Measures

Descriptive data for the study group included maternal demographic variables (age at midgestation, race, education, employment, marital status) and clinical characteristics (parity, smoking status, alcohol intake) and depression symptom score on the 29-item Structured Interview Guide for the Hamilton Depression Rating Scale With Atypical Depression Supplement (SIGH-ADS)¹⁶ at 20 weeks antepartum and 2 weeks postpartum. The delivery and infant data included the rate of preterm birth (< 37 weeks gestation), 1-minute and 5-minute Apgar scores, infant sex, growth measurements (weight, length, and head circumference) at birth, infant age at assessment, and admission to the neonatal intensive care unit (NICU).

Primary outcome analyses focused on a variation of the modified Finnegan Scale that ranged from a score of 0 to 41 such that higher scores reflected more severe signs. As analysis results will illustrate, the primary outcome of interest required categorization into more meaningful groups due to the highly skewed nature of the scores.

The primary predictor of interest was the study group to which the mother belonged: (1) Comparison, (2) SRI, or (3) Mood Disorder. Preterm birth was a secondary predictor of interest because it is more frequent in both SRI- and depression-exposed than unexposed newborns.⁵ Additionally, the behavior of preterm infants also differs from that of full-term infants. Two weeks after birth, preterm infants have significantly more autonomic, motoric, state, attentional, and regulation disorganization than full-term infants.¹⁷

Descriptive statistics (frequency and percent for categorical variables; mean and standard deviation for continuous variables) were used to summarize patient characteristics. To compare the Finnegan scores across the 3 study groups, we used the nonparametric Kruskal-Wallis test followed by a series of Wilcoxon rank sum tests for pairwise comparisons. However, since approximately two-thirds of the Finnegan scores were either 0 or 1, we dichotomized the primary outcome variable into presence of signs (Finnegan score ≥ 2; score of 2 or more) or absence of signs (Finnegan score ≤ 1; score of 0 or 1).

We employed the Pearson χ² test to assess associations between group and categorical measures (symptom presence, group and preterm birth status, and preterm status and symptom presence). Fisher exact test was used when expected cell counts were less than 5 for a given cross-tabulation. Continuous variables were compared across groups with 1-way analysis of variance (ANOVA) for normally distributed data and the Kruskal-Wallis test when statistical assumptions (ie, normality) were questionable. The association between Finnegan score and demographic and clinical measures were examined using similar methods. Post hoc comparisons between exposure groups employed Bonferroni corrections for significant baseline characteristics and clinical measures. Given the non-normal distribution of SIGH-ADS scores in our sample, we employed the nonparametric Wilcoxon rank sum test to compare postpartum SIGH-ADS scores across symptom groups. Additionally, we performed multiple linear regression on postpartum SIGH-ADS controlling for antepartum SIGH-ADS at 20 weeks gestation to assess the effect of exposure groups.

The Cochran-Mantel-Haenszel (CMH) test was used with study group as the stratification factor to examine the association between preterm birth status and symptom presence. We summarized the effects with odds ratios (ORs) and corresponding 95% confidence intervals (95% CIs) for comparison between preterm status and symptom presence. In addition, we used the Breslow-Day test to examine the homogeneity of these ORs.¹⁸ We conducted multiple logistic regression for symptom presence using preterm birth and study group as predictors. Analyses were performed using IBM SPSS Statistics for Windows, Version 23.0 (Armonk, NY; IBM Corp) and R (R Foundation for Statistical Computing; Vienna, Austria; Version 3.2.0). All analyses assumed a 5% level of significance.