The risk of water, sanitation and hygiene on diarrhea-related infant mortality in eastern Ethiopia: a population-based nested case-control

This study was conducted in six randomly selected districts in West Hararghe administrative zone of Oromia region and Zone 3 of Afar region, which are situated in Eastern part of Ethiopia. The districts include Chiro, Mieso, Gemechis and Tullo districts of West Hararghe administrative zone of Oromia while Amibara and Awash Fentale districts that comprised in zone 3 administration of Afar region. Based on the 2007 National Population and Housing Census of Ethiopia [33], the projected population estimate for the year 2019 to each districts; Chiro (243,151), Mieso (191,978), Gemechis (263,615), Tullo (212,234), Amibara (85,964) and Awash Fentale (40,448). Of the total population, 3.4% are under the age of one year in the national context [33]. The study area has two main climate seasons - a dry season and a rainy season: The dry season ranges from October to February while the rainy seasons have two periods – the period from June to September is the main rainy season and some rainy weather period usually from March to May [34]. As of 2019, the health facilities available to each districts; Chiro (One hospital, 7 health centres and 39 health posts), Meiso (One hospital, 3 health centres and 33 health posts), Gemechis (6 health centres and 36 health posts), Tullo (8 health centres and 30 health posts), Amibara (one hospital, 3 health centres and 13 health posts) and Awash Fentale (2 health centres and 11 health posts). As a common, the most frequently cases observed in the districts are Acute upper respiratory tract infection, Lower respiratory infections such as pneumonia and bronchitis, Diarrheal diseases, Malaria, Urinary tract infections and skin infections. Due to inconsistency of secondary data, access coverage to improved water supply, sanitation and good hygiene practices is challenging in each districts. However, as per the expert’s opinion, most of the districts population have inadequate access to improved drinking water and sanitation, and insufficient hygienic practices were still a common problem.

The study employed community based nested unmatched case-control study, which was nested from longitudinal survey conducted to identify cause of infant mortality from September, 2016 to August, 2018 in Eastern Ethiopia.

The study population (mother-infant pair) consisted of cases and controls drawn from the longitudinal survey conducted in the randomly selected study area. Cases were infants who died from diarrheal disease (all 61 cases that have been identified in the study area during the study period were included) while controls were those who survived their first year of life.

The sample size was computed using Openepi version 3.03a, and calculated with the sample size for unmatched case control study. The sample size was calculated for each variable taking the result of the study conducted previously that would allow to taking the largest sample size. Accordingly, the sample size was determined based on the assumption that the proportion of latrine of the household as a risk factor for diarrhea-related infant death among the household of controls to be 15.4 and 5.93% for cases [30], two-sided significance level (1-alpha) = 95%, Power (1-beta, % chance of detecting) = 90, Ratio of Unexposed to Exposed in sample = 1:4, to detect an odds ratio of 0.29. Thus, the minimum sample size required for the study estimated to be 290 (58 cases and 232 controls). Because of many reasons of uncertainty and possible non-response, 10% were added, thus the final sample size resulted as 319 (64 cases and 255 controls).

The sampling techniques for this particular study exploit all diarrheal deaths of infants, which were selected as cases while density sampling used for the control groups. The entire cases (infant died from diarrhea) that ascertained by electronics verbal autopsy were directly taken from the longitudinal survey. Infants who had survive their first year were eligible for selection as control group, which were randomly selected from the adjacent area linked to that of the infants died due to diarrhea. For each case, four neighborhood controls were randomly selected from the longitudinal survey database and compared in terms of socio-demographic and several risk components of Water, Sanitation and Hygiene.

Data were collected by trained data collectors using structured and pre-tested questionnaire and conducted under closely overseen by supervisors. The questionnaire was prepared based on WHO, UNICEF and national standards as well as adopted from relevant literatures [35, 36]. This questionnaire was first developed in English, then translated into both “Amharic” and “Afan Oromoo”, and back translated into English to ensure its meaning of questions is retained and checked consistency. The respondents were mothers or primary caretaker of the infants (< 1 year of age).

Data quality was assured by controlling both random and systemic error. Data quality was maintained through properly designed data collection tool, and data collectors as well as supervisors recruited with the relevant educational background and language proficiency. Training was held and pre-test carried out in a community with similar characteristics. The data collection procedures developed and the collected data were reviewed by principal investigator. Any identified errors were discussed and immediate measure has been taken.

