Factors associated with the use of antibiotics for children presenting with illnesses with fever and cough obtained from prescription and non-prescription sources: a cross-sectional study of data for 37 sub-Saharan African countries

Secondary data analysis was conducted based on the most recent Demographic and Health Survey (DHS) data of 37 sub-Saharan African (SSA) countries that have data on the sources of prescriptions for antibiotics used for cough and fever among under-five children (Angola, Burkina Faso, Benin, Burundi, Dr Congo, Congo, Cote d’Ivoire, Cameroon, Ethiopia, Gabon, Ghana, Gambia, Guinea, Kenya, Comoros, Liberia, Lesotho, Madagascar, Mali, Mauritania, Malawi, Mozambique, Nigeria, Niger, Namibia, Rwanda, Sierra Leone, Senegal, Sao Tome, Swaziland, Chad, Togo, Tanzania, Uganda, South Africa, Zambia, and Zimbabwe). DHS is conducted every five years to generate updated health and health-related indicators. A two-stage stratified sampling technique was employed to recruit the samples with Enumeration Areas (EAs) and households as primary and secondary sampling units, respectively. The detailed methodology is available athttps://​dhsprogram.​com/​Methodology/​index.​cfm. The data were extracted using the Kids Record dataset (KR file) after we obtained an authorization letter from the measure DHS program for data access. Mothers or caregivers of under-five children were asked about whether their children experienced fever and/or cough within 2 weeks within 2 weeks preceding the survey. Then, for those who experienced cough and/or fever were asked for the source of the prescription. A total weighted sample of 18,866 under-five children in SSA who took antibiotics for fever/cough were considered for this study (Table 1).

The outcome variable was antibiotics taken for fever/cough prescribed from qualified sources. It was generated using three DHS questions such as "had fever/cough within 2 weeks?", "antibiotic taken for fever/cough?" and "source of prescription?". Then, we classified the source of prescription as prescribed from formal healthcare settings and prescribed from informal healthcare settings. Antibiotics prescribed from governmental hospitals, clinics, Non-governmental Organizations (NGOs), private hospitals, and public health sectors were classified as prescribed from formal healthcare settings, and shops, churches, traditional practitioners, drug sellers, friends, relatives, supermarkets, and others were classified as prescribed from informal healthcare settings [18]. Independent variables included in this study were the child's age, sex of the child, household wealth status, maternal age, maternal level of education, number of under-five children in the household, mothers' healthcare decision-making autonomy, media exposure, mother working status, delivery by cesarean section, duration of breastfeeding, residence and ever had vaccination.

All the analyses reported in this study were based on the weighted data. STATA version 17 statistical software was used for the data management and analysis. As the proportion of antibiotics taken for fever/cough prescribed from formal health care settings among under-five children in SSA was 67%. In this case, using Odds Ratio (OR) as an effect measure could exaggerate the relationship between the dependent and independent variables. Therefore, we applied the multilevel modified poisson regression model to obtain the prevalence odds ratio.

We preferred this model because of the following reasons. For start, when the magnitude of the outcome variable is common, the odds ratio obtained using the binary logistic regression approach overestimates the strength of the relationship. Second, because the DHS data is hierarchical, mothers were nested within cluster/EA. As a result, our model considers data dependencies as well as the problem of overestimation.

The Likelihood Ratio (LR) test, Intra-class Correlation Coefficient (ICC), and Median Odds Ratio (MOR) were computed to measure the variation between clusters. The ICC quantifies the degree of heterogeneity between clusters (the proportion of the total observed individual variation of antibiotics taken for fever/cough prescribed from qualified sources among under-five children that is attributable to cluster variations).

\(ICC= {\delta }^{2}/{(\partial }^{2}+ {\pi }^{2})\) [19], but MOR quantifies the variation or heterogeneity in outcomes between clusters and is defined as the median value of the odds ratio between the cluster at more likely to have antibiotics taken for fever/cough prescribed from formal health care settings and cluster at lower risk when randomly picking out two clusters (EAs) [20].

