status: mean SRQ-20 score, per SD1.04** [0.13, 1.96]0.65* [−0.02, 1.33]0.66** [0.06, 1.25]0.22* [−0.03, 0.47]−0.10* [−0.22, 0.02]1.63*** [1.04, 2.22]0.41*** [0.20, 0.61]−0.21* [−0.44, 0.01]
Difference over time: change in SRQ-20 score, per SD0.68* [−0.02, 1.38]0.17 [−0.28, 0.62]0.28 [−0.31, 0.87]0.06 [−0.09, 0.21]0.00 [−0.10, 0.10]1.31*** [0.86, 1.76]0.23*** [0.08, 0.38]−0.09 [−0.22, 0.04]
Baseline characteristics
Sex (female)2.56* [−0.04, 5.16]1.00* [−0.02, 2.02]1.02* [−0.12, 2.17]0.02 [−0.50, 0.54]−0.17 [−0.43, 0.09]−0.47 [−1.52, 0.58]−0.23 [−0.58, 0.13]0.02 [−0.42, 0.46]
Lifetime Trauma
1–2 events1.36* [−0.17, 2.88]0.87 [−0.32, 2.06]0.58 [−0.41, 1.57]0.35 [−0.18, 0.89]−0.2 [−0.61, 0.22]0.94** [0.03, 1.86]0.53* [−0.11, 1.17]−0.19 [−0.47, 0.10]
3+ events5.06*** [2.17, 7.95]0.22 [−1.47, 1.92]−0.47 [−1.76, 0.82]0.17 [−0.37, 0.71]−0.10 [−0.49, 0.29]0.26 [−0.89, 1.41]−0.01 [−0.55, 0.53]0.01 [−0.39, 0.40]
Site (Peshawar)−0.16 [−1.98, 1.67]1.37 [−1.12, 3.85]3.03*** [1.69, 4.37]−0.98*** [−1.36, −0.59]−0.21* [−0.43, 0.00]2.81*** [1.35, 4.28]−0.67** [−1.16, −0.18]−0.20 [−0.75, 0.34]
Maternal literacy1.08 [−1.23, 3.40]0.35 [−0.59, 1.30]0.58 [−0.39, 1.56]0.20 [−0.36, 0.76]−0.17 [−0.53, 0.20]0.91** [0.00, 1.81]0.12 [−0.20, 0.44]0.21 [−0.06, 0.47]
Past year events
Domestic violence−1.34 [−3.71, 1.04]0.52 [−1.01, 2.05]0.84*** [0.34, 1.34]0.47 [−0.13, 1.07]0.03 [−0.27, 0.34]1.87*** [0.66, 3.08]0.42 [−0.17, 1.00]−0.24 [−0.67, 0.18]
Family in a major conflict5.16* [−0.98, 11.31]2.31* [−0.16, 4.78]0.06 [−1.80, 1.93]0.38 [−1.17, 1.94]−0.01 [−0.56, 0.55]−0.51 [−3.05, 2.02]0.12 [−1.05, 1.30]0.20 [−0.42, 0.81]
Family member seriously ill0.53 [−1.18, 2.25]1.18** [0.06, 2.29]−0.15 [−1.53, 1.23]0.10 [−0.25, 0.44]0.42** [0.08, 0.77]−0.50 [−1.47, 0.47]0.19 [−0.12, 0.50]0.25* [−0.03, 0.54]
Better family unity−1.67 [−4.88, 1.55]0.39 [−0.49, 1.27]−0.19 [−1.08, 0.71]−0.44* [−0.93, 0.06]0.53*** [0.31, 0.75]0.30 [−0.85, 1.45]0.03 [−0.39, 0.44]0.30 [−0.10, 0.69]
Better family life at home−0.35 [−2.19, 1.49]−0.82 [−1.92, 0.27]−0.73** [−1.41, −0.05]−0.34* [−0.70, 0.02]0.00 [−0.33, 0.33]0.07 [−1.04, 1.18]−0.25 [−0.59, 0.08]−0.14 [−0.58, 0.31]

Thus for post-traumatic stress, a one SD change in mean SRQ-20 score for caregivers was associated with a 1.04 [0.13, 1.96] unit change on CRIES for children (p ≤ .05) at follow-up, whereas a SD change in SRQ-20 differences over time was associated with a 0.68 [−0.02, 1.38] unit change on CRIES (p ≤ .10). To put this finding in perspective, associations between caregiver and child MH health were comparable with the child experiencing one or two lifetime trauma events [1.36 (−0.17, 2.88), p ≤ .10] but less impactful than 3 +  trauma events [5.06 (2.17, 7.95), p ≤ .01]. These findings were corrected for other risk factors, such as female gender [2.56 (−0.04, 5.16)] and past-year occurrence of a major family conflict [5.16 (−0.98, 11.31)].

