Article Text
Abstract
Background As a consequence of the global spread of HIV infection in sub-Saharan Africa, the impact of the AIDS pandemic has contributed to more than 40 million orphaned and vulnerable children (OVC) in sub-Saharan Africa. The children's vulnerability is hypothesised to increase the risk of acute respiratory infection (ARI), which remains one of the major infectious causes of child mortality and morbidity in this region. This study examines the linkages between children's vulnerability and ARI symptoms, with a focus on the individual as well as community levels.
Methods Using population-based Demographic and Health Surveys of preschool age children (0–59 months old) from the Democratic Republic of Congo (n=7755), Malawi (n=17 559) and Nigeria (n=28 049), multilevel logistic models were conducted to establish the relationships between ARI symptoms among OVC and community-level vulnerability, adjusting for children's and household characteristics and a season variation.
Results The child's OVC status and community vulnerability were shown to be significantly associated with an increased likelihood of the child suffering from ARI symptoms in three countries. In contrast, the relationship between ARI symptoms and some selected variables associated with community development was not consistent across the 3 countries.
Conclusions Community vulnerability demonstrates a significant influence on whether or not the child exhibits ARI symptoms. Being OVC contributes to differences in children's likelihood of ARI symptoms between communities. In addition to industrial pollution, specific components, such as community development, vary in their associations with the ARI symptoms across different countries. The large differences in childhood ARI symptoms across countries suggests that policies aimed at promoting child health need to be context specific.
- CHILD HEALTH
- Neighborhood/place
- SOCIAL FACTORS IN
- SOCIAL INEQUALITIES
- MULTILEVEL MODELLING
Statistics from Altmetric.com
Introduction
Despite international support aimed at improving child health over the past few decades, the health consequences associated with the AIDS pandemic in sub-Saharan Africa remain a concern; this is particularly true for children who have a vulnerable status in general, and specifically true for those living in vulnerable communities.1 The UNICEF estimates that about 12 million children and adolescents in sub-Saharan Africa are orphans who have lost one or both parents due to AIDS.2 These orphaned and vulnerable children (OVC) are among the groups most likely to suffer from health problems during childhood, including acute respiratory infection (ARI).
ARI is often related to an increased likelihood of childhood morbidity and mortality.3 ,4 Previous research on ARI has primarily focused on the characteristics of the children. Jackson et al5 conducted a meta-analysis using 36 published studies and identified young age, low birth weight, and a large number of siblings as major risk factors for ARI symptoms. A handful of studies have gone beyond assessing child-related and mother-related factors that correlate with ARI symptoms and these have suggested a potential role of household factors in a child's ARI symptoms.1 For example, a child who comes from a household where there is a lack of social and economic resources seems to be more likely to have ARI symptoms because he or she may have to live in a crowded environment5 ,6 and not have completed all relevant immunisation.1 ,5 Several studies have indicated that one of the significant predictors of ARI symptoms is household wealth.7 Risks of ARI symptoms are thus hypothesised to be linked to household characteristics such as household socioeconomic status (SES) in rural nations with low income levels.
