Predictors of Maternal Depressive Symptom Trajectories Over the First 18 Months in Home Visiting

Abstract

Maternal depression negatively impacts maternal functioning and parenting behaviors. Mothers participating in home visiting programs are at particularly elevated risk for depressive symptoms due to demographic and associated risk factors. Moreover, additional empirical evidence has demonstrated that mothers with depression do not benefit from home visiting interventions to the same extent as their peers without depression. The purpose of this study was to identify predictors of depression course in mothers participating in home visiting over the first 18 months of service. Participants were 220 low income mothers participating in a home visiting program who completed the Beck Depression Inventory-II (BDI-II) at enrollment and 9 and 18 months later. Measures of childhood maltreatment history, social support, and locus of control were also collected at enrollment. Group-based trajectory modeling revealed 3 groups labeled as minimal (63.6%), mild (30.5%), and moderate-severe (5.9%). Although a slight decrease in depressive symptoms was observed over time in the minimal and mild groups, mothers in the moderate-severe group exhibited a large increase from enrollment to 9 months that persisted through 18 months. Membership in the mild and moderate-severe groups was predicted by history of childhood maltreatment, low levels of social support, and an external locus of control. Implications of these findings for home visiting programs are discussed.

Depression is common among new mothers. Research indicates prevalence rates of clinically significant depressive symptoms ranging from 12.0% (Bennett, Einarson, Taddio, Koren, & Einarson, 2004) to 23.1% (Mayberry, Horowitz, & Declercq, 2007) in pregnancy and postpartum. In their summary of epidemiologic studies, O’Hara and Swain (1996) concluded that 13% of women experienced depression during the postpartum period. In a more recent study examining the prevalence of depressive symptoms postpartum, Wisner et al. (2013) found that a positive depression screen (defined as a score of ≥10 on the Edinburgh Postnatal Depression Scale) was associated with a number of factors including lower educational attainment, being single, minority status, and having public insurance. A sizable majority (91.1%) of this sample of mothers with positive depression screens were later diagnosed with a mood disorder. Predictors of postpartum depression or a chronic course of depression include younger age, low levels of social support, childhood maltreatment, and financial stress (Vliegen, Casalin, & Luyten, 2014).

The symptoms of clinically significant depression (e.g., sadness, fatigue, anhedonia) impact multiple domains of maternal functioning, most notably parenting. Depressive symptoms can negatively impact maternal life course and child development. The impact of depression on children is particularly salient in the first years of the child’s life which constitute a sensitive period for biological and psychological development (Goodman et al., 2011). A large body of evidence indicates that, in contrast to nondepressed mothers, depressed mothers are less sensitive to child cues (Field, 2002), have more negative and less positive interactions with their children (Pelaez, Field, Pickens, & Hart, 2008), and use physical discipline more often (Coyl, Roggman, & Newland, 2002). Children of depressed mothers are at greater risk for emotional regulation deficits (Maughan, Cicchetti, Toth, & Rogosch, 2007), insecure attachment (Teti, Gelfand, Messinger, & Isabella, 1995) and diminished cognitive functioning compared to children of nondepressed mothers (Hay, Pawlby, Waters, & Sharp, 2008). Greater severity of depressive symptoms, longer depressive episodes, and chronicity of symptoms are associated with increased impairments in parenting (Lyons-Ruth, Wolfe, & Lyubchik, 2000).

Research has also documented the significant impact of less severe symptoms of depression on maternal psychosocial functioning. These symptoms, characterized by lowered maternal self-esteem (Weinberg et al., 2001) and high levels of parenting stress (Wade, Giallo, & Cooklin, 2012), can also interfere with sensitive and nurturing caregiving. A persistent course of elevated but subclinical levels of maternal depression can impede normal child development. In a 12-year longitudinal study of mothers who completed self-reported measures of depression at 10 time points, Campbell, Morgan-Lopez, Cox, & McLoyd (2009) found that adolescent offspring of mothers who exhibited a persistent course of subclinical depressive symptoms were more likely than their counterparts to exhibit elevated internalizing and externalizing behavior problems at age 15 years.

