OBJECTIVES

: Low birthweight is a term used to describe babies who are born weighing less than 2,500 grams. Over 8 percent of all newborn babies in the United States have low birthweight. This study aims to examine the predictors of low birth weight and build a predictive model for low birth weight using artificial neural network and compare its performance to logistic regression model.

METHODS

: The National Survey of Family Growth (NSFG) 2011-2015 data were used for this study. All the participants who were eligible were randomly assigned into 2 groups: training sample and testing sample. We used artificial neural network and logistic regression to predict the risk of low birth weight. Receiver operating characteristic (ROC) were calculated and compared for these two models for their discrimination capability and a curve using predicted probability versus observed probability were plotted to demonstrate the calibration measure for these two models.

RESULTS

: About 10.2% (n=1339) of 13159 births were low birth weight. According to the logistic regression, duration of completed pregnancy in weeks, wantedness of pregnancy of the respondent, education, race, Whether R received public assistance in prior calendar year, Poverty level income, born outside of US were significant predictors for low birth weight. According to the neural network, the top 5 most important predictors were Duration of completed pregnancy in weeks, Birth order, Formal marital status at pregnancy outcome, Age at time of conception, Labor force status.

CONCLUSIONS

: We identified several important predictors for low birth weight e.g., duration of completed pregnancy in weeks, birth order, formal marital status at pregnancy outcome, age at time of conception, labor force status. This provided important information for social works and healthcare providers for early intervention. We built a predictive model using artificial neural network as well as logistic regression to provide a tool for early detection.