Multilevel and interaction modelling of factors associated with pneumonia among children under-five years in Uganda

Abstract

The study aimed at determining the risk factors associated with pneumonia among children under-five years in Uganda using secondary data from 2016 Demographic and Health Survey. Data was applied to 4417 children under- five years nested within the mother. Frequency distribution tables were used to describe the data. Chi-square tests were done at bivariate level to identify risk factors for pneumonia and multilevel mixed effects complementary log log regression model was fitted to assess the determinants of pneumonia. Results indicate that child’s age, mother’s age, and region were significant predictors of childhood pneumonia. Children aged 12 to 23 months had an increased risk of pneumonia compared to children aged 0 to11 months (OR=1.433, p=0.012). Children aged 36 to 47 months (OR= 0.662, p=0.047) and 48 to 59 months (OR=0.498, p=0.001) had a reduced risk of pneumonia compared to children aged 0 to 11 months. Children born to mothers aged 20 to 24 years (OR=0.552, p=0.011), 25 to 29 years (OR=0.622, p=0.046), 30 to 34 years (OR=0.514, p=0.007) and 40 to 44 years (OR=0.443, p=0.005) had a reduced risk of pneumonia compared children born to mothers aged 15 to 19 years. Children born to mothers from the western region had a reduced risk of pneumonia compared to children born to mothers from the central region (OR=0.225, p=0.000). Children born to mothers from eastern region had a reduced risk of pneumonia compared to children born to mothers from central region (OR=0.335, p=0.001). This study recommends that governments’ educational programs aimed at creating public awareness especially those that are related with childhood pneumonia should be put in place targeting mothers from central region. This is because children born to mothers from central region are the ones at higher risk of getting pneumonia. This study also recommends that other researchers should use random effects models when demographic and health survey data are used for childhood pneumonia analysis due to the hierarchical structure of the data. This is because multilevel modelling has the advantage of taking the hierarchical structure of such nested data into account to identify possible risk factors, unlike classical regression models.