Home visiting is a widely used prevention approach that seeks to optimize child developmental outcomes and maternal life course by providing new mothers with education and support around child health and development and positive parenting (Adirim & Supplee, 2013). Most of the national program models of home visiting seek to enroll mothers with characteristics such as low income, young maternal age, or being unmarried, which may place them at risk for poor parenting outcomes. These demographically at-risk mothers are also at high risk for developing depression. Studies have found that low income women are at twice the risk of developing clinically elevated symptoms of depression compared to mothers with adequate resources (Holzman et al., 2006). Ammerman et al. (2009) found that 45.3% of first-time mothers participating in a home visiting program had elevated scores on the BDI-II (Beck, Steer, & Brown, 1996) at enrollment, 9 months later, or both time points. Similarly, a study of the prevalence of maternal depression in mothers of children enrolled in Early Head Start (Chazan-Cohen et al., 2007) revealed that 57.2% exceeded the clinical cutoff on the Center for Epidemiologic Studies Depression Scale (CES-D; Ross, Mirowsky, & Huber, 1983). Easterbrooks, Katake, Raskin and Bumgarner (2016) found that 35.5% of first time, adolescent mothers participating in a randomized control trial of a home visiting program were above the clinical cutoff for the CES-D at the start of the study. Participation in the home visiting program was associated with remaining in the nondepressed range, and depressed mothers who reported greater father support were more likely to show improvements over time. High rates of depression are consistently reported in home visiting programs, and are typically attributed to the cumulative risk factors present in the high-risk mothers who enroll in home visiting (Ammerman, Putnam, Bosse, Teeters, & Van Ginkel, 2010).

Depression is a significant clinical issue for home visiting programs. Depressed mothers are difficult to engage in home visits (Stevens, Ammerman, Putnam, Gannon, & Van Ginkel, 2005), require more telephone contact with home visitors (Stevens, Ammerman, Putnam, & Van Ginkel, 2002), and appear to be less motivated (LeCroy & Whitaker, 2005). Furthermore, home visitors report feeling they are inadequately trained to meet the needs of mothers with mental health issues, including depression (Tandon, Parillo, Jenkins, & Duggan, 2005). Maternal depression can limit the effectiveness of home visiting programs to improve outcomes. In contrast to nondepressed mothers in home visiting, depressed mothers are more likely to report more maladaptive parenting beliefs and attitudes (Jacobs & Easterbrooks, 2005) and use physical discipline (Jacobs & Easterbrooks, 2005; Mitchell-Herzfeld, Izzo, Greene, Lee, & Lowenfels, 2005). Further, these mothers are more likely to have child abuse and neglect reports (Easterbrooks et al., 2013) and impaired parent-child interactions (Green, Tarte, Harrison, Nygren, & Sanders, 2014) in comparison to their nondepressed counterparts. To the extent that maternal depression challenges service delivery, poorer outcomes may result.

Although several studies have found high rates of depression in home-visited mothers, these have typically relied on measurement at a single point in time. In addition, these studies solely reported proportions of respondents above or below a predetermined cutoff, thereby overlooking the range of symptom severity. Although commonly used and sometimes theoretically informed and empirically derived, these predetermined cutoffs are open to question and can differ widely between populations (Venkatesh, Zlotnick, Triche, Ware, & Phipps, 2014). Furthermore, relying on measurement at a single point in time fails to capture the continuity or discontinuity of depression. In contrast, multiple measurements over time more clearly elucidate the course of depressive symptoms and delineate the exposure of young children to depression. To our knowledge, no studies have empirically derived trajectories of depressive symptoms in mothers participating in home visiting programs. Given that maternal depression can undermine the effectiveness of home visiting, there is a critical need to identify patterns of depressive symptoms over time as well as identify the demographic and associated clinical features associated with these trajectories in order to guide intervention efforts. Multiple measurements permit documentation of patterns and severity which, in turn, can greatly inform treatment and prevention. Differentiating groupings of mothers by persistence of symptoms and severity allows for separate examination of these two distinct but also overlapping parameters of depression. To date, extant research has not provided a clear distinction between predictors of these features of depression over time.

Because known risk factors for maternal depression (e.g., poverty, younger age, social isolation, and history of childhood maltreatment) are common among home visiting participants, further research is needed to identify specific predictors of depression course and severity in this population (Mayberry et al., 2007). Maltreatment experiences in childhood, as an example, play an etiologic role in subsequent depression in adulthood through disruptions in the physiological and psychological processes essential for emotional and behavioral regulation (Gibb, Chelminski, & Zimmerman, 2007). It is possible that history of childhood maltreatment contributes not only to the occurrence of depression in home-visited mothers but also to a more persistent course of elevated symptoms. Locus of control, reflecting a sense of self-efficacy and belief that one has control over one’s environment and can affect change, has also been linked to depression (Cheng, Cheung, Chio, & Chan, 2013) and has been found to be an important feature of response to home visiting services (Olds & Korfmacher, 1998). Locus of control has also been identified as a mediator between experiences of maltreatment and subsequent emotional and behavioral outcomes (Bolger & Patterson, 2001). Social support is a well-documented protective factor for depression in mothers. Access to people who provide emotional and tangible support is a robust determinant of the occurrence, severity, and length of depressive episodes (Cohen & Wills, 1985). Each of these constructs is theoretically linked and potentially synergistic in their possible impacts on depression. Taken together, understanding how these factors relate to patterns of depression and symptom severity in mothers in home visiting over time can inform development of interventions that may alter trajectories.

