Introduction
Diarrhoea remains an important cause of morbidity and mortality among children aged five and below in most developing regions of the world. According to an estimate, diarrhoea is reported to be responsible for close to 2 million deaths annually in this age-group[
1]. In sub-Saharan Africa (SSA), the occurrence of diarrhoea like other infectious disease has been associated with poverty [
2‐
4]. Timely administration of oral rehydration salt (ORS), and more recently Zinc tablets have proved to be both more cost effective and efficacious as primary interventions for preventing diarrhoea morbidity [
5‐
11]. Despite the availability of these interventions; there have been no decline to diarrhoea incidence as many children are not using the interventions[
12]. As a result, diarrhoea disease continues to be a serious threat to children in many countries in SSA [
13,
14].
The socio-economic gradient of diarrhoea occurrence is well established in literature. However, it is important to know the extent to which it also determines where care is sought when managing diarrhoea. Although several studies have document the role of socioeconomic factors in care seeking for childhood diarrhoea [
15‐
17], many of these studies have not been able to explore the role of socio-economic characteristics of the neighbourhood in shaping or hindering the treatment options by caregivers. There is evidence suggestive of major influence of socio-economic characteristics of neighbourhoods in access to, and utilization of care [
18]. In particular, living in socio-economic deprived neighbourhoods has been well documented to be associated with less likelihood of seeking medical care[
19]. Thus, examining various perspectives by which socio-economic factors can influence selection of treatment options for childhood diarrhoea is of high importance.
So therefore, to bridge the paucity of knowledge on the wider influencing role of socioeconomic status in childhood diarrhoea treatment, this study used multilevel modelling [
20‐
22]. Multilevel modelling technique permits simultaneous investigation of the role that individual socio-economic position and neighbourhood socio-economic status has on selection of treatment options. Highlighting such would contribute to a greater understanding of what is needed to be done to mitigate less uptake of the recommended intervention.
Discussion
The central aim of this study is to investigate the association between socio-economic status and selection of treatment for childhood diarrhoea among care givers in sub-Saharan Africa using multilevel multinomial regression analysis. The result shows that, caregiver's choice of treatment for childhood diarrhoea depends on several individual and neighbourhood measures of socio-economic status. Specifically, at the individual level, the analysis indicate that choice of medical centre for managing childhood diarrhoea was highly associated with the caregiver's and her partner's educational attainments. Highly educated caregivers had a higher odd of utilizing medical centre for managing childhood diarrhoea; this finding is in contrast to a study conducted in another developing region of the world with high prevalence of childhood diarrhoea [
30]. Whereas, our findings are compatible with those of many others [
31‐
33], that have documented positive association between maternal education and choice of medical centre for managing childhood diarrhoea. The finding that partners education is associated with choice of medical centre is in consonance with what had been reported earlier [
33,
34].This finding, further confirms the protective role of fathers education[
35], as an additional reinforcement factor for mothers decision to seek appropriate care when managing childhood illness. The positive association between parental education and choice of medical centre as noted in this study further reverberates its importance for child survival in developing world [
36].
In sub-Saharan Africa, fathers are the overall head of the household and sometimes decides where care is sought [
37]. Hence, it is not surprising as noted in this study that, caregiver's partner's education, is associated with patronage of pharmacy store and medicine vendors and to some extent home care for managing diarrhoea episode. In addition, the decision of the educated caregivers to use home treatment, have been attributed to their ability to utilize health information wisely[
38].
The influence of wealth status on caregiver's propensity to choose private and public medical centres when managing childhood illness has been documented in the literatures [
30,
39‐
41]. This is however, not noticed in this study. Although this study shows that, care givers from poorer households compared with those from poorest household engaged in home care as an alternative option for managing childhood diarrhoea. This finding is not surprising, it had been reported elsewhere that households sometimes do engaged in self medication especially when the cost of treatment in medical centre is high [
42,
43]. On the other hand this study joined other studies in documenting care givers' occupation, as another important factor influencing choice of treatment [
33,
39,
41]. In this study, being a manual worker is closely associated with selection of medical centre, while being a professional working class is associated with patronage of pharmacy store or medicine vendors.
