Caregiver Strain, Social Support and Mental Health Service Use Among Urban African American Mothers

Abstract

Caregiver strain and social support have been identified as both facilitators and deterrents to parental mental health service use on behalf of their children. This study focused on the relationship between caregiver strain, social support, and mental health service use among African American mothers of children at-risk or meeting criteria for a disruptive behavioral disorder and living in urban communities of concentrated poverty. Mothers (n = 89), participating in a five-year NIMH funded study of school-based community mental health services, completed measures at baseline of caregiver strain and both perceived and received social support. Service use was calculated as the sum total of services (sessions) received. Associations between caregiver strain and service use were examined, and perceived and received social support were explored as potential moderators. Baseline covariates included child’s age, gender, symptom severity, and maternal employment status. Findings highlighted child symptom severity as the strongest predictor of caregiver strain and perceived social support as moderating the association between caregiver strain and service use. Mothers were more likely to utilize services when experiencing relatively high levels of perceived support or high caregiver strain but not both, highlighting the importance of their interrelationship. Received support did not moderate the association between strain and service use. In addition, mothers utilized services more often for sons than daughters and when unemployed. Implications for research and practice are discussed.

Racial and ethnic disparities in mental health service access and use are long-standing (Alegría, Vallas, & Pumariega, 2010; Kataoka, Zhang & Wells, 2002). Many individual and societal factors explain lower rates of initiation and early termination of treatment. Among them, social support and caregiver strain have been identified as both facilitators and deterrents to parental service initiation and ongoing mental health service (MHS) use for children across racial/ethnic groups (Angold, Messer, Stang, Farmer, Costello, & Burns 1998; Pescosolido, Wright, Alegría, & Vera, 1998; Shin & Brown, 2009). The current study sought to advance a more nuanced understanding of the complex associations among caregiver strain, social support, and MHS use among low income African American families already receiving services for their children meeting criteria for a Disruptive Behavior Disorder (DBD). Figure 1 synthesizes and illustrates the complex relationships among caregiver strain, social support and MHS use identified in prior research informing the current inquiry and described in the following sections.

Figure 1.

Visual Framework Synthesizing Prior Research Informing Current Study

Social Support

Social support (hereinafter: support) is linked to reduced risk for mental and physical health problems, underscoring the importance of examining individuals’ social circumstances and connectedness to better understand treatment participation and adherence (e.g., Reblin & Uchino, 2008; Shim, Ye, Baltrus, Daniels, & Rust, 2012). However, Reblin and Uchino (2008) note that although the link between support and improved health is strong and robust, a more comprehensive understanding of the context and mechanisms of these associations remains limited. In fact, support is a complex construct that encompasses various types of assistance available to individuals and groups through social ties and interactions in their social network (Barrera, Sandler, & Ramsay, 1981; Harber, Cohen, Lucas & Baltes, 2007). Perhaps reflecting this diversity, support has been operationalized by the structural aspects of social embeddedness or connections with others (e.g., size of network, density, dispersion), and by its function as instrumental (material, tangible) or emotional (affirmational, affective) assistance across diverse samples (e.g., Barrera, 1986; Ceballo & McLoyd, 2002; McCabe, Yeh, Garland, & Hough, 2003; Nurullah, 2012; Williams, Barclay & Schmied, 2004).

Several decades of literature document the importance and strength of extended family, fictive-kin (i.e., very close friends considered family), and friendship networks within African American communities (e.g., Boyd-Franklin, 2003; Jarrett, Jefferson, & Kelly, 2010; Snowden, 1998). Toward the end of the millennium, Chatters and colleagues, using data from the National Survey of Black Americans data set (n = 1,322), found that nine out of 10 African American adults reported seeking assistance from their support network when experiencing a serious personal problem (Chatters et al., 1989). Qualitative research highlights that strong support networks in communities experiencing high rates of poverty buffer African American families from neighborhood adversities (Jarrett et al., 2010). In addition, social support networks in low income, urban African American families have been found to be women-centered and to include the pooling of resources and the exchange of tangible (e.g., food and clothing) and non-tangible (e.g., ride to the doctor) supports to stretch limited assets and cope with the stressors associated with living in marginalized and socio-disadvantaged communities (Jarrett et al., 2010). Extended family support also has been found to serve a protective role for low income urban, African American mothers and their children by promoting their psychological adjustment through the provision of social and emotional support (Taylor, Seaton, & Dominguez, 2008). However, studies also document that support networks available to African American families vary in how helpful they are to members (Jarrett & Burton, 1999; MacDonald & Armstrong, 2001; Roschelle, 1997). For example, lower-income African American parents report smaller social networks than higher-income African American and European American parents (Bussing et al., 2003; Chatters et al., 1989). Low income African American mothers also report more aggravation from their support networks than European American and Latinx mothers (Miller-Loncar, Erwin, Landry, Smith, and Swank, 1998).

Perceived and received support.

The current study focuses on received and perceived support. Perceived support encompasses subjective, cognitive appraisals of support, including satisfaction and perceptions of support availability, and is often described by its perceived function – for example, emotional and/or instrumental (aid) support - and the social network that provides it (e.g., Melrose, Brown, & Wood, 2015; Norbeck, Lindsey, & Carrieri, 1983). Received support typically includes specific supportive actions and quantity of support received (e.g., Barrera et al., 1981; Melrose, Brown, & Wood, 2015). Perceived and received support have been weakly to moderately correlated and thus conceptualized as distinct constructs (Harber et al., 2007; Melrose et al., 2015). Furthermore, while perceived support is consistently linked to positive health outcomes, the relationship between received support and positive health outcomes is mixed (e.g., Nurullah, 2012; Reinhardt, Boener, & Horowitz, 2006). This seemingly dichotomous relationship (association/no association) with positive health outcomes may contribute to a greater focus on perceived support than other types of support (e.g., Reinhardt et al., 2006). Others note that received support is often measured as the amount of support received without considering whether the support was experienced as needed by the recipient (e.g., Melrose et al., 2015) which may explain low correlations between received and perceived support. Melrose et al. (2015) found, in a predominantly European American sample, that when received support was identified as needed by the recipient, the association between perceived and received support doubled on average. Hence, perceived and received support need to be understood within the context in which they occur (Harber et al., 2007).

Support has been associated with help-seeking from informal and formal helpers among African American parents (e.g., Harrison, McKay, & Bannon 2004). However, social network characteristics vary by socioeconomic status and race/ethnicity (e.g., Bussing et al., 2003) and the positive influence of support on parental stress and parenting appears to vary by contextual factors. For example, neighborhood quality (maternal perception of quality, crime and poverty rates) has been found to moderate the relationship between support - instrumental and emotional - and parenting (e.g., warmth/nurturing and punitive discipline) for African Americans families living in neighborhoods experiencing high rates of poverty and crime (Ceballo & McLoyd, 2002). This suggests that the positive effect of social support on parenting may be lower for mothers living in neighborhoods experiencing high rates of poverty and crime. In another study, African American mothers belonging to economically diverse kinship or familial networks reported receiving more assistance from their network than those belonging to networks in which all members had limited economic resources (Jarrett & Burton, 1999).

Social support and service use.

