Study design, data source, and participants

The data used for this study were accessed from the Ethiopian Demographic and Health Survey (EDHS) 2005–2016 data, which was used in a community-based cross-sectional study. Ethiopia has nine regional states (Afar, Amhara, Benishangul-Gumuz, Gambella, Harari, Oromia, Somali, the Southern Nations Nationalities and People’s Region (SNNPR), and Tigray), as well as two city administrations (Addis Ababa and Dire Dawa). The emerging regions of Ethiopia were Afar, Somali, Benishangul-Gumuz, and Gambella [14]. The source population was all under-five children having diarrhoea who lived with their mother, whereas the study population was all chosen or sampled living under-five children having diarrhoea who lived with their mother in the selected areas. Mothers or carers with under-five children who were sick or had hearing impairments during enumeration time were excluded from the study.

The standard EDHS data set has a large sample size, which helps to generate parameters [15]. To collect data from nine regional states and two municipal administrations, the EDHS employed a two-stage sampling technique. In the first phase, clusters of enumeration areas (EAs) (urban and rural areas) were chosen with a probability proportionate to their size using the 2005–2016 primary health care (PHC) frameworks, with independent selection in each sample stratum. Details were described elsewhere [10, 16, 17]. The second step is a systematic sampling of residences within each cluster, or EA, followed by interviews with selected mother-child pairs. The EDHS findings, published on the Measure DHS website, provide a comprehensive sample procedure (www.dhsprogram.com).

Data collection instruments

Child sex, maternal age, marital status, religion, maternal education, wealth index, and media exposure were individual-level factors, whereas residence and regions were community-level factors [12]. Regions of the country were classified as developed and emerging based on existing evidence [18]. Women’s media exposure was assessed using three variables: frequency of reading newspapers or magazines, listening to radio, and watching TV. Women were classified as yes if they were exposed to all or any of the three, and no if they were not exposed to at least one [15]. The wealth index was computed using principal component analysis (PCA). Details are described elsewhere [12, 16, 17].

Outcome assessment

The zinc utilisation was assessed by asking the mother if zinc was administered to her child at any point since the start of diarrhea. The response variable was dichotomised as yes (1) or no (0) [9, 19].

Data processing and analysis

Stata^TM 14 was used to conduct the data analysis. Descriptive statistics, such as frequency and percentage, were employed to describe informants. The proportions and frequencies were weighted. All analyses used the individual sample weight (v005/1,000,000) to account for over- and under-sampling. The EDHS dataset is hierarchical, with children nested in households and households within clusters. A correlation matrix was used to assess multicollinearity among predictors.

There were four models used: the null model (no predictors), model 1: individual-level variables, model 2: community-level factors, and model 3: individual and community-level factors. Multilevel mixed-effects logistic regression analyses were carried out to determine the relationship between individual and community-level characteristics and zinc utilisation (yes = 1, no = 0). The data were analysed using a complex sample survey multilevel mixed-effects logistic regression analysis (melogit [pweight = swt] || v001]. The Stata command "svyset" was used to establish survey data and estimate the percentage of zinc utilisation. The strength of the link between predictors and endpoint variables was evaluated using the adjusted odds ratio (AOR) along with a 95% confidence interval (CI). The model’s goodness-of-fit was evaluated using the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and the log-likelihood ratio (LLR). The model with the lowest AIC and BIC, as well as the highest LLR, was chosen as the best fit [20].

The Median Odds Ratio (MOR), defined as the median value of the odds ratio between the areas at the lowest and highest risk when two clusters are randomly selected, was used to quantify variation. MOR = e0.95√VA or exp. [√ (2 × VA) × 0.6745], where VA represents the area-level variation [21]. The median odds ratio (MOR) is a statistical metric that evaluates the relationship between an exposure and an outcome in a logistic regression model. Its interpretation is similar to that of the ordinary odds ratio, but it focusses on the odds of a randomly picked individual from one group versus a randomly selected individual from another group. MOR = 1: This result implies that the chances of the outcome are the same between the two groups being compared. In other words, the exposure doesn’t change the chances of the result. MOR > 1: A MOR larger than 1 indicates that the exposure is associated with an increased likelihood of the result happening. This suggests a positive relationship, which means that as exposure grows, so does the chance of the consequence. If the MOR is less than 1, it indicates that the exposure is associated with a decreased likelihood of the outcome [22]. The proportional change in variance (PCV) measures the variance in zinc utilisation among under-five children, which is explained by several variables. The PCV is calculated as Vnull-VA/Vnull* 100. Where Vnull is the initial model’s variance and VA is the model’s variance with added terms. The intraclass correlation coefficient (ICC) estimates the variation in response variable between clusters. It is computed as ICC  =  VA ÷ VA + 3.29 * 100%, where VA  =  area/cluster level variance [21]. A p value less than 0.05 was declared statistically significant.

