The study aims to evaluate the impact of perinatal common mental disorders (CMDs) on child mortality up to 3.5 years in a demographic surveillance site at Butajira, Ethiopia.
The study aims to evaluate the impact of perinatal common mental disorders (CMDs) on child mortality up to 3.5 years in a demographic surveillance site at Butajira, Ethiopia.
One thousand sixty-five eligible women were assessed for CMD in the third trimester of pregnancy and at 2 months post-delivery using the Self-Reporting Questionnaire. We derived a four-level categorical exposure variable for the course of perinatal CMD. The outcome measure was child death recorded from 1 month after the postnatal assessment up to 3.5 years. Potential confounders and mediators were evaluated.
The cumulative child mortality rates were 62.6/1000 at 1 year and 82.5/1000 at 3.5 years, respectively. Exposure to perinatal CMD did not significantly affect child survival at 3.5 years, with results showing fully adjusted hazard ratio (HR) and 95% confidence interval (95% CI) of 1.85 (0.43, 7.88) for CMD in pregnancy only, 1.47 (0.14, 15.66) for CMD in postnatal period only and 0.41 (0.02, 7.38) for persistent CMD (both in pregnancy and postnatal). Only using soap less frequently than daily (HR 5.67, 95% CI 1.58–20.30) and episode of malaria in pregnancy (HR 5.02, 95% CI 2.15–11.72) were associated with child mortality in multivariable analysis.
Maternal health, health behaviours and family structure appear to be the most important factors affecting post-neonatal child mortality in this Ethiopian birth cohort, with little evidence for an effect of maternal perinatal CMD.
Although the child mortality rate in Ethiopia has been reported to be among the highest in the world, recent reports showed a marked decline with the infant and under-5 child mortality rates dropping from 124 to 67/1000 births and 210 to 104/1000 births, respectively, between 1990 and 2009. However, further efforts are needed in order to achieve the Millennium Development Goal of a two-thirds reduction in child mortality by 2015.
In sub-Saharan Africa, studies have shown that premature birth, low birthweight and infant illnesses are the main proximate causes of child death, with undernutrition underlying 50% of the mortality associated with these conditions.[2, 3] It has been postulated that all social and economical determinants of child mortality necessarily operate though a common set of biological mechanisms, or proximate determinants, to exert an impact on mortality.
Perinatal common mental disorders (CMDs), characterised by significant levels of depressive, anxiety and somatic symptoms occurring in pregnancy and the postnatal period, appear to be as prevalent in Ethiopia as in the West. There is accumulating evidence that maternal mental ill-health is associated with some of the established proximal determinants of child mortality. In South Asia, postnatal CMD is consistently associated with child under-nutrition,[6-9] but the situation in sub-Saharan Africa is less clear-cut. Although clinic-based studies have found an association between maternal CMD and infant under-nutrition in Nigeria and Malawi.[10, 11] This was not confirmed by our earlier findings in Ethiopia,[12, 13] and other population-based studies from Ethiopia and South Africa.[14, 15]
Earlier studies from Ethiopia have identified poor socio-economic indexes, poor maternal health practices, short birth interval and early cessation of breastfeeding[16-18] as factors associated with child mortality. Maternal depression has been reported recently as a risk factor for child mortality in Ethiopia, but only in the presence of intimate partner violence.
The Perinatal Maternal Mental disorder in Ethiopia (P-MaMiE) study is a population-based birth cohort located in a predominantly rural area of Ethiopia. We previously showed that maternal antenatal CMD was not associated with stillbirth or neonatal mortality in this cohort. In this paper we evaluated the impact of perinatal CMD on post-neonatal infant and child mortality up to 3.5 years.
A population-based prospective cohort of pregnant women was established within the framework of the demographic surveillance site (DSS) in Butajira, a predominantly rural setting approximately 135 km south of Addis Ababa, the capital city of Ethiopia. The DSS enumerators identified pregnant women during their routine surveillance. Eligible women were then interviewed by female data collectors employed to work full time on the P-MaMiE project. Eligibility criteria include: (i) having been identified in the third trimester of pregnancy between July 2005 and February 2006; (ii) aged between 15 and 49 years; (iii) ability to communicate in Amharic, the official language of Ethiopia; (iv) being a resident of the DSS site; and (v) consenting to participate in the study. Participating women were followed up, together with their newborn, until 3.5 years postpartum.
