Longitudinal Bidirectional Relations Among Parenting Quality, Parenting Stress, and Child Functioning in HIV-Affected Families

Abstract

Mothers living with HIV (MLH) face unique stressors impacting parenting, parenting stress, and child psychosocial functioning, but longitudinal, bidirectional relations among family processes have not been examined in this population. This study examined relations among parenting quality, parenting stress, and child functioning in 174 MLH-child dyads (aged 6–14, M_age = 9.65, SD = 2.49; 51% female; 57% Black/African American; 35% Latinx). Families completed self-report questionnaires over four waves spanning 15 months. Cross-lagged panel analysis revealed unidirectional and bidirectional relations between parenting stress and child functioning; parenting quality and child functioning; and parenting quality and parenting stress. The findings suggest that prevention and intervention efforts with HIV-affected families should target both parent factors (e.g., communication skills) and child factors (e.g., emotion regulation), emphasizing parenting stress reduction in order to bolster family outcomes.

Introduction

Nearly two million people in the U.S. are living with HIV, and women of color are disproportionately affected by the epidemic (Centers for Disease Control and Prevention, 2019). As the life expectancy for people living with HIV increases, mothers affected by the virus are increasingly able to raise their children. HIV presents unique challenges for families, which can result in elevated maternal parenting stress, compromised parenting quality, and child functioning difficulties (Chi et al., 2015). In the general population, bidirectional effects have been observed among parenting quality, parenting stress, and child functioning (Paschall & Mastergeorge, 2016). However, these bidirectional relations have not been examined among families affected by HIV. Considering the unique stressors faced by this population, this study investigates the longitudinal interplay of family processes in order to identify modifiable targets for prevention and intervention efforts.

Impact of HIV on Mothers and Children

Maternal HIV often has a profound impact on family processes and child psychological well-being, in addition to introducing unique stressors. The process model of parenting (Belsky, 1984) posits that parenting is primarily determined by three domains of influence: parents’ own personalities, psychological resources or lack thereof, and psychological well-being (e.g., depression, coping skills, parenting stress); child factors (e.g., temperament and behavior); and contextual sources of stress and support (e.g., social support, employment/financial stress, HIV-related stigma and other stressors). The first domain includes factors that are internal or proximal to the parent, including psychological functioning, mental health, and coping skills. In this study, parenting stress is conceptualized as an aspect of psychological functioning within the first domain. The second domain refers to the child’s contribution, including temperament and emotional/behavioral functioning. Child internalizing and externalizing difficulties are conceptualized as child factors that may reciprocally contribute to parenting. The third domain refers to the supports and stressors within the ecological context, such as social support (or lack thereof), stress related to employment or financial strain, and other contextual challenges. Though this theory was not developed specifically in reference to mothers living with HIV (MLH), contextual HIV-related stressors can be conceptualized in the third domain, including HIV stigma, difficulty accessing social support due to nondisclosure of HIV status, and other stressors related to living with a highly stigmatized illness.

Consistent with this conceptualization, research has demonstrated that HIV has a multifaceted impact on MLH and their children, both directly and indirectly through associated challenges, such as compromised parenting and increased parenting stress (Chi et al., 2015; Rochat et al., 2018). HIV may have deleterious effects on children’s psychological functioning regardless of whether children are aware of their mothers’ HIV status. A recent review concluded that children affected by HIV—including those who were aware and unaware of their parents’ status—exhibited more depression, anxiety, loneliness, suicidal ideation, and hopelessness, as well as more disruptive behaviors, sexual risk behaviors, delinquency, and conduct problems, compared to their non-affected counterparts (Chi & Li, 2013). Children living with an HIV-positive parent experience similar or even greater difficulties compared with children orphaned by HIV/AIDS, including similar levels of loneliness, self-esteem, and other emotional challenges compared to non-HIV affected children (Chi & Li, 2013), suggesting that not only potential HIV-related loss, but also the stress of living with an ill parent, may negatively impact children. These difficulties may be even more pronounced among children whose mothers have not disclosed their HIV serostatus, as many suspect that something is wrong but do not have an explanation for the stress in the home (Qiao et al., 2013). It is also likely that the impact of parental HIV on children shifts and varies across developmental stages (Stein et al., 2014), as children experience significant social, emotional, and cognitive changes in middle childhood and early adolescence. For example, given that many women are first diagnosed with HIV during the perinatal period (Momplaisir et al., 2015), disease progression often occurs during middle childhood, which may correspond with an amplification of the impact of HIV coinciding with significant developmental changes (Rochat et al., 2017). Accordingly, based on the timing of MLH’s diagnosis and the course of their illness, families’ needs and the impact of HIV on children may ebb and flow across time, highlighting the need for longitudinal studies examining family processes across these developmental periods.