In the context of formal help seeking, support networks can hinder or promote use of services across ethnic/racial groups (Harrison et al., 2004; Martinez & Lau, 2011; Pescosolido et al., 1998). Support is also postulated to function as deterrent or substitute for service engagement due to its positive influence on mental health status across diverse samples (Martinez & Lau, 2011; Pescosolido et al., 1998). Furthermore, different characteristics and types of social support vary in their association to service engagement. For example, in a diverse sample of youths (27.5% African American, 53.6% European American, 19% Latinx; 4–14 years old), perceived size of support networks did not predict MHS use for parents without a high school education; however, low perception of support was linked to higher MHS use regardless of parental educational level (Martinez & Lau, 2011). Research also points to differing influences of type and source of support on formal help-seeking. For example, in one study of European (67%) and African American (33%) parents of children with a DBD, perceived high instrumental support was associated with lower MHS use (Bussing et al., 2003). In another study, kinship or familial support was associated with higher parental MHS use in a predominantly African American (79%) sample (Harrison et al., 2004). Hence, the strength and direction of association between social support and MHS use seem to depend on the type and quantity of support, contextual variables, and recipient and social network characteristics (e.g., poverty, socio-economic status, race/ethnicity).

Caregiver Strain

Caregiver strain (hereinafter: strain) is defined by the experience of burden and responsibility, worry and guilt, and disruptions in routines and relationships related to a child’s significant behavioral difficulties (Brannan, Heflinger, & Bickman, 1997). Hence, severity of child symptoms is the strongest predictor of strain for African American and European American parents (Brannan & Heflinger, 2006). Relatedly, strain is positively associated with a DBD diagnosis, particularly for male children, controlling for child and family demographics and internalizing disorders (Tsai, Yeh, & Slymen, 2015). Strain and children’s emotional and behavioral difficulties are also negatively associated with maternal employment in diverse samples (Brannan, Brennan, Sellmaier, & Rosenzweig, 2018; Richard, Gaskin, Alexandre, Burke, & Younis, 2014), reflecting competing demands for mothers’ time and attention. Notably, African American and Latinx mothers parenting a child with emotional and behavioral difficulties are likely to work less and miss more workdays than European American mothers (Brannan et al., 2018). However, African American parents also have been found to report less strain due to children’s behavioral difficulties related to relationship with family and neighbors, finances and the police than European American families (Brannan & Heflinger, 2006).

Caregiver strain and service use.

Across ethnic and racial groups, strain, particularly in reaction to their children’s psychopathology, prompts many parents to seek services (Angold et al., 1998; Brannan & Heflinger, 2006; Shin & Brown, 2009). Longitudinal research within a predominantly rural European American sample found that 89% of children with parents reporting strain received services compared to 13% of children whose parents reported no strain (Farmer et al., 1997). However, counter-intuitively, high levels of parenting stress also are associated with lower use of MHS, premature termination of treatment and greater child symptom severity in community mental health settings for low income African American and European American parents ( Pellerin, Costa, Weems, & Dalton, 2010). This suggests that high levels of stress may hinder parents’ MHS use. Nonetheless, qualitative research suggests strain influences when and how African American and European American parents seek help from both informal social networks and formal services (Mayberry & Heflinger, 2013). Notably, the confluence of low perceived instrumental support and high strain are strong predictors of families’ service utilization by African American and European American parents (Bussing et al., 2003).

Social Support and Caregiver Strain

Findings from the Patterns of Youth Mental Health Care in Public Service Systems, a longitudinal study of service need and use among children with emotional and behavioral difficulties indicate that although high rates of perceived support are associated with low strain, the endorsement of strain and the relationship between support and strain differs by race/ethnicity (e.g., MacCabe et al., 2003; Shin & Brown, 2009). For instance, Shin and Brown (2009) found that African Americans parents reported lower strain, while McCabe et al., (2003) found they perceived less support from their network than European American parents. In addition, African American parents’ lower rates of strain became more pronounced after controlling for perceived support (McCabe et al., 2003), indicating support did not explain their reported experience of lower strain. However, others have found that African American parents experiencing strain report greater levels of perceived emotional and instrumental support, and more contact and use of their smaller networks than European Americans (Bussing et al., 2003).

Research examining low self-reported rates of strain among African American caregivers compared to their European American counterparts suggest they bring unique group experiences and coping skills (e.g., McCabe et al, 2003; Shin & Brown 2009) as a result of and in response to historical oppression and adversity, ongoing social injustice, prejudice and distinct sub-group risks. Scholars point to an interdependent and complex set of race-related, gender-related and generic stress experiences, coalescing to influence experiences of overall stress for African American women (e.g., Beauboeuf-Lafontant, 2009; Watson & Hunter, 2015; Woods-Giscombé & Lobel, 2008). This research also points to pressures for African American women to be strong and independent, and to suppress emotion and vulnerability (Beauboeuf-Lafontant, 2009; Woods-Giscombé, 2010). Moreover, research highlights parenting that includes gendered socialization and fear of negative societal perceptions, stereotypes and risks for males, with parents perceiving African American boys as more vulnerable and girls more capable resulting in lower behavioral and academic expectations for sons (Rowley, Ross, Lozada, Williams, Gale & Kurtz-Costes, 2014; Varner & Mandara, 2013). This fear may be exacerbated by more frequent school disciplinary referrals (Kaufman et al., 2010) and services for externalizing problems among African American boys compared to children identifying with other racial/ethnic groups (Gudiño, Lau, Yeh, McCabe, & Hough, 2008). Importantly, research also points to discordance between teacher and parent interpretations of African American children’s behavior, as early as preschool; specifically, teachers are more likely than parents to report behaviors as problematic and aggressive, suggesting that parental fear of negative societal perceptions is activated early in their children’s educational experience (Munzer, et al., 2018).

Current Study

Research on the interrelationship of strain, support, and use of services by low income African American parents on behalf of their children points to some equivocal and counterintuitive findings and underscores the importance of considering context. Extant research with low income African American parents frames lower reported rates of strain within a personal social context composed of smaller support networks (e.g., Shin & Brown, 2009) that are leveraged at greater rates than in European American samples (Bussing et al., 2003). Furthermore, this appears to be experienced within a larger services context characterized by higher rates of unmet mental health needs (e.g., Alegría et al., 2010) and treatment dropouts (De Haan et al., 2018), reflecting perhaps distrust and doubt about the helpfulness of formal treatment seeking (Richardson, 2001). In particular, research primarily has focused on barriers to accessing and engaging in services, patterns of service initiation and help-seeking (e.g., Alegría et al., 2010; Godoy, Mian, Eisehower, & Carter, 2014), attendance (Staudt, 2007), and premature treatment termination (e.g., De Haan et al., 2018). Notably, the literature cited above included systematic reviews and meta-analyses – which inconsistently report ethnic/racial composition in study samples – and individual studies that included diverse samples inclusive of African Americans, and a smaller set of studies with only African American samples, highlighting the limited research focused on the unique experiences of this population (e.g., Harber, et al., 2007). The current study contributes to the scholarship on strain and support and their relationship with MHS use in low-income African American families already engaged in services for their children.