For depression, a one SD change in mean SRQ-20 was associated with a 0.65 [−0.02, 1.33] unit change in children's DSRS scores at follow-up, which was two thirds the impact attributed to female gender [1.00 (−0.02, 2.02), p ≤ .10] for this age group. Depression scores independently changed by 2.31 [−0.16, 4.78] points at follow-up with the occurrence of a major family conflict, and by 1.18 [0.06, 2.29] points where a family member fell seriously ill.

For SDQ total difficulties scores, impact on social life, and prosocial strength, there was good concordance in respondent-specific data (regression coefficients signs are consistent for child and caregiver ratings). Interestingly, caregiver—child MH associations were strongest for caregiver than child ratings: for example, a one SD change in caregiver SRQ-20 was associated at follow-up with a 1.63 [1.04, 2.22] unit change (p ≤ .01) in caregivers' proxy reports of child TDS when compared with a 0.66 [0.06, 1.25] unit change (p ≤ .05) in children's self-reports.

The associations between past-year domestic violence and follow-up SDQ total difficulty scores were highly significant (p ≤ .01) for both caregiver-rated [1.87 (0.66, 3.08)] and child-rated data [0.84 (0.34, 1.34)]. By contrast, the protective factors of family life were noted for children ratings, not caregiver ratings. For children, better ‘family life at home’ was associated with better MH at follow-up, namely, with −0.73 [−1.41,−0.05] unit change (p ≤ .05) in TDS and −0.34 [−0.70, 0.02] unit change (p ≤ .10) in Impact scores. Similarly for children, better ‘family unity’ was associated with beneficial changes in Impact scores (−0.44 [−0.93, 0.06], p ≤ .10) and Prosocial strength (0.53 [0.31, 0.75], p ≤ .01).

Sensitivity analyses by caregiver

Restricting the cohort to dyads with the same caregiver at both time-points (Table 3aa) strengthened the observed associations: we observed higher adjusted coefficients for all child-rated outcomes, and similar findings for caregiver-rated outcomes. For example, each SD change in mean caregiver SRQ-20 was associated at follow-up with a 0.85 unit change (p ≤ .05) in child-rated TDS in Table 3aa (n = 201, same caregiver at T1 and T2) when compared to a 0.66 unit change (p ≤ .05) in Table  2 (n = 331, all caregivers). We found that families in which caregivers were different at baseline and follow-up were more likely to report overcrowding (p = .040) and the occurrence of a family member getting married (p ≤ .001) in the intervening year, relative to families in which caregivers remained the same. Comparison of dyads in which caregivers were parents or extended relatives yielded similar findings (data not tabulated).

Table 3a. Sensitivity analyses for predictors of child mental health, with caregivers the same at baseline and endline (n = 201 caregiver—child dyads)
Caregiver mental healthSelf-rated child outcomes at follow-upCaregiver-rated child outcomes at follow-up
CRIESDSRSTDSImpactProsocialTDSImpactProsocial
  1. Regression models for child outcomes at follow-up were adjusted for all covariates and dependent variables present in Table  2. Adjusted regression coefficients with 95% confidence intervals shown in brackets. Significance level shown as: *.05 < p ≤ .10; **.01 < p ≤ .05; ***p ≤ .01.

Overall status: mean SRQ-20, per SD1.45* [−0.14, 3.05]0.83* [−0.16, 1.83]0.85** [0.20, 1.50]0.34** [0, 0.69]−0.19*** [−0.31, −0.07]1.49*** [0.87, 2.11]0.48*** [0.25, 0.72]−0.14* [−0.31, 0.03]
Difference over time: change in SRQ-20, per SD0.68 [−0.52, 1.89]0.37 [−0.81, 1.55]0.48 [−0.59, 1.56]−0.05 [−0.26, 0.16]−0.03 [−0.22, 0.17]1.19** [0.30, 2.09]0.24** [0.04, 0.44]−0.06 [−0.27, 0.15]

