The authors have declared that no competing interests exist.
Conceived and designed the experiments: RA JS JT AvdV. Performed the experiments: RA JS JT AvdV. Analyzed the data: RA JS. Wrote the paper: RA JS JT AvdV.
Short birth intervals are independently associated with increased risk of adverse maternal, perinatal, infant and child outcomes. Anemia in children, which is highly prevalent in Africa, is associated with an increased risk of morbidity and mortality. Birth spacing is advocated as a tool to reduce anemia in preschool African children, but the role of gender differences and contextual factors has been neglected. The present study aims to determine to what extent the length of preceding birth interval influences the hemoglobin levels of African preschool children in general, as well as for boys and girls separately, and which contextual factors thereby play a crucial role.
This cross-sectional study uses data from Demographic and Health Surveys (DHS) conducted between 2003 and 2011 in 20 African countries. All preschool children aged 6–59 months with a valid hemoglobin measurement and a preceding birth interval of 7–72 months as well as their corresponding multigravida mothers aged 21–49 years were included in the study. Hemoglobin levels of children and mothers were measured in g/l, while birth intervals were calculated as months difference between consecutive births. Multivariate analyses were done to examine the relationship between length of preceding birth interval and child hemoglobin levels, adjusted for factors at the individual, household, community, district, and country level. A positive linear relationship was observed between birth interval and the 49,260 included children’s hemoglobin level, whereby age and sex of the child, hemoglobin level of the mother, household wealth, mother’s education and urbanization of place of residence also showed positive associations. In the interaction models, the effect of a month increase in birth interval is associated with an average increase of 0.025 g/l in hemoglobin level (P = 0.001) in girls, while for boys the effect was not significant. In addition, for girls, the effect of length of preceding birth interval was highest in young mothers and mothers with higher hemoglobin levels, while for boys, the highest effect was noticed for those living in more highly educated regions. Finally, significantly higher hemoglobin levels of girls compared to boys were observed at birth but with increasing age, the sex difference in hemoglobin level gradually becomes smaller.
A longer birth interval has a modest positive effect on early childhood hemoglobin levels of girls, and this effect is strongest when their mothers are in their early twenties and have a high hemoglobin level. Remarkably, although the physiological iron requirement is higher for boys than girls, birth spacing has little influence on hemoglobin levels of preschool boys. We speculate that the preference for male offspring in large parts of Africa significantly influences nutritional patterns of African preschool boys and girls, and as such also determines the different effect of birth spacing. Finally, gender aspects should be considered in intervention programs that aim to improve anemia in African children.
Anemia is a major public health problem, affecting 1.62 billion people worldwide. The prevalence of anemia in low-income countries is around 43% in contrast to 9% in high-income countries [
The etiology of childhood anemia in Africa is multifactorial, including nutritional iron and folate deficiencies [
Delaying pregnancy until the mother’s current child reaches at least its second birthday increases the likelihood that the child gets adequate care during its first vulnerable years. It also may lengthen the lactating period, which is favorable as breast milk continues to be an important source of nutrition during the second year of life [
Despite the general consensus about the positive effects of birth spacing, the optimal duration of the interpregnancy interval is not clear. Some studies suggest that a birth interval of 2–3 years reduces child mortality while on the other hand recent studies from the United States Agency for International Development (USAID) recommend a birth interval of 3–5 years [
Socioeconomic status, lifestyle, stress, and adequacy of prenatal care, which are all associated with birth interval length, may also influence maternal, prenatal, infant, and child outcomes [
Although there is some circumstantial evidence suggesting a positive association, large-scale studies exploring the relation between birth spacing and anemia in pre-school children are still lacking, despite the circumstantial evidence suggesting a positive relation. The aim of the present study is therefore to determine to what extent the length of preceding birth interval influences the hemoglobin levels of African pre-school children in general, as well as for boys and girls separately and which contextual factors thereby play a crucial role.