On the other hand, an equally important but also less investigated aspect of ARI in sub-Saharan Africa is its community context, with particular focus on the global health threats of HIV and AIDS. The present study attempts to explain variations in ARI symptoms by assessing community influences as suggested by the exposure-disease-stress model.8 This model focuses on individuals in social context and includes biological, psychosocial and structural components and considers the extent and intensity of associational links to ARI symptoms. Accordingly, I hypothesise that social and community risk factors work to create community stressors that serve as a foundation that exacerbates effects on the child's health such as ARI symptoms.8 Particular attention needs to be paid to the idea of the vulnerability of a community in the context of sub-Saharan African countries where the low status of women, high rates of economic expansion, rapid urbanisation and various biological mechanisms lead women and children in this region to be categorised as the most vulnerable groups in terms of HIV and AIDS.9 I propose community OVC and community sexual violence against women are able to acts as indicators of community vulnerability. Community development is another crucial component that influences the children's risk of suffering from ARI.8 Using the community development elements suggested by Chiao et al,10 I examined the effects of community structure opportunities such as SES, economic opportunities and infrastructural development. In addition, we also included various aspects of the physical environment by assessing which cooking fuel is used by the family and using this as an indicator of ARI risk.5 ,11–13
The data set to be analysed here does not have the capacity to test every aspect of this exposure-disease-stress framework;8 instead, it examines some major variables. I therefore examine the components of the community environment (community vulnerability and community development) that may either exacerbate or facilitate the risk of ARI symptoms. I specifically assess six community factors: community OVC, community sexual violence by male partners, electricity coverage, community cooking fuel, community education of women and community employment of women. I specifically hypothesise that young children living in communities with a higher level of community vulnerability and a lower level of community development will have a higher likelihood of suffering from ARI symptoms. Furthermore, this study investigates the extent to which ARI symptoms are homogeneous within communities after controlling for characteristics of the children and household. Since the majority of OVC occur in sub-Saharan Africa,2 this study focused on young children in three sub-Saharan African countries: the Democratic Republic of Congo (DRC), Malawi and Nigeria.
Methods
Sample and data collection
This study used recent Demographic and Health Surveys (DHS) data to examine the linkage between children's vulnerability and ARI symptoms in various contexts of sub-Saharan Africa. In DHS surveys during/from 2007 to 2014, information about OVC was only available for six countries: the DRC (2007), Malawi (2010), Namibia (2007), Nigeria (2013), Swaziland (2007) and Zambia (2007). In addition to OVC, this study also explores whether community vulnerability is associated with ARI symptoms. I thus excluded two countries (ie, Namibia and Swaziland) that had no data on sexual violence of women. In order to gain a better understanding about ARI symptoms in context, different levels of ARI symptoms prevalence were of focus here ranging from <2.5% in Nigeria, 2.5–12.5% in Malawi, to above 12.5% in the DRC.
DHS data were nationally representative surveys of children that used a two-stage sampling strategy. These standard surveys were designed to examine population health and included detailed information on infant, child and maternal mortality as well as examining child health. Further information on the DHS can be found at http://www.measuredhs.com. The DHS defined a community by sample cluster (census track), usually a village or an urban census block.
Given that the current research aims are to gain a better understanding of the relationship between the risk of ARI and community context, my analyses focused on surviving children aged 0–5 with completed information on ARI status, OVC status and the indicators of community context. This selection yielded a total of 7705 children from 300 communities in the DRC, 17 931 children from 849 communities in Malawi and 27 642 children from 896 communities in Nigeria. The study protocol was approved by the Ethical Committee of National Yang-Ming University, Taiwan.
Measures
Outcome measure
Symptoms among the children with ARI were assessed based on mothers' reports, whether or not their children had been ill with a cough accompanied by short, rapid breathing that was chest-related and/or had had difficulty in breathing that was chest related; the time period examined was the two weeks preceding the DHS children survey. Responses are categorised into two groups; where young children had suffered from ARI symptoms, this is coded as Y=1 or ‘yes’ and young children not suffered from ARI symptoms, this is coded Y=0 or ‘none’.
Children's orphaned and vulnerability status
On the basis of prior studies,14 vulnerable children were identified as those who had experienced the death of a family member, who had been ill for at least 3 months during the past 12 months, or who were a member of a household with a member who had been ill for at least 3 months during the past 12 months. We selected a number of variables that assessed children's orphaned and vulnerability status. Vulnerability was measured by several questions, including living with a chronically ill adult or living in a household in which one or more household members have died during the past 12 months period preceding the survey, with those who died having been sick for at least 3 months before they died. In addition, questions were also included to identify children whose parents were not living in the same household at the time of the survey to determine if these parents had been severely sick during the 12 months before the survey. After a series of recodes, responses were categorised into two groups: having an orphaned and vulnerable status (coded as 1), and not having an orphaned and vulnerable status (coded as 0). Table 1 shows the categorisation and distribution of the community characteristics from the DHS data obtained in DRC, Malawi and Nigeria.
