The Neonatal Environment and Health Outcomes (NEHO) Birth Cohort Study: Behavioral and Socioeconomic Characteristics and Drop-Out Rate from a Longitudinal Birth Cohort in Three Industrially Contaminated Sites in Southern Italy

Abstract

Pregnant women living in industrially contaminated sites (ICSs) are exposed to environmental contaminants through different pathways, and thus children's health may be affected by pollutants. We created the Neonatal Environment and Health Outcomes (NEHO) longitudinal birth cohort in three ICSs in the Mediterranean area of southern Italy, collecting comprehensive information on personal data and lifestyles by questionnaire. Through multiple correspondence analysis, we identified possible clusters of enrolled women, and a neural network classifier analysis (NNCA) was performed to identify variables capable of predicting the attrition rate of the study. NEHO recruited 845 mother-child pairs over two years. The mothers' mean age was 31.1 ± 5.2 SD years. We found significant differences in socioeconomic status (SES) among the three evaluated ICS, and an overall 11.1% prevalence of mothers who actively smoked during pregnancy. Active smoking during pregnancy was strongly associated with the lowest socioeconomic level (p < 0.0001). By means of the NNCA, we found that smoking during pregnancy and the lowest education level characterized the cluster with the highest attrition rate (p < 0.001). Our results demonstrate that reason for public health concern still exists regarding smoking during pregnancy and that SES influences both lifestyles, producing negative pregnancy outcomes and a higher survey attrition rate.

Keywords: birth cohort study; fetal exposure; industrially contaminated sites; lifestyles; maternal exposure; socioeconomic status.

Figure 1

Distribution of weight gain in the categories “superior”, “recommended”, and “inferior” with respect to body mass index (BMI) baseline values within each national priority contaminated site.

Figure 2

Tukey’s post hoc multiple comparisons among the three national priority contaminated sites (NPCSs). Six variables relevant to gestational period and birth outcomes are evaluated. MIL: Milazzo NPCS; AUG: Augusta-Priolo NPCS; CRO: Crotone NPCS.

Figure 3

Distribution of dropouts within each national priority contaminated site.

Figure 4

Multiple correspondence analysis (MCA): four dimensions described the sample, and three clusters were identified. (a) Dimensions describing the sample variability; (b) Distribution of drop out in the three clusters in MCA classification.

Figure 5

Neural network classifier analysis including the following input variables: Industrially Contaminated Site (ICS), age, educational level, marital status (never married, married, separated/divorced), working condition, active smoking, and pregnancy wanted. (a) The neural network plot with 10 input nodes, a three-node hidden layer, and the output node; (b) importance of the variables in the estimated model.

Figure 6

Confusion matrix to describe the performance of the neural network classifier model for total testing sample and within each evaluated site. (a) Confusion matrix including the number of true and false positive/negative dropouts generated by the network; (b) confusion matrix for each National Priority Contaminated Site.