Maternal Depressive Symptoms and Infant Health Practices Among Low-Income Women

Study Design

This research was part of a larger, prospective, community-based study on maternal stress, birth outcomes, and infant health. Our study included data from 3 surveys, the first of which was conducted with pregnant women at their first prenatal clinic visit after written consent. During a 2-year period, from February 2000 to November 2001, we consecutively recruited women who were receiving prenatal care at a consortium of public health centers in Philadelphia, Pennsylvania. Our inclusion criteria at enrollment were a singleton intrauterine pregnancy and the ability to speak English or Spanish. Ninety percent of all eligible women enrolled in the study. The second and third surveys were conducted with these women in their homes, at ∼3 to 4 months and 9 to 12 months after parturition, respectively. Eighty-four percent of all enrolled study participants with known live-birth outcomes completed the first postpartum survey, and 85% of those women subsequently completed the second postpartum survey. Eligible study participants were lost to follow-up after the initial survey for the following reasons: 4% refused, 3% moved out of the Philadelphia area, and 9% could not be contacted. Similarly, study participants were lost to follow-up after the second survey for the following reasons: 2% refused, 3% moved out of the Philadelphia area, 9% could not be contacted, and 1% were dropped because of death or severe mental illness. There were no significant differences with respect to race/ethnicity, education, or marital status between study participants who were interviewed and those who were lost to follow-up. However, study participants who were interviewed were on average ∼1 year younger than those who were lost to follow-up.

Structured surveys were conducted in English and Spanish by trained female interviewers using standardized questionnaires. At the first interview, we assessed chronic stress, sociodemographic factors, and behavioral practices. The second and third interviews contained information on the health of the child, including maternal reports on the cognitive, social, and behavioral development of the child, and the use of infant health services, parenting practices, and the use of injury-prevention measures. We also asked questions regarding parenting attitudes, the mother's social support network, and her neighborhood and financial conditions. As part of all 3 surveys, we used the Center for Epidemiologic Studies Depression Scale (CES-D) to measure maternal depressive symptoms. Details of the larger aforementioned study were published previously.²⁷

For this analysis, we included women living with the index child at the time of both postpartum surveys (98% of all study participants) and those who completed the CES-D at all 3 time points. A total of 778 mothers completed all 3 surveys and were still living with their infants. Four mothers were dropped because they did not complete the CES-D at all 3 time points, resulting in a total of 774 subjects.

Study Variables

Maternal depressive symptoms were measured with the CES-D, a reliable valid scale that has been used widely for more than a decade.^28–30 We defined depressive symptoms as a CES-D score of ≥16. This cutoff point corresponds to 80th percentile scores for community samples and has a 95% sensitivity for diagnosing major depressive disorder (MDD) among low-income women, although the specificity and positive predictive value for MDD are low (70% and 0.28, respectively).^29,30 The cutoff point of 16 has been used by many investigators assessing depressive symptoms in a variety of cohorts, including pregnant women.^28–33 When studying depressive symptoms during pregnancy, some investigators chose to use a higher CES-D cutoff point (eg, the 90th percentile) to account for the possibility that symptoms of normal pregnancy may overlap with symptoms of depression.^9,18 There is no evidence that this approach is more accurate or preferable to using the cutoff point of 16, and the use of higher cutoff points increases specificity but decreases sensitivity for MDD.²⁸ We used a consistent cutoff point of 16 to define depressive symptoms before and after parturition. Mothers were identified as having persistent depressive symptoms if their scores on the CES-D were ≥16 at all 3 time points, as ever having depressive symptoms if their scores were ≥16 at 1 or 2 time points, and as never having depressive symptoms if their scores were always <16.

We studied 6 dependent variables, which were clustered into 3 categories: 1) infant health services use (adequate well-child care and ever being hospitalized); 2) parenting practices (breastfeeding for ≥1 month and using corporal punishment); and 3) injury-prevention measures (having a smoke alarm and using the back sleep position). The dependent variables used in our study were chosen because they are health-related topics that have been discussed widely in the lay press and medical literature and they have all been shown to be related to the well-being of young infants.^34–39

Adequate well-child care was defined as the appropriate number of well-child care visits for the age in months, as recommended by the American Academy of Pediatrics,³⁶ and was based on the answer at 3 to 4 months after parturition (second survey) to the following question: “How many times has [child] been to a doctor or nurse for routine well-infant care? Don't count the times you took your infant for care when he or she was sick.” For this variable, adequate well-child care was determined for each child on the basis of his of her age, providing a 1.5-month allowance. The American Academy of Pediatrics recommends at least 1 visit by 1 month of life, and our criterion for adequate well-child care was having at least 1 visit by age 2.5 months, 2 visits by age 4.5 months, and so forth.

Two other dependent variables were assessed by using data from the second survey, when the infants were an average of 3 to 4 months of age. Breastfeeding for ≥1 month was determined with the following questions: “How long did you breastfeed [child] for? How many days, weeks, or months?” Similar to surveys conducted by the Centers for Disease Control and Prevention,⁴⁰ use of the back position was determined with the following questions: “In what position do you usually put [child] down to sleep? On his or her side, back, or stomach?”

Ever being hospitalized in the first year of life, use of corporal punishment in the past week, and having a smoke alarm in the house were determined with information obtained at 9 to 12 months after parturition (third survey), with the following questions. “How many times has [child] been admitted to a hospital since he or she was born?” “About how many times, if any, have you had to spank your child in the past week?” “Please answer ‘yes’ or ‘no’ to: your home has a working smoke alarm.” For the question on spanking, we dichotomized responses into “0” and “≥1” episode of spanking, with the latter defining use of corporal punishment in the past week.