Perinatal CMD was measured during the third trimester of pregnancy and at two months postnatal using the Self-Reporting Questionnaire (SRQ-20). The 20-item scale asks about depressive, anxiety and somatic symptoms in the preceding month, and generates a continuously distributed scale score indicating overall mental morbidity. The Amharic version of the SRQ-20 validated previously for use in perinatal women in the Butajira population was used in this study. It was administered by trained female data collectors as interviewers.
Child deaths were recorded from a month after the postnatal assessment of perinatal CMD up to 3.5 years. This limitation to 1 month after the postnatal CMD assessment was imposed because we believed that the death of a baby between 28 days and 2 months post-delivery could have led to the development of maternal postnatal CMD, rather than our hypothesised direction of association from postnatal CMD to child mortality, The cumulative child mortality, time-to-death and loss to follow up were recorded. Children whose mother died during the follow-up period were excluded.
Means and proportions were used to describe continuous and categorical characteristics, respectively. χ2, t-test, and analysis of variance and log-rank tests were used to evaluate bivariate associations.
Post-neonatal child mortality up to 3.5 years was the main outcome. The main exposure variable was the four-level categorical variable for perinatal CMD (SRQ-20 ≥ 6) derived as follows: no CMD in pregnancy or postnatal 2 months (no CMD), CMD in pregnancy only (CMD-pregnancy), CMD at 2 months postnatal only (CMD-postnatal), and persistent perinatal CMD both during pregnancy and at 2 months postnatal (CMD-persistent). For multivariable analysis, we used Cox proportional hazard regression to evaluate the effect of different variables on the time to death. Three criteria were used to determine the variables to be included in the final model: theoretically potential cofounders, associations with increasing categories of perinatal CMD and associations with child mortality. All data analysis was carried out using STATA version 10 and hazard ratios were presented with 95% confidence intervals (CIs).
Women were identified by the enumerators and, after giving informed consent, were interviewed. The study was conducted with the approval of the Ethiopian Science and Technology Agency (National Research Ethics Committee) and King's College London Research Ethics Committee.
While 1065 women were recruited into the study, excluded from the final analysis were those with stillbirths (n = 40), multiple births (n = 16), or who migrated or died during pregnancy (n = 3), leaving a sample of 1006 births. The largest ethnic group was Meskan (46.7%). The mean age was 26.9 years (SD 6.3) with a range of 15 to 44 years. Almost all women (99.1%) were married, of whom 175 (18.2%) were in a polygamous marriage.
There were 40/1046 stillbirths (38.2/1000 births) and out of the remaining sample of 1006, there were 35 neonatal deaths (34.8/1000 live births), with 63 children dying in the first year (62.6/1000) and 83 dying by 3.5 years (82.5/1000). With data limited to singleton live births, and mortality limited to those from 1 month after the postnatal assessment of maternal CMD up to 3.5 years (for reasons earlier explained), there were 38 child deaths occurring among the 969 children remaining at risk.
A survival analysis was done with the maternal CMD groups (no CMD, CMD-pregnancy, CMD-postnatal and CMD-persistent). The survival estimates for the main exposure groups (Fig. 1) showed that exposure to perinatal CMD did not significantly affect child survival at 3.5 years. This was supported with incidence rate ratios (IRRs) of 0.71 (0.14, 2.22) for CMD-pregnancy, 2.11 (0.25, 8.30) for CMD-postnatal and 1.65 (0.19, 6.37) for CMD-persistent, respectively. A log-rank test of equality of survival function further confirmed this findings for CMD-pregnancy (P = 0.803), CMD-postnatal (P = 0.772) and CMD-persistent (P = 0.980), respectively.