The current study uses data from the Caregiver and Social Support (CareSS) project. CareSS was part of a larger five-year NIMH-funded grant that examined a school-based community MHS model for elementary students at-risk and referred for a DBD (Atkins et al., 2015). CareSS builds on prior research by exploring associations among perceived and received social support, strain and MHS use for African American mothers receiving services on behalf of their children. Three research questions and hypotheses guided this study: (1) What is the relationship between strain and maternal MHS use on behalf of their children? (Hypothesis 1: Baseline caregiver strain will positively predict MHS use), (2) How are different types of baseline social support (perceived and received) associated with MHS use? (Hypothesis 2: Baseline perceived and received social support will differentially predict MHS use), and (3) How does social support (perceived and received) moderate the association between strain and service use? (Hypothesis 3: Baseline social support will moderate the association between caregiver strain and MHS use).

Methods

University and school district institutional review board approvals were obtained prior to initiating study procedures. Informed consent was obtained from all individual participants.

Research Design

The NIMH-funded grant was conducted in a large Midwestern city, with four community mental health agencies and seven public elementary schools identified by school district public reports as serving predominantly low income (98%) African American (97%) families living in segregated communities (Atkins et al., 2015). Schools were randomly assigned to experimental (n = 4) or comparison conditions (n = 3). Families in the experimental condition received classroom, school, and home services provided by agency treatment teams composed of mental health professionals and parent advocates. Treatment teams were majority female (76%) and African American (64.7%, M = 5.7 years providing services, SD = 3.96, range 2–16). Families in the comparison condition received MHS-as-usual. Participating schools and agencies did not have preexisting relationships. The current study examines relationships between baseline strain and support, and MHS use, for African American mothers in the treatment condition.

Procedures

A three-step multiple gating recruitment procedure to protect families’ confidentiality was followed. Teachers completed the Systematic Screening for Behavior Disorders (SSBD; Walker & Severson, 1990; e.g., not following classroom rules, displaying aggression) for students about whom they had behavioral concerns. School mental health teams reviewed teacher-completed SSBD forms to identify and contact potentially eligible families about the study (for further details see Atkins et al., 2015). With families’ permission, research staff received contact information for families that expressed interest in learning more about available services and study procedures. Families that declined study participation but expressed interest in services were referred to nonparticipating community mental health agencies. Consented families completed a packet of surveys (read aloud by research staff) at baseline and five additional time points (beginning and end of each school year) over a three-year period and in locations identified as most convenient by parents (primarily in schools). Eligibility was determined by parent or teacher report on the DBD Rating Scale indicating elevated disruptive behaviors (Pelham, Gnagy, Greenslade, & Milich, 1992) characteristic of one or more DBDs: 171 children and families were eligible and consented to participate in the parent study (n = 104 experimental, n = 67 comparison). See Child Symptom Severity below for more details on the DBD ratings by mothers in the CareSS sample.

Participants

The present study focused on the experiences of 89 African American mothers randomized to the experimental condition (see Table 1). Fifteen participants from the experimental condition were excluded from the present analyses: seven that did not self-identify as African American or did not respond to the query, six that did not complete a baseline DBD rating, the only father in the sample and an outlier in MHS use (range of MHS use was 0–63; outlier: 120, 3 times greater than the interquartile range). On average, mothers and their children participated in the study for approximately 1.5 years (M = 17.88 months, SD = 8.1 months, range = 4 to 29 months).

Table 1.

Sample Demographic Characteristics and Key Variables by Maternal and Teacher Endorsement of DBD Criteria

Characteristics and Variables	Only Maternal DBD Endorsement	Only Teacher DBD Endorsement	Maternal and Teacher DBD Endorsement	No DBD Endorsement**	Total Sample n (%)
n	22	19	44	4	89
Baseline Education n (%)
GED	2 (9.1)	3 (15.8)	3 (6.8)	---	8 (9%)
High School	9 (40.9)	9 (47.4)	23 (52.3)	4	45 (50.9)
Community College	1 (4.5)	---	---	---	1 (1.1)
Some college	5 (22.7)	5 (26.3)	9 (20.5)	---	19 (21.3)
Bachelor’s degree	2 (9.1)	---	1 (2.3)	---	3 (3.4)
Graduate Degree	---	---	1 (2.3)	---	1 (1.1)
Other***	3 (13.6)	2 (10.5)	15.9	---	12 (13.5)
Baseline Marital status n (%)
Single	20 (90.9)	12 (63.2)	35 (79.5)	4 (100)	71 (79.8)
Married/Domestic Partner	2 (9.1)	4 (21.1)	4 (9.1)	---	10 (11.2)
Widowed	---	---	2 (4.5)	---	2 (2.2)
Separated/Divorced	---	3 (15.8)	3 (6.8)	---	6 (6.7)
Baseline Household Income n (%)
$0,000–$20,000	14 (63.6)	12 (63.2)	26 (60.5)	3+(75)	74 (85.0%)
$21,000–$30,000	5 (22.7)	7 (36.8)	7 (15.9)	---	8 (9.2)
$31,000–$40,000	1 (4.5)	---	7 (15.9)	---	2 (2.3)
$41,000–$50,000	1 (4.5)	---	2 (4.5)	---	2 (2.3)
Over $50,000	---	---	1 (2.3)	---	1 (1)
Baseline People in household (including mother)	5.14 (2.00)	4.83 (1.82)	5.21 (2.66)	6.67 (2.08)	5.16 (2.32)
Baseline people in social support network ++	7.68 (5.00)	7.39 (6.26)	5.28 (3.36)	8.5 (6.45)	6.47 (4.72)
Baseline Employment n (%)
Employed	14 (63.6)	12 (63.2)	23 (52.3)	2 (50)	51 (57.3)
Unemployed	8 (36.4)	7 (36.8)	21 (47.7)	2(50)	38 (42.7)
Baseline Age n (SD)
Children	7.54 (2.02)	7.31 (2.0)	7.45(1.80)	7.00 (1.54)	7.42 (1.8)
Mother	31.32 (7.64)	34.95 (7.59)	31.86 (8.22)	29.33 (6.50)	32 (7.92)
Child Gender n (%)
Male	20 (90.9)	13 (68.4)	28 (63.6)	3 (75)	64 (71.9)
Female	2 (9/1%)	6 (31.6)	16 (36.4)	1(25)	25 (28.1)
Baseline Maternal DBD Endorsement n (%)*
No DBD endorsement	0	19 (100)	0	4 (100)	23 (25.84)
Endorsement of 1 DBD	8 (36.4)	0	11 (25)	0	19 (21.35)
Endorsement of 2+ DBDs	14 (63.6)	0	33 (75)	0	47 (52.81)
Baseline Caregiver Strain n (SD)	2.30 (.78)	1.71 (.65)	2.38(.76)	1.43 (.54)	2.17 (.79)
Baseline Social Support n (SD)
Perceived Functional Support	149.13 (108.55)	148.11 (132.14)	102.76 (72.88)	168.25 (111.04)	127.76 (99.96)
Perceived Network Support	78.85 (53.54)	76.72 (64.54)	55.51 (35. 60)	87.25 (62. 70)	67.13 (49.12)
Received Support	107.28 (38.83)	101.94 (36.50)	115.64 (39.34)	119.00 (16.55)	110.73 (37.99)
Services n (SD)
Total Months Could have received services	20.49 (8.16)	18.36 (8.99)	16.35 (7.63)	18.14 (8.198)	17.88 (8.1)
Total number of services received	11.00 (14.53)	13.26 (18.46)	11.66 (13.59)	18.50 (14.34)	12.15 (14.83)

^*DBD endorsement based on mother ratings only.