Control variables that were selected for possible inclusion in ≥1 of the multivariate models were as follows: sociodemographic characteristics (maternal race/ethnicity, education, age, marital status, parity, income, nativity, Women, Infants, and Children Supplemental Nutrition Program (WIC) use, employment, housing stability [measured by address changes and homelessness], and child's age and insurance status), maternal health behaviors and health status (smoking status, physical health, and postpartum hospitalization), and infant health characteristics (gestational age at birth, birth weight, admission to the intensive care nursery at birth, health status as perceived by the mother, and special medical needs).

Statistical Analyses

Unadjusted group differences were tested by using analysis of variance with Bonferroni posthoc tests (to adjust for multiple comparisons) for continuous variables and the χ² test for categorical variables. To estimate the risk of adverse infant health practices associated with maternal depression, odds ratios were calculated as approximations of relative risks. To adjust for potential confounding variables and to derive maximal likelihood estimates of combined relative odds with 95% confidence intervals, multivariate, unconditional, logistic regression analyses were performed for each of the 6 dependent variables. The α level was set at .05 (2-sided), and Stata 6.0 software (Stata Corp, College Station, TX) was used for all analyses.⁴¹

For possible inclusion in these models, candidate risk factors and confounders were identified a priori on the basis of our review of the literature and theoretical considerations. The following variables were included in all 6 multivariate models: maternal race/ethnicity, education, age, income, and nativity. Other variables (maternal parity, housing stability, hospitalization, perceived health status, employment, use of the Women, Infants, and Children Supplemental Nutrition Program, and cigarette smoking; whether the mother was living with a partner; and infant gestational age, birth weight, need for transfer to an intensive care nursery, health insurance, special needs, health status as perceived by the mother, and age at the time of the survey) were included if the adjusted odds ratio differed from the crude odds ratio by at least 10%, which is a well-accepted method of confounder selection when the decision of whether to adjust is unclear.^42,43 Any variable associated with both the predictor (depression) and the outcome (infant health services use, parenting practices, or injury-prevention measures) at P < .25, as suggested by Mickey and Greenland,⁴² was also included. The α level was set at .25 to provide sufficient statistical power to detect important confounder effects.⁴³ Interactions (none of which were identified) were assessed by stratification and by testing of interaction terms for statistical significance. Table 3 provides details of the final variables in each of the 6 models. The goodness-of-fit χ² statistic described by Lemeshow and Hosmer⁴⁴ was calculated to assess the fit of each multivariate model. This research was approved by the institutional review boards at Thomas Jefferson University and the University of Pennsylvania as part of a larger study.

Subject Characteristics and Maternal Depression

Sociodemographic characteristics for the 774 subjects are presented in Table 1. The majority of participants were single, United States-born, low-income, young, African American women. Approximately one half were primiparous. Sixty-three percent were uninsured at the time of the first prenatal visit, and 83% did not have an education beyond high school. Most infants were born at term and were of normal birth weight (Table 1).

On the basis of CES-D scores at all 3 time points, participants were divided into 3 symptom categories; 40% of all participants never had depressive symptoms, 48% ever had depressive symptoms, and 12% had persistent depressive symptoms. Of those who ever had depressive symptoms, 25% had symptoms in the antepartum and postpartum periods, 26% in the antepartum period only, and 49% in the postpartum period only.

The 3 symptom groups were similar with respect to race/ethnicity and marital status. Women who never had depressive symptoms were younger, on average, than women who ever or persistently had symptoms. Those with persistent symptoms were more likely to have less than a high school education, be multiparous, and smoke cigarettes and were less likely to describe their child's health as excellent (Table 2).

Dependent Variables

For our study population, we assessed the overall prevalence of each of the 6 outcomes. Ninety percent of participants reported adequate well-child care for their young infants, and 21% reported their infant ever being hospitalized in the first year of life. Thirty percent breastfed for ≥1 month, and 16% reported the use of corporal punishment in the week before the survey. Ninety-four percent reported having a smoke alarm, and 52% reported using the back sleep position (results not shown).

Figure 1 shows the prevalence of infant health practices for each of the 3 symptom groups, ie, never, ever, and persistent. Women with persistent symptoms were less likely to obtain adequate well-child care, have a smoke alarm, and use the back sleep position but were more likely to have an infant ever hospitalized and use corporal punishment than those who never had symptoms. When the 3 groups were compared, there was no statistical difference in the prevalence of breastfeeding for ≥1 month, but the overall trend showed a decline in breastfeeding rates with increasing depressive symptoms.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Prevalence of infant health practices among the various depressive symptom groups (N = 774). Statistical significance was determined with χ² analyses (*P < .05; **P ≤ .01; ***P ≤ .001).

Associations between the dependent variables and depressive symptoms, after adjustment for confounding variables, are shown in Table 3. Women with persistent depressive symptoms were nearly 3 times as likely to have their infant ever hospitalized than were those who never had symptoms (Table 3). Although Fig 1 shows that infants of women with persistent symptoms had a lower prevalence of adequate well-child care, no difference in adequate well-child care was found between women with persistent or ever symptoms and those without symptoms, after adjustment for confounding variables. In the adjusted analysis, there was no association between maternal depressive symptoms and breastfeeding for ≥1 month.

Women with persistent depressive symptoms were nearly twice as likely to use corporal punishment, compared with women who never had symptoms. Those who ever had depressive symptoms were 39% less likely to use corporal punishment than were those who never had symptoms. Mothers with persistent symptoms were approximately one-fourth as likely to have smoke alarms in their homes and nearly one-half as likely to use the back sleep position than mothers without symptoms.