Figure 3.

Summary of the relationships between low GA, neurodevelopment of preterm infants (non-optimality at 2 years) and the risk of parental separation. OR and HR were estimated using two different models (because of the absence of direct effect of low GA on the risk of parental separation). Model 1: logistic regression with outcome=non-optimality at 2 years and exposure=gestational age. Model 2: Cox model with outcome=parental separation and exposure=non-optimality at 2 years. Adjustment variables for both models: gender, multiple pregnancies, Z-score of birth weight, socioeconomic level, social security benefits for those with low incomes and urbanicity of the residential municipality. Only significant adjustment variables were reported in this figure. GA, gestational age; *SSB: social security benefits due to low income; ZS, Z-score