HIV and parenting quality.

In addition to the direct link between MLH’s physical health and child functioning, HIV impacts children indirectly via disruptions in positive parenting quality. Two important aspects of parenting quality include relationship quality and communication. Parent-child relationship quality is defined as the support parents provide children through warmth, affection, and nurturing (Barber et al., 2005). Parent-child communication refers to the content, frequency, and quality of conversations between parents and youth (Markham et al., 2010). Though parent-child relationship quality and communication involve dyadic processes between parent and child, they are conceptualized as facets of parenting quality and, in large part, driven by parent behavior (Murphy et al., 2015). Research has revealed that maternal HIV is associated with worsened parenting quality, which is in turn related to child emotional and behavioral problems (Murphy et al., 2015). Among a sample of South African MLH, maternal health predicted child externalizing problems indirectly via compromised parent-child relationship quality (LeCroix et al., 2019), suggesting that mothers’ health-related difficulties may spill over to impact their parenting and, in turn, child functioning. A study conducted among a sample similar to that of the current study—MLH with 6- to 14-year old children—found that MLH reported considerable difficulty with certain aspects of parenting, such as communication (Murphy et al., 2015). Specifically, descriptive analyses in the study demonstrated that a majority of MLH reported low levels of at least one aspect of parenting, including monitoring, relationship quality, or communication. The researchers posited that MLH may experience low confidence in their parenting skills (e.g., as evidenced by MLH’s relatively low parenting self-efficacy ratings), and this lack of efficacy, combined with lack of resources and physical limitations due to HIV, may hinder mothers’ ability to employ effective parenting strategies. Similarly, a large-scale study in South Africa demonstrated that HIV may impede caregivers’ efforts at positive parenting (Lachman et al., 2014). The study found that the negative association between parental HIV and positive parenting was mediated by other factors including poverty, parental depression, and child behavior problems. Though the study was cross-sectional, the association between child behavior problems and parenting suggests that there may be bidirectional effects. Longitudinal examinations of these associations are needed in this population given the complex stressors introduced by parental HIV.

HIV and parenting stress.

Distinct from parenting quality, parenting stress arises when caregivers perceive that the demands of childrearing outweigh their resources and ability to meet these demands (Morgan et al., 2002). This mismatch results in parents’ negative feelings toward themselves and/or their children. MLH face multiple risk factors for elevated parenting stress, including the challenges associated with chronic illness in families (Cousino & Hazen, 2013), low socioeconomic conditions (Gleeson et al., 2016), stigma and discrimination due to intersecting marginalized identities (Sangaramoorthy et al., 2017), and low parenting self-efficacy (Murphy et al., 2015). However, few studies have examined parenting stress among this group. One study demonstrated that MLH experienced clinically significant parenting stress at much higher rates compared to the general population and compared to parents living with a chronically ill child (Johnson et al., 2015). A cross-sectional study among MLH and their children in rural China demonstrated direct and indirect linkages among parenting stress, parenting quality, and child functioning (Chi et al., 2015). Moreover, within the sample of the current study, parenting stress was cross-sectionally negatively associated with multiple aspects of family functioning, including family cohesion, family routines, and parent-child communication, from both mothers’ and children’s perspectives (Schulte et al., 2017). Current literature with HIV-affected families highlights important impacts of HIV on families, but longitudinal transactional relations among family processes have not been examined in this unique population. In the absence of this data, theoretical and empirical work conducted in the general population can serve as a guide to understand potential bidirectional effects.