The purpose of this study was to empirically identify groups of mothers based upon severity and course of depression over the first 18 months of participation in home visiting. Specifically, 220 first-time mothers were recruited following enrollment in home visiting. Mothers completed the BDI–II at enrollment and at 9 and 18 months later. Group-based trajectory modeling was used to identify discrete groups of mothers, reflecting patterns of depressive symptoms across the three time points and informed by demographic and clinical features at enrollment. These included age, race, history of childhood maltreatment, social support, and personal agency (locus of control). It was hypothesized that distinct groups, reflecting depressive severity and stability over time, would emerge. Greater experiences of maltreatment and lower levels of social support and locus of control were also hypothesized to predict patterns reflecting greater depression severity and persistence.

Method

Sample

Participants consisted of 220 first-time mothers enrolled in Every Child Succeeds (ECS), a regional home visiting program for new mothers and their infants. ECS uses two models of home visiting: Healthy Families America (HFA; Holton & Harding, 2007) and the Nurse-Family Partnership (NFP; Olds, 2010). Mothers meet one of four criteria: low income (<300% of poverty), unmarried, 17 years of age or younger, or inadequate prenatal care. Mothers are enrolled prior to 28 weeks gestation in the NFP program and from 20 weeks through the child reaching 3 months of age in HFA, as per model guidelines. Mothers were referred to home visiting from prenatal clinics, birth hospitals, social service agencies, and community organizations. The majority of mothers enrolled prenatally (81.8%). In terms of home visiting models, 55% were enrolled in HFA while 45% were enrolled in NFP. At the 18-month assessment, mothers were in the program a mean of 360.6 (SD = 206.1) days (11.9 months) and had received an average of 24.5 (SD = 14.8) home visits.

Table 1 presents the demographic characteristics of the sample. Mothers ranged in age from 16 to 42 years of age (M = 21.3, SD = 3.9), and were largely Caucasian (80.5%; African American = 15.5%, Biracial = 3.6%, Asian American = 0.5%). The majority were unmarried (84.1%), reported annual household income primarily below $20,000 (79.3%) and had a high-school degree (51.4%) or less (33.2%). At the 18-month assessment, mean age of the children in the sample was 453.0 days (SD = 83.2) (14.9 months).

Table 1.

Demographic Characteristics of Participants (n = 220)

Variable	M (SD) or N (%)
Mother age (years)	21.3 (3.9)
Mother race
White	177 (80.5)
African American	34 (15.5)
Native American	0 (.0)
Asian American or other Pacific Islander	1 (.5)
Biracial	8 (3.6)
Mother ethnicity
Appalachian	22 (10.0)
Latina	2 (.9)
None	196 (89.1)
Marital status
Single, never married	185 (84.1)
Married	27 (12.3)
Separated	3 (1.4)
Divorced	5 (2.3)
Education (years)	11.9 (1.8)
Income
US$0–9,999	82 (37.3)
US$10,000–19,999	60 (27.3)
US$20,000–29,999	33 (15.0)
US$30,000–39,999	18 (8.2)
>US$40,000	21 (9.5)
Unknown	6 (2.7)
Child age (days)
9-month assessment	172.3 (88.4)
18-month assessment	453.0 (83.2)

Procedure

Mothers were assessed at three time points: enrollment in home visiting and 9 and 18 months later. Home visitors presented newly enrolled mothers with information about the study and contacted study staff with contact information for interested participants. All assessments were completed in participants’ homes by independent research staff. Mothers were paid $25 for the enrollment assessment, $30 for the 9-month assessment, and $35 for the 18-month assessment. All study procedures were approved by the Institutional Review Board at Cincinnati Children’s Hospital Medical Center.