The collected data were entered into CSPro version 6.1 then transformed to SPSS version 23 for analysis. Descriptive statistics such as frequency distribution and cross tabulation were made to summarize the study variables. Both bivariate and multivariable conditional logistic regression was used to estimate crude and adjusted odds ratios with 95% confidence intervals for the association between risk factors and diarrhea-related infant death. Bivariate conditional logistic regression for each variable was analyzed and any risk factors that showed marked association (p-value < 0.25) were considered as candidate for multivariable logistic regression. The Hosmer-Lemeshow goodness-of-fit test was checked and indicated as this model is valid seeing that the p-value is greater than 0.05. Variables that resulted P value < 0.05 in the multivariable logistic regression were affirmed as significantly associated with the outcome variable (diarrhea-related infant death).

A total of 305 study subjects (61 cases and 244 controls) were included in the study, which yields the non-response rate 4.7 and 4.3% for case and control, respectively. Among these studied subjects, 61 were infants who died as the result of diarrheal disease (cases),and 244 were those who survived their first year of life (controls),which was nested in a longitudinal survey database (from September 2016 to August 2018) residing in Eastern Ethiopia.

Of the Interviewed infants’ mothers/caretakers, the mean age (± SD) of the respondent among cases (infant who died due to diarrhea) was 26.8 (±3.9) years old and 27.8 (±4.5) for controls (infant who survived their first year of life). The majority of the infants’ mothers of the cases fall within the youth age group of 20-34 (90.2%), while it was (82.0%) for the controls. About 44.3% of cases and 48.4% of controls had history of having borne two to four viable offspring (Parity). The mean (± SD) family size of households with infants in the cases was 4.70 (±1.98) and 4.99 (±1.96) for controls. Majority of the study participants were married at 96.7% of cases and 98.0% of controls. Oromo ethnic group comprises the largest proportion of the study subjects (90.2% of cases and 87.3% of controls). Muslim followers were larger in the study participants at 98.4% for cases and 88.5% for the control group. The majority of the respondents (88.5% of cases and 74.2% of controls) were not educated. Likewise, most of the spouse of the cases (78.7%) was uneducated as compared with the controls (65.2%). Almost equal proportion of the cases (90.2%) and the controls (90.6%) were housewives by occupation. Spouse’s occupational status between the two-study groups indicated that about (90.2%) cases and (85.7%) controls were found to be farmers/own farm labor. High proportion of controls (50.8%) compared with that of cases (45.9%) had an average household monthly income of more than and equal to 570 ETB.

The bivariate analysis between socio-demographic characteristics and diarrhea-associated infant death indicated that mothers and spouse’s level of education were significantly associated with infant’s diarrheal death. In this analysis, it was estimated that the infants whose mothers were not educated had less likely for risk of infant death as a result of diarrhea than those whose mothers had reached at some level of schooling (P = 0.021, COR: 0.37, 95% CI: 0.16, 0.86). Likewise, the death of infants due to diarrhea was less likely to occur for the uneducated spouse than educated (P = 0.046, COR: 0.51, 95% CI: 0.26, 0.98) (see Table 1).

The distribution of cases and controls as well as bivariate analysis in the different categories of Water supply, Sanitation and Hygiene presented as follows:

Almost equal proportion of case (78.7%) and control (78.3%) group of infant’s households were used improved water sources. The household’s time to access water source resulted with 30 min or less (65.6% of case and 62.7% of control). About 85.2% of cases and 82.0% controls fetched water within 1 km radius from their dwelling. Most of the infants in the households with less than 25 l water consumption per capita per day among case and control group appears to be 88.5 and 75.4%, respectively. The vast majority of cases (91.8%) and controls (79.1%) among infants in the households found with water inaccessibility (i.e households not access to at least 25 l per capita per day within a distance up to 1 km radius). High proportion of cases 65.6% compared with that of control (40.2%) found to have unsafe drinking water storage. Majority of controls (65.2%) compared with case (52.5%) reported to know at least one and more households point-of-use drinking water treatment methods. However, about 80.3% of the cases and 61.5% of controls’ group have ever practiced water treatment at household’s point-of-use.

