Background
Intestinal parasitic infections are one of the neglected tropical diseases (NTD) which is mainly concentrated in developing countries [
1]. Soil-transmitted helminths like
roundworms, whipworms and hookworms are among the most familiar infections to human beings and which found mainly in areas with warm and moist climates where sanitation and hygiene are poor [
2,
3]. Those soil-transmitted helminths occupy the intestine of human beings and their eggs are passed in the feces of infected persons [
3]. If an infected person defecates outside (near bushes, in a garden, or field) or if the faeces of an infected person are used as fertilizer, eggs are deposited on the soil [
3]. Among the most common cause of intestinal parasitic infections are
worms, Enterobius vermicularis, Amebiasis trichuriasis, Giardia lamblia, Ancylostoma duodenale, Necator americanus, and Entamoeba histolytica are the most common species [
4,
5]. Gastrointestinal parasitic infections are diseases of poverty, which mainly affects children living in tropical and subtropical regions like Ethiopia [
6]. Worldwide, more than one billion people are influenced by gastrointestinal parasitic infections and majority of them are children [
7]. Children under the age of 5 years are more prone to intestinal parasitosis because of their delicate immunity system and mouthing of contaminated soil or object [
8]. Those infections are more common in countries with low level of Water, Sanitation and Hygiene (WASH). In Ethiopia, the availability of WASH services are poor. For example, research done using the 2016 Ethiopian demographic health survey’s datasets revealed that the proportion of households having access to improved drinking water sources and toilet facilities were relatively low(i.e.70% for drinking water and 25.4% for toilet facilities) [
9]. Another pocket study done in the Tigray region of Ethiopia also showed a similar occasion (i.e. Poor quality of WASH in the country) that revealed that only 53, 42.4 and 36.2% households utilize latrine, hand washing, and water facilities properly [
10].
Additionally, the prevalence of gastrointestinal parasitic infection is still high in Ethiopia. A systematic review in Ethiopia showed that the pooled prevalence of intestinal parasitosis was 48%. Intestinal parasitic infection can lead to under-five mortality by inducing bloody diarrhea, chronic loss of blood that ends up with anemia, impair intestinal food absorption, and loss of appetite [
11,
12]. The influence of helminthic infections goes beyond these obvious health effects to include economic and social effects resulting from lost school attendance and effective work time. Deworming to children aged 24–59 months of age is one of the strategic initiatives to halt the global burden of intestinal parasitosis among under-five children. Deworming (sometimes called Preventive chemotherapy) is the administration of albendazole or mebendazole drug to a high-risk group of population to treat soil-transmitted helminthiases (STH) and is also administrated to minimize future parasitic related morbidity [
13]. Besides, a drug like praziquantel is used to treat schistosomiasis [
14]. World Health Organization (WHO) recommended annual single-dose administration of either Albendazole (400 mg)or Mebendazole (500 mg) to regions that have a baseline prevalence of any soil-transmitted infection of 20% and Biannual if the baseline prevalence raised above 50% [
2]. Despite Ethiopia is launching deworming programs to intervene against the burden of intestinal parasitic infection among preschool-age children and the coverage of preventive chemotherapy (deworming) is still not satisfactory.
Existing studies are believed to embody different limitations, which this paper is designed to address. Firstly, prior research in Ethiopia was conducted by the standard binary logistic regression to find factors associated with deworming uptake among children aged 24–59 months. Ordinary/standard binary logistic regression doesn’t account for the hierarchical nature of Demographic Health Survey (DHS) data, and this might have biased the model estimation. But this study applied multivariable multilevel logistic analysis for such kind of hierarchical data to increase the statistical power and to get the appropriate estimation. Secondly, previous studies in Ethiopia didn’t take into consideration the spatial pattern of deworming uptake among children aged 24–59 months. Sub-regional estimates of poor deworming uptake using spatial modeling could give information to local decision-makers about what is happening at local (geographic) and administrative levels. The information generated at local levels like hotspot areas (clusters with a high proportion of poor deworming uptake) could help decision-makers to develop location-based interventional strategies. Therefore, this study was aimed to assess the spatial variation and factors associated with poor deworming uptake among children aged 24–59 months in Ethiopia using evidence from the 2016 Ethiopian Demographic Health Survey (EDHS).