Ethical consideration

The study did not require ethics clearance or participant consent. Ethical permission was received from Measure DHS using a data access request form. The EDHS data is available to the general public in various formats upon request from the Measure DHS website (www.measuredhs.com). All approaches adhered to the key concepts of the Helsinki Declaration [23].

Sociodemographic and economic factors

This study employed a weighted sample of 29,525 mother-child pairs from the EDHS 2005–2016 (EDHS-2005 = 9,808, EDHS-2011 = 11,469, and EDHS-2016 = 8,248) datasets. Most of the children’s mothers (EDHS-2005 = 7,603 (77.16%), (EDHS-2011 = 8130 (69.86%), and (EDHS-2005 = 6,830 (64.25%)) were uneducated. The majority (EDHS-2005 = 2,525 (25.63%), (EDHS-2011 = 3,620 (31.11%), and (EDHS-2016 = 3,989 (37.52%)) of the study subjects were in the poorest wealth quintile. Most (EDHS-2005 = 8,497 (86.24%), (EDHS-2011 = 9,655 (82.96%), and (EDHS-2016 = 8,657 (81.43%) respondents were residents in the Oromia region of Ethiopia (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic and economic factors of study subjects, Ethiopia, EDHS 2005–2016 (N = 29,525).

https://doi.org/10.1371/journal.pone.0314127.t001

Proportion and trends of zinc utilisation

The pooled proportion of zinc utilisation among under-five children with diarrhoea in Ethiopia was 8.96% (95% CI: 7.44, 10.76%). In Ethiopia, the proportion of zinc utilisation by under-five children decreased from 0.22% (95% CI: 0.07, 0.74%) in EDHS 2005 to 0.04% (95% CI: 0.00, 0.22%) in EDHS 2011, and sharply increased to 33.60% in EDHS 2016.

Individual and community-level predictors of zinc utilisation among under-five children

The interclass correlation coefficient (ICC) of the null model was 30.25%. This suggested that under-five children had varying zinc utilization across clusters. The full model had the lowest AIC and highest LLR; hence, it was selected as the best-fit model. The multivariable multilevel mixed-effects logistic regression analyses revealed that having mothers complete primary education [AOR = 3.16, 95% CI: 1.57, 6.35], EDHS-2005 [AOR = 0.001, 95% CI: 0.0001, 0.004], and EDHS-2011 [AOR = 0.0001, 95% CI: 0.00002, 0.001] were significantly associated with zinc utilisation among under-five children with diarrhoea (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Multivariable multilevel mixed-effects logistic regression analyses of zinc utilisation and predictors among under-five children with diarrhoea, Ethiopia, 2005–2016 (N = 29,525).

https://doi.org/10.1371/journal.pone.0314127.t002

Policy implications

This study’s findings have significant implications for clinical and public health intervention. Despite the benefits of zinc, limited zinc utilization indicates a policy gap in Ethiopia’s maternal and child nutritional status. The findings have implications for improving food and nutrition policy, national nutrition strategy, and national nutrition programme by increasing mothers’ educational levels in order to boost zinc utilization among under-five children [33–35].

To improve zinc use in under-five children with diarrhoea, numerous policy strategies might be implemented: Governments and health organisations should engage in mass media efforts to educate parents on the benefits of zinc supplementation in addition to oral rehydration salts (ORS). Such campaigns can target less educated communities to enhance understanding and habits related to child health [19, 32]. Establishing mechanisms to track zinc supplementation rates can aid in identifying gaps in service delivery and informing future initiatives.

Strengths and limitations of the study

This study was based on nationally representative large weighted data and used appropriate statistical analysis techniques. However, the data were collected via self-reports, which may have contributed to recall bias. Moreover, because the study is cross-sectional, a cause-and-effect link may not be established. Furthermore, the EDHS survey does not include some data on dietary intake of zinc (dietary sources of zinc), which affects zinc utilisation. Women do not always offer zinc tablets because they believe that zinc-rich food is a substitute for zinc tablets.