^**Mothers completed the DBD Rating scale but did not endorse a DBD; teachers did not complete a baseline DBDRS; children were retained in the parent study (and hence eligible for the present study) because of risk indicated on the teacher’s initial referral (SSBD) screener

^***Alternative to secondary school or non-high school-completion

⁺Missing data for 1 parent

⁺⁺Number of people identified by mothers in their social support network is included in the Perceived Network Support score

Mothers (Age: M = 32 years, SD = 7.92 years, range = 20 to 53 years), reported being part of a household composed of an average 5.6 members (SD = 2.32), the majority of which were identified as single-parent households (79.8%). The majority of mothers (85%) reported an annual income below the federal poverty level ($0 to $20,000); the other 15% of mothers, based on family size, were low income (Wight & Chao, 2009). More than half of mothers (59.6%) reported having a GED/high school diploma while 22.4% reported some college level education, 13.5% reported high school alternative/non-high school completion, and 3.4% indicated having earned a college degree. Thirty-eight (42.7%) were unemployed at baseline. Most children were male (71.9%) and African American (98.9%) ranging from 5 to 11 years old (M = 7.42) and attending Kindergarten to Fourth grade (Kindergarten to 2^nd Grade: n = 51, 57.3%; 3^rd to 4^th grade: n = 38, 42.7%) at baseline.

Measures

Perceived social support.

Mothers completed the Norbeck Social Support Questionnaire at baseline (NSSQ; Norbeck et al., 1983). The NSSQ examines multiple components of perceived social support. Two of its subscales were used in this study, perceived network support and perceived functional support. Perceived network support (three items) assesses the perceived structural framework from which perceived functional support is obtained. Perceived network support is the sum of the total number of members (up to 24) mothers identified in their support network (size of network), frequency of contact with each member on a 5-point scale (1= once a year or less to 5 = daily), and longevity of the relationship with each member on a 5 point-scale (1 = less than 6 months to 5 = more than 5 years). Perceived functional support (six items) assesses perceptions of emotional and aid support from members identified in their social support network (via the perceived network support). Perceived functional support is the sum of perceived affective/emotional (n = 4 items; e.g., How much does this person make you feel liked or loved?) and instrumental/tangible support (n = 2 items; e.g., If you needed to borrow $10, a ride to the doctor, or some other immediate help, how much could this person usually help?) on a 5-point scale (0 = not at all to 4 = a great deal) items. Scores were first anchored on respondents’ social network members and then aggregated. Summary scores, rather than average scores, were used based on developer’s recommendations (Norbeck, 1995) in order to measure quality and quantity in perceived social support. Baseline alpha coefficients for the NSSQ in the current sample were high (.98 to .99).

Received social support.

Mothers completed the Inventory of Socially Supportive Behaviors (ISSB; Barrera, 1986; Barrera, Sandler, & Ramsay, 1981) at baseline. The ISSB includes 40 items rated on a 5-point scale (1= not at all to 5 = about every day) to assess how often individuals received social support or assistance in the past month in the form of emotional, tangible (e.g., support mobilization and aid provision), and guidance support (e.g., was right there with you (physically) in a stressful situation, loaned you over $25, pitched in to help you do something that needed to be done). Respondents reported on support received from their entire support network as a group and did not report information about individual members in their network. Hence, due to its focus on emotional support and aid, it aligns conceptually with the perceived functional support subscale of the NSSQ. The Total ISSB total summary score was used, and baseline reliability was high (α = .96).

Caregiver strain.

The Caregiver Strain Questionnaire is a 21-item measure (1 = not at all to 5 = very much) for parents of children with emotional and behavioral disorders (Brannan et al., 1997; Brannan et al., 2018). The baseline summary score used herein estimates how much strain was experienced by mothers in the last six months across six areas: economic burden, impact on family relations, impact on psychological adjustment of family members, stigma, anger, and worry/guilt (e.g., Your missing work or neglecting other duties because of your child’s emotional or behavioral problem; Your child getting into trouble with the neighbors, the school, the community, or law enforcement; how worried did you feel about your child’s future). Baseline reliability was high (α = .92).

Service use.

The frequency of maternal MHS use on behalf of their children was calculated as the sum of the total number of sessions received at the end of each time point. The sum total of services (sessions) received across all six time points was used for all analyses. Services available to families participating in the current study (based on state Medicaid service definition and reimbursement guide for community services) represented a range of therapeutic services, including the parenting intervention being tested in the larger study (i.e., parent training to support children’s learning at home and improve parent-teacher communication), as well as case management, evaluation/assessment, community support, and crisis assistance.

Child symptom severity.

Parents and /or teachers completed the 45-item Disruptive Behavior Disorder (DBD) Rating Scale (Pelham, et al., 1992; 0 = not at all to 3 = very much) at baseline. Scores were computed by symptom count (pretty much or very much) based on the DSM-IV: (1) Attention Deficit Disorder-Inattentive (n = 9 items; e.g., is often easily distracted by extraneous stimuli); (2) Attention Deficit Disorder-Hyperactivity (n = 9 items; e.g., is often “on the go” or acts as if “driven by a motor”); (3) Oppositional Defiant Disorder (n = 8 items; e.g., is often angry and resentful), and (4) Conduct Disorder (n = 15 items; e.g., often bullies, threatens or intimidates others). At baseline, all children had a DBD rating scale completed by their mothers and only maternal perspective as reflected by their scores – endorsing or not endorsing a DBD- were used in the current study. The CareSS sample (Table 1) is composed of four subgroups reflecting both maternal and teacher ratings, endorsing or not endorsing a DBD: only maternal endorsement (n = 22, 24.7%), only teacher endorsement (n = 19, 21.3%), both maternal and teacher endorsement (n = 44, 49.4 %), and four students that did not have maternal DBD endorsement and their teacher DBD was missing. (This latter group was retained for three reasons: 1. they were identified at-risk on the teacher-completed SSBD screener with borderline or subthreshold symptomatology (Walker & Severson, 1990); 2. their average length of time participating in the project was consistent with the other subgroups; and 3. they engaged in greater MHS use.) Most children (87.4%) also had DBD ratings scales completed by their teachers. Seventy-four percent (n = 66) of children were endorsed for one or more DBDs by their mother, while 70.7% (63) were endorsed for one or more DBD by their teachers; About one fourth of children were not endorsed for a DBD by their mothers (25.8%, n = 23). More than one half of mothers (n = 47, 52.8%) endorsed symptoms characteristic of more than one DBD. For the current study, maternal DBD ratings were used to compute child symptom severity using a 0 to 2 scale (0 = no endorsement of a DBD, 1 = 1 DBD, 2 = 2+ DBDs).

Covariates.