Sensitivity analyses by literacy

As an independent baseline variable, maternal literacy was not robustly associated with child MH outcomes (Table  2). We found, however, significant maternal literacy*MH interactions for several child-rated outcomes (CRIES, p ≤ .10; TDS, p ≤ .05, Prosocial, p ≤ .10; data not tabulated). Thus, maternal literacy moderated the strength of caregiver—child MH associations, where children (but not caregivers) reported on their own MH. This is shown in Table 3bb. In households where mothers were literate (n = 93), a change in caregiver mean SRQ-20 was associated with changes in all child-rated outcomes at follow-up (in the expected direction, being positively associated with CRIES, DSRS, TDS, and Impact scores, and negatively with Prosocial strength); moreover, adjusted regression coefficients were higher than those tabulated in the first two rows of Table  2). In households where mothers were nonliterate (n = 233), no such associations were detected. Father literacy made no robust impact on data variation, and results for mother and father literacy were consistent in both Kabul and Peshawar (data not shown).

Table 3b. Sensitivity analysis for predictors of child mental health for households with and without literate mothers (n = 326 caregiver—child dyads)
Caregiver mental healthSelf-rated child outcomes at follow-upCaregiver-rated child outcomes at follow-up
CRIESDSRSTDSImpactProsocialTDSImpactProsocial
  1. Regression models for child outcomes at follow-up were adjusted for all covariates and dependent variables present in Table  2. Adjusted regression coefficients with 95% confidence intervals shown in brackets. Significance level shown as: *.05 < p ≤ .10. **.01 < p ≤ .05; ***p ≤ .01.

Literate mothers (n = 93)
Overall status: mean SRQ-20, per SD2.92** [0.64, 5.19]1.46*** [0.49, 2.43]1.49*** [0.60, 2.38]0.56** [0.02, 1.09]−0.34** [−0.62, −0.07]1.25** [0.22, 2.29]0.36*** [0.11, 0.60]−0.24** [−0.45, −0.03]
Difference over time: change in SRQ-20, per SD1.38 [−0.86, 3.63]0.52** [0.01, 1.04]1.09*** [0.42, 1.75]0.30* [−0.05, 0.64]0.10 [−0.08, 0.27]1.89*** [0.99, 2.79]0.57** [0.13, 1.01]0.07 [−0.20, 0.35]
Non-literate mothers (n = 233)
Overall status: mean SRQ-20, per SD0.23 [−1.00, 1.47]0.48 [−0.23, 1.19]0.41 [−0.23, 1.05]0.10 [−0.17, 0.37]−0.03 [−0.13, 0.07]1.80*** [1.17, 2.43]0.48*** [0.23, 0.72]−0.21 [−0.50, 0.08]
Difference over time: change in SRQ-20, per SD0.65 [−0.31, 1.61]0.10 [−0.40, 0.60]0.06 [−0.51, 0.63]0.01 [−0.19, 0.20]−0.04 [−0.17, 0.09]1.21*** [0.41, 2.02]0.11* [−0.01, 0.23]−0.17** [−0.30, −0.04]

Having a literate mother entailed different caregiving arrangements and family dynamics. Relative to counterparts (n = 74), children in households with literate mothers (n = 54) were twice as likely to be looked after by their biological mothers (58% vs. 32%, p = .04), rather than under the care of grandparents, siblings, uncles, or aunts; they also lived in smaller-sized households (p < .01). By contrast, we detected no literacy-related differences in father supervision (fathers were the primary caregivers in 26% of father-literate households vs. 28% of father-illiterate households), nor differences in socioeconomic status.

Data on life events experienced between baseline and follow-up showed that our 331 households had experienced substantial demographic and socioeconomic stressors: 57 families (17%) had moved home, 16 (5%) were threatened with eviction, 223 (67%) had incurred substantial debt, 70 (21%) had a loss of salaried job, 231 (70%) had someone fall seriously ill, and 44 (13%) had someone move out. Households with maternal literacy were more likely to report better life at home (p = .002), better progress at school (p = .010), and for adults, a better situation at work (p = .003). They were more likely to report as severely stressful past-year events such the loss of a wage earner (p < .001) and an impending marriage (p = .026), and to report a major conflict with family members (p = .030), but not domestic violence.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Our work is one of the rare studies able to examine caregiver—child associations in humanitarian and refugee contexts, and the first to assess Afghan family dynamics and mental health outcomes prospectively. Despite the challenges of working in areas experiencing chronic conflict and in a culture where access to randomly selected girls and women is usually denied, we were able to draw a gender-balanced sample of Afghan school-aged children, and obtain family-level data by interviewing both children and their primary caregivers. Our contribution to the existing literature in global mental health, and specifically child and adolescent mental health, lies in documenting the strength of crossgenerational associations. We will discuss four main findings.

