Factors associated with diarrhoea and acute respiratory infection in children under-5 years old in Ghana: an analysis of a national cross-sectional survey

Our study population was children under-5 years old (0–59 months) in Ghana. We analysed data from the 2017–2018 MICS conducted in Ghana. The MICS is a national representative household survey conducted in many countries in the world with assistance from United Nations Children’s Fund (UNICEF), and provides robust data on women and children [21]. The MICS is a cross-sectional design, which employs a two-stage sampling technique which selects census enumeration areas from each sampling strata proportional to the number of households in an enumeration area. The second stage involves selection of households from each enumeration area to form survey clusters using systematic random sampling. Eligible mothers with children under-5 years old in selected households were interviewed [20, 22]. The response rate for eligible mothers with children under-5 years old was 99.7%. Details of MICS sampling procedures have been published [21].

Our primary outcomes of interest were diarrhoea and acute respiratory infection (ARI). A child had diarrhoea if the mother or primary caretaker reported that the child had three or more loose or watery stools per day, or blood in stool two weeks prior to the survey. A child had ARI if the mother reported the child was ill with cough or difficulty in breathing [23], in the last two weeks. The term “mother” in our study refers to a biological mother or a female primary caretaker of the child under-5 years living in the same household.

We assessed whether child, maternal and household level factors were associated with diarrhoea and ARI in children under-5 years of age. These factors included: child’s age (0–5, 6–11, 12–23, 24–59 months); gender (boy, girl); child’s health insurance (yes, no); mother’s education (no formal education, primary, secondary, college or higher education); source of drinking water (improved, unimproved); sanitation (improved, unimproved); floor material (improved, unimproved); household wealth (poorest, poor, middle, rich, richest); place of residence (rural, urban); and presence of hand washing station in dwelling (yes, no).

We used the WHO/ UNICEF Joint Monitoring Programme for Water Supply, Sanitation and Hygiene (JMP) to categorize our source of drinking water [24]. An improved water source included source of drinking water from boreholes, piped water or tube wells, protected dug wells, protected springs, tanker-truck, rain water and packaged water, whilst an unimproved water source included unprotected springs, unprotected dug wells, and surface water collected directly from river, dam, lake, pond, stream, canal and irrigation channels. A household sanitation was considered unimproved if members of the household used pit latrines without a slab or platform, hanging latrines or bucket latrines, whilst improved sanitation included flush/pour flush toilet, pit latrine with slab, ventilated improved latrine and composting toilet according to JMP [25]. A household floor was considered improved if it was made up of cement, ceramic tiles, vinyl asphalt strips, parquet, polished wood, whilst floor materials such as earth, sand, dung, wood planks, palm, bamboo were considered unimproved [26]. Household wealth was categorized into wealth quintiles (poorest, poor, middle, rich, richest), and was determined using principal component analysis [27]. Our variable selection was guided based on previous studies [11, 16, 28], and data available in MICS.

Data were analysed using descriptive statistics and complex survey multivariable logistic regression. Descriptive statistics were used to describe our study sample, and to assess the prevalence of our primary outcomes (diarrhoea and ARI) in our study sample.

Complex survey regression models (i.e. model for diarrhoea and a separate model for ARI) were used to assess the relationship between our primary variables of interest and primary outcomes. We conducted both univariate and multivariable logistic regression analyses for each of the primary outcomes. Univariate analysis was conducted for all primary variables in relation to their respective outcomes. Univariate analysis with a P-value of less than 0.2 was used to select variables into the multivariable regression models. Variables that had a P-value of greater than 0.2 at univariate analysis but were clinically relevant or might have some biological plausibility with the outcome were included in their respective multivariable regression model. We also tested for multicollinearity for each of the multivariable regression models using pairwise correlation matrix, variance inflation factor and tolerance, and eigensystem analysis of correlation matrix [29], to ensure that there was no multicollinearity issues. We assessed the fitness of each of our models using the global null hypothesis test. A P-value of less than 0.05 was considered statistically significant. Missing data was not a problem in our study as only the sanitation variable had missingness of 0.01% (i.e. only one child), which was dropped.

In all our descriptive statistics and regression analyses, we applied sample weights, stratification and clustering to account for the complex survey design, and to ensure the representativeness of the data. Data were analysed using SAS version 9.3 (SAS Institute, Cary, NC).

The study population and analytic sample was made up of 8879 children under-5 years old. The mean age of a child was 30 ± 7.3 months. One in 2 children were females (50.8%), and more than half of the number of children in our study had a health insurance (58.4%). Many of the children were living in the rural setting (56.9%) (Table 1). The overall prevalence of diarrhoea in our study was 17.0% (95% CI: 15.70, 18.24%), and the overall prevalence of ARI was 33.3% (95% CI: 31.72, 34.82%) [Results not shown].

The multivariable regression analysis on diarrhoea showed that children aged 12–23 months and 6–11 months had 2.37 and 2.06 respectively, times the odds of diarrhoea compared to children aged 0–5 months. Children whose mothers had a secondary education, and a college or higher education had 34% and 59% lower odds of diarrhoea respectively, compared to children whose mothers had no formal education. Children from the richest households had 42% lower odds of diarrhoea compared to children from the poorest households [Adjusted prevalence odds ratio (aPOR): 0.58, 95% CI: 0.39, 0.86]. Children resident in rural settings had 22% lower odds of diarrhoea compared to those in urban settings (aPOR: 0.78, 95% CI: 0.63, 0.98) (Table 2).

Our multivariable regression analysis on ARI revealed that children aged 6–11 months (aPOR: 1.43, 95% CI: 1.06, 1.93), and 12–23 months (aPOR: 1.41, 95% CI: 1.10, 1.82), had higher prevalence of ARI compared to children aged 0–5 months. All other variables were not associated with ARI (Table 3).