Covariates included child’s age and gender, child symptom severity (count of maternal- reported disruptive behavior disorder symptoms), and maternal unemployment at baseline.

Analytic Plan

An initial ordinary least squares (OLS) multiple regression was conducted first to examine the relationships at baseline between perceived functional and network support, received support, strain and covariates. Hypothesis 1 (Baseline caregiver strain will positively predict MHS use) was tested by a second OLS multiple regression. Hypothesis 2 (Different types of baseline social support- perceived and received- will differentially predict MHS use) was tested with three additional OLS multiple regressions. To test Hypothesis 3 (Baseline social support will moderate the association between baseline caregiver strain and MHS use) a series of OLS multiple regressions with 2-way interactions were conducted to determine if baseline perceived and received support moderated the association between strain and service use. Each moderator (perceived functional support, perceived network support, and received support at baseline) hypothesized to predict MHS use and modify the “when and for whom” of the association between strain and MHS use, was analyzed in a separate model (Wu & Zumbo, 2008). The moderators met criteria (Wu & Zumbo, 2008) because they preceded the dependent variable (MHS use), were baseline characteristics of the sample, and were not correlated with strain, the independent variable of interest. To further explore moderation patterns (De Coster, Iselin, & Galluci, 2009), enhance interpretability, and cross verify the consistency of findings (i.e., strength of association among variables), strain was dichotomized using a median split (low or high). Strain was analyzed in two ways (i.e., continuous and dichotomous) to shed light on its relationship with the other variables and to gain a deeper understanding of its complexity while also capitalizing on the strengths and offsetting the potential weakness of the analytical lenses afforded by each approach (DeCoster, et al., 2009). Hence, three sets of moderation analyses were conducted. The first of each set used continuous variables and the second used strain as a dichotomized variable (De Coster, et al., 2009). Finally, post-hoc analyses further verified findings and examined possible conditional MHS use patterns between high and low levels of support, covariates, and strain. Analyses were conducted using Stata 15 (Stata Corp., 2017).

Results

Descriptive statistics are summarized in Tables 1 and 2. Bivariate correlations reveal expected, strong correlations between the two subtypes of perceived support, functional support and network support (r = .948), and indicate that received support and perceived functional support are distinct constructs (r = .246). Mothers reported a mean social support network size of 6.47 (SD = 4.72) and a household size of 5.16 (including mother). About 62% of mothers reported knowing individuals within their network for more than five years and indicated their availability from weekly (43%) to daily (51%). Mothers’ average strain score was 2.17 (Median = 1.97, 51% below median; range = 1.00 to 4.23) placing their experiences closer to a little (2) than to the somewhat (3) anchor of the scale. At study’s end, mothers had received an average of 12.15 services (SD = 14.83, range = 0 to 63).

Table 2.

Correlations between variables at baseline and Service Use

Variable	range	1	2	3	4	5	6	7	8
1. Caregiver Strain	1.00–4.23
2. Perceived Functional Support	12.00–525.00	−.199
3. Perceived Network Support	10.00–244.00	−.101	.948**
4. Received Support	44.00–195.00	−.082	.246*	.122
5. Maternal Service Use	0.00–63.00	.143	.097	.128	−.083
6. Child Symptom Severity	0.00–4.00	.498**	−.157	−.116	−.011	−.014
7. Child’s age at baseline	5.00–11.00	.159	.166	.209*	.132	−.006	.053
8. Child’s gender		−.033	−.155	−.181	.111	−.291**	−.063	−.104
9. Parent Unemployment		−.033	.095	−.013	.162	.176	.060	.071	.016

Notes. N = 89,

^*p <.05,

^**p <.01

Caregiver Strain

A linear regression, including both types of perceived support, received support and covariates, established the association of the independent variables to strain at baseline. The model explained 31% of the variance in mothers’ strain. Significant predictors included child symptom severity, perceived network support, and perceived functional support (R² = .315, adjusted R² = .251, F(7, 87) = 4.93, p < .001; Table 3). Child symptom severity and mothers’ perceived network support (i.e., size, contact frequency, and relationship length) were positively related to strain, while mothers’ perceived functional support (i.e., emotional and tangible, instrumental aid) was negatively related to strain. Received support was not significantly related to strain.

Table 3.

Predictors of Caregiver Strain at Baseline and Baseline Caregiver Strain as Predictor of Service Use

Variable	b a	S.E.	β b	t
Predictors of Caregiver Strain at Baseline 1
Child’s Gender	−.0430	.5168	−.0083	−0.08
Child’s Age	.1402	.1269	.1102	1.10
Child symptom severity	.1.259	.2721	.4510	4.63***
Parental Unemployment	−.2681	.4707	−.0564	−.57
Perceived Functional Support	−.0173	.0071	−.7324	−2.42*
Perceived Network Support	.3289	.1518	.6483	2.17**
Received Support	.0001	.0064	−.0013	-0.01
Constant	4.0743	1.1866		3.43**
Caregiver Strain as Predictor of Service Use 2
Caregiver Strain	1.533	.744	.245	2.060*
Child’s Gender	−9.701	3.336	−.297	−2.910**
Child’s Age	−.753	.824	−.095	−.914
Child symptom severity	−3.105	2.076	−.176	−1.496
Parental Unemployment	6.640	3.057	−222	2.172*
Constant	11.314	7.250	.245	1.56

¹Notes. Dependent Variable: Baseline Caregiver Strain Global Score

N = 83; R² = .315 adjusted R² = .251, F(7, 75) = 4.93, p < .0001

^aUnstandardized coefficient

^bStandardized coefficient

^*p <.05,

^**p <.01,

^***p <.001

²Notes. Dependent Variable: Service Use. Predictor variables at baseline.

N = 87; R² = .169 adjusted R² = .118, F(5, 81) = 3.292, p < .01

^aUnstandardized coefficient

^bStandardized coefficient

^*p <.05,

^**p <.01

Hypothesis 1: Baseline caregiver strain will positively predict mental health service use.

Although the baseline bivariate correlation was not significant (Table 2), strain positively predicted MHS use, accounting for baseline covariates (Table 3). The model explained 17% of the variance and strain was positively associated with MHS use (β = 1.53, p < .01). However, covariates were the strongest predictors of MHS use, specifically, child gender (β = −9.701, p < .01) and maternal unemployment (β = 6.64, p < .05). Mothers of sons attended 9.7 more sessions than mothers of daughters, and unemployed mothers attended 6.6 sessions more than employed mothers.

Hypothesis 2: Baseline perceived and received social support will predict mental health service use.

The bivariate correlations (Table 2) between perceived network support, perceived functional support and received support at baseline with MHS use were not significant. OLS multiple regression models similarly revealed no bivariate relationship between perceived network support (R² = .12, adjusted R² = .07, F(5, 79) = 2.246, p = .058) or perceived functional support (R² = .11, adjusted R² = .05, F(5, 80) = 1.962, p = .09) and MHS use. The model with received support (R² = .13, adjusted R² = .07, F(5, 79) = 2.399, p < .05) was significant, but received support itself was not significantly related to MHS use (β = −.029, p = .50).