\(\partial^2\) indicates that cluster variance

Variables with a p-value < 0.2 in the bi-variable multilevel modified poisson regression analysis were considered for the multivariable analysis. Deviance was used to verify model fitness and a model with the lowest deviance was considered the best-fit model. Finally, the Adjusted Prevalence Odds Ratio (APOR) with its 95% confidence interval (CI) was reported, and variables with a p-value < 0.05 in the multivariable analysis.

A total weighted sample of 18,866 under-five children in SSA who took antibiotics for fever/cough were included. Among them, 11,664 (61.83%) were residing in rural areas. About 9,618 (50.98%) were males and 2,899 (27.01%) were aged 12–23 months. Regarding household wealth status, 3,618 (19.18%) and 3,829 (20.29%) of the children belonged to the poorest and richest households, respectively. Among 18,866 under-five children who experienced fever and took antibiotics, about 7048 (37.36%) of the children's mothers attained a primary level of education (Table 2). The proportion of use of antibiotics prescribed from unqualified sources for under-five children who had a fever/cough and took antibiotics was 67.19% (95% CI: 66.51%, 67.85%). It ranged from 40.34% in Chad to 92.67% in Sao Tome (Fig. 1).

In the multivariable multilevel modified poisson regression analysis; residence, age of the child, mother's level of education, household wealth status, number of under-five children, and mother's healthcare decision-making autonomy were significantly associated with the use of antibiotics prescribed from formal health care setting for fever/cough among under-five children. The prevalence odds of the use of antibiotics prescribed from formal health care settings for fever and cough of under-five children were increased by 1.23 times (AOR = 1.08, 95% CI: 1.04, 1.12) compared to urban children. The prevalence odds of the use of antibiotics prescribed from formal health care setting for fever/cough among under-five children born to mothers whose husband and someone else decided on their child health issues were decreased by 10% (AOR = 0.90, 95% CI: 0.86, 0.94) and 5% (AOR = 0.95, 95% CI: 0.90, 0.99) compared to those who made their child health issues alone, respectively.

Children belonging to middle household wealth status had 1.23 times (AOR = 1.07, 95% CI: 1.02, 1,11) higher prevalence odds of taking antibiotics prescribed from formal health care setting for fever/cough compared to those belonging to the poorest households. The prevalence odds of taking antibiotics prescribed from formal health care settings among children in households that had two, three, and four under-five children were decreased by 6% (AOR = 0.94, 95% CI: 0.91, 0.97), 11% (AOR = 0.89, 95% CI: 0.85, 0.93) and 10% (AOR = 0.90, 95% CI: 0.83, 0.98) compared to having one under-five children, respectively.

Maternal exposure to news/electronic media increased the prevalence odds of taking antibiotics prescribed from formal health care settings for their children with fever/cough by 1.06 times (AOR = 1.06, 95% CI: 1.02, 1.10). The prevalence odds of taking antibiotics prescribed from formal health care setting for fever/cough among under-five children whose mothers had primary, and secondary education lowered by 4% (AOR = 0.96, 95% CI: 0.79, 0.93), and 5% (AOR = 0.95, 95% CI: 0.92, 0.99) compared to those whose mother had no formal education, respectively (Table 3).

Our study has notable strengths, including a large sample size encompassing 37 countries in the SSA region. In addition, we used a robust statistical approach to obtain the prevalence odds ratio to prevent the overestimation of the association between variables. However, it is important to acknowledge several limitations. Firstly, the cross-sectional design of the study restricts our ability to establish causal relationships between variables. Secondly, our classification of antibiotic sources was limited to formal and informal healthcare settings, and future studies could explore the associations between different factors and various sources of antibiotics. Further research is necessary to provide a comprehensive understanding of prescription patterns for antibiotics in other common childhood illnesses such as diarrhea, Acute Respiratory Infection (ARI), and others.

The study has shown that the majority of the antibiotics used for fever and cough in under-five children were prescribed from formal healthcare settings. Overall, our study provides valuable insights into the factors associated with use of antibiotics prescribed from formal healthcare settings for fever and cough in under-five children. The findings underscore the significance of addressing healthcare disparities, improving access to qualified healthcare providers, promoting maternal education, and empowering mothers in healthcare decision-making to ensure appropriate antibiotic use in this vulnerable population. Further research and interventions targeted at these factors are warranted to optimize antibiotic prescribing practices and promote responsible antibiotic use in the management of fever and cough in under-five children.