Community-level variables
A mean score for each community was created by averaging all individual responses within each tract. The study sample within each tract was assigned to have a computed mean score. Two sets of community factors are hypothesised to affect the children in terms of ARI. These were, first, community vulnerability and, second, community development. The measures used in the DHS to assess community vulnerability consisted of community OVC and community sexual violence by male partners. Using the DHS household surveys, the community OVC was computed by averaging children under 18 years of age and measuring the proportion of OVC that had occurred at the ages of 18 years or younger among these children within the communities (quartiles). A measure of community sexual violence against women was constructed using a derived indicator of ever suffered from sexual violence by a male partner obtained from the DHS women surveys and measured the proportion of women who had ever experienced sexual violence by a male partner in the community (tertiles).
Community development
Community development is another set of derived measures and consisted of electricity coverage, the education attainment of women within the community, community cooking using solid fuel in the house, and employment of women within that community. The former three variables were constructed by aggregating the household data at the cluster level, for example, average household electricity coverage within communities (range 0–1) and the proportion of households using solid fuel for cooking in the house within the community, the latter being a derived indicator of two variables, namely cooking fuel and cooking place (range 0–1). Community women's education was assessed as the per cent of women aged 18 and above within a community block who had received no education (quartiles). The final variable, women's employment, was an integral community variable computed directly from the census tract that gave the percentage of women being employed aged 15–49 within the communities (0 to <50%; 50% to <75%; 75% and above).
Several measures in terms of child and household characteristics were used as potential confounding variables that might affect the relationships between the main explanatory variables and ARI. These characteristics of the children consisted of age, gender (male or female), birth weight and birth order (1–3, 4–6, >6). Each child's age was categorised as one of the following, newborn (<1 year old), infant (1–2 years old) or toddler (3–5 years old), based on physical and cognitive developmental stages. Birth weight was reported by the mother or the main caregiver and coded categorically as low birth weight (<2500 g), not low birth weight (2500 g or greater) and unknown. The analysis also included various household characteristics with four aspects of a child's household being salient to ARI: sex of the household head, education of the household head, place of residence (urban or rural) and household wealth. These constructs were also closely linked to the measures assessing SES. The sex of the household head was dichotomised into male-headed and female-headed families. Household wealth was categorised into quintiles: those who were the poorest, the second quintile, the middle quintile, the fourth quintile and the richest quintile of households. These were based on the assets-based wealth index quintiles for each country.15 Finally, this study adjusted for a seasonal variation. The season factor was dichotomised into the dry season and others, which were based on the months of interviews and country-specific seasonal variation.
Analytical strategy
Analysis began with bivariate tabulations that characterised the distribution of individual characteristics of the sample population in terms of country. Then, to address our research questions, we used multilevel modelling techniques to study the simultaneous associations between community context, OVC and ARI symptoms, adjusting for individual and household backgrounds. A two-level (level 1=children and level 2=census tract/community) random intercept logit model was used for a binary outcome of ARI risk. The random intercept is shared by all sampled children from the same census tract and this model incorporates the simultaneous effects of individual-level and community-level variables on the odds of ARI risk.
We assessed whether ARI risk varied across children and communities using a two-level multilevel model with no observed covariates (the null model). The percentage of the total variance in ARI risk that was related to the community (the intracommunity correlation or the intraclass correlation coefficient) was used as a measure of the contextual influences. The intracommunity correlation was approximated as σµ2/(σµ2+π2/3), where σµ2 denotes community-level variance.16 In the null models, the percentage of the total variance in ARI risk in the community explained by the community in which the child resided was 17% in the Congo Democratic Republic (p<0.01), 12% in Malawi (p<0.01) and 31% in Nigeria (p<0.01). I then used this multilevel model to further examine whether or not the community environment independently influenced the likelihood of ARI risk after taking the characteristics of the children and the household into consideration. All analyses were carried out separately for DRC, Malawi and Nigeria and conducted using Stata V.13.0 (Stata Corporation. Stata statistical software: release 13. In. Release 13. College Station, Tex.: Stata Press. 2013).
