A Comparison of the Effective Factors of Preterm Birth Versus Low Birth Weight in Southern Iran Using Artificial Neural Network

Infants are one of the most vulnerable social groups whose mortality is considered as the development index of a community and family health status. Since preterm birth (PTB) and low birth weight (LBW) are two most important causes of death in infants and are affected by social and economic conditions and geographical living area, investigation of their important risk factors was the attempt in the present study.
The variables during pregnancy of 1102 newly delivered mothers referred to Shiraz (southern Iran) University’s hospitals were gathered to analyze their effects on birth weight and gestational age of their infants. Artificial neural network (ANN) method was utilized to determine and rank the effective factors on PTB and LBW separately. The performance of ANN model was evaluated by sensitivity, specificity, accuracy and the area under the receiver operating characteristic (ROC) curve. In addition, the amount of increase of mean squared errors (MSEs) in the trained network was considered as the ranking criterion.
Some differences in effective factors of these two pregnancy outcomes appeared. The first three important risk factors were consumption of iron, abortion history and hyperthyroidism for PTB and gestational age, consumption of iron and number of pregnancies for LBW respectively.
The results confirmed the proper performance of ANN method. PTB may be more dependent on the mothers’ habits or internal factors while LBW depends on the mothers’ history and external factors.