Study design and data source
The study is cross-sectional in nature and based on data from the population-based Ethiopian Demographic and Health Survey (EDHS). Datasets used in this study were collected from the 1st, 2nd, 3rd and 4th rounds of EDHS conducted in 2000, 2005, 2011 and 2016 respectively, which are used to carry out the analysis. These surveys were conducted based on nationally representative sample households that provide estimates at the national and regional levels [37, 43-45].
EDHS was carried out by the Ethiopian Central Statistical Agency (CSA) and ICF International and provide quality information on a wide range of socio‐demographic, health and health-related indicators. In general, a DHS sample is stratified, clustered and selected in two stages. At the first stage of sampling, enumeration areas (EA) were selected using systematic sampling with probability proportional to size. In the second stage of sampling, a systematic sample of households per EA was selected in all the regions to provide statistically reliable estimates of key demographic and health variables. The 1994 Population and Housing Census (PHC), conducted by the CSA, provided the sampling frame from which the 2000 and 2005 EDHS sample was drawn. Whereas, the sampling frame used for the 2011 and 2016 EDHS is the 2007 PHC. A representative sample of 11,645 households from 539 clusters (138 in urban areas and 401 in rural areas) in 2000 EDHS; 14,500 households from 540 clusters (145 urban and 395 rural) in 2005 EDHS; 17,817 households from 624 clusters (187 in urban areas and 437 in rural areas) in 2011 EDHS, and 16,650 households from 645 clusters (202 in urban areas and 443 in rural areas) in 2016 EDHS were selected for the surveys and the response rates were 99, 98, 94, and 98%, respectively. Details of the survey are described elsewhere [37, 43-45]. The present study included all youngest children under age five living with the mother and mothers were asked about the disposal practice of the last passed feces for the youngest child. All respondents who responded to the outcome variable were included in the analysis for this study.
Study variables
Outcome variable
The outcome variable for this study was unsafe child feces disposal practices. The outcome variable was constructed based on the recent WHO definition, response categories such as ‘child used toilet or latrine' and ‘put/rinsed into toilet or latrine' were combined and coded as "safe disposal of child feces (coded as ‘0')". And the others were coded as "unsafe disposal of child stool (coded as ‘1')". Unsafe disposal of child feces was defined as the disposal of feces in any site other than a latrine, such as '‘put/rinsed into drain/ditch" '‘thrown into the garbage, '‘buried, '‘left in the open,' and ‘other' [7].
Explanatory variables
The explanatory variables include; sex of children (male, female), age of the child (0-12 months, 13-24 months, ≥ 25 months), mother's age (< 24, 24-34, ≥ 34 years ), mother educational level (no education, primary, secondary, higher), mother's working status (not working, working), partner educational level (no education, primary, secondary, higher), partner occupational status (working in agriculture, work in non-agriculture, not working), household size (<5, ≥ 5), number of children 5 and under (≤ 2,≥ 3), main floor material (cement, earth), sex of household head (male, female), place of residence (urban, rural), mother's exposure to media (yes, no), toilet facility (improved, unimproved), sources of drinking water (improve, unimproved) and presence of diarrhea in the last two weeks (yes, no) [7, 14, 20, 23, 29, 33].
The variable on media exposure includes exposure to the radio and television. The mothers who were not exposed to radio/television were coded as "no" and those who have frequent exposure were coded as "yes". Also, the toilet facility and source of drinking water were categorized into ‘improved' and ‘unimproved' following the WHO/UNICEF definition [46].
Operational definitions
Unsafe child feces disposal: refers to disposing of child feces in open areas or not disposing of them at all; those left in the open, thrown into the garbage, put/washed/rinsed into open drains, buried, or any other methods are considered unsafe disposal [6, 7].
Safe child feces disposal: safe disposal refers to a child use a toilet or latrine or, for very young children, to put or rinse their feces into a toilet or latrine [6, 7].
Statistical analysis
Data from the four waives of EDHS (2000-2016) is used to carry out the analysis. First, data were examined how outcome and explanatory variables were defined in each survey and, if necessary, create new "variables" that are as identical as possible over the survey years. Next, the four datasets (ETKR41FL, ETKR51FL, ETKR61FL, and ETKR70FL) were merged into a single data and analyzed using a complex sample analysis, taking into accounts for the strata, clusters, and weight variable. A complex sample analysis is a two-step process in SPSS, (1) create a complex sample “plan file” after computing a weight variable (V005) and (2), run analyses using the plan file through the complex sample package to account for sample design. DHS strongly recommends that weights be included in any statistical analysis that conducts with DHS data and complex sample command must be considered: for analyses of significance testing or a confidence interval (CI) [47]. A detailed explanation of the weighting procedure can be found in the EDHS methodology report [37, 43-45].
Descriptive summaries (weighted frequency and percentage) were used to explain the number of study participants in the analysis. A complex sample binary logistic regression model was employed and presented the crude odds ratio (COR) with 95% CIs to identify the relationship between the outcome variable and explanatory variables. Those variables with a p-value of < 0.25 were then entered into a multivariable logistic regression to control the effect of confounder's and to estimate the independent factors of unsafe child feces disposal [48]. Finally, significant variables were identified based on the adjusted odds ratio (AOR) with 95% CIs and p-value < 0.05. Multicollinearity effect was assessed with a cut of off point of variation inflation factor (VIF) of greater than ten [49] and finally, to check the correctness of the final formulated model, the Hosmer–Lemeshow test for overall goodness of fit was used [50]. All statistical analysis was carried out using SPSS version 20.0 (IBM Corp., Armonk, NY, USA).
Data quality assurance
In all rounds of EDHS, the data collection tools were pretested and data collectors were passes through extensive training. The training consisted of in-class training, biomarker training, and field practice days. Following the field practice, a debriefing session was held with the pretest field staff, and modifications to the questionnaires were made based on lessons drawn from the exercise [37, 43-45]. In this specific paper, I have greatly worked on data quality assurance by cleaning data before performing analysis.
Ethical consideration
The DHS surveys are anonymous surveys that do not allow any potential identification of any single household or individual in the data file. The analysis presented in the paper is based on EDHS (2000-2016) which is a publically available dataset with no identifiable information on the survey participants; no further effort was made to trace back the subjects. All the ethical concerns, including informed consent, are strictly followed in all rounds of EDHS. Given these, no ethical approval or informed consent was required for the current study. The data used in this analysis were obtained via online registration to measure the DHS program and downloaded after the purpose of the analysis was communicated and approved.