Measures

The BDI-II (Beck et al., 1996) is a widely used self-report measure of depressive symptoms. Respondents endorse each of the BDI-II’s 21 items on a 4-point scale: 0 (not present) to 3 (severe) to assess the severity of depressive symptoms over the past two weeks. Item responses are summed with subsequent scores ranging from 0 to 63. The total score was used in the analyses. Beck et al. also recommend clinical cutoffs with the following categories: minimal depression (0–13), mild depression (14–19), moderate depression (20–28), and severe depression (29–63). These categories were used to develop labels for the derived subgroups reflecting symptom levels and trajectories. The BDI-II is one of the most widely used self-report measures of depression and demonstrates strong reliability and validity (Beck, Steer, & Carbin, 1988).

The Childhood Trauma Questionnaire (CTQ; Bernstein et al., 2003) is a 28-item measure of maltreatment experienced in childhood and adolescence. It is comprised of 28 items that are rated on a 5-point Likert scale reflecting the degree to which each item is true for the respondent: 1 (never true) to 5 (very often true). Responses to the CTQ yields raw scores that can be used to determine categories of exposure (none, low, moderate, severe) for five subscales reflecting physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect. In the current study, the total raw score was used. We elected to use the total score rather than examine each type of maltreatment separately, given the growing evidence that maltreatment contributes broadly to the development of psychopathology (Vachon, Krueger, Rogosch, & Cicchetti, 2015). Cronbach’s alpha for the sample was 0.95. The CTQ has demonstrated excellent psychometric properties in prior research (Scher, Stein, Asmundson, McCreary, & Forde, 2001). For purposes of describing the maltreatment experiences of the sample, mothers reported types of childhood maltreatment that were categorized as low, moderate, or severe as follows: physical abuse = 36.4%, emotional abuse = 42.7%, sexual abuse = 26.8%, physical neglect = 30.5%, and emotional neglect = 35.9%.

The Interpersonal Support Evaluation List (ISEL; Cohen & Hoberman, 1983) is a self-report measure of perceived social support and has been used with both clinical and nonclinical populations. It is comprised of 40 statements assessing the availability of emotional and tangible social supports and uses a 4-point scale: 0 (definitely false) to 3 (definitely true). While the ISEL yields four subscales (Appraisal, Belonging, Tangible, and Self-Esteem), the total score was used in the current study, given our interest in social support as a broad construct. Cronbach’s alpha for the sample was 0.93. The ISEL has been found to have strong psychometric properties (Brookings & Bolton, 1988).

The Rotter Locus of Control Scale, abbreviated (LOC; Rotter, 1966) is a self-report measure of locus of control and is comprised of seven items from Rotter’s LOC. This abridged version was used by Olds and Korfmacher (1998). Respondents use a 4-point Likert scale to indicate how much they agree or disagree with statements asking them to reflect on their perceptions of how much they are in control of events in their lives. The LOC yields a total score reflecting the degree to which respondents see themselves as having control over events. In contrast to the full measure, coding in this variable was such that higher scores indicated greater feeling of control of events. Cronbach’s alpha for the seven item version in this sample was 0.70. The LOC has received extensive psychometric evaluation and has been found to have acceptable reliability and validity properties (Marsh & Richards, 1986).

Data Analytic Strategy

The primary aim of this study was to identify risk factors at enrollment that predict membership in various trajectories of maternal depressive symptoms throughout 18 months of follow-up. Trajectories of maternal depressive symptoms were identified using group-based trajectory modeling (GBTM; Nagin, 1999) via SAS PROC TRAJ (Jones, Nagin, & Roeder, 2001). GBTM is a useful analytic tool when the goal is to characterize individual variability on an outcome of interest, in this case maternal depressive symptoms, by creating subgroup trajectories that represent similar patterns of individual variability across time within a particular population. Through maximum likelihood estimation, GBTM used the observed series of responses collected on an outcome, in this case, BDI-II total scores collected at enrollment, 9-month follow-up, and 18-month follow-up, to generate probabilities of each individual belonging to each subgroup trajectory. These posterior probabilities then allowed for the assignment of each individual to the subgroup where she has the highest likelihood of belonging. Models estimating one through six subgroup trajectories based on a censored normal probability distribution of BDI-II total scores were compared to identify the number of subgroup trajectories that optimally represent the observed data. The model with the optimal number of subgroup trajectories was chosen based on established criteria (Nagin, 2005; Nagin & Odgers, 2010): (a) the Bayesian information criteria, where lower scores and more parsimonious models represent a better overall fit to the observed data, (b) average posterior probabilities ≥ .70 for each subgroup trajectory, (c) odds of correct classification ≥5 for each subgroup trajectory, (d) close agreement between actual (determined by posterior probabilities) and estimated (determined by model estimation) subgroup proportions, (e) subgroup proportions that are sufficiently large (>5%) to have statistical and clinical relevance, and (f) statistical significance of trend parameters (intercept, linear, quadratic) that best represent the shape of each subgroup trajectory. GBTM retains participants with missing data under the assumption that these data are missing at random. However, it is recommended that cases containing less than two points of observation on the outcome of interest be omitted from the analysis (Nagin, 2005). Therefore, any individual case with less than two points of observed data on BDI-II scores was removed from all GBTMs.