This study used cross-sectional data derived from the Demographic and Health Surveys (DHS) [
Hemoglobin levels of children were available for children of up to 59 months. Within the 20 countries, 142 subnational regions (henceforth called ‘districts’), and 1999 clusters were distinguished. Information at district and cluster level was used to study effects of context factors. In all analyses, the household weights provided by DHS were used to obtain national representative samples. These weights were recoded to a mean of one, so that their application did not increase the sample sizes. For descriptive purposes, also national weights were constructed that made the sample representative of all countries together. To construct these national weights, the household weight factors were adjusted based on the World Bank population estimates for children aged 0–14 years within their respective country and year of survey [
Inclusion criteria were valid hemoglobin measurement, preceding birth interval of more than six months and less than six years and having a mother aged 21 and over. Mothers aged under 21 were excluded to avoid potential bias that might arise due to adverse outcomes of under-age pregnancies. A preceding birth interval of less than 7 months was considered implausible and of over six years as an indication of unusual household situation. The total number of children satisfying these criteria was 50,264. Of these children, 197 had to be removed because mothers BMI was missing, 147 because mother’s education was missing, 659 because the child’s age was missing and one because mother’s work was missing. The final study population therefore constituted of 49,260 children.
To measure hemoglobin levels, the blood of the women and children, obtained through a finger or heel prick, was tested for hemoglobin level. This testing was voluntary [
Preceding birth interval was expressed in months and defined as the period between the previous childbirth and the index child birth. The difference in months between two consecutive live births was considered a birth interval.
Control factors were included at the level of the children, the mother, the household and the context in which the household lived. Characteristics of the children that are associated with hemoglobin level include age (months), sex, birth order and twin status (singleton or multiple birth) [
Contextual factors were included at the cluster as well as at the district level. At cluster level, we included mean years of education of adults; preventive health measures (proportion of children having received vaccination against measles) and cluster altitude (in meters) [
Characteristic | Category | Sample |
---|---|---|
(N = 49,260) | ||
N/A | 34 (13.2) | |
N/A | 32.1 (15.5) | |
Male (%) | 50.5 | |
Female (%) | 49.5 | |
N/A | 4.6 (2.3) | |
N/A | 106.7 (17.2) | |
N/A | 124.9 (18.2) | |
N/A | 30.9 (6.1) | |
N/A | 22.4 (4.3) | |
N/A | 7.8 (7.1) | |
mean (SD) | ||
N/A | 3.5 (4) | |
N/A | 1.7 | |
Not in a union | 6.2 | |
First union | 78.9 | |
Second or later union | 14.9 | |
N/A | 62.2 | |
No breastfeeding | 38.7 | |
Index child breastfeeding | 35.1 | |
Other child breastfeeding | 26.2 | |
Rural | 79.3 | |
Urban | 20.7 | |
Poorest | 23.6 | |
Poorer | 22.2 | |
Middle | 21.9 | |
Richer | 19.1 | |
Richest | 13.2 | |
N/A | 936 (745) |
Bivariate and multivariate regression analyses were performed to determine the associations between the selected independent variables and the children’s hemoglobin level. All regression models contained fixed-effects dummy variables at the district level, with separate dummy variables for the urban and rural parts of the districts, to control for clustering and achieve complete control for direct effects of confounders at those levels. Given that all variation at the district level was controlled by these dummy variables, no direct effects of the context factors at the district level could be estimated. These context factors are therefore only used in the interaction analysis. We tested for nonlinearities by including quadratic terms for all interval variables, but only for the altitude variable the quadratic effect was significant. For altitude, this term was therefore included in the models.
Interaction analysis was used to examine the influence of characteristics of the children, their mothers, the households and the context on the strength of the preceding birth interval effect. We tested for significant interactions between the birth interval effect and the children’s age, birth order, their mother’s age, mother’s hemoglobin level, mother’s BMI, and mother’s age difference with her partner, household wealth, cluster educational level, district level of development and availability of health facilities. To adjust the significance level for multiple interaction testing, the Bonferroni correction was applied [
The baseline characteristics of children and mothers are described in
Coefficients of the bivariate regression analyses are presented in the second column of
Variable | Bivariate Model | Multivariate Model |
---|---|---|
Coefficient (SE) | Coefficient (SE) | |
Intercept | - | 69.603 (0.954) |
0.007 (0.005) | 0.015 (0.005) |
|
Sex is female (reference = male) | 0.872 (0.134) |
0.847 (0.127) |
Age of child, months | 0.232 (0.004) |
0.207 (0.006) |
Birth order | 0.118 (0.03) |
0.000 (0.047) |
Maternal Hemoglobin, g/l | 0.147 (0.004) |
0.139 (0.004) |
Age of mother, years | 0.181 (0.011) |
0.071 (0.018) |