Transactional Parent-Child Interactions in the General Population

Consistent with systems theory, emphasizing the transactional nature of parent-child behaviors across development (Sameroff, 2009, 2010), parenting quality, parenting stress, and child psychosocial functioning appear to mutually influence one another over time among families not affected by HIV (Harrison et al., 2019). Although the field remains largely limited by its reliance on cross-sectional research, the body of longitudinal research examining these family processes is growing. Previous studies have provided support for longitudinal relations between parenting quality and externalizing behaviors such as aggression and delinquency (Yoo, 2017). Although reciprocal effects of parenting and youth functioning appear to be stronger when examining externalizing behaviors (Reitz et al., 2006), research has found that parenting quality and internalizing difficulties also mutually predict one another over time. For example, a study among disadvantaged French Canadian families demonstrated that mother-reported positive parenting quality and teacher-reported child internalizing problems were significantly reciprocally related, such that more internalizing problems at Wave 1 predicted more positive parenting three years later at Wave 2, which in turn predicted fewer internalizing problems at Wave 3 (Serbin et al., 2015). Parents may be responding to their children’s internalizing problems by increasing their use of effective parenting strategies, which in turn confers benefits for children’s emotional functioning. In a large sample of girls assessed between ages 7 and 12, child depression and parental warmth predicted one another across time, though the effect of parental warmth on depression was stronger than the reverse (Hipwell et al., 2008). To a lesser extent, a few studies have examined transactional relations between parenting stress and child or family outcomes, revealing that parenting stress and child behavior problems are both antecedents and consequences of one another (Neece et al., 2012). However, much of the longitudinal work with parenting stress is limited to younger samples (e.g., infancy and early childhood; Crnic & Ross, 2017), despite important developmental changes during middle childhood and early adolescence. In sum, though there is a growing body of literature demonstrating reciprocal relations among parenting, parenting stress, and child functioning among the general population, few longitudinal studies have examined family processes among HIV-affected families, and none have examined these processes transactionally over time. Understanding reciprocal effects can have important implications for prevention and intervention efforts by highlighting the relative value of targeting parent factors vs. child factors based on which factors are driving longitudinal effects.

Research in the general population suggests that the reciprocal nature of parent-child interactions is particularly important during middle childhood and early adolescence (Yoo, 2017). During this period, important developmental changes may influence the bidirectional relations among family processes and child outcomes. Children in middle childhood (age 6–10) experience an increased sense of responsibility, expansion of social relationships outside the family, cognitive development that facilitates self-reflection, and heightened social comparison with peers (Eccles, 1999). Adversity in middle childhood, including maternal HIV or other stressors, can lead to a range of emotional and behavioral challenges, such as sadness, worry, low self-esteem, or behavioral noncompliance. In the context of maternal HIV, children in this developmental stage may show more readiness to learn of their mother’s HIV status. However, they also may experience parentification due to their growing sense of responsibility. Early adolescents (age 11–14) experience significant cognitive, biological, psychological, and social changes that facilitate their development of an autonomous sense of self (Eccles, 1999). Early adolescents who face adversity may experience internalizing problems and/or exhibit rule-breaking behavior. Family relationships shift as youth increasingly look to peers for support, which may lead adolescents to internalize misinformation or stigmatized messages related to HI from their peers. MLH are in a unique position to educate their early adolescents about sexual health and risk reduction, as youth explore sexual and romantic relationships (Tarantino & Armistead, 2016). Middle childhood and early adolescence is also the developmental period when many MLH choose to disclose their HIV status to their children (Murphy et al., 2011). Despite important developmental changes occurring during middle childhood and early adolescence, coupled with unique impacts of parental HIV on families, no studies have examined the longitudinal, transactional associations among family processes and child outcomes in this population.

Current Study

Guided by Belsky’s process model of parenting and Sameroff’s systems theory, the purpose of this study was to examine the longitudinal, transactional relations among parenting stress, parenting quality, and child functioning among a sample of MLH and their 6 to 14-year-old children. Figure 1 displays the conceptual model for the study. The overall research question was: How are parenting stress, parenting quality, and child functioning reciprocally related to one another over time among mothers living with HIV and their children? Applying cross-lagged panel analysis, the current study included three major hypotheses. First, it was expected that better parenting quality (i.e., better communication and relationship quality) would predict fewer child functioning difficulties over time, and that the effect would be reciprocal, such that better child functioning would predict better parenting quality over time (Hypothesis 1). Second, it was hypothesized that parenting stress and child functioning would be reciprocally related over time, such that higher levels of parenting stress would predict subsequently more child functioning difficulties, and conversely, child functioning difficulties would predict higher parenting stress over time (Hypothesis 2). Finally, it was expected that more parenting stress would predict worse parenting quality (parent-child communication and relationship quality; Hypothesis 1). In addition to these three hypotheses related to bivariate associations, a multivariate model was also explored to examine transactional relations among aspects of parenting quality (relationship quality), parenting stress (perceptions of dysfunctional parent-child interactions), and child functioning (depressive symptoms).

Figure 1.

Conceptual model displaying bidirectional relations among parenting practices, parenting stress, and child functioning.

Note. Data collected across four timepoints spanning 15 months. Covariates, concurrent associations at each timepoint, and residual variances are not displayed.