Once the optimal number of subgroup trajectories was identified, the set of predictors measured at enrollment into home-visiting were then entered simultaneously into the GBTM to determine whether one or more risk factors predicted membership in one or more trajectory subgroups. These predictors were time-invariant risk factors and were related to membership in trajectory subgroups through a generalized logit function where the resulting log-odds can be exponentiated into odds ratios (ORs). These ORs can then be interpreted as the odds of belonging to a specific trajectory subgroup relative to the reference trajectory subgroup when there is a one-unit change in the level of a particular risk factor. Both log-odds and ORs were reported for each predictor. In GBTM, missing data on individual predictors are dropped from analysis automatically, although in our sample there were no missing data on predictor variables.

Results

Consideration of Covariates and Correlations Among Predictor Variables

Approximately 97% of mothers (n = 213) completed the BDI-II on at least two occasions (92% completed the measure at all three occasions) and were therefore included in GBTMs estimating the optimal number of subgroup trajectories and the shapes of each of those trajectories. In constructing the final model, we considered several covariates (e.g., demographics, home visiting model, number of home visits). None were significantly related to the outcomes or to the predictor variables and were not included in the final model. Table 2 presents the correlations among the three predictor variables. Significant correlations were found in all combinations.

Table 2.

Zero-Order Correlations Among Predictor Variables

Variable	CTQ	ISEL	LOC
CTQ
ISEL	−.36**
LOC	−.16*	.44**

Note. CTQ = Childhood Trauma Questionnaire; ISEL = Interpersonal Support Evaluation List; LOC = Locus of Control (abbreviated).

^*p < .05.

^**p < .001.

Trajectories of Maternal Depressive Symptoms

Preliminary models identifying the optimal number of subgroup trajectories indicated that a 3-group model provided the best fit to the observed BDI-II data (Tables 3 and 4). Figure 1 depicts the 3-group trajectory model and the trends that best represent their shapes over time. Mothers in the lowest depressive symptom subgroup (63.6%) had an average initial value of 9.32 on the BDI-II at enrollment, representing lower levels of severity of depressive symptoms. Guided by the categories recommended by Beck et al. (1996) based on clinical cutoffs, this group was labeled “minimal.” Change over time for this subgroup was best represented by a linear decline in maternal depressive symptoms over the 18-month follow-up. Mothers in the middle subgroup trajectory (30.5%) had an average initial value of 16.43 on the BDI-II at enrollment, indicating that, on average, these mothers were experiencing clinically significant but otherwise a subclinical severity of depressive symptoms consistent with subthreshold depression (Cuijpers, Smit, & van Straten, 2007). This group was labeled “mild.” Their change over time was also best represented by a small linear decline over time. The third and final subgroup trajectory (5.9%) had an initial value of 22.13 on the BDI-II at enrollment, suggesting more severe levels of depressive symptoms at enrollment. Change over time for this subgroup indicated an increase in the severity of depressive symptoms from enrollment to the 9-month follow-up assessment followed by a relatively stable pattern of severity from the 9-month follow-up to the 18-month follow-up. This rate of change was best represented by a significant quadratic trend over time, resulting in the label “moderate-severe.”

Table 3.

Identification of the Optimal Number of Trajectory Groups

		Predicted group proportions
Trajectory groups	BIC	Group 1	Group 2	Group 3	Group 4	Group 5	Group 6
1	−2176.80	1.00
2	−2118.09	.91	.09
3	−2102.85	.59	.34	.06
4	−2110.90	.00a	.59a	.34a	.06a
5	−2109.55	.25a	.38a	.26	.06	.06
6	−2114.91	.25a	.00a	.38a	.26	.06	.06

^aModel estimate of group probability not significantly different from zero (p < .05), indicating poor model fit.

Table 4.