First, we showed that caregiver mental health was significantly associated with eight dimensional measures of child mental health (Table  2). A one SD change in caregiver MH (mean SRQ-20) was associated with changes in children's post-traumatic stress at follow-up (CRIES, 1.04 points, p ≤ .05); adjusted for baseline and other risk factors, it was equivalent to the predictive impact of a child's lifetime exposure to one or two trauma events (CRIES, 1.36 points, p ≤ .10). A one SD change in caregiver MH was also associated with changes in children's symptoms of depression (DSRS, 0.65 points, p ≤ .10) at follow-up, equivalent to two thirds of the effect attributed to female gender. Second, a SD change in caregiver MH was robustly associated with changes in children's strength and difficulties SDQ data (caregiver-rated TDS, 1.63 points, p ≤ .01), Impact scores (0.41 points, p ≤ .01), and Prosocial strength (−0.21 points, p ≤ .10). Such associations were found across a range of caregivers, whether biological parents or other close relatives, and whether the child is placed under the care of the same or different caregivers one year apart. They were also detected despite cohort-level improvements in mental health (for all outcomes except posttraumatic stress); this was unexpected, given the absence of a mental health intervention and the persistence of area-level violence, as discussed in more detail elsewhere (Panter-Brick et al., 2011). Our sample consisted of Afghan families who evidenced significant resilience, anchoring their children in school despite challenges including severe illness in the family, threats of eviction, loss of employment, debt, or major household conflict. These data point to the value of children's education as an engine of hope for socioeconomic advancement, a significant driver of family-level wellbeing in Afghanistan. They also point to a mental health cascade across generations, linking caregiver and child mental health, for better or for worse, over and above individual-level drivers of mental health such as gender and lifetime trauma exposure, or area-level differences between Kabul and Peshawar. As such, our findings strengthen those of other longitudinal studies with conflict-affected youth (Betancourt et al., 2013b) that highlight a compelling need for interventions to address family and community-level pathways of risk and resilience.

Second, we explicitly compared SDQ data across multiple respondents. This speaks to the issue of which informant might most reliably rate child mental health, and whether proxy reports by caregivers are likely to be biased by their own health status (Thielman et al., 2012). Caregiver—child associations were strongest for caregiver ratings - consonant with the view that adults in poor mental health might regard children under their care as having poor mental health also. For example, one SD change in caregiver mental health was associated with a change of 1.63 points in caregiver-rated TDS (p ≤ .01), but a change of only 0.66 points for child-rated TDS (p ≤ .05). By implication, exclusive reliance on caregiver data might inflate the strength of caregiver—child associations regarding mental health. We also implemented the SDQ while respecting a specific cultural construction of caregiving: in the highly gendered Afghan society, primary caregivers involve both fathers and mothers, and there is a common delegation of care to relatives other than biological parents. In contrast to other SDQ surveys in which mothers are nearly always selected as ‘parent’ respondents, reflecting the dominant construction of parenting (as mothering) in Western contexts, our survey involved a wide range of relatives who were designated as having primary caregiving responsibilities.

Third, assessing both child and adult ratings of life events and MH allows for greater depth of understanding with respect to which factors are critical for risk and resilience. Our multirespondent analyses point to family-level violence and quality of home life playing key roles for child mental health. Reports of domestic violence were associated with changes in both caregiver-rated TDS (1.87 points, p ≤ .01) and child-rated TDS (0.84 points, p ≤ .05) at follow-up. ‘Better home life’ was associated with changes in child-rated TDS and Impact scores, whereas, independently, better ‘family unity’ (Dari: ittifaq-i-famil) was negatively associated with Impact scores and positively with Prosocial scores (Table  2). Both adults and children readily identified the quality of proximate family dynamics for getting ‘better’ or ‘worse’ in terms of wellbeing, especially children for reports regarding family-level protective factors. This would argue that violence prevention and peace-building at family-level are critical leverage points for effective child mental health interventions – even in conflict settings, where attention to war violence often outweighs attention to domestic violence, and in refugee settings, where child-focused mental health interventions often outnumber family-level treatment or prevention. In the Afghan context, resilience-building interventions for promoting child mental health would include multisectoral interventions to alleviate the everyday stressors of overcrowding, improper housing, insecure jobs, and family conflict, stressors which Afghans readily identify as deleterious to family wellbeing.