In the bivariate analysis, the exposure variables that showed significant association with diarrheal-related infant death were households Water Consumption Per Capita per day with less than 25 l/c/day (P = 0.031, COR: 0.40, 95% CI: 0.17, 0.92), households water inaccessibility (P = 0.028, COR: 0.34, 95% CI: 0.13, 0.89), households with unsafe drinking water storage (P < 0.001, COR: 0.35, 95% CI: 0.20, 0.63), households reported as did not practices water treatment at Point-of-use (P = 0.007, COR: 0.39, 95% CI: 0.20, 0.77). However, all of these variables as the risk were less likely to be occurred infant death due to diarrhea (see Table 2).

In the bivariate analysis, households with unimproved sanitation (P = 0.011, COR: 0.39, 95% CI: 0.19, 0.80), households with unsafe disposing of child feces (P < 0.001, COR: 0.28, 95% CI: 0.15, 0.51), households with improper management of solid waste (P < 0.001, COR: 0.21, 95% CI: 0.12, 0.39) were less likely to occur infant death as a result of diarrhea. However, infants in the households who disposed liquid waste unsafely were 3.4 times more likely at risk of diarrheal-associated infant death as compared to infants in the households who properly managed it (P = 0.001, COR: 3.38, 95% CI: 1.68, 6.80) (see Table 3).

In the bivariate analysis, infants in the households did not practice hand washing in any critical time at all were less likely to occur infant diarrheal death as compared to households practiced three and more critical time of Hand Washing (P = 0.025, COR: 0.38, 95% CI: 0.16, 0.89). In contrary, infants in the households practiced less than three critical time of hand washing was three times more likely to have diarrhea related death than households practiced three and more critical time of Hand Washing (P = 0.002, COR: 3.02, 95% CI: 1.48, 6.16). The occurrence of diarrhea death among infant’s households who did not use agents (water with soap or ash/abrasives) during hand washing was two times higher than those who used agents (P = 0.030, COR: 1.91, 95% CI: 1.07, 3.43) (see Table 4).

In the multivariable conditional logistic regression, the risk factors that were significantly associated with infant diarrheal death identified includes age of mother with < 20 years old, unsafe drinking water storage, infants in households without point-of-use water treatment practices, households with unimproved sanitation status, households with unsafe disposing of child feces, households with improper management of solid and liquid waste, households with lesser hand washing practices at critical time.

Infants in households with unsafe drinking water storage and households treating their drinking water at point-of-use showed significant association with infant diarrheal death. However, households exposed to unsafe drinking water storage were less likely to be at risk of infant death from diarrhoea as compared to that of safe drinking water storage (P = 0.013, AOR: 0.4, 95% CI: 0.18, 0.81). Similarly, those households who did not treat their drinking water at point-of-use were less likely to be at risk of having infant death from diarrheal than those treated their drinking water (P = 0.004, AOR: 0.21, 95% CI: 0.08, 0.61).

The occurrence of diarrheal death among infants in households with unimproved sanitation status was less likely than households with those improved sanitation (P = 0.050, AOR: 0.36, 95% CI: 0.13, 1.00). Compared to households with the safe disposing of child faeces, those disposing unsafely were found with an increased odds of infant death due to diarrhoea (P = 0.014, AOR: 0.34, 95% CI: 0.15, 0.81). Infant diarrheal death were less likely to happen in the households disposing solid wastes improper than those properly managed (P = 0.003, AOR: 0.29, 95% CI: 0.13, 0.66).

Households with improper management of liquid waste management found with strong association with more than three times more likely to occur diarrhea-associated infant death as compared to those with proper liquid waste management (P = 0.010, AOR: 3.43, 95% CI: 1.34, 8.76). Infants whose mother/caretaker practiced hand washing with less critical time was three times greater risk to infant death from diarrhea than those who had practice more than three critical times of hand washing (P = 0.027, AOR: 3.04, 95% CI: 1.13, 8.17) (see Table 5).

The strengths of this study include as it used nested case-control study design which is valid and efficient design and can minimized both selection and recall bias. Despite such studies is available insufficiently at population level, in particular, this study could provide useful information for building evidence based health policy, better insights in planning the best solutions, and generate ideas for further research.

This study has limitations as respondents might not give their exact observable fact towards some given questions, which leads to social desirability biases. This could be minimized by providing training to data collectors and conducted pre-test before actual data collection launched. A very shortage of previous similar studies was made comparison difficult.