Method

Participants and Recruitment

Data for this study were drawn from a larger NIMH-funded randomized controlled trial evaluating an HIV disclosure intervention called Teaching, Raising, and Communicating with Kids (TRACK; Schulte et al., 2021). MLH and their children were recruited via HIV-related community agencies at sites in Georgia and California. Potential participants contacted study staff and completed a phone screen to determine eligibility. Families were eligible if a female primary caregiver (e.g., mother, aunt, grandmother, or other female primary caregiver) had a confirmed HIV/AIDS diagnosis and was the caregiver for a child between the ages of 6 and 14 who was not living with HIV. Though caregivers in the study did not have to be biological mothers, term “mothers living with HIV” is used to refer to all female caregivers in the sample to be consistent with previously published manuscripts from this study. Families were English- or Spanish-speaking, and children must have been living with the caregiver. A total of 174 dyads (N = 348) were enrolled across both sites.

Roughly half (51.4%) of the children in the sample were girls. The average age of children was about 10 (M = 9.65, SD = 2.49), and the average age of mothers was 39 (M = 39.24, SD = 7.90). One-third of the families identified as Latino or Hispanic. Families reported a range of racial identifications, including Black/African American (56.1% of mothers and 57.2% of children), White (34.4% of mothers and 31.1% of children), multiracial (6.7% of mothers and 9.4% of children), American Indian or Alaska Native (1.7% of mothers and 1.1% of children), Asian (0.6%), and Native Hawaiian or other Pacific Islander (0.6%). As anticipated, the ethnic breakdown of the sample reflects the demographic impact of the HIV epidemic in Georgia and California (CDC, 2014). Most mothers were single parents; mothers’ relationship status included never married (44%), married (20%), separated (14.9%), divorced (13.1%), and widowed (8%). About half (52.6%) of mothers reported having other adults living in the household. Mothers reported a wide range of education and socioeconomic levels; 16.6% of mothers completed eighth grade or less, 28% completed some high school, 19.4% finished high school, 2.9% completed a GED, 5.7% attended vocational or technical school, 22.9% completed some college, and 4.6% earned a college degree or higher. At baseline, approximately one-third (30.3%) of mothers were employed in the last 30 days. The median and modal reported household monthly income, including income from other adults living in the household, was $1,200, with a range from $0 (n = 2) to $6,000 (n =1). MLH had been living with HIV for an average of 12.04 years (SD = 6.91), and the majority (80.9%) reported having an undetectable viral load at their most recent laboratory test. Additional information about the physical health functioning of the sample can be found in the primary RCT outcome paper (Schulte et al., 2021).

Procedures

Study procedures were approved by Institutional Review Boards at both sites. Mothers and children provided informed consent and assent, respectively, prior to interviews. Data were collected at four waves, including baseline, 3-month follow-up, 9-month follow-up, and 15-month follow-up. Of the 174 dyads who enrolled, 100% completed baseline assessments, 95% completed 3-month follow-up, 93% completed 9-month follow-up, and 79% completed 15-month follow-up. Two trained interviewers conducted the interviews separately with mothers and children, using Computer-Assisted Personal Interviewing (CAPI) software. Mothers and children were reimbursed $60 in cash and $30 in gift cards, respectively, for each interview. For the larger study, half the mothers were randomized to receive a three-session, individual intervention focused on HIV disclosure. The other half were in a waitlist control group and received the intervention in a group format after completing all interviews. The sample for the current study includes both the intervention group and waitlist control group, and group assignment was included as a covariate for all bivariate pairs with significant cross-lagged effects. Intervention content focused on child development psychoeducation, pros and cons of disclosure, family routines, parent-child communication, prior experiences and quotes from MLH who had disclosed and their children, preparing for disclosure-related questions, and behavioral practice of disclosure. Other published papers describe the TRACK intervention in more detail as well as effects of the intervention on disclosure rates and child and family outcomes (Schulte et al., 2021). The results of the efficacy trial revealed that the intervention resulted in higher disclosure rates, improved family communication, reduced parenting stress, and reduced child internalizing symptoms (Schulte et al., 2021). Though an examination of the longitudinal effects of disclosure was beyond the scope of this paper, additional longitudinal analyses demonstrated that HIV disclosure was related to improvements in parenting stress, parent-child relationship quality, and communication over time (Goodrum et al., 2021).

Measures

In addition to the assessment of HIV disclosure relevant to the larger TRACK study, families reported on parenting quality, parenting stress, and child internalizing and externalizing outcomes. Both MLH and children reported on aspects of parenting and child functioning, and MLH reported on their parenting stress and HIV disclosure. To validate the use of scale scores, measurement models for each construct were established using confirmatory factor analysis to evaluate longitudinal measurement invariance, with scalar invariance established for each measure. Based on evaluation of measurement models, subscale scores, rather than total scores, were used for the Child Behavior Checklist and Parenting Stress Index.