BMI of mother, Kg/m2 | 0.003 (0.000) |
0.001 (0.000) |
Breastfeeding mother | ||
- Not breastfeeding and index child | Reference group | Reference group |
- Breastfeeding and index child | -5.317 (0.157) |
-1.253 (0.206) |
- Breastfeeding mother and other child | 0.638 (0.170) |
-0.390 (0.169) |
Multiple birth (Ref = Singleton) | 0.003 (0.520) | -0.404 (0.497) |
Wealth index | 1.140 (0.062) |
0.751 (0.062) |
Age difference mother and partner, years | -0.038 (0.010) |
-0.028 (0.010) |
Education year of mother, years | 0.225 (0.022) |
0.119 (0.022) |
Number of unions | ||
- First union | Reference group | Reference group |
- Not in a union | -0.398 (0.282) | -0.452 (0.269) |
- Second or later union | -0.403 (0.195) |
-0.303 (0.188) |
Employed mother (Ref = no) | 0.905 (0.166) |
0.295 (0.159) |
Altitude | 0.08 (0.001) |
0.006 (0.001) |
Altitude2 | -0.000 (0.000) |
-0.000 (0.000) |
* P<0.05
** P<0.01
BMI = Body Mass Index
The outcomes of the multivariate regression analysis are shown in the third column of
The results again show that the hemoglobin level of girls is significantly higher than that of boys. We also observe that mother’s age, maternal BMI, maternal hemoglobin level, maternal educational level and her age difference with her partner all have significant positive effects on hemoglobin level of the index children. Household wealth has also a strong effect. Children living on higher altitude had significantly higher hemoglobin levels but the effect is nonlinear, with the effect starting to stabilize when the residence of the child was above 4,000 meters.
To analyze whether the effect of length of preceding birth interval affected hemoglobin levels of boys and girls differently, an interaction term was added to the multivariate model. This term turned out to be significant and positive (effect size 0.023 P = 0.019), which means that the preceding birth interval effect was substantially stronger for girls than for boys. Separate multivariate analyses for boys and girls made clear that the effect was only significantly present in girls (effect size 0.021, P = 0.003) and not in boys (effect size 0.004, P = 0.579).
To analyze whether the effect of length of preceding birth interval depended on the circumstances in which the children were living, interactions between this effect and the other variables in the model were studied. Given the substantial sex difference in the birth interval effect, the interaction analysis was performed separately for boys and girls. Because little knowledge of such interactions was available, this analysis is largely explorative in nature. We tested all interactions between preceding birth interval and the major other variables in the model (children’s age, mother’s age, mother’s hemoglobin level, mother’s BMI, mother’s age difference with her partner, household wealth, cluster educational level, district level of development and availability of health facilities) and added the significant interactions to the model.
Coefficient (SE) | ||
---|---|---|
Variable | Boys | Girls |
Intercept | 98.175 (1.039) |
95.854 (1.059) |
-0.000 (0.008) | 0.025 (0.007) |
|
0.003 (0.001) |
-0.004 (0.001) |
|
0.001 (0.000) |
||
0.001 (0.000) |
||
Multivariate models for the relationship of hemoglobin level of children (age 6–59 months, n = 49,260) from 20 African countries (DHS database) and their preceding birth interval adjusted for all predictors.
* P<0.05
** P,0.01
# P<0.0056 (Bonferroni correction)
Hb = Hemoglobin level
The interaction effects are visualized in
Several important conclusions can be drawn from the present study with regard to the birth interval effect. First, we found that the length of the preceding birth interval has a modest positive effect on hemoglobin levels of African preschool girls, but not of that of boys. Second, having a long preceding birth interval is particularly important for African girls with a young mother. Third, girls were also found to profit more if their mother has a higher hemoglobin level. Fourth, boys were found to profit from a longer birth interval only under specific circumstances, namely if they lived in an area where the educational level of mothers is high.
Positive effects of increasing birth intervals on health outcomes of children and their mothers have been noticed in several studies [
The focus on gender differences turned out to be important too. When studying differences between boys and girls, while controlling for confounding factors, a significant effect of longer birth interval was particularly found for girls but turned out to be insignificant for boys. The positive effect observed in girls was in the order of 0.025 g/l for a one month increase in the birth interval, or 0.3 g/l for an increase of a year. In addition, we found that the highest hemoglobin increments were noticed in girls with young mothers and in girls with mothers with a high hemoglobin levels. Compared to girls with the oldest mothers in our data (late forties) and the mothers with the lowest hemoglobin levels (under 60), for those with the youngest mothers (early twenties) and with the mothers with the highest hemoglobin levels (over 160) the effect of a year increase in birth interval is associated with an approximate increase of 1.4 g/l in hemoglobin level. Given that the effects are additive, for girls with a young mother with a high hemoglobin level the birth interval effect is in the order of 2.8 g/l higher than for those with an old mother with a low hemoglobin level.