Results
Table 2 shows the percentage distribution of young children with selected individual and household characteristics obtained from DHS by country. There is a substantial difference in ARI prevalence: 16% in DRC, 7% in Malawi and 2% in Nigeria. Nevertheless, the average age and the percentage birth order of the children are very similar for these three countries. However, other characteristics differ markedly between these countries. For instance, approximately one-fifth (19%) of young children were OVC in DRC, but this number was much fewer in Malawi (6%) and Nigeria (3%).
Table 3 presents the adjusted ORs and 95% CIs of the predictor variables for ARI risk when both community-level and individual-level factors are simultaneously included for the selected three countries. Community vulnerability (community OVC and community sexual violence) showed a significant positive association with ARI symptoms among children in the three selected countries. In DRC and Nigeria, young children living in tracts with the highest quartile of percentage of OVC were more likely to report ARI symptoms, compared with children living in communities in the lowest quartile of percentage of OVC (DRC: OR=1.97, p<0.01; Nigeria: OR=2.29, p<0.01). This statistically significant relationship was also found in the relationship between community sexual violence and ARI. In Malawi and Nigeria, children living in a tract where a middle percentage of women reported sexual violence by their male partner reported 1–2 times higher odds of ARI symptoms, compared with children in living in low violence communities.
Several measures that are associated with the concept of community development are significant; however, some measures act in contradictory directions in different countries. In terms of electricity coverage, children in the communities with a higher per cent of electricity coverage had significantly lower odds of ARI symptoms in Nigeria (OR=0.60, p<0.05), but this effect was non-significant in DRC and Malawi. In Nigeria, children living in a tract with a higher percentage of solid fuel cooking in the house had 1.64 times higher odds of ARI symptoms, compared with children in communities with a lower percentage of solid fuel cooking in the house. However, this statistically significant relationship was not found in Malawi and DRC. Furthermore, Nigerian children in the communities with higher women's employment, as measured by tracts where at least 75% of women were employed, had significantly lower odds of ARI symptoms, compared with children in communities with lower employment of women (OR=0.67, p<0.05); this relationship was reversed in the other two countries.
Turning to the individual level, OVC is significant at predicting the risk of ARI symptoms. The odds for ARI risk among children who were OVC were significantly higher than for those who were not OVC in DRC (OR=1.32, p<0.01) and in Malawi (OR=1.48, p<0.01), although this significant relationship was absent in Nigeria.
Discussion
Researchers have often hypothesised that one mechanism that affects children's health is community context, but this hypothesis has rarely been examined in the context of whether community context can independently affect a child's health, such as suffering from ARI symptoms. This study uses national surveys from DRC, Malawi and Nigeria, and has demonstrated the importance of independent contextual influences on children's ARI risk using multilevel logit models.
Consistent with the exposure-disease-stress model,8 our analysis has empirically demonstrated that community vulnerability is significantly related to ARI risk among preschool age children, although the magnitude of this effect differs across three countries. The findings suggest that preschool age children living in communities with a larger proportion of OVC have a significantly high risk of ARI. In addition, children living in a community with a higher proportion of women who ever experienced sexual violence by their male partners also have a significantly greater likelihood of having ARI symptoms across the three countries.
Another community context affecting ARI symptoms is community development. Analysis indicated that an increase in electricity coverage decreased children's ARI risk across three countries and supports the common hypothesis that ARI risk is likely to be reduced as community SES increases. Analyses of the Nigeria DHS data set also provide evidence supporting the hypothesis that there is a negative association between employment of women from the community and child ARI risk. However, an opposite association was observed in DRC and Malawi. One possible interpretation on combining these two pieces of information involves the degree of community development and the types of occupation in which the women are employed in these three countries.17 When community development reaches a certain level and traditional agricultural work begins to decrease, women living in developed communities do not need to provide as much help as previously with agricultural work. As a result of this decrease in agricultural work participation and the resulting change in community development, there may be a decrease in the risk of ARI among preschool age children. Given that the distribution of types of women's occupation depends on the community's SES, as ARI risk decreases in developed communities, women with a low status occupation in less developed communities are likely to become much more vulnerable to poor outcomes in terms of their child's health. In DRC, a large proportion of employed women worked for the mining sector and communities with a higher rate of women's employment seemed to be in mining sites. Discharges from mining are major sources of air pollution, and consequently the socioeconomic and environmental complexity behind childhood ARI symptoms is warranted. Thus, national programmes, strategies and policies aimed at socioeconomic development should take specific components of community SES into consideration, especially with respect to issues of child health.