Diagnostics for 3-Group Trajectory Model of Maternal Depressive Symptoms

	Posterior probabilities				Prevalence
Trajectory group	Mean	Range		OCC	Expected	Actual
Minimal	.91	.50	.99	7.03	59.47	56.81
Mild	.87	.54	.99	12.99	34.37	37.09
Moderate-severe	.96	.75	1.00	363.64	6.15	6.10

Note. OCC = odds of correct classification.

Figure 1.

Trajectories of maternal depressive symptoms using the Beck Depression Inventory-II following enrollment in home visiting. The percentages in each group from the total sample are: minimal (63.6%), mild (30.5%), and moderate-severe (5.9%).

Predictors of membership in these three subgroup trajectories were then added to the GBTM to identify risk factors for maternal depressive symptoms at enrollment. The minimal subgroup served as the reference group for this analysis. Results indicated that several risk factors assessed at enrollment predicted membership in the elevated depressive symptoms trajectories over time (Table 5). Specifically, experiencing maltreatment in childhood significantly increased the risk for membership in both the mild (OR = 1.04; 95% CI: 1.00–1.08) and moderate-severe (OR = 1.05; 95% CI: 1.01–1.10) subgroup trajectories for mothers in home visiting. Conversely, increases in self-reported social support served as a protective factor, significantly reducing the risk of being assigned to either the mild (OR = 0.95; 95% CI: 0.90–0.99) or moderate-severe (OR = 0.94; 95% CI: 0.89–0.99) subgroup trajectories. Increased belief in having control over one’s life (LOC) also served as a protective factor, significantly reducing the risk for assignment in either the mild (OR = 0.71; 95% CI: 0.57–0.88) or moderate-severe (OR = 0.62; 95% CI: 0.45–0.84) subgroup trajectories. Table 6 presents the means and standard deviations of the three predictor variables for each of the subgroups.

Table 5.

Predictors of Maternal Depressive Symptoms Trajectories

Trajectory group	Estimate	SE	p	OR	95% CI
Minimal
Mild
Childhood maltreatment history	.04	.02	.041	1.04	1.00	1.08
Social support	−.05	.03	.045	.95	.90	.99
Locus of control	−.34	.11	.002	.71	.57	.88
Moderate-severe
Childhood maltreatment history	.05	.02	.020	1.05	1.01	1.10
Social support	−.06	.03	.030	.94	.89	.99
Locus of control	−.48	−.16	.003	.62	.45	.84

Note. Minimal served as reference group.

Table 6.

Means and Standard Deviations for Each Predictor by Trajectory Group

Predictor	Trajectory group
	Minimal (n = 131)		Mild (n = 69)		Moderate-severe (n = 13)
	Mean	SD	Mean	SD	Mean	SD
CTQ	38.16	16.67	45.06	18.00	55.38	29.26
ISEL	99.30	12.26	89.51	17.85	80.38	23.28
LOC	23.51	2.72	21.54	2.82	20.15	3.44

Note. CTQ = Childhood Trauma Questionnaire; ISEL = Interpersonal Support Evaluation List; LOC = Locus of Control (abbreviated).

Discussion

Findings from the current study identified three trajectories of maternal depressive symptoms characterized by patterns of persistent depressive symptoms at three time points over 18 months. Using group-based trajectory modeling, three distinct groups emerged: minimal, mild, moderate-severe. While results indicated that there were differences in mean BDI-II scores across each of the time points from enrollment to 18 months, mothers’ depressive symptom levels tended to cluster within distinct and clinically narrow ranges. Thus, it is likely that those mothers who enter home visiting programs with higher depressive symptoms will continue to display those symptoms over time.