Parenting quality.

Two aspects of parenting were assessed, including parent-child communication (child-reported) and parent-child relationship quality (both mother- and child-reported).

Parent-child communication.

Parent-child communication was assessed using the child-reported Parent-Child Communication Scale (PCC; Loeber, Farrington, Stouthamer-Loeber, & Van Kammen, 1998). The PCC is a 10-item measure using a Likert-type scale that ranges from 1 = Almost Never to 5 = Almost Always. A sample item is, “Do you think that you can tell your mother how you really feel about things?” The PCC has demonstrated adequate psychometric properties among similar samples (Murphy, Armistead, Payne, Marelich, & Herbeck, 2017). Items were averaged, and higher scores indicated better communication. Cronbach’s alpha in this sample was .86 at baseline.

Parent-child relationship quality.

Parent-child relationship quality was assessed using both child and mother reports of the Conflict Behavior Questionnaire, short form (CBQ; Prinz, Foster, Kent, & O’Leary, 1979). The CBQ includes 20 dichotomous True/False items, such as “We almost never seem to agree.” The measure has been previously used with other samples of families affected by HIV (Armistead et al., 2014; Goodrum et al., 2017). Items were averaged to create a scale score with higher values indicating stronger relationship quality. Cronbach’s alpha was .68 in this sample.

Parenting stress.

MLH completed the Parenting Stress Index, short form (PSI; Abidin, 1990), which demonstrates good psychometric properties (Haskett et al., 2006). The PSI is a 36-item scale assessing three domains of parenting stress: Parental Distress (PSI-PD; e.g, “I often have the feeling that I cannot handle things very well”); Parent-Child Dysfunctional Interaction (PSI-DI; e.g., “My child rarely does things for me that make me feel good”); and Difficult Child (PSI-DC; e.g., “My child does a few things which bother me a great deal.”). Based on a CFA of the PSI-PD, one item (“Having a child has caused more problems than I expected in my relationship with my spouse”) was removed from the scale due to a lack of both empirical validity and sociocultural relevance, as the majority of MLH in this sample were single parents. Each PSI subscale was computed as an average of the items, with higher scores representing more stress. Cronbach’s alpha at baseline was adequate for all three subscales, ranging from .82 to .85.

Child functioning.

Child psychosocial functioning included child-reported depressive symptoms, mother-reported depression/anxiety symptoms, and mother-reported aggressive behavior.

Depression/anxiety.

Two measures of internalizing symptoms were used. Children reported on their depressive symptoms over the past two weeks using the Children’s Depression Inventory (CDI; Kovacs, 1985), a 27-item scale with each item consisting of three statements on a scale from 1 to 3 (e.g., 1 = I am sad once in a while; 2 = I am sad many times; 3 = I am sad all the time). The CDI has previously been used among families affected by HIV (Murphy, Armistead, Marelich, Payne, & Herbeck, 2011) as well as with other samples of children of MLH (Gamarel et al., 2017). Items were summed to create a scale score, with higher scores indicating more depressive symptoms. Cronbach’s alpha for the baseline measure was .83.

MLH reported on children’s comorbid depression and anxiety using the Anxious/Depressed subscale of the Child Behavior Checklist (CBCL; Achenbach, 1991). The Anxious/Depressed subscale consists of 10 items with a 3-point response scale from 1 = Not True to 3 = Very True or Often True. Sample items include “Feels worthless or inferior,” and “Feels he/she has to be perfect.” Though this measure assesses comorbid anxiety and depression, the term “anxiety” is used to refer to this construct to distinguish this measure from the CDI. The CBCL demonstrates good psychometric properties and has been extensively used among diverse samples (Goodrum et al., 2012) Items were averaged to create a scale score, and higher scores indicated more anxiety symptoms. Cronbach’s alpha was .81 at baseline.

Aggression.

MLH reported on children’s aggressive behaviors using the Aggression subscale of the Child Behavior Checklist (CBCL; Achenbach, 1991), which comprises 17 items on a 3-point response scale from 1 = Not True to 3 = Very True or Often True. Sample items include “Disobedient at home,” and “Physically attacks people.” As with the CBCL anxiety scale, items were averaged to create a scale score, with higher scores representing more aggression. Cronbach’s alpha was .86 in this sample at baseline.