Moderation Analyses

The results for multiple regression models testing Hypothesis 3 are presented in Table 4. The left side of the table presents the models using all variables as continuous, and the right side presents parallel models using strain as a dichotomous variable to facilitate interpretation of the interaction. Each regression tests a different type of support as moderator and MHS use as the dependent variable in all of the models.

Table 4.

Moderation Analyses by Continuous Caregiver Strain and Dichotomized (High vs Low) Caregiver Strain Associated with Service Use

	Caregiver Strain Continuous				Caregiver Strain Dichotomized
	b a	S.E.	β b	t	b a	S.E.	β b	t
Perceived Support
Strainc	3.094	1.303	.499	2.74**	15.178	4.889	.523	3.10**
Functional Support 1 ,4	.0893	.0471	.605	1.9++	.0381	.0208	.258	1.83+++
CGSc × Functional Support1,4	−.0134	.0077	−.554	−1.73+++	−.0652	.0315	−.380	−2.07*
Child’s Gender	−7.878	3.424	−.244	−2.30*	−8.448	3.374	−.262	−2.50*
Child’s Age	−.7905	.8417	−.100	−.94	−.7100	.8244	−.090	−.86
Child Symptom Severity	−1.856	2.084	−.107	−.89	−1.143	1.965	−.065	−.58
Parent Unemployed	5.213	3.064	.177	1.70+++	5.396	3.048	.183	1.77+++
Constant	1.255	9.097		−.14	9.71	7.152		1.36
Strainc	2.702	1.230	.436	2.20*	15.257	5.101	.524	2.99**
Network Support 2 ,5	.1334	.0901	.441	1.48	.0890	.0426	.295	2.09*
CGSc × Network Support2,5	−.0167	.0142	−.384	−1.18	−.1293	.0638	−.396	−2.02*
Child’s Gender	−8.423	3.434	−.261	−2.45*	−8.607	3.365	−.267	−2.56**
Child’s Age	−.7222	.8493	−.092	−.85	−.7189	.8282	−.091	−0.87
Child Symptom Severity	−2.110	2.086	−.122	−1.01	−1.098	1.958	−.063	−0.56
Parent Unemployed	5.761	3.109	.195	1.85++	5.904	3.047	.200	1.94+
Constant	2.999	9.693		.03	8.944	7.166		1.25
Received Support
Strainc	4.794	2.101	.750	2.28*	18.812	10.177	.626	1.85++
Received Support 3 ,6	.1857	.132	.470	1.41	.0132	.0535	.033	.25
CGSc × Received Support3,6	−.0332	.020	−.716	−1.66	−.0961	.0854	−.381	−1.13
Child’s Gender	−9.165	3.497	−.276	−2.62**	−9.882	3.516	−.297	−2.81**
Child’s Age	−.8445	.8556	−.104	−.99	−.7729	.8530	−.095	−.91
Child Symptom Severity	−3.345	2.091	−.189	−1.60	−3.279	2.001	−.185	−1.64
Parent Unemployed	5.906	3.204	.194	1.84+++	6.821	3.188	.225	2.14*
Constant	−5.391	14.670		−.37	16.294	8.657		1.88

Note. Dependent variable = Service use. Remaining variables are baseline.

^aUnstandardized coefficient

^bStandardized coefficient

^cCaregiver Strain

¹Perceived Functional Support: N = 85; R² = .193, adjusted R² = .119, F(7, 77) = 2.63, p < .05

²Perceived Network Support: N = 84; R² = .189, adjusted R² = .114, F(7, 76) = 2.54, p < .05;

³Received Support; N = 84: R² = .207, adjusted R² = .134, F(7, 76) = 2.84, p < .01;

⁴Perceived Functional Support: N = 85; R² = .212, adjusted R² = .140, F(7, 76) = 2.96, p < .01;

⁵Perceived Network Support: N = 84; R² = .222, adjusted R² = .150, F(7, 89) = 3.10, p < .01;

⁶Received Support; N = 84: R² = .207, adjusted R² = .134, F(7, 76) = 2.85, p < .01;

^*p <.05,

^**p <.01,

^***p <.001,

⁺p = .05,

⁺⁺p=.06 ,

⁺⁺⁺p= .07–.09

Hypothesis 3: Baseline social support will moderate the association between baseline caregiver strain and mental health service use.

Perceived functional support as moderator.

The continuous model explained 19% of the variance and strain was positively associated with MHS use (β = 3.09, p < .05), such that for every unit increase in strain, mothers attended approximately 3.09 more sessions. The moderated association between strain and perceived functional support was negative and not significant (β = −.0134, p < .08). Child male gender was significantly related to MHS use (β = −7.87, p < .05). Specifically, mothers of boys attended, on average, 7.9 more sessions than mothers of girls. Maternal employment status was positive and not significant (β = 5.21, p = 09).

The model using a dichotomized strain explained 21.2% of the variance. Strain (β = 15.18, p < .01) was positively associated with MHS use. The moderated association between caregiver strain and perceived functional support was negative and significant (β = −.065, p < .05). As before, male gender was significantly associated with MHS use and the association between maternal unemployment and MHS use was not significant.

Perceived network support as moderator.

The continuous model explained 18.9% of the variance. Strain was positively associated with MHS use (β = 2.70, p < .05). Specifically, for every unit increase in strain, mothers attended approximately 2.70 more sessions. There was no evidence of moderation in the association between strain and perceived network support (β = −016, p < .24) . Child gender was significantly related to MHS use (β = −8.42, p < .05), such that mothers of boys attended, on average, approximately 8.4 more sessions than mothers of girls. Maternal unemployment was positively associated with MHS use and not significant (β = 5.76, p = .068).

The model using a dichotomized strain explained 22.2% of the variance, and strain was positively associated with MHS use (β = 15.25, p < .05). Moreover, there was a significant, negative moderated association between strain and perceived network support (β = −.129 p < .05). As before, boys received significantly more services than girls, and the association between maternal unemployment and MHS use was significant (p = .05).

Received support as moderator.

The continuous model explained 20.7% of the variance. Strain was positively associated with maternal MHS use (β = 4.79, p = .05), and the relationship was consistent to that observed in the prior models. There was no moderation by received support (β = −.033, p = .102). Child gender significantly predicted MHS use (β = −9.165, p < .01) with mothers of boys attending, on average, 9 more sessions than mothers of girls. Maternal unemployment was not significant (β = −5.90, p = .069).

The model with a dichotomized strain explained the same 20.7% of the variance. Strain was marginally and positively associated with MHS use (β = 18.81, p = .068). Received support was not a moderator. Child gender and maternal unemployment significantly predicted service use (β = −9.88, p <.01; β = −6.82, p <.05, respectively).