This is one of the few studies examining predictors of depressive symptoms over time in home-visited mothers. Although 5.9% of home-visited mothers were in the moderate-severe range of depressive symptoms over time, an additional 30.5% experienced significant but subthreshold levels of depression that would be expected to impair maternal functioning and parenting. Subthreshold depression (defined as having significant depressive symptoms that approach but do not meet full diagnostic criteria for major depressive disorder [MDD]) is associated with functional impairment, increased risk for developing MDD, and increased health care costs (Cuijpers et al., 2007; Rucci et al., 2003). Our findings are consistent with previous research that has found similar patterns of maternal depressive symptoms over time in mothers recruited from specialty clinics or the general population. Luoma, Korhonen, Salmelin, Helminen, & Tamminen (2015) identified four trajectory groups in 329 mothers recruited from maternity centers who completed self-report depression measures at six time points over 18 years. They found that 30% of mothers were in the higher ranges of severity characterized by both stable and intermittent patterns. Matijasevich et al. (2015) found five groups emerging from 4,231 mothers followed from birth through six years after delivery. Although 5.9% of mothers exhibited a pattern labeled high-chronic, an additional 18.9% fell into two groups of significant levels of depression that were increasing or decreasing over time. In a sample of 1,807 mothers recruited from maternity wards, van der Waerden, Galéra, Saurel-Cubizolles, Sutter-Dallay, and Melchior (2015) identified groups of mothers in the severe range (5.0%) and in a range reflecting persistent but intermediate levels of depression (25.2%). Interpreting results from these and other studies is complicated by their use of different depression measures, populations sampled, frequency and timing of assessments, length of follow-up, and methods used to identify trajectory groups. However, synthesizing these findings, including results from the current study, indicate that about one third of mothers exhibit clinically significant levels reflecting mild to severe levels of depressive symptoms over time. This group of mothers is likely to have onset of symptoms during pregnancy, greater comorbidity, and poorer outcomes (Postpartum Depression: Action Toward Causes and Treatment (PACT) Consortium, 2015). Our findings, in conjunction with prior research, underscore the importance of considering both patterns and symptoms of depression as each reflect important features of its manifestation over time. Intervention is needed to prevent or treat these mothers, preferably early in parenthood to optimize potential beneficial impacts on children.

As hypothesized, trajectory groups were differentiated by maternal characteristics that are associated with the occurrence, severity, and persistence of depression. Maltreatment in childhood is a consistent associated feature of depression that has emerged in multiple studies as an important contributor. In this sample, in which the majority of mothers experienced abuse and neglect in childhood, maltreatment was related to an increased likelihood of being in the mild and moderate-severe trajectories. In contrast, less severe levels of maltreatment were associated with a persistent course of minimal depressive symptoms. The experience of abuse and neglect in childhood can have profound negative effects on emotional and behavioral regulation, which, in turn, are manifested as mood disturbances in adulthood (Maniglio, 2010). Locus of control, which also emerged as a strong predictor of mild and moderate-severe levels of depression over time, may partially mediate the relationship between maltreatment history and depression (Bolger & Patterson, 2001). To the extent that history of childhood maltreatment contributes to a sense of lowered personal agency, these two variables may combine to both worsen and prolong depressive symptoms. In the high-risk population of mothers enrolled in home visiting, these experiences can lead to stable levels of depressive symptoms across the initial 18 months of service. This, in turn, puts offspring at risk for poor social, emotional, and behavioral outcomes. An impaired caregiver contributes to “toxic stress” with resulting disruptions in developmental processes essential for healthy development (Shonkoff, Boyce, & McEwen, 2009).

Social support was also a significant determinant of group membership. Mothers reporting lower levels of tangible and emotional support were more likely to have trajectory levels reflecting mild and moderate-severe depressive symptoms. Those with increased levels of social support, on the contrary, were more likely to be in the minimal trajectory group. Social support is one of the most robust protective factors for depression in new mothers. Even among low income new mothers, who are at elevated risk for small social networks, the presence of supportive individuals in their lives can buffer the effects of stress and reduce risk of subsequent depression. Easterbrooks et al. (2016) attributed the reduction in depressive symptoms in their sample of depressed adolescents in home visiting to the support provided by fathers. Although not the primary focus of home visiting programs, increasing the size and support of maternal social networks is a secondary outcome and one that is seen as important for facilitating a successful life course and self-sufficiency. Enhancing home visiting curricula to more explicitly assist mothers in building strong social networks has the potential to decrease depression and improve home visiting outcomes more broadly.

There are several implications of these results for home visiting programs. There has been a sizable increase in the number of home visiting programs that routinely screen mothers for depression (Segre & Taylor, 2014). With 36.4% of mothers having trajectories reflecting mild or moderate-severe levels of depressive symptoms over the first 18 months of service this practice is clearly warranted, although patterns were slightly declining between time points in two of the three groups (minimal and mild). For these mothers, the range of scores between time points is restricted and overall the patterns indicate a stable persistence of depressive symptoms over time. For the moderate-severe group, however, there is a marked increase between enrollment and the 9-month assessment, levelling off but remaining elevated through 18 months. Because most mothers enrolled prenatally, findings indicate that high scores in the moderate-severe group prior to the child’s birth are associated with an increase in severity during the first year postpartum. It is possible that the stress of raising a child in the first year of life is particularly overwhelming for those mothers already experiencing clinically elevated levels of depressive symptoms during pregnancy, thereby contributing to subsequent symptom exacerbation and persistently severe depression. Systematic screening with multiple administrations after 18 months would be needed to determine if this pattern is maintained or remits.