Data Analysis

Cross-lagged panel analysis (CLPA) in a structural equation modeling (SEM) framework was employed to estimate the longitudinal reciprocal interrelations among child psychosocial functioning, parenting quality, and parenting stress. CLPA is advantageous in its ability to test reciprocal effects, measure interindividual variation, and estimate over-time effects of one variable on another while accounting for their concurrent correlation (Selig & Little, 2012). To examine bidirectional associations between pairs of variables, a series of four CLPA models was specified. First, an autoregressive model was specified with no cross-lagged paths. In the autoregressive model, each variable was regressed on the same variable at the previous timepoint. Additionally, within-time correlations were specified across constructs. Second, a model was estimated with the addition of cross-lagged paths from one variable at each timepoint to the other variable at the subsequent timepoint. Third, a model was specified with cross-lagged paths in the reverse direction. Finally, a bidirectional model was specified with cross-lagged paths from each variable at one timepoint to the other variable at the subsequent timepoint. In all models, autoregressive and cross-lagged paths were constrained to be invariant across time; exceptions were made when model fit was significantly improved by freely estimating paths across timepoints. Models were compared using nested model likelihood ratio tests. Models were initially run without covariates due to sample size constraints; group assignment (intervention or control), child age, child gender, and study site (Georgia or California) were added as covariates for all models that had significant cross-lagged effects. Race/ethnicity was not included as a covariate because of its high collinearity with study site; the majority of MLH (89%) at the Georgia site were non-Hispanic Black/African American, and most MLH (72%) at the California site were Hispanic/Latina.

Results

Descriptive Results

Descriptive statistics for all variables are displayed in Table 1, and bivariate correlations are presented in Table 2. On average, children exhibited moderate levels of aggression, anxiety, and depressive symptoms across time. Children and mothers endorsed relatively positive perceptions of parent-child communication and relationship quality. Across time, mothers reported moderate levels of parenting stress, with slightly lower stress reported on the Parent-Child Dysfunctional Interaction subscale compared with the Parental Distress and Difficult Child subscales. More detail about longitudinal trajectories of child functioning, parenting stress, and parenting quality in this sample can be found in Goodrum et al. (2021).

Table 1.

Descriptive Statistics for All Study Variables

	Baseline			3-Month			9-Month			15-Month
	Mean (SD)	Min	Max	Mean (SD)	Min	Max	Mean (SD)	Min	Max	Mean (SD)	Min	Max
CDI	35.93 (6.95)	27.00	58.00	34.45 (7.09)	27.00	61.00	33.11 (5.66)	27.00	57.00	33.11 (6.29)	27.00	70.00
CBCL-Agg	1.48 (0.35)	1.00	2.71	1.44 (0.36)	1.00	2.71	1.45 (0.36)	1.00	2.71	1.44 (0.35)	1.00	2.76
CBCL-Anx	1.44 (0.37)	1.00	2.80	1.38 (0.35)	1.00	2.90	1.38 (0.37)	1.00	2.60	1.36 (0.35)	1.00	3.00
PCC	3.66 (0.73)	1.50	5.00	3.81 (0.72)	1.70	5.00	3.82 (0.69)	2.00	5.00	3.94 (0.69)	1.50	5.00
CBQ-C	1.83 (0.19)	1.05	2.00	1.84 (0.20)	1.05	2.00	1.86 (0.17)	1.30	2.00	1.88 (0.16)	1.05	2.00
CBQ-M	1.75 (0.23)	1.05	2.00	1.78 (0.22)	1.10	2.00	1.78 (0.22)	1.05	2.00	1.79 (0.22)	1.05	2.00
PSI-PD	2.48 (0.77)	1.00	4.55	2.25 (0.72)	1.00	4.18	2.18 (0.75)	1.00	4.30	2.21 (0.84)	1.00	4.82
PSI-DI	2.02 (0.73)	1.00	4.60	1.91 (0.64)	1.00	3.40	1.90 (0.68)	1.00	4.00	1.83 (0.70)	1.00	4.20
PSI-DC	2.29 (0.79)	1.00	4.92	2.25 (0.79)	1.00	4.17	2.15 (0.77)	1.00	4.50	2.20 (0.76)	1.00	4.08

Note. CDI = Children’s Depression Inventory. CBCL-Agg = Child Behavior Checklist - Aggression Subscale. CBCL-Anx = Child Behavior Checklist - Anxiety Subscale. PCC = Parent-Child Communication Scale. CBQ-C = Conflict Behavior Questionnaire - Child Report. CBQ-M = Conflict Behavior Questionnaire - Mother Report. PSI-PD = Parenting Stress Index - Parental Distress Subscale. PSI-DI = Parenting Stress Index - Dysfunctional Interaction. PSI-DC = Parenting Stress Index - Difficult Child.

Table 2.

Bivariate correlations among all study variables at baseline.