Methods
Study area, data source and study period
This study utilized a dataset of the Ethiopian fourth demographic health survey which was conducted from January 18, 2016, to June 27, 2016. The first administrative level of the country is composed of nine Regional States: Tigray, Afar, Amhara, Oromia, and Somali, Southern Nation Nationalities and Peoples Region (SNNPR), Benishangul-Gumuz, Gambela, and Harari; and two City Administrations council of Dire Dawa and Addis Ababa. Every first- level administration are divided into Woredas (districts) and Kebele (sub-districts) [
15].
Sampling technique and study population
The 2016 EDHS utilized Enumeration areas of the Population and Housing Census (PHC) as a sampling frame. The 2016 EDHS sampling design was a multistage stratified sampling strategy. The sampling procedure was two-staged ways of selecting the study sample. In the first stage of selection, 645 primary sampling units (PSU) (202 in urban areas and 443 in rural areas) or EAs (enumeration areas) were from the 11 first administrative levels based on the proportion of EAs they contributed. In the second stage of selection, a fixed number of 28 households per each EAs were selected, yielding a total of 15,683 women were eligible for interview. Finally, total weighted sample of 5949 children aged 24–59 months were embodied in this study.
Study variables
The dependent variable was deworming status, which was dichotomized as “poor” and “good”. Poor deworming uptake (i.e. a child who had not taken deworming medication), which was labeled as “poor” and coded 1. A child who has taken supplementary deworming medication was said good with deworming drug and labeled as “good” and coded 0.
Data management and analyses
Cross tabulations and summary statistics were done using STATA version 16. The forest plot technique was utilized to display the magnitude of poor deworming medication uptake across the administrative regions. To plot the 95% confidence interval (CI) of the coefficient of each variable of the best-fitted model, the STATA command “coefpot” was applied. We applied survey commands to consider the complex nature of the DHS design and to restore representativeness.
Spatial analyses
To explore, create, visualize and edit the spatial information of poor uptake of deworming medication, ArcGIS version 10.8 software was used. To declare whether the spatial configuration of poor deworming uptake was scattered, clustered, or uniformly distributed across the study area using the global spatial autocorrelation model, Moran’s I value was calculated [
16]. Moran’s I value close to − 1 indicates that poor deworming medication uptake is spatially dispersed or unrelated, whereas the closer the Moran’s I value to a positive one demonstrates poor deworming drug uptake are spatially related (clustered). But, if the Moran’s I value is more proximal to 0 it denotes that random distribution. Besides, a statistically significant (
p-value < 0.05) and Moran’s I result with a positive z-score value indicates that clustering of deworming drug uptake, while a negative z-score value indicates spatial dispersion of deworming drug uptake. Moreover, the ordinary Kriging method of spatial interpolation was done to predict the proportion of poor deworming uptake among children aged 24–59 months in unmeasured locations by forming weights from surrounding measured values.
To quantify and explore spatial dependency (spatial autocorrelation) of poor deworming uptake, the Semivariograph model was used. In addition to this, Getis-Ord Gi* statistics were computed to measure how spatial autocorrelation varies over the study location by calculating the GI* statistic for each area. Z-score and
p-value were calculated to declare the statistical significance. Statistical output with high GI* indicates a “hotspot” area, whereas, low GI* means a “cold spot”. Statistical significance of clustering was certified at
p-value of less than 0.05 and 95% CI. If the
p-value < 0.05 and the z-score is less than − 1.96, it was declared as cold spot and if Z-score is greater than + 1.96 it was declared as a hotspot areas [
17].
