QUANTILE REGRESSION MODELLING OF FACTORS INFLUENCING NUTRITIONAL STATUS OF UNDER-FIVE CHILDREN IN NIGERIA

Abstract

BACKGROUND Body Mass Index (BMI) is a measure of under-five children nutritional status and a proxy for a nation’s economic future productivity. BMI may not always satisfy the normality assumptions when studied with parametric models, restraining comprehensive assessment of the explanatory variables. Limiting health challenges has been associated with extreme values of the BMI. The study investigated the factors associated with the nutritional status of under-five children measured by BMI and explored its full distributional effect using quantile regression model. METHODS Nationally representative records of under-five children was obtained from the Multiple Indicator Cluster Survey round 5 (MICS5 – 2016/2017). The outcome variable was BMI defined as weight in kilograms per height in metres squared (kg/m2). The explanatory variables were age (months), sex, location of residence, mothers’ level of education, geopolitical region, parent’s wealth status and household size. Descriptive statistics such as frequency tables and proportions were used to summarize categorical variables, median and IQR for continuous variables, skewness and kutorsis to measure the spread and peakedness of BMI, the test of normality was performed BMI and the pseudo R2 to measure the model goodness of fit. The Quantile regression model was fitted using selected conditional quantiles, with their respective 200 resample bootstrapped standard error at 95% confidence level. RESULTS Of the 27,766 total records, 14,048 (50.6%) were male children. The median BMI of male children was slightly higher (15.44 kg/m2, IQR=1.92) than the median BMI of female children (15.11 kg/m2, IQR=1.91). The skewness and kurtosis of the total participants BMI were 1.20 and 15.13 respectively and the test of normality on BMI was statistically significant. Male children were more likely to have increased BMI compared to the female children with coefficients of 0.23, 0.33 and 0.27 at the 5th, 85th and 95th quantile respectively. Location of residence was associated with BMI at the 85th quantile for female participants. By geopolitical region, North Central was associated with BMI at the 25th, 50th, and 75th quantile for all children. Age, sex, mothers’ level of education, geopolitical region, and parent’s wealth status of both male and female children were associated with BMI. The model summary revealed that the pseudo R2 decreased with increasing quantiles. CONCLUSION Quantile regression provides a comprehensive insight to how the covariates vary across the BMI distribution. The inherent characteristics of the children’s BMI was characterized by a significant test of normality, revealing the great importance of the quantile regression model. The findings of this study further showed a relatively low BMI of the children indicating that the nutritional status of children is still a problem of public health importance.