Intriguingly, maternal literacy moderated the strength of caregiver—child mental health associations. Globally, this is a well-known risk factor for common mental health disorders, especially for the poor, who face the insecurity of income flow, the cognitive humiliation of social marginalization, and vulnerabilities to violence and uprootedness (Patel & Kleinman, 2003; Zournazi, 2002). The literature has shown that mothers' schooling is strongly related to child survival and other wellbeing outcomes (LeVine et al., 2012). Our data provide a twist on such findings, in that the beneficial effect of maternal literacy is only consistent in moderating the strength of caregiver—child associations, and then only for child ratings.

In our study, we systematically recorded educational data for the child's parents, but not for caregivers other than parents (who constituted 37% of the cohort). At cohort-level, maternal literacy was therefore a household marker, rather than a respondent attribute, and it was found to have a bearing on childcare and family dynamics. In Afghan households with literate mothers, children were twice as likely to be cared for by their own biological mothers (58% vs. 32%, p = .04), and more likely to report improved ‘home life’ (p = .002) and progress at school (p < .01). This would suggest that the caregiver's mental health had a stronger predictive impact on child mental health in households where the mother was literate. By contrast, father literacy was not consequent for the designation of the child's primary caregiver, and played no moderating effect.

Study limitations

Our data have several limitations, common to work in conflict settings. Our sample was restricted to school-attending children: we could not randomly sample those children who did not attend school due to economic constraints or cultural dictates (Panter-Brick et al., 2009). Our measures were also based on reported data, although we corroborated self-reports with stress biomarkers in a separate study (Panter-Brick, Eggerman, Mojadidi, & Mcdade, 2008). At follow-up, we suffered significant attrition in one of our two cohorts (without refusals to participation). With a 64% retention rate, the Kabul cohort represents a randomly selected sample with high external validity. By contrast, the 31% retention rate gives the Peshawar cohort lower external validity. Given the dearth of longitudinal data on refugee mental health (Fazel, Reed, Panter-Brick, & Stein, 2011; Reed et al., 2011), we featured both datasets in this study, but checked that results were robust to including Peshawar. There was decisive push by local government to close the Peshawar refugee camps, while remaining in line with the UNHCR-led policy of voluntary refugee repatriation. In 2006, Afghan refugee schools were closing because teachers were no longer receiving pay. By 2007, one refugee settlement, housing nearly 64,000 Afghans, had been entirely leveled by bulldozers, to make room for commercial and residential urban housing (Peshawar Weekly, 2007). The fact that refugee populations must often move in response to significant political and socioeconomic hardship makes them inherently difficult to include in research that aims for greater rigor, in drawing systematic rather than convenience samples (Reed et al., 2011).

Our most significant attrition bias was related to literacy: the cohort retained to follow-up included more literate fathers (p < .001), more literate mothers (p < .001), and children with one extra year at school (p ≤ .002). This suggests that relatively educated families (most of which had just primary school literacy) anchored their children at school, and that nonliterate families found it harder to prioritize child education given the many stressors of everyday life. The retention bias is indisputable in terms of education, but weak or absent in terms of mental health. We have no data to assess whether families lost-to-follow-up experienced similar changes over time, or the same strength of ‘caregiver impact’ on child mental health outcomes. Our data do indicate, however, that results are unlikely to be biased by attrition, given that caregiver—child mental health associations at baseline were either stronger (for child-rated DSRS and TDS) or similar (for all other outcomes) in families subsequently lost to follow-up.

In the Afghan context, a child is handed over from one caregiver to another, in cases where a family member migrates, falls ill, or secures employment. We found that in 39% of cases, caregivers were not the same relative at baseline and follow-up. To better reflect local caregiving contexts, our analyses included all caregiver—child dyads, even dyads in which children lived under the care of substitute caregivers. Statistically, we found that caregiver mental health status (SRQ-20), relative to difference over time (change in SRQ-20), was a stronger and more consistent predictor of child outcomes at follow-up, controlling for past-year risk and protective factors that might impact both child and caregiver mental health. Importantly, our findings are similar – even strengthened – where we restrict analyses to caregivers who were the same person at both time-points.