Purely spatial scan statistic was used to identify statistically significant hotspot areas using SaTScan™ version 9.7 software. This purely spatial scan statistic was the Bernoulli model- based with 1/0 event data, such as cases (1) (respond “no” to uptake of deworming medication) and controls (0) (respond “yes” to uptake of deworming medication). In this dataset, the controls are calculated by subtracting the cases of each cluster from total number of 24–59 months of children of the respective cluster. To allow maximum-sized clusters, the default setting (50% of the population at risk) was used. While running spatial analysis, 21 clusters with missing coordinates and 5 clusters with zero observations were excluded from the study.
Mixed model
Both bivariable and multivariable multilevel logistic regression models were fitted to identify factors associated with poor deworming uptake in Ethiopia. Multilevel analysis is useful for nested data like DHS. For instance, there are individual-level characteristics, such as each mother’s education and her household wealth status. We expect that, the greater the mother’s income and education, the higher her uptake of deworming medication for her children aged 24–59 months. But, the mother’s level characteristics i.e. income and Education might be predicted by enumeration area-level/community level characteristics like place of residency, community level media exposure, community-level poverty or regions in which they reside. For example, mothers from wealthy community might have higher enthusiasm to bring here child to health facility for deworming. In DHS data, the scores of individual level characteristics of children are more likely to be more correlated within cluster/ enumeration area than with the scores from another clusters. Similarity of score of individual characteristics within cluster might consequence violation of observation independency assumptions of classical regressions. Multilevel analysis can address the lack of independence of the observations while analyzing nested/hierarchical data like EDHS.
Two-level binary logistic regression (i.e. individual and community level) was fitted to identify factors that associated with poor deworming medication uptake. Four models were fitted. Of the four models, null model (model not including independent variables) also called random intercepts model was fitted to calculate the extent of cluster variability on poor uptake of deworming medications. Model fitness was assessed using different fitness parameters like the Likelihood Ratio test (LR), deviance, Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC). The model with the lowest of the four fitness parameters was selected as best fitted model. Cluster variability was assessed through calculating Intra-Class Coefficient (ICC), median odds ratio (MOR) and Proportional Change in Variance (PCV). ICC measures the percentage variation attributed by the community-level variables and while PCV measures the proportional change in the community-level variance between the null model and the succeeding models [
18]. The MOR describes the area-level variance as an Odds Ratio (OR), as the median value of the distribution of ORs obtained when two children with the same covariate values are picked from two different areas, comparing the one from the higher poor deworming uptake with the one from the area with lower poor uptake of deworming medication. In the absence of any area-level variation, the MOR is equal to 1. The value of ICCs and MORs was estimated from intercept-only models (null model) to examine the presence of clustering and heterogeneity between areas in the outcomes of poor deworming medication uptake. Finally, Variables with
p-value< 0.2 were considered for multivariable analysis. Adjusted odds ratio with 95% CI and
p-value < 0.05 were used to declare statistical significances of factors.
Ethical consideration
This study used a dataset of national representative demographic health surveys. Therefore, ethical approval is not required. But, datasets utilized for this study were requested by providing a clear explanation of the objectives and necessity of this study. I registered and requested the DHS dataset to the online database (
www.dhsprogram.com) and received an authorization letter to download the requested dataset.
Discussion
The overall magnitude of poor deworming uptake among children aged 24–59 months was high (i.e. 85%). The spatial pattern of poor deworming uptake among children in Ethiopia is not random (clustered). Employment status, perceived distance, mother’s education, place of delivery, region of residency and diarrhea were the main factors associated with deworming uptake among children. The spatial analysis identified five clusters of poor deworming uptake in eastern Ethiopia.