This study contributes to our understanding of how community context and community vulnerability, in particular, are associated with the risk of ARI symptoms among preschool age children; nonetheless, my findings should be interpreted within the context of the study's limitations. First, in addition to the common limitations associated with using a self-reported measure of ARI symptoms, the association of community-level and individual-level factors with children's ARI risk might be influenced by concerns regarding endogeneity problems. Given the cross-sectional nature of the data, we cannot disentangle or establish directionality (cause vs effect) from the results. However, our analyses do show convincing associations between selected community/OVC variables and ARI risk. Second, since the DHS surveys were conducted in different languages within a country and between three countries, some measures related to vulnerability concepts may not be the same vulnerability concept in various contexts. Third, the assessment of other important factors of ARI symptoms is limited by the use of existing data in the DHS, such as country-specific environmental factors. For instance, mining discharges in the south-eastern provinces of DRC were left with polluted air in these neighbourhoods; as a result, a high prevalence of ARI in children was observed.18 Fourth, my analysis is based on sets of simple multilevel logit models with random intercept and fixed coefficients only. As a result, my findings cannot provide evidence as to the effects of individual factors that vary across communities. However, by using a multilevel analytical approach, my study does provide important insights and is able to identify the multilevel determinants of ARI symptoms in three countries in sub-Saharan Africa that have suffered from serious HIV/AIDS epidemics.
Overall, it is important to take into account all known factors, including those that go beyond the life of the household such as industrial pollution. This study extends prior research using multilevel analysis and contributes important knowledge to the current literature where there is only limited evidence concerning community context and ARI risk in sub-Saharan African countries. Further research is essential to continue an exploration of how community context affects a child's ARI risk. In addition to the horizontal aspect of community vulnerability, research should also target how the vertical aspect of the community context impacts ARI symptoms, such as whether the bonding between community members and the government affects the cared-for child's ARI symptoms. Further investigation could also consider the cross-level interactions between individual and community characteristics. For instance, OVC children may suffer the most when a community context creates a particular vulnerability to childhood ARI risk.
What is already known on this subject
Studies have suggested that community factors are associated with differences in acute respiratory infection (ARI) symptoms among preschool age children. However, only limited research has empirically addressed the idea of vulnerability at individual and community levels with respect to children's ARI risk. Even fewer studies have examined whether community vulnerability shapes the risk of ARI symptoms among children in the context of the impact of the AIDS pandemic in regions of sub-Saharan Africa.
What this study adds
Being an orphaned and vulnerable child (OVC) increases the odds of ARI symptoms in three sub-Saharan African regions.
Community vulnerability, as measured by community OVC and community sexual violence against women by their male partners, is significantly associated with increased odds of a child suffering from ARI symptoms, even after adjusting for the children's backgrounds, household characteristics and a seasonal variation.
Specific components within community development vary in their associations with ARI symptoms across different countries.
Acknowledgments
The author would like to thank the editor and anonymous reviewers for their useful comments and suggestions for improvement in the quality of this manuscript. Special thanks to Yu-Yun Chen for her assistance in data management.
References
Footnotes
Funding The study was supported by the Ministry of Science and Technology (MOST) in Taiwan under grants 104-2628-H-010-001 and 104-2511-S-010-004-MY3.
Disclaimer The conclusions here are those of the authors and do not represent the viewpoint of the Ministry.
Competing interests None declared.
Ethics approval The study was approved by the Research Ethics Committee of National Yang-Ming University.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The Demographic and Health Surveys (DHS) data sets are publicly available and can be downloaded from the DHS Measures website: www.measuredhs.com.