Demographic characteristics are presented in . By design, the sample was comprised of young women, with 49% of participants aged younger than 20 years at study entry. The majority (80%) were African American. Only 31% were currently employed; most had completed high school or were currently in school (75%). Two-thirds were nulliparous. Seventy percent of participants were in a relationship with the infant's father, with 38% cohabitating.
Baseline Demographic Characteristics of the Study Group, Including Parity and Poverty Level
Participants reported relatively high levels of social support and moderate levels of social conflict: mean scores, respectively, were 4.17 (standard deviation [SD]±.85; median 4.46, interquartile range 3.71−4.86) and 2.94 (SD ±1.03; median 3; interquartile range 2.14−3.71). The Pearson r correlation between social support and social conflict was −.41, P<.001. Based on cutoff scores for denoting “probable depression” in research using community samples, 33% of participants reported elevated levels of depressive symptoms (Center for Epidemiological Studies-Depression Scale more than 16; mean 12.74, median 11; SD ±8.45, interquartile range 6−18; range 1−43).
presents results from hierarchical regression predicting depressive symptoms from sociodemographic factors, social support, and social conflict. Together, sociodemographic factors accounted for only 5% of the variance in depressive symptoms, with education and relationship status emerging as the only uniquely significant factors to predict depressive symptoms. Those with less education and not in a relationship with the infant's father were more likely to have depressive symptoms. Both social support (β=.21, P<.001) and social conflict (β=.47, P<.001) had independent effects on depressive symptoms and together accounted for an additional 34% of the variance in depressive symptoms. Social conflict emerged as a stronger predictor of depressive symptoms than social support. Finally, the interaction between social support and social conflict was non-significant; thus, these results do not provide support for an interactive or multiplicative effect of the two interpersonal factors on depressive symptoms.
Hierarchical Regression Analysis Predicting Depressive Symptoms From Sociodemographic Factors, Social Support, and Social Conflict
Post hoc analyses were conducted to examine whether the effect of social support and social conflict on depression differed depending on site, age, race, parity, relationship status with infant's father, education, employment, or community poverty. No significant interactions were found, indicating that within this sample of low-income, young women, the effect of social support and social conflict on depressive symptoms is similar across various sociodemographic characteristics.
Having established that both social support and social conflict have independent effects on depressive symptoms, the next objective of these analyses aimed to identify a specific clinical assessment that could be used by clinicians to best identify pregnant women with interpersonal risk factors for depression. For these analyses, the 1−5 response scales for the social support and social conflict items were dichotomized to reflect high-risk or low-risk responses, with the two least favorable responses reflecting high risk. For example, on items such as, “Do you have someone to talk to when you are upset?” responses of “Definitely not” or “Probably not” were collapsed to reflect high risk, and responses of “Possibly, Probably yes, and Definitely yes” were collapsed to reflect low risk.
The associations between dichotomized items and elevated depression were first computed with unadjusted odds ratios. Then, multivariable logistic regression was used to estimate the unique effect of these items on elevated depressive symptoms, controlling for sociodemographic characteristics. Four items (SS4, PSC1, PSC4, PSC7) were associated with an odds ratio of 2 or higher in multivariate analysis (). None of the other items on the social conflict or social support scale had an odds ratio greater than 2.
Unadjusted and Adjusted Odds Ratio for Four Key Social Support and Social Conflict Items Predicting Elevated Depression in Logistic Regression Analysis (n=1,047)
A variable reflecting a count of high-risk responses on these four key items was computed. This interpersonal risk composite was entered as a predictor of elevated depression in logistic regression, controlling for sociodemographic factors. Relative to women who gave no high-risk responses (ie, women with a 0 on the interpersonal risk composite; n=346), the odds of depression were significantly higher in women with one risk response (adjusted odds ratio [AOR] 3.37, 95% confidence interval [CI] 2.21−5.15; n=357), two risk responses (AOR 9.12, 95% CI 5.87−14.18; n=243), three risk responses (AOR 28.34, 95% CI 14.78−54.34; n=78), and four risk responses (AOR 28.20, 95% CI 9.60−82.75; n=23). illustrates the percentage of women with elevated depression for each grouping on the interpersonal risk composite. As shown, there was a “dose–response” effect, with each increase in interpersonal risk resulting in consequent risk for probable depression based on symptom reports (ie, Center for Epidemiologic Studies-Depression Scale 16 or more). A test of linear increase was significant, χ2 test of linear association was 199.2, df 1, n=1,047, P<.001). Specifically, although only 11% of women with a 0 on the interpersonal risk composite reported depressive symptoms, 29% of those with one risk factor reported depressive symptoms, 51% of those with two risk factors reported depressive symptoms, and more than 76% of those with three or more risk factors reported depressive symptoms.
Fig. 1 Percent of women with elevated depressive symptoms among women with different scores on the interpersonal risk composite. Scores on the interpersonal risk composite reflect how many of the four key interpersonal risk factors were endorsed (SS4, PSC1, (more ...)