Finally, we used relatively straightforward regression models, rather than latent class trajectory analysis that can be applied to multiwave data (Betancourt et al., 2013b). Mindful of controversies regarding the analysis of change (Glymour, Weuve, Berkman, Kawachi, & Robins, 2005; Willett, 1989) and attentive to our main research question (Fitzmaurice, 2001), namely, whether caregiver mental health was prospectively associated with child mental health, we presented results with child outcomes at follow-up, adjusted for baseline. Results were similar in analyses that instead used the controversial yet intuitively appealing measure of adjusted difference scores. Nonetheless, the caregiver—child mental health associations identified in this study do not establish causality, nor rule out the absence of confounding factors influencing both adult and child wellbeing.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Our study has three significant implications for global mental health policy and practice. First, we support – with longitudinal data – arguments for moving from individual-level to family-level service provision. This entails adding assessments of caregiver mental health to community-based measures of child health, and the ‘establishment of family-based clinics in which both children and caregivers receive attention' (Thielman et al., 2012). Second, we extend recommendations to address the ‘unique adversities of gender and trauma’ in war-affected settings (Betancourt et al., 2013a) with consideration of the mental health cascade across generations. Third, we demonstrate the importance of dovetailing adult with child interviews in psychiatric epidemiology to identify (from risk and protective factors) the culturally relevant leverage points that shape mental health trajectories over time. The identification of culturally meaningful interventions that resonate with locally perceived needs is a highly prioritized research issue in humanitarian settings (Tol et al., 2011b, 2012). Our data show that family-level leverage points encapsulated by family unity, domestic violence, child schooling, and maternal literacy, deserve careful consideration within a mental health prevention agenda.

Prevention and intervention should not be designed, nor should they be evaluated, solely in terms of their impact ‘across the board’ for all sites and family contexts. They are far more likely to be successful if targeted in ways that address a cluster of adversities related to both caregiver and child wellbeing, rather than limited to the consideration of gender-specific differences, individual-level trauma exposures, and area-level prevalences. The recent emphasis placed upon the importance of pre-intervention assessments in humanitarian settings (World Health Organization & United Nations High Commissioner for Refugees, 2012) is a positive development in supporting this effort. In addition to being targeted, initiatives to promote mental health will be more impactful if they work across economic, social, educational, and health sectors to have synergistic impact on family life. In brief, culturally meaningful, targeted, synergistic, and resilience-building approaches are key pathways to promote child development and human wellbeing.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

This research was funded by the Welcome Trust. The authors have declared that they have no competing or potential conflicts of interest.

