Analysis of Factors Affecting Weight of Child at Birth in Pakistan: Multiple Imputation of Missing Data

Abstract

The problem of missing data arises very often in studies based on secondary data. Similar problem was encountered while identifying the potential risk factors of Low Birth Weight infants in Pakistan. Low Birth Weight (LBW) is an important indicator of newborn’s health status at the time of birth and during infancy. LBW is also identified as one of the significant risk factor of mortality during neonatal period. It would be useful for health sector to identify the significant risk factors associated with LBW infants for a particular community. Data of LBW infants was obtained from Pakistan Demographic and Health Survey (2006-07). It was found that data contained 76.8% missing observations for various potential risk factors. Missing values cause reduction in sample size leading to reduction in statistical power. Multiple Imputation is used to deal with the problem of missing values which is a powerful and flexible technique and is relatively easy to implement for handling missing data. To handle the problem, Multiple Imputation was used before fitting the Ordinal Regression model. It is a powerful and flexible technique and is relatively easy to implement for dealing with missing data. After Multiple Imputation of missing data, an Ordinal Regression model was fitted to determine the associated risk factors of LBW for infants. The analysis revealed that place of residence of the mother, wealth index, gender of the child, number of antenatal visits and multiple births have significant effect on birth weight of child in Pakistan.