We did not find a significant overall effect of birth interval on hemoglobin level among boys. However, the interaction analysis made clear that in areas where mothers have relatively high educational levels, boys profit more from a longer birth interval than in areas where mothers have relatively low educational levels. This difference is in the order of 1 g/l for an increase of a year.
Childhood anemia is very prevalent in Sub-Sahara Africa but gender differences are thereby rarely considered. Our results suggest that increasing hemoglobin levels in mothers may positively affect hemoglobin levels in their daughters. Furthermore, it seems important to target the interventions especially to young mothers, as their female offspring profits most from longer birth intervals.
Our data were derived from the DHS, including 49,260 children in 20 African countries whereby many biological, maternal, socioeconomic and contextual factors were collected as well. Because of the large sample size, internal validity was secured and variance of estimates reduced. Adjustment for covariates was performed using multivariate analysis with country and region fixed effects. In these analyses, it was noticed that biological, maternal, socioeconomic, environmental and contextual factors were related to hemoglobin levels of African preschool children and that most of the factors affect hemoglobin levels of African boys and girls rather similarly.
Apart from the separate effect of the various factors, a common denominator may play a significant role in the different sex response to birth spacing. We hypothesize that the preference for male offspring in many African countries leads to more maternal investments in boys compared to girls, making them less susceptible for the effect of birth spacing. Studies from Ethiopia (Quisumbing,2003) [
The maternal depletion syndrome could be another factor that contributes to hemoglobin levels of their offspring. Maternal depletion syndrome is characterized by a deterioration of the maternal nutritional status during the reproductive cycle, especially when the period of depletion is long while the time for repletion is short and the mother has a marginally inadequate food intake [
Given that no significant association between preceding birth interval and the children’s age was found, our study also indicates that the positive effects of a longer birth interval persist across the first 6 years of life. This may be explained by the fact that the age difference between the two succeeding children remains, which may create the opportunities for more maternal investments over the whole period. In addition, also biological factors, like epigenetic changes may contribute to the constant positive effect of a longer birth interval. The clinical relevance of our finding is difficult to give, as the effects on hemoglobin levels are only modest and mostly seen in girls that have already a greater chance of survival while boys are biologically weaker and more susceptible to disease and premature death [
Apart from the birth interval effect, our analyses made clear that hemoglobin levels of African children steadily increase during the first 59 months of life (effect size 0.207, P<0.001) and that hemoglobin levels of girls were significantly higher than those of boys (effect size 0.847, P<0.001). An additional test for sex differences in the age effect revealed a significant interaction (effect size 0.045, P<0.001), whereby the increment with age is significantly stronger in boys (effect size 0.229, P<0.001) than in girls (effect size 0.183, P<0.001). Hence, with increasing age, the sex difference in hemoglobin level becomes smaller. In
Various other studies have also reported higher hemoglobin levels of girls [
Our study is limited by the lack of information on major co-morbidities that affect hemoglobin status, such as malaria, tuberculosis and Human Immunodeficiency Virus (HIV) infections as well as non-communicable diseases, like nutritional deficiencies, including those for folate, vitamin B12 and vitamin A. The same applies to inherited or acquired disorders that affect hemoglobin synthesis, red blood cell production or red blood cell survival. The overestimation effect of miscarriage and underestimation effect of stillbirth in the length of birth interval cannot be discussed upon due to restricted information of maternal pregnancy in DHS data. Further, the study uses a cross-sectional design, which prevents the inference of cause and effect relationship.
In conclusion, increasing birth interval has a moderate positive effect on hemoglobin levels of preschool African girls. Remarkably, although iron requirement of boys seem higher than girls, birth spacing did not influence hemoglobin levels of preschool boys. The preference for male offspring in large parts of Africa may not only significantly contribute to the sex differences in hemoglobin levels but may also determine the different effect of birth spacing on hemoglobin levels of African boys and girls. Finally, gender aspects should be considered in interventions programs that aim to improve anemia in African children.
(TIF)
(DOC)
The content of the work presented here is solely the responsibility of the authors and does not necessarily represent the official views of any health organizations. The authors are grateful for the resources of DHS. Without the DHS data, this work would not have been possible; we thank the respondents, field staff and management of the site for sharing their indispensible data and allowing us to use it indefinitely.