Key points
  • We conducted a two-wave survey with 331 caregiver—child dyads in Kabul and Peshawar to assess how strong and consistent were the associations between caregiver and child mental health.
  • Caregiver mental health was prospectively associated with eight measures of child mental health. One SD change in caregiver mental health was equivalent to the predictive impact of a lifetime exposure of one or two trauma events for child posttraumatic stress, and approached the association of child depression with female gender. Caregiver—child associations were stronger for caregiver-rated than child-rated data for data on prosocial strength and mental health difficulties.
  • We identified culturally meaningful leverage points for building family-level resilience, such as family unity and maternal literacy, relevant to a mental health prevention agenda.
  • Improving child mental health requires family-level interventions to address a cluster of adversities that impact family dynamics and wellbeing across generations.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  • Betancourt, T.S., Borisova, I., Williams, T.P., Meyers-Ohki, S.E., Rubin-Smith, J.E., Annan, J., & Kohrt, B.A. (2013a). Research review: Psychosocial adjustment and mental health in former child soldiers – a systematic review of the literature and recommendations for future research. Journal of Child Psychology and Psychiatry, 54, 1736.
  • Betancourt, T.S., Mcbain, R., Newnham, E., & Brennan, R. (2013b). Trajectories of internalizing problems in war-affected Sierra Leonean youth: Examining conflict and post-conflict factors. Child Development, 84, 455470.
  • Betancourt, T., Yudron, M., Wheaton, W., & Smith-Fawzi, M.C. (2012). Caregiver and adolescent mental health in Ethiopian Kunama refugees particpating in an emergency education program. Journal of Adolescent Health, 51, 357365.
  • Cardozo, B.L., Bilukha, O.O., Crawford, C.A.G., Shaikh, I., Wolfe, M.I., Gerber, M.L., & Anderson, M. (2004). Mental health, social functioning, and disability in postwar Afghanistan. Journal of the American Medical Association, 292, 575584.
  • Donini, A. (2007). Local perceptions of assistance to Afghanistan. International Peacekeeping, 14, 158172.
  • Dupree, N.H. (2004). The family during crisis in Afghanistan. Journal of Comparative Family Studies, 35, 311331.
  • Eggerman, M., & Panter-Brick, C. (2010). Suffering, hope, and entrapment: Resilience and cultural values in Afghanistan. Social Science & Medicine, 71, 7183.
  • Fazel, M., Reed, R., Panter-Brick, C., & Stein, A. (2011). Mental health of displaced and refugee children resettled in high-income countries: Risk and protective factors. The Lancet, 379, 266282.
  • Fitzmaurice, G.M. (2001). A conundrum in the analyses of change. Nutrition, 17, 360361.
  • Glymour, M., Weuve, J., Berkman, L.F., Kawachi, I., & Robins, J. (2005). When is baseline adjustment useful in analyses of change? An example with education and cognitive change. American Journal of Epidemiology, 162, 267278.
  • Goodhand, J., Dennys, C., & Mansfield, D. (2012). A dangerous peace? Drugs, post-conflict state building and horizontal inequalities in Afghanistan. In A. Langer, F. Stewart & R. Venugopal (Eds.), Horizontal inequalities and post-conflict development (pp. 249274). London: Palgrave Macmillan.
  • Goodman, A., & Goodman, R. (2009). Strengths and difficulties questionnaire as a dimensional measure of child mental health. Journal of the American Academy of Child and Adolescent Psychiatry, 48, 400403.
  • Husain, N., Chaudhry, I.B., Afridi, M.A., Tomenson, B., & Creed, F. (2007). Life stress and depression in a tribal area of Pakistan. British Journal of Psychiatry, 190, 3641.
  • Kessler, R.C., McLaughlin, K.A., Green, J.G., Gruber, M.J., Sampson, N.A., Zaslavsky, A.M., … & Williams, D.R. (2010). Childhood adversities and adult psychopathology in the WHO World Mental Health Surveys. The British Journal of Psychiatry, 197, 378385.
  • Leckman, J., & Leventhal, B. (2008). Editorial: A global perspective on child and adolescent mental health. Journal of Child Psychology and Psychiatry, 49, 221225.
  • LeVine, R., LeVine, S., Schnell-Anzola, B., Rowe, M., & Dexter, E. (2012). Literacy and mothering: How women's schooling changes the lives of the world's children. Series: Child development in cultural context. New York: Oxford University Press.
  • Miller, K.E., & Rasmussen, A. (2010). War exposure, daily stressors, and mental health in conflict and post-conflict settings: Bridging the divide between trauma-focused and psychosocial frameworks. Social Science & Medicine, 70, 716.
  • Oldham, P. (1962). A note on the analysis of repeated measurements of the same subjects. Journal of Chronic Diseases, 15, 969977.
  • Oxfam (2006). Free, quality education for every Afghan child. Oxford, UK: Oxfam International.
  • Panter-Brick, C., Eggerman, M., Gonzalez, V., & Saftar, S. (2009). Violence, suffering, and mental health in Afghanistan: A school-based survey. The Lancet, 374, 807886.