Geographic location, place of residence in particular, has been shown to be another form of disparity[
44,
45], which could prevent access to utilization of care[
46]. This study shows that, residing in rural area, though statistically not significant, is associated with likelihood of patronizing traditional healers. Of the main interest in this study, is to examine the effect of neighbourhood socio-economic disadvantaged after controlling for individual SEP on clustering of selected treatment options around the neighbourhoods. The multilevel multinomial regression models indicate that, with all other factors being held constant, living in highly socio-economically disadvantaged neighbourhood is associated with less likelihood of using medical centre, pharmacy or vendors. The results of the between caregivers variation in the choice of treatment at the community-level indicates that several other factors which might be in part due to the caregivers neighbourhoods play a greater role in the individual choices. This finding suggests that, compositional characteristics of the caregivers are less important than that of the community with regards to individual choices.
Study limitations and strengths
This study is without limitations and should be mentioned. First, the findings from this study are based on data from cross-sectional survey and the initiation of the caregivers to the health system for managing childhood diarrhoea. It is however, possible for care to be sought from more than one provider upon the failure of the initial treatment. Second, we used an indirect measure of household wealth status. However, as DHS surveys do not collect data on income, the use of household possession has been shown to be relevant in developing country settings[
47]. Finally, the analysis was based on self reported diarrhoea morbidity has reported by the care giver which could be subject to recall bias. In spite of these limitations, the strength of our study lies in the unique characteristics of the DHS. The DHS are nationally representative, and allows for findings to be generalized across the entire country. In addition, the design and the variables included in the survey are the same across countries, and thus, making it possible for us to be able to pool the data from these countries.
Conclusions
In sum, this study has revealed that regardless of where individual care giver resides, treatment choices would be similar based on their level of educational attainments (compositional effects). On the other hand, the characteristics of the neighbourhoods level of economic development, accounts for variation from caregiver to caregiver in their choice of treatment at the community level (contextual effects). Hence, interventions aimed at improving appropriate care seeking for managing childhood diarrhoea must take into consideration care giver's SEP and the level of socio-economic development of the community in which each care giver resides.
Appendix A. Statistical method
In this analysis, no treatment was used as the reference category and a set of
t- 1 logistic regressions were computed for the four remaining treatment options. These options are then, contrasted one after the other against no treatment (reference). Thus we specified a multilevel multinomial regression model as follows:
Where
S take values from 1 to t- 1, the subscript
s denotes a separate set of intercepts for the reference and the four remaining set of categories. The caption
β
0j
(s)
depict the fixed part of the model and was interpreted as the effect of a 1-unit increase in
X(that the set of predators variables (socio-economic variables) on the log odds of selecting category
s(i.e medical centre or any other categories) other than the reference category
t (no treatment). The terms in the brackets in the equation represents the random effects associated with the primary sampling units at the community level. These are assumed to be normally distributed with a mean value of 0 and different variances. The random effects are measures of variations, and are contrast specific as indicated by the subscript
s, since different unobserved factors at each level may affect each contrast. For this analysis, the intra-clusters correlation (ICC) was used as a measure of random effects [
48]. We based the regression and variance parameters on penalized quasi-likelihood (PQL) estimation, using second order Taylor series linearization.
Acknowledgements
The authors wish to acknowledge the support of Professor Anders Hjern of National Board of Health and Welfare and Centre for Health Equity Studies, Stockholm. The authors thank the following researchers Dr Asli Kulane and Dr Anna Måndsdotter (Department of public health, Karolinska institute) and Dr Niklas Zethraeus (Centre for Health Economics&Medical Management Centre, Karolinska Institute).
The authors are grateful to Measure DHS for providing them with the DHS data for countries.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
OA, SL, KD and TM were involved in the conception of the study. OA carried out data extraction. OA conducted statistical analysis and was checked for correctness by SL and KD. OA drafted the paper with contributions from the co-authors. All authors read and approved the final manuscript prior to submission.