	1. Group	2. Site	3. Child Age	4. Child Gender	5. CDI	6. CBCL-Agg	7. CBCL-Anx	8. PCC	9. CBQ-C	10. CBQ-M	11. PSI-PD	12. PSI-DI
1. Group	.
2. Site	0.00	.
3. Child Age	0.05	0.04	.
4. Child Gender	0.02	−0.10	0.00	.
5. CDI	−0.06	−0.05	−.21**	−0.02	.
6. CBCL-Agg	0.05	0.10	−.20**	−0.04	.24**	.
7. CBCL-Anx	−0.10	−0.07	−0.12	0.08	.31**	.52**	.
8. PCC	0.10	.178*	0.10	−0.05	−.46**	−0.10	−.17*	.
9. CBQ-C	0.12	0.08	0.14	−0.04	−.54**	−.25**	−.27**	.59**	.
10. CBQ-M	−0.11	0.03	−0.01	0.03	−.17*	−.66**	−.37**	0.13	.17*	.
11. PSI-PD	0.06	0.05	0.01	−0.05	0.08	.46**	.40**	−0.07	−0.10	−.50**	.
12. PSI-DI	0.04	0.00	0.08	0.06	.22**	.40**	.35**	−.20**	−0.15	−.59**	.54**	.
13. PSI-DC	0.05	0.03	−0.10	0.02	.23**	.65**	.46**	−0.13	−.16*	−.70**	.58**	.70**

Note. CDI = Children's Depression Inventory. CBCL-Agg = Child Behavior Checklist - Aggression Subscale. CBCL-Anx = Child Behavior Checklist - Anxiety Subscale. PCC = Parent-Child Communication Scale. CBQ-C = Conflict Behavior Questionnaire - Child Report. CBQ-M = Conflict Behavior Questionnaire - Mother Report. PSI-PD = Parenting Stress Index - Parental Distress Subscale. PSI-DI = Parenting Stress Index - Dysfunctional Interaction. PSI-DC = Parenting Stress Index - Difficult Child.

Dichotomous variables are coded as follows: Group (0 = control, 1 = intervention). Site (0 = CA, 1 = GA). Child gender (1 = female, 2 = male).

^*p < .05

^**p < .01.

A few site differences were observed. Parenting stress—specifically parental distress and difficult child subscales—at 9-month and 15-month follow-up was higher for mothers at the Georgia site compared to the California site. Children at the California site reported better parent-child communication than those at the Georgia site at the first two timepoints. Site differences were not observed for child outcomes. Within child outcomes, aggression and depressive symptoms were positively correlated at baseline. Between child outcomes and parenting, parent-child communication and relationship quality were each negatively correlated with depressive symptoms across all timepoints. However, communication was not significantly correlated with aggression, and relationship quality was negatively correlated with aggression only at baseline. For parenting stress and child outcomes, all three aspects of parenting stress were positively correlated with child functioning at all timepoints (for aggression) or some timepoints (for depressive symptoms).

Bivariate Cross-Lagged Panel Models

Model fit for all bivariate models are displayed in Table 3, and standardized parameter estimates and standard errors are presented in Table 4. Throughout the text, unless otherwise indicated, parameter estimates are displayed for cross-lagged effects between baseline and 3-month follow-up, and estimates for other timepoints are found in Table 4.

Table 3.