In our Cox regression analysis (mortality = 38), when compared with the no-CMD group, the crude hazard ratio (HR) and 95% CI were 0.86 (0.26, 2.80) for CMD-pregnancy, 1.32 (95% CI 0.18, 9.66) for CMD-postnatal and 1.01 (95% CI 0.14, 7.38) for CMD-persistent. Table 1 showed that adjusting separately for maternal factors, socio-economic/living situation, social support, maternal functioning, and health and child factors did not affect our finding of no association between maternal CMD and child mortality at 3.5 years. The fully adjusted HRs were 1.48 (95% CI 0.37, 6.00) for CMD-pregnancy, 1.04 (95% CI 0.12, 8.78) for CMD-postnatal and 0.61 (95% CI 0.06, 5.85) for CMD-persistent.
|HR (95% CI)|
|CMD in pregnancy only||CMD in postnatal only||Persistent CMD||SRQ-20-scores in pregnancy*||SRQ-20 scores postnatal|
|Crude HR||0.86 (0.26–2.80)||1.32 (0.18–9.66)||1.01 (0.14–7.38)||0.96 (0.87–1.06)||1.03 (0.91–1.16)|
|Adjusted separately for|
|Maternal factors||0.86 (0.26–2.82)||1.34 (0.18–9.89)||1.09 (0.15–8.10)||0.96 (0.86–1.06)||1.03 (0.91–1.17)|
|Socio-economic/living situation||0.87 (0.26–2.99)||1.21 (0.16–9.12)||0.67 (0.09–5.23)||0.93 (0.83–1.04)||1.00 (0.88–1.13)|
|Social support||0.98 (0.29–3.26)||1.26 (0.17–9.34)||1.06 (0.14–8.04)||0.97 (0.87–1.08)||1.03 (0.92–1.16)|
|Functioning and health||1.22 (0.33–4.48)||1.29 (0.16–9.41)||0.61 (0.07–5.07)||0.94 (0.82–1.06)||1.00 (0.85–1.16)|
|Child factors||1.00 (0.30–3.30)||1.76 (0.24–13.08)||1.12 (0.14–8.72)||0.97 (0.86–1.09)||1.04 (0.91–1.19)|
|Fully adjusted model||1.85 (0.43–7.88)||1.47 (0.14–15.66)||0.41 (0.02–7.38)||0.97 (0.83–1.15)||0.98 (0.81–1.19)|
|Fully adjusted without child factor||1.48 (0.37–6.00)||1.04 (0.12–8.78)||0.61 (0.06–5.85)||0.96 (0.83–1.11)||0.99 (0.84–1.17)|
Exploratory analyses were further conducted using the SRQ-20 scores as a continuous exposure variable. The results as revealed in Table 1 showed that SRQ-20 in pregnancy HR (uncensored data, total mortality figure = 123) was 0.96, 95% CI 0.87, 1.06 (fully adjusted HR 0.97, 95% CI 0.83, 1.15); and SRQ-20 in postnatal HR was 1.05, 95% CI 0.95, 1.16 (fully adjusted HR 1.00, 95% CI 0.90, 1.19).
Table 2 shows the crude and adjusted HRs between child mortality at 3.5 years and maternal factors, socio-economic factors, living situation, social support and child variables. In the final multivariable model, cumulative child mortality at 3.5 years was only significantly associated with less than daily use of soap (5.67, 95% CI 1.58, 20.30) and malaria in pregnancy (HR 5.02, 95% CI 2.15–11.72).