Research suggests that home visitors typically feel insufficiently prepared to work effectively with depressed mothers (Tandon et al., 2005) and additional training in this area might increase their comfort and confidence. Recent informational resources, such as Depression in Mothers: More than the Blues—A Toolkit for Family Service Providers (Substance Abuse & Mental Health Services Administration, 2014), are designed to provide home visitors (and other professionals working with low income, depressed mothers) with information and strategies to support mothers struggling with depression.

Both the moderate-severe group and a sizable proportion of the mild group are at significant risk for MDD. Depressed mothers in home visiting face formidable barriers to accessing mental health care in the community, with most studies suggesting that less than 20% are able to obtain any form of mental health treatment and an even smaller percentage receive evidence-based treatments for perinatal depression (Ammerman et al., 2010). Interventions to prevent the emergence of depression in home-visited mothers (Perry, Tandon, Edwards, & Mendelson, 2014) and to provide home-based treatment for depressed mothers delivered concurrently with home visiting (Ammerman, Putnam, Teeters, & Van Ginkel, 2014) provide empirically demonstrated options for serving this high-risk population.

The study had several strengths. First, the sample was large compared to other research in home visiting. Second, the home-visiting program and population from which the sample was drawn is typical of such programs, and used two models of home visiting, thus supporting generalizability of findings. Much of the research in home visiting has been model-specific. Examination of common issues that are experienced in all models holds promise for the accelerated development and rapid dissemination of new community-based approaches for supporting all mothers and children who participate in home visiting. However, variability between models in demonstrated outcomes and variation in implementation also warrant caution in generalizing our findings to other programs. Replication in other programs and populations is recommended. Third, depression was measured at three time points over an 18 month period. The longitudinal design and multiple administrations permitted examination of course and continuity/discontinuity of depression across time in contrast to a single point in time or pre-post administration that is most commonly reported in the home visiting literature. Fourth, the raw score of the BDI-II was used rather than clinical cutoffs, thereby allowing more fine-grained reflection of severity of depressive symptoms. Fifth, the study focused on the first year of the child’s life, a critical period in which the deleterious impact of maternal depression is likely to be most acute (Bagner, Pettit, Lewinsohn, & Seeley, 2010).

The study also had several limitations that warrant caution in interpreting findings. First, all measures were self-report which may have led to shared method variance. Second, some have suggested that the BDI-II is vulnerable to inflated scores in postpartum mothers due to the overlap between some symptoms of depression and common physical experiences of new mothers. Research indicates that this concern is modest in that the BDI-II and measures thought to be less susceptible to this symptom overlap are highly correlated (Jia & Schoppe-Sullivan, 2011) and, as a result, the BDI-II remains a widely used measure with this population. Third, although three administrations of the BDI-II over 18 months is an advance over prior research in this area, more frequent administrations over a longer period of time would allow for identification of more differentiated subgroups of trajectories which could be additionally informative. However, implications of frequent administrations of depression screens to validity of reporting have not been examined, and additional research in this area is needed.

Findings from this study should inform future research. There is a need to replicate these findings in a larger sample that is followed over a longer period of time. The impact of different trajectories on longer-term child and mother outcomes is needed. The degree to which depression is moderated by participation in home visiting would allow delineation of where home-visiting strategies need to be strengthened or changed to meet the needs of depressed mothers. Depressed mothers may benefit from home visitors who receive increased and enhanced training in working with maternal mental health problems. Yet, the professional backgrounds of home visitors will also limit the extent to which clinical support of mothers is possible. Furthermore, home visiting curricula are large and require sufficient time to implement effectively, and the possible burden associated with asking home visitors to do more with depressed mothers needs to be considered. Understanding what home visitors can, cannot, and should do to support depressed mothers is a priority for future research and program development. Additionally, documenting the differential impacts of enhanced training, mental health consultation, supervisor support, and prevention and treatment approaches is needed. Indeed, there are empirically supported effective prevention (Tandon, Leis, Mendelson, Perry, & Kemp, 2014) and treatment options (Ammerman et al., 2013) for depressed mothers in home visiting, although their long-term impacts are unknown. Finally, self-reported screens are useful in identifying clinically elevated depressive symptoms; examination of how these measures relate to actual diagnosis of major depressive disorder is an important area for future inquiry.