  • Panter-Brick, C., Eggerman, M., Mojadidi, A., & Mcdade, T. (2008). Social stressors, mental health, and physiological stress in an urban elite of young Afghans in Kabul. American Journal of Human Biology, 20, 327641.
  • Panter-Brick, C., Goodman, A., Tol, W., & Eggerman, M. (2011). Mental health and childhood adversities: A Longitudinal Study in Kabul, Afghanistan. Journal of the American Academy of Child and Adolescent Psychiatry, 50, 349363.
  • Panter-Brick, C., & Leckman, J.F. (2013). Editorial Commentary: Resilience in child development: Interconnected pathways to wellbeing. Journal of Child Psychology and Psychiatry, 54, 333336.
  • Patel, V., Flisher, A.J., Hetrick, S., & Mcgorry, P. (2007). Mental health of young people: A global public-health challenge. The Lancet, 369, 13021313.
  • Patel, V., Flisher, A.J., Nikapota, A., & Malhotra, S. (2008). Promoting child and adolescent mental health in low and middle income countries. Journal of Child Psychology and Psychiatry, 49, 313334.
  • Patel, V., & Kleinman, A. (2003). Poverty and common mental disorders in developing countries. Bulletin of the World Health Organisation, 81, 609615.
  • Peshawar Weekly (2007). UN closes oldest Afghan refugee camp in Pakistan. Peshawar: Peshawar Weekly.
  • Rahman, A., & Hafeez, A. (2003). Suicidal feelings run high among mothers in refugee camps: A cross-sectional survey. Acta Psychiatrica Scandinavica, 108, 392393.
  • Reed, R., Fazel, M., Jones, L., Panter-Brick, C., & Stein, A. (2011). Mental health of displaced and refugee children resettled in low and middle-income countries: Risk and protective factors. The Lancet, 379, 250265.
  • Schmeidl, S. (2009). Repatriation to Afghanistan: Durable solution of responsibility shifting? Forced Migration Review, 33, 2022.
  • Thielman, N., Ostermann, J., Whetten, K., O'Donnell, K., & Team, T. P. O. F. O. R. (2012). Correlates of poor health among orphans and abandoned children in less wealthy countries: The importance of caregiver health. PLoS ONE, 7, e38109.
  • Tol, W.A. (2013). Annual Research Review: Resilience and mental health in children and adolescents living in areas of armed conflict: A systematic review of findings in low- and middle-income countries. Journal of Child Psychology and Psychiatry, 54, 44560.
  • Tol, W.A., Barbui, C., Galappatti, A., Silove, D., Betancourt, T.S., Souza, R., … & van Ommeren, M. (2011a). Mental health and psychosocial support in humanitarian settings: Linking practice and research. Lancet, 378, 15811591.
  • Tol, W.A., Komproe, I.H., Jordans, M.J.D., Gross, A.L., Susanty, D., Macy, R.D., de Jong, J.T.V.M. (2010). Mediators and moderators of a psychosocial intervention for children affected by political violence. Journal of Consulting and Clinical Psychology, 78, 818828.
  • Tol, W.A., Patel, V., Tomlinson, M., Baingana, F., Galappatti, A., Panter-Brick, C., … & van Ommeren, M. (2011b). Research priorities for mental health and psychosocial support in humanitarian settings. PLoS Med, 8, e1001096.
  • Tol, W.A., Patel, V., Tomlinson, M., Baingana, F., Galappatti, A., Silove, D., … & Panter-Brick, C. (2012). Relevance or excellence? Setting research priorities for mental health and psychosocial support in humanitarian settings. Harvard Review of Psychiatry, 20, 2536.
  • Tol, W., Rees, S., & Silove, D. (2013). Editorial: Broadening the scope of epidemiology in conflict-affected settings: Opportunities for mental health prevention and promotion. Epidemiology and Psychiatric Sciences, 22, 17.
  • Tol, W.A., & van Ommeren, M. (2012). Evidence-based mental health and psychosocial support in humanitarian settings: Gaps and opportunities. Evidence-Based Mental Health, 15, 2526.
  • Trani, J.-F., Bakhshi, P., Noor, A.A., Lopez, D., & Mashkoor, A. (2010). Poverty, vulnerability, and provision of healthcare in Afghanistan. Social Science & Medicine, 70, 17451755.
  • Van Ommeren, M., Sharma, B., Thapa, S., Makaju, R., Prasain, D., & Bhattarai, R. (1999). Preparing instruments for transcultural research: use of the translation monitoring form with Nepali-speaking Bhutanese refugees. Transcultural Psychiatry, 36, 285301.
  • Ventevogel, P. (2005). The psychiatric epidemiological studies in Afghanistan: A critical review of the literature and future directions. Journal of Pakistan Psychiatric Society, 2, 912.
  • Waldman, R., Strong, L., & Wali, A. (2006). Afghanistan's health system since 2001: Condition improved, prognosis cautiously optimistic. Kabul, Afghanistan: Afghanistan Research and Evaluation Unit.
  • Willett, J. (1989). Questions and answers in the measurement of change. Review of Research in Education, 15, 345422.
  • World Health Organization & United Nations High Commissioner for Refugees (2012). Assessing mental health and psychosocial needs and resources: Toolkit for humanitarian settings. Geneva: WHO.
  • Zournazi, M. (2002). Hope: New philosophies for change. Annandale, Australia: Pluto Press.