The most likely cluster occurred in the Somali regional state (area: 153.22 km, RR: 1.17 and
p-value< 0.0001) (Table
3). This high proportion of poor deworming uptake in Somali and Afar region might be related to low maternal and child service utilization rate [
19]. There were also documented higher perceived health care access challenges in Somali and Afar regions [
20]. Those two regions are also among the regions with the lowest Human Development Index [
21]. Besides, the observed cluster of poor deworming uptake in Somali region might be explained by poor health-seeking behavior toward child health [
22]. Relative to other regions, Gambela, Addis Ababa and Benishangul gumuz region have good deworming medication uptake (i.e. no clustering of cases was observed there) [
23]. This might be due to good maternal health uptake in those areas. Having good maternal health services uptake might lay the foundation to take deworming medication for their children. Focused interventions in these high-burden clusters (high proportion of poor deworming uptake areas) can optimize resources and achieve the WHO targeted soil-transmitted helminthiases coverage, i.e. reaching up to 75% of preschool children in need of treatment [
24].
The spatial Kriging interpolation assured that a higher proportion of predicted poor uptake of deworming was found, concentrated mainly in Afar and Somali regions (Fig.
5). This can be explained by the pastoralist’s nature of the people in those areas might lead theme to confront difficulty to access health facilities because of no permanent residency [
25]. Another reason for this finding might be related to the shortage of trained health professionals in those areas [
22]. In Somali, prior research documented that husbands were reported as the predominant decision-maker with regard to children’s health [
26]. This restricted autonomy of power over their child’s health has been reported as a major barrier to women’s ability to access child health care services [
27,
28]. The interpolated poor deworming uptake could help decision-makers and local authorities in evaluating the performance of prior installed deworming programs, allocate resources and develop targeted intervention plans without conducting additional research.
The multivariable multilevel logistic analysis revealed that employment status, perceived distance, mother’s education, place of delivery and diarrhea were the individual-level factors associated with poor deworming uptake. In addition to this, the region of residency was the only significant variable among the community-level factors. This study showed that children born to uneducated mothers are more likely not to take the deworming drugs. This result is in agreement with another study [
29]. Another research in Pakistan [
30] also support this finding, which found that the odds of child health care practice were lower among children born to not educated mother. This might be explained by; educational status is more likely to correlate with wealth status, which leads to lack of money to afford cost of transportation and cost of health services.
The current study revealed that regardless of other variables, being a child born to not working/unemployed mother increases the odds ratio of not deworming to children from working women. This finding is in line with a study conducted in Ghana [
29], which disclosed that employed women are more likely to utilize deworming for their children compared to unemployed women. This might be due to not working women are less granted to afford child health services and cost of transformation.
This study revealed that being a child born at home was associated with higher odds of not taking deworming medication. A supportive study in Kenya [
31] reached a similar conclusion to this finding. This might be justified by; women who deliver at home are expected to have unfavorable awareness towards the health of their child compared to their counterparts. In addition to this, women who were not given birth at health facilities are less likely to receive advice about their future health of the child specifically advice about the importance of deworming drug supplementation during early childhood.
Furthermore, children who had diarrhea 2 weeks before the survey were less likely to take deworming drugs compared to those who had no diarrhea episode. This might be the relationship between diarrhea and intestinal parasitic infection [
32]. Children who had encountered diarrhea are more likely to seek health services, and the diarrhea episode could have created an opportunity to take deworming tablets.
From the community level factors, group regions of residency have increased the odds of not taking deworming drugs. Children residing in regions Afar, Somali, Harari, SNNP and Amhara has found increased odds of not taking deworming tablets. A similar supportive study from Ghana revealed that, residents from the western part of the country were less likely to take deworming tablets [
29]. This might be justified by the unequal distribution of health services and other Infrastructures throughout the country.
Conclusion
To conclude, the distribution of not taking deworming drug among 24–59 months of children was unevenly distributed across space and major public health concern in Ethiopia. Maternal education, perceived distance to health facility, and place of delivery to the last child, maternal working/employment status, region of living and diarrhea in the last 2 weeks prior to date of interview were statistically significant factors associated with unfavorable deworming drug uptake in Ethiopia. Developing targeted interventional programs towards the clusters identified by scan statistics are recommended to minimize this unfavorable deworming uptake. Additionally, enhancing mothers through education and employment, and increasing access towards facility delivery are also recommended to halt this public health problem.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.