This is one of a few African studies in the published literature that has examined the prevalence and risk factors of childhood depressive disorders in a community sample. The principal finding of this study is that north-eastern Uganda had a high point prevalence of childhood depressive disorders. Secondly, that the ecological factors (represented by district in this study), markers of the quality of the child-principal caregiver relationship (nature of living arrangements and domestic violence) and the presence of psychiatric co-morbidities were the important independent determinants of childhood depression in this study.
This study was carried out in four disadvantaged districts of north-eastern Uganda two of which were suffering war conflict at the time of the study. Most of respondents were children of poor peasant farmers with little formal education and largely operating at subsistence level outside the formal cash economy. These study communities had a heavy burden of adverse psychosocial factors, domestic violence was reported by a fifth of the respondents, war trauma was reported by a third of the respondents, orphan hood was reported by a third of the children and only half of the children stayed with both parents.
The point prevalence of childhood depressive disorder syndromes (DDS) of 8.6% in this study is much higher than the rate of about 1% that has been reported in both rural and urban Ethiopia [12, 13] and of less than 2% reported in western studies [15, 19]. The prevalence of depression in this study (where half of the study districts were war affected) is however lower than that of 19.6% reported in war affected Sri Lanka . Attanayake and colleagues (2009)  in a meta-analysis involving 17 studies on mental disorders among children exposed to war reported a 43% elevation in the risk of depression associated with exposure to war. The results from the above mentioned meta-analysis suggest that the elevated prevalence of childhood depression observed in this study relative to the results from Ethiopia may have been due to the fact that half of the study districts were suffering ongoing war conflict at the time of the research.
Ecological, socio-demographic, adverse psychosocial and psychiatric factors were each independently significantly associated with childhood depression in this study. District in this study representing ecological factors was independently significantly associated with depression with the non-war affected districts of Kaberamaido (3.5%) and Tororo (3.9%) reporting lower rates of depression than the war affected districts of Gulu (5.0%) and Lira (23.1%). However the marked difference between the prevalence of depression in Gulu and Lira which were both war exposed suggests that there were additional area level factors at play influencing the district prevalence for depression. In this study it was not possible to investigate further the factors underlying these area level differences in the rates of depression. Among adults, Vinck and colleagues (2007)  in 4 war affected districts in north-eastern Uganda and Kinyanda and colleagues (2009)  in a 14 district study in Uganda observed similarly wide variations in district rates of depression. Among the investigated factors to account for these area level differences, Kinyanda and colleagues (2009) reported that literacy rates were inversely significantly associated with the district rates of depression .
In this study, the socio-demographic factor, ‘nature of living arrangement’ was independently significantly associated with depression. Children not living with both parents had an increased risk of developing depression which was higher in those staying with a single parent (either mother or father) and highest in those staying with neither parent (staying with grandparents or with non-relatives). The fact that in this study domestic violence was also independently significantly associated with childhood depression but not orphan hood seems to suggest that the important underlying construct was the quality of the ‘child-principal caregiver relationship’ but not just the absence of a biological parent(s). Indeed emotional unavailability of both mothers and fathers has consistently been reported to be a risk factor of childhood depression [30, 31]. In central Uganda, Minde (1975)  observed that polygamous family set-ups were associated with psychological problems in school going children. Sander and McCarty (2005) on parental risk factors for depression reported that these included parents who were: emotionally unavailable or uninvolved, who lacked warmth in their interactions, are overcontrolling, or who use harsh physical discipline . On domestic violence, Nicodimous and colleagues (2009)  among college going adolescents in Ethiopia reported that witnessing parental violence was associated with more than a 2 fold increased risk for depression.
In this study, co-morbidity with the psychiatric disorders/problems of emotional distress (assessed by the SDQ scale), suicidality, eating disorders and marginally anxiety disorder syndromes were independently significantly associated with childhood depression. Co-morbidity between depression and behavioral and developmental disorder syndromes was significantly protective against depression in this study. Pataki (2000)  on the epidemiology of childhood and adolescent depression observes that between 40–70% of children and adolescents with depression will have a co-morbid psychiatric disorder usually anxiety disorder, dysthmic disorder, disruptive behavioral disorder or a substance abuse disorder. In this study, surprisingly the rate of DDS was lower among subjects with a history of mental illness. Two plausible explanations can be offered for this observation. Firstly, given that only four participants had a history of mental illness, the small sample size may have led to spurious results. Secondly, professional help provided during attendance at the mental health facility may have effectively resolved the psychological problem hence lower levels of DDS.
In this study we found no significant independent effect for measures of socio-economic deprivation (nature of housing, family income, highest educational attainment of parent/guardian) and some of the adverse psychological factors (experience of war trauma and orphan hood). It is possible that these could be risk factors for childhood depression, acting indirectly through affecting the quality of the child-principal care-giver relationship or leading to psychiatric comorbidities.
In conclusion, there is a high prevalence of childhood major depressive disorders in disadvantaged north-eastern Uganda. Ecological factors, quality of the child-principal care-giver relationship and the presence of psychiatric co-morbidities were the important independent determinants of childhood depression in this study. We found that socio-economic deprivation, experience of war trauma and orphan hood were not significantly associated with depression, but it is possible that these factors had an indirect effect on depression through affecting either the child-principal care-giver relationship or leading to associated psychiatric co-morbidities.
On limitations, firstly, given that this study was cross-sectional in nature the causal direction between the dependent and the independent factors could not be ascertained, hence there is need for longitudinal studies to disentangle the casual directions between these factors. Secondly, the psychological assessment tools used in this study have never been formally validated in the Ugandan culture settings, however the SDQ has previously been validated in Congo-DRC (a socio-cultural setting similar to Uganda) with good results . In addition, to ensure accurate translation of the SDQ and the M.I.N.I.-KID from English into the 4 local dialects used in this study, a process of forward and back translation was undertaken using teams of local mental health professionals. Thirdly, because 4 dialect versions of the questionnaire were used in this study, reported district differences could be due to bias introduced by the translation process, this effect was however thought to be minimal as the process of forward and backward translation was undertaken carefully. Fourthly, this paper presents the results of a secondary analysis of data, and important data relating to the sample selection was not available, in particular it would have been helpful to have data on the age and sex composition of children in the sampled households, including children who were not included in the sample, in order to ascertain whether the age and sex distributions of sampled children were representative of children in each district. Korn and Graubard (1995) suggest that weighting the data to reflect the underlying age and sex distribution of the districts would impact on the estimated prevalence of depressive disorders but not on the estimates of association for the risk factors .
As recommendations, addressing childhood depressive disorder in this community requires both social interventions to improve the quality of the child-principal caregiver relationship (through interventions such as training in parenting skills and domestic violence counseling) and the development of comprehensive child and adolescent mental health services (to address the entire spectrum of childhood psychopathology due to a tendency for co-morbidity). There is a need for the formal validation of the SDQ and M.I.N.I.-KID in the African socio-cultural setting of Uganda.