|Variables||Crude HR at 3.5 years||Adjusted HR|
|Age in years||1.02 (0.98–1.07)||1.05 (0.96–1.16)|
|Not functionally literate||0.68 (0.34–1.37)||0.73 (0.24–2.23)|
|Polygamous marriage||1.70 (0.90–3.22)||2.37 (0.98–5.72)|
|Parity||1.20 (0.77–1.88)||1.34 (0.53–3.40)|
|Number of children under 5 years||0.91 (0.61–1.37)||0.64 (0.34–1.19)|
|Hungry in the last 1 month||1.59 (0.81–3.12)||1.30 (0.49–3.43)|
|In debt that cannot pay||1.28 (0.51–3.24)||0.47 (0.11–1.92)|
|No resources to deal with emergency||0.86 (0.49–1.52)||0.95 (0.43–2.08)|
|Report of lower relative wealth||1.40 (0.78–2.53)||2.39 (0.98–5.80)|
|Standard of Living Scale||0.94 (0.68–1.46)||0.99 (0.68–1.46)|
|Using soap less frequently than daily||2.34 (1.13–4.83)||5.67 (1.58–20.30)|
|Urban residence||0.75 (0.30–1.89)||1.05 (0.18–6.32)|
|Unprotected water supply||0.69 (0.35–1.34)||0.57 (0.21–1.58)|
|Lowland residence||1.78 (1.00–3.17)||1.15 (0.41–3.20)|
|Sees family monthly or less||1.10 (0.59–2.01)||1.47 (0.66–3.30)|
|Sees friends monthly or less||0.83 (0.41–1.67)||0.70 (0.26–1.87)|
|Can't rely on family for help||0.72 (0.41–1.27)||0.61 (0.28–1.34)|
|Not enough help with housework||1.18 (0.65–2.12)||1.84 (0.78–4.29)|
|Score on Marital Discord Scale||0.84 (0.56–1.27)||0.55 (0.30–1.01)|
|Categories of empowerment||0.88 (0.59–1.32)||1.07 (0.62–1.85)|
|Maternal functioning and health|
|Poor functioning in last month||0.96 (0.38–2.41)||1.10 (0.30–4.00)|
|Life events since pregnancy||0.87 (0.59–1.29)||0.72 (0.41–1.27)|
|Fever lasting >1 day in pregnancy||1.16 (0.52–2.59)||0.64 (0.19–2.09)|
|Fever lasting >1 day postnatal||2.21 (1.24–3.94)||1.49 (0.66–3.33)|
|Any episode of malaria in pregnancy||2.68 (1.47–4.89)||5.02 (2.15–11.72)|
|Any episode of malaria postnatal||1.35 (0.63–2.89)||1.28 (0.47–3.51)|
|Using alcohol at least weekly||0.85 (0.21–3.50)||1.98 (0.41–9.64)|
|Using khat at least weekly||1.35 (0.63–2.89)||0.72 (0.23–2.24)|
|Diarrhoea lasting >1 day in pregnancy||2.06 (0.50–8.48)||1.19 (0.22–6.56)|
|Diarrhoea lasting >1 day postnatal||1.12 (0.35–3.61)||0.92 (0.25–3.43)|
|BMI (kg/m2) in pregnancy||1.01 (0.89–1.15)||1.12 (0.94–1.33)|
|No vaccination (at 2 months)||1.11 (0.60–2.06)||0.67 (0.31–1.49)|
|Delayed initiation of breastfeeding||0.90 (0.28–2.89)||0.25 (0.03–2.05)|
|Non-exclusive breastfeeding <6 months||1.29 (0.60–2.80)||0.92 (0.31–2.71)|
|Stunting at 2 months||0.46 (0.14–1.49)||0.24 (0.06–0.93)|
|Underweight at 2 months||1.52 (0.67–3.47)||1.90 (0.64–5.61)|
|Perceived severe infant illness||1.38 (0.69–2.76)||1.26 (0.57–2.86)|
Our research is one of the few community-based studies from a low- or middle-income country to investigate the possible long-term effect of maternal perinatal CMD on child survival. This is a follow-up to our earlier study which found no association between antenatal CMD and stillbirth or neonatal survival in Ethiopia. In this paper we concentrated on infants and children surviving beyond the first 28 days of life. The essential findings in our study were the 6.26% and 8.25% cumulative child mortalities at 1 year and 3.5 years, respectively. Our study failed to establish any association between perinatal CMD and child mortality at 3.5 years, with maternal self-reported episodes of malaria in pregnancy and not using soap daily the only factors independently associated with cumulative child mortality at 3.5 years.
Our child mortality rate of 8.25% may seem lower than the UNICEF figure of 10.4%, but it must be noted that we provided a 3.5-year rate instead of 5-year rate. Other possible reasons for the lower child mortality rate in our study are: (i) our moderate sample; (ii) the women were recruited from a DSS which may affect the level of health care and maternal awareness; and (iii) as a condition of the research ethics committee, the project paid for all health-care costs of participating mothers and babies – this may have led to improved health outcomes in the cohort. The health-care cost covered from 0 to 12 months and then from 2.5 to 3.5 years. The possibility of this contributing to mortality rate is being explored in a separate paper. However, there is evidence that infant mortality is decreasing in Ethiopia.