Post-hoc Analyses: Caregiver Strain and Service Use Patterns

Post-hoc pairwise comparisons of means were used to interpret interaction terms in the regression models. We created simple 2×2 tables (see Table 5) of mean MHS use across dichotomized levels (low/high) of strain and support, resulting in four groups in each table: low support-low strain (low-low), low support-high strain, high support-low strain, high support-high strain (high-high). The first set of results relate to perceived (functional and network) support. Patterns were similar across types of perceived support, such that mothers in the perceived low support-low strain groups were least likely to use MHS (M = 4.55, M = 3.35), while MHS use by the other groups was notably higher. Mothers in the perceived high support-high strain groups were the second least likely group to use MHS (M = 11.85, M = 9.95). Mothers in the perceived high support -low strain groups (i.e., M = 16.35, M = 18.36) had the highest MHS use, followed by mothers in the perceived low support-high strain groups (M = 12.61, M = 14.27). The second set of results relate to received support, and mirror those just described for perceived support. Specifically, mothers in the received high support-low strain group received the most services (M = 18.39); however, the means for the remaining three received support-strain groups were comparable and notably smaller (range = 9.72 to 9.85).

Table 5.

Mean Number of Maternal Service use

	Caregiver Strain
	Low	High	t
Perceived Functional Support
			(.90)
Low	4.55	12.61	1.87+
High	16.35	11.85	−1.04
t	2.73**	−.18	(1.63)
Perceived Network Support
			(.99)
Low	3.35	14.27	2.58**
High	18.36	9.95	−2.01*
t	3.55***	−1.03	(1.54)
Received Support
			(2.04*)
Low	9.85	9.58	−.06
High	18.39	9.72	−1.87+
t	1.89+	.03	(−.03)
Child Gender
			(−1.27)
Female	5.0	5.54	0.09
Male	11.47	17.84	1.78
t	−1.35	−2.66**	(2.66**)
Workforce Involvement
			(−.13)
Employed	8.95	10.0	.26
Unemployed	10.52	20.88	2.21**
t	.36	2.44*	(2.59*)

Note. N = 89,

^*p < .05,

^**p < .01,

^***p < .001,

⁺p = .06;

All variables are baseline, except service use.

Table 5 also includes the mean frequency distribution of MHS use when strain is dichotomized by gender and employment status. Mothers of boys received significantly more services than mothers of girls, the difference ranged from more than twice as much for those experiencing low levels of strain to more than three times as much for those experiencing high levels of strain. The greatest difference was between mothers of girls reporting low strain and mothers of boys reporting high strain (t = 2.66, p < .01). Unemployed mothers experiencing high levels of strain used more than twice as many services than employed mothers with high strain and unemployed mothers with low strain and these differences were statistically significant. The greatest difference was between employed mothers experiencing low strain and unemployed mothers experiencing high strain (t= 2.59, p < .05)

Discussion

The current study examined the relationship between strain and support and their impact on MHS use by low income African American mothers on behalf of their children referred for disruptive behavior problems. We tested three hypotheses: 1) Baseline caregiver strain will positively predict MHS use, 2) Baseline perceived and received social support will differentially predict MHS use and 3) Baseline social support will moderate the association between baseline caregiver strain and MHS use. Findings supported Hypothesis 1, did not support Hypotheses 2 and partially supported Hypothesis 3. Results support and extend prior findings on the complex relationships among social support, strain, and MHS use by advancing our understanding about “when and for whom” these relationships may apply for low income African American mothers (e.g., Reblin & Uchino, 2008; Wu & Zumbo, 2008). Prior research points to child symptom severity as a predictor of both strain and MHS use. Current findings suggest that child symptom severity doesn’t trigger service use above and beyond its association with strain, which itself relates to MHS use. Findings have informed a modified conceptual framework, for future hypothesis testing, of these complex associations for low income African American mothers of children with a DBD (Figure 2).

Figure 2.

Modified Conceptual Framework Relating Caregiver Strain and Social Support to Mental Health Service Use for Low-Income African American Mothers of Children with DBD

Unpacking Caregiver Strain

Consistent with prior research, mothers’ reports of higher strain corresponded with more severe child symptoms (Brannan & Heflinger, 2006) and lower perceived functional support (affective, emotional and instrumental; McCabe et al., 2003). Of particular interest, perceived network support - the structure of mothers’ support - was also related to strain, suggesting that a larger network with increased contact may not facilitate more perceived support or reduced caregiver burden. Overall levels of strain reported in the present study were relatively low compared to previous samples, including a diverse sample (Tsai et al., 2015) and an African American sample (McCabe et al., 2003). Findings highlight the complexity and nuance of social networks and social support and the need for careful, precise and comprehensive measurement. Notably, consistent with prior research, mothers reporting higher strain also used more services for their children (e.g., Angold et al., 1998; Farmer et al., 1997; Shin & Brown, 2009).

Perhaps current findings can be explained by the cultural and social context afforded by prior research drawing attention to African American mothers’ intensive use of smaller networks (Bussing et al., 2003) that may in turn create insularity and access to a social network with limited and strained resources (Jarrett & Burton, 1999) and increase opportunities for friction and vexation among members (Miller-Loncar et al., 1998). The proverb bigger is not (necessarily) better may apply; growing the size of one’s network (adding new members resembling old ones) may not result in more support. Relatedly, received support’s weak association with perceived support may be explained by a possible disconnect between support offered and the support needed to reduce caregiver burden (Melrose et al., 2015). That is, within a small network of limited resources, perceiving someone is there for you may be more reassuring and comforting than the assistance they can provide.

Social Support Moderates Strain

Findings shed light on the complex interaction between perceptions of social circumstances and connectedness, particularly in highlighting support-strain conditional relationships. Mothers’ MHS use varied as a function of perceived support-strain combinations, suggesting an important but complex picture of when-and-for-whom associations among strain, support, and MHS use. Specifically, mothers endorsing high strain or high support engaged in high levels of MHS, whereas mothers endorsing high strain and high support engaged in low levels of MHS. Of particular note, mothers experiencing a combination of low strain and high support received the most services. These findings lend support to prior research indicating that strain at high levels, counterintuitively, may predict lower MHS use (Angold et al., 1998; Bussing et al., 2003; Pellerin et al., 2010; Shin & Brown, 2009), and that support can serve as either a facilitator or a deterrent (e.g., Martinez & Lau, 2011; Pescosolido et al., 1998). Furthermore, the conditions in which more services were received (e.g.., high support-low strain, low support-high strain) are most likely influencing maternal MHS use differently. That is, while a mother with high perceived support combined with low strain may engage in MHS use with the perception that she has folks in her court, this may not be the case for a mother with low perceived support and high strain. Perhaps, for the former group, perceived support may have served as a facilitator in her MHS use by socially anchoring mother with social stability and integration within her support network (Grzymala-Kazlowska, 2016). Whereas, for the latter group, the lack of this presumed social anchoring or support from her network combined with high strain may have facilitated MHS use. Finally, results suggest that low income African American mothers’ reports of strain and MHS use reflect perceptions of their social world, including availability of emotional/affective and instrumental/tangible support and the social structural framework that sustains this support.

Boys Receive More Treatment than Girls

Findings reflect a pattern in prior research indicating boys with behavioral difficulties are more likely to receive treatment than girls (e.g., Bussing et al., 2003; Tsai et al., 2015). However, child gender and caregiver strain were not related, suggesting that mothers’ concerns about their sons may have become more pronounced once engaged with formal services, as has been reported in previous studies (Gudiño et al., 2008; Kaufman et al., 2010; Rowley et al., 2014). Perhaps the same reason that school based MHS are appealing – that they are made available in school – also raises maternal concerns about school and/or teacher biases toward African American boys. That is, if mothers perceive racial bias toward their sons in their schools, then by association, they may distrust MHS offered in school, even if services are provided by staff from an external agency (Varner & Mandera, 2012). Findings advance understanding of MHS use patterns among low income African American mothers on behalf of their children with a DBD by providing a window into parents’ accessing of MHS on behalf of their sons presenting with behavioral difficulties.

Unemployment Facilitates Services but at What Cost?

Unemployment by itself did not predict MHS use. Nonetheless, mothers who were both unemployed and experiencing high strain used more MHS than any other group. Current findings afford a snapshot of mothers grappling with unemployment and high strain that may help to explain previously reported negative relationships between children’s emotional and behavioral difficulties and maternal employment (Brannan et al., 2018; Richard et al., 2014). Perhaps unemployment creates more time and opportunity to engage in MHS, for instance by reducing competing demands and inserting more flexibility for scheduling appointments. However, we do not know to what extent unemployment may have been an unwanted direct result of maternal strain associated with parenting children with DBD in this sample (Brannan et al., 2018). We do know that single mothers of children with severe emotional and behavioral problems are more likely to reduce their work hours rather than stop working compared to married mothers, underscoring the increased financial vulnerability experienced by single parent households (Richard et al., 2014). We also know that unemployment in primarily single-parent households in low income, urban communities can interfere with healthy family functioning (e.g., maternal psychological functioning and parenting) and child well-being (e.g., socioemotional functioning) (Chung & McGuire-Jach, 2020; McLoyd, Jayaratne, Ceballo, & Borquez, 1994). Hence, higher participation in MHS by low income, unemployed African American mothers could be interpreted as increasing risk of family financial instability and functioning, and as clinical red flag that children’s behavioral difficulties are so pronounced that its effect may be spilling over to mothers’ workforce engagement. Future research to further understand the impact of caregiving demands on maternal employment and how these are impacted by the African American experience of parenting male children with behavioral difficulties is needed.

Maternal Endorsement Patterns

The majority of children were endorsed for a DBD by their mothers, suggesting maternal concern may have been an important driver of study participation and MHS use. Nevertheless, parent-teacher endorsement patterns also point to discordant perceptions of children’s behaviors, a common finding in psychosocial functioning measures (De Los Reyes, Cook, Gresham, Makol, & Wang, 2019). Teacher reports of disruptive classroom behaviors in the absence of maternal concerns may reflect implicit teacher biases toward African American students (Munzer et al., 2018); on the other hand, discordance may reflect meaningful contextual variation. Recent scholarship on inter-rater discordance on children’s behavior moves away from psychometric explanations, toward understanding disagreement as clinically informative and reflecting the contextual differences in which children’s behaviors are observed and experienced (De los Reyes, et al., 2019; Dirks, De Los Reyes, Briggs-Gowan, Cella, & Wakschlag, 2012).

Limitations

Findings should be interpreted with caution in light of several methodological limitations. First, although dichotomization can produce meaningful findings, it may also risk a study’s analytic power, decrease the strength of association among variables, and lead to false positive results (DeCoster, et al., 2009). Prior to dichotomizing, analyses were conducted using continuous variables, confirming consistency of the effect of strain and support on MHS use; the strength of association among the variables increased and moderations were significant with dichotomization. Moreover, the reduction of the data to mean MHS use across dichotomized levels of strain and support explained more variance and clarified important patterns distinguishing mothers that were more or less likely to engage in services. Second, the sample represents African American families, commonly excluded from research, from a small number of multiply stressed and under-resourced communities in one urban center. While also a strength of this study, the heterogeneity and complexity of findings highlight risks to simplifying conclusions or generalizing to other African American mothers of children with DBD, living in other cities, or in rural or suburban communities. Third, mother-teacher discordance relating to endorsing symptoms of DBD may reflect implicit teacher race-based biases toward African American children, and/or maternal perceptions of bias, which may have influenced behavior ratings and/or MHS use (e.g., Munzer, et al., 2018). Mother-teacher DBD endorsement discrepancies also could be explained by different symptom presentations at school and home (Dirks et al., 2012). What we are able to ascertain is that although eligible children were identified in the first recruitment step of the study via teacher-completed screeners by school mental health teams, maternal endorsement of a DBD was comparable if not a bit higher than that of teachers. Fourth, we opted to include data in the analyses from four children that were not endorsed for a DBD by their mother and their teacher had not completed the measure. Although they did not endorse their children for a DBD, these mothers engaged in MHS use at a rate higher than the other subcategories of DBD endorsement and maternal MHS use.

Fifth, some potentially important demographic variables were assessed but not examined as covariates because of limited variability (e.g., socio-economic status and family composition). Moreover, other factors not included in the current analyses (e.g., therapist, service setting home or school) may also influence MHS use. Sixth, we relied on self-report Likert-type scales, common in social support research, and participating mothers endorsed a wide range of educational levels. To address possible literacy concerns, questionnaires were read aloud to participants and included an example item describing how to utilize the response anchors. Seventh, it is possible that results indicating received support did not predict strain nor moderate MHS may be due to the measure used. This measure did not also inquire if the received help was wanted by the recipient as suggested by prior research (Melrose et al., 2015). Relatedly, we measured strain with a widely utilized self-report tool. Perhaps the consistent report of low strain by African American mothers reflects how strain has been conceptualized and measured, absent considerations of race, culture or context; perhaps, a culture- or context-specific measure that is more discerning may yield higher rates of reported strain. However, this would not negate the value of understanding the relationship of (herein reported) low strain levels to service use and support. If African American mothers tone-down reporting – perhaps as a general reaction to historical and ongoing prejudice – then low strain may still reflect their experiences.

Conclusions

This study sheds light on the complex interrelationships of strain, support and MHS use for low income, urban African American mothers parenting children with (or at risk for) a DBD. Findings suggest conditional relationships in which different strain-support combinations lead to notably different levels of MHS use for mothers engaged in services. Notably, the context most conducive to maternal MHS use was the reported baseline combination of low strain and high support. The conceptual framework (Figure 2) informed by these findings illustrates the moderated relationships between strain and perceived support while also highlighting the primacy of different variables for strain (i.e., child symptom) and for MHS use (i.e., child gender). Future studies with larger samples are warranted to further examine the complex and heterogeneous experiences of caregiving, strain-support and MHS use among low income African American mothers. Findings highlight implications for research and practice. Attending to mothers’ social connectedness and experiences of strain when developing and delivering MHS interventions to reduce children’s disruptive behaviors may be more responsive to the needs of some African American families, in particular families living in under-resourced and marginalized urban communities. Additionally, attending to mothers’ concerns for their children, particularly their sons, using sociohistorical lenses that acknowledge racial biases and individual experiences may also help to build trust, and engage and retain families in services, for instance by including or leveraging families’ social support networks and/or addressing caregiver strain directly.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (Office for the Protection of Research Subjects, Institutional Review Board, 2005–0133) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Highlights.

• Mothers reporting high caregiver strain or high perceived social support, but not both, were more likely to use services

• Received support did not moderate the association between strain and service use

• Mothers were more likely to receive services for their sons than their daughters