Results from our study failed to support the hypothesis that survival was lower in children whose mothers had high levels of CMD during the perinatal period. Our finding is in line with our earlier reports of no association between maternal CMD and neonatal mortality and infant undernutrition in this setting. A recent study in the same study area found an association between maternal depression and child mortality, but only in the presence of intimate partner violence. This study had a smaller sample size with few deaths in the relevant groups. We were not able to test for an interaction with intimate partner violence in our study. We only asked if the mothers had ‘experienced violence’ without specifying who had carried out the violence but the levels were very low (≤2.5%), and so testing for interaction was not possible.
Only maternal malaria episodes in pregnancy and use of soap less frequently than daily were independently associated child mortality at 3.5 years. Malaria in pregnancy may lead to child mortality by increasing risk of low birthweight, premature labour, intra-uterine growth retardation, placenta infection and stillbirth probably via placental malaria. Also, because the prevalence of HIV is very low in this population, the possible link between placental malaria and increased risk of mother to child HIV transmission, leading to reduced early life development of affected children and seen as indirect cause of infant and child mortality, is not likely to be the case in this study. Earlier studies have shown that infrequent use of soap may be associated with child mortality, either directly or indirectly via helminth infection in mothers and their infants. However, because this study is underpowered (due to low rate of child mortality) and the fact that we provided helminthic treatment at 12 months, we may not be able to ascertain this.
Other studies have previously investigated some of the putative causal mechanisms whereby perinatal CMD could lead to a reduction in child survival and only found associations with delayed initiation of breastfeeding (antenatal CMD) and increased risk of infant diarrhoeal episodes (postnatal CMD).[12, 33] Our study, however, differs from these earlier studies in that whereas we examined mortality up to 3.5 years and excluded early neonatal period, the previous studies have examined early neonatal mortality on a large sample. Also, rather than infant diarrhoea, we presented ‘severe infant illness’ and did not have enough data to separate out the cause of infant illness. Polygamous marriage was common in the cohort and close to being correlated with risk of child mortality in the multivariable model, so probably underpowered. Our earlier work showed that polygamy could promote competition between wives and less input from the father and all these may have a negative effect on the child.
Our study is not without limitations. The SRQ-20 used to measure CMD is a scale-based measure and not a definitive diagnostic assessment of mental disorders. However, it had been validated for perinatal women in this population and extensively used in the study area for assessments of CMD in the general population. Most importantly, the lower-than-expected child mortality and the low prevalence of maternal CMD observed in perinatal periods would have significantly reduced the power to detect any effect of maternal CMD on child mortality, and the wide CIs in our results is a strong evidence of this. Although to minimise the type II error in our study, we selected an overall sample size to give adequate power to investigate an effect of postnatal mental morbidity on infant mortality (at 1 year). Assuming 20% prevalence of postnatal mental disorder, we required a sample size of 700 women to detect a 2.6-fold increased risk in infant mortality with 80% power and α = 0.05. Despite having 1046 women in our study, both the maternal CMD and child mortality were much lower than anticipated.
It should be noted that although it did not reach a significant level, the lowest survival proportion was just above 0.8 and the highest just above 0.9; a difference in survival of about 10% should not be totally ignored in our subjects. It should also be noted that although the IRR for each group spans 1, their width is really large, which may suggest that a higher sample size may give clinically significant values.
The strength of our study included the relatively large population-based community sample, prospective design, high response and retention rate, and adjustment for a large number of cofounding variables.
In conclusion, our study failed to find a significant association between levels of maternal exposure to CMD and child mortality. The factors independently associated with child mortality were maternal malaria in pregnancy and less frequent use of soap. The small number of deaths in our study may have masked true association and there is a need for a larger study in order to achieve greater precision. However, the babies born to the mothers in the P-MaMiE cohort are still being followed up and we will be in a position to report any future possible associations between maternal CMD in the later postnatal life and child mortality.
This study was funded through a Wellcome Trust research Training Fellowship for C.H. (GR071643).