Model fit indices for final retained bivariate cross-lagged panel models

X with Y Bivariate Pair	Retained Model	CFI	RMSEA	χ2	df	p
PCC with CDI	X ←→ Y	0.93	0.12	40.68	12	0.00
CBQ-C with CDI	Y on X	0.84	0.13	97.53	24	0.00
CBQ-M with CDI	Autoregressive	0.92	0.10	47.88	18	0.00
PSI-PD with CDI	Autoregressive	0.87	0.11	74.41	25	0.00
PSI-DI with CDI	X ←→ Y	0.87	0.12	78.24	23	0.00
PSI-DC with CDI	Autoregressive	0.88	0.12	83.68	25	0.00
PCC with CBCL-Agg	Autoregressive	0.94	0.10	51.77	18	0.00
CBQ-C with CBCL-Agg	Autoregressive	0.91	0.10	72.32	25	0.00
CBQ-M with CBCL-Agg	Autoregressive	0.98	0.07	33.72	18	0.01
PSI-PD with CBCL-Agg	X on Y	0.88	0.12	88.70	24	0.00
PSI-DI with CBCL-Agg	Autoregressive	0.93	0.10	66.85	25	0.00
PSI-DC with CBCL-Agg	X ←→ Y	0.89	0.14	105.12	23	0.00
PCC with CBCL-Anx	Autoregressive	0.89	0.12	83.50	25	0.00
CBQ-C with CBCL-Anx	Autoregressive	0.86	0.13	98.00	25	0.00
CBQ-M with CBCL-Anx	Autoregressive	0.95	0.10	49.98	18	0.00
PSI-PD with CBCL-Anx	X on Y	0.87	0.13	98.02	24	0.00
PSI-DI with CBCL-Anx	X on Y	0.89	0.12	86.68	24	0.00
PSI-DC with CBCL-Anx	X ←→ Y	0.87	0.15	114.07	23	0.00
PSI-PD with PCC	Autoregressive	0.84	0.12	90.67	25	0.00
PSI-DI with PCC	Autoregressive	0.88	0.11	78.35	25	0.00
PSI-DC with PCC	Autoregressive	0.90	0.11	76.79	25	0.00
PSI-PD with CBQ-C	Autoregressive	0.78	0.14	112.39	25	0.00
PSI-DI with CBQ-C	Y on X	0.81	0.14	103.86	24	0.00
PSI-DC with CBQ-C	Autoregressive	0.84	0.13	102.15	25	0.00
PSI-PD with CBQ-M	X on Y	0.90	0.13	66.64	17	0.00
PSI-DI with CBQ-M	X on Y	0.92	0.12	56.62	17	0.00
PSI-DC with CBQ-M	X ←→ Y	0.95	0.11	50.33	16	0.00

Note. CDI = Children's Depression Inventory. CBCL-Agg = Child Behavior Checklist - Aggression Subscale. CBCL-Anx = Child Behavior Checklist - Anxiety Subscale. PCC = Parent-Child Communication Scale. CBQ-C = Conflict Behavior Questionnaire - Child Report. CBQ-M = Conflict Behavior Questionnaire - Mother Report. PSI-PD = Parenting Stress Index - Parental Distress Subscale. PSI-DI = Parenting Stress Index - Dysfunctional Interaction. PSI-DC = Parenting Stress Index - Difficult Child.

Autoregressive models refer to models with no retained cross-lagged paths. X on Y models refer to models with retained unidirectional cross-lagged paths with the X variable regressed on the Y variable. Y on X models refer to models with retained unidirectional cross-lagged paths with the Y variable regressed on the X variable. X ←→ Y models refer to models with retained bidirectional cross-lagged paths.

Table 4.

Standardized parameter coefficients for bivariate cross-lagged panel models.

X, Y Bivariate Pair	X1 with Y1	X2 with Y2	X3 with Y3	X4 with Y4	Y2 on X1	Y3 on X2	Y4 on X3	X2 on Y1	X3 on Y2	X4 on Y3
X ←→ Y Models
PCC, CDI	−0.45 (0.06)*	−0.24 (0.08)*	−0.14 (0.09)	−0.48 (0.08)*	−0.11 (0.04)*	−0.14 (0.05)*	−0.12 (0.04)*	−0.09 (0.04)*	−0.10 (0.05)*	−0.08 (0.04)*
PSI-DI, CDI	0.05 (0.03)	0.07 (0.04)	0.12 (0.07)†	0.08 (0.05)†	0.20 (0.07)*	0.03 (0.08)	0.24 (0.07)*	0.03 (0.07)	0.26 (0.06)*	−0.01 (0.08)
PSI-DC, CBCL-Agg	0.64 (0.05)*	0.46 (0.07)*	0.43 (0.07)*	0.34 (0.09)*	0.14 (0.05)*	0.15 (0.05)*	0.14 (0.05)*	0.11 (0.04)*	0.13 (0.05)*	0.14 (0.05)*
PSI-DC, CBQ-M	−0.70 (0.04)*	−0.53 (0.06)*	−0.40 (0.08)*	−0.59 (0.06)*	−0.10 (0.05)*	−0.12 (0.06)*	−0.11 (0.05)*	−0.11 (0.04)*	−0.12 (0.05)*	−0.12 (0.05)*
Y on X Models
CBQ-C, CDI	−0.35 (0.04)*	−0.43 (0.04)*	−0.53 (0.04)*	−0.63 (0.04)*	−0.22 (0.04)*	−0.25 (0.05)*	−0.25 (0.05)*	--	--	--
PSI-DI, CBQ-C	−0.15 (0.07)*	−0.31 (0.08)*	−0.24 (0.09)*	0.15 (0.10)	−0.11 (0.04)*	−0.12 (0.05)*	−0.12 (0.05)*	--	--	--
X on Y Models
PSI-PD, CBCL-Agg	0.46 (0.06)*	0.32 (0.08)*	0.11 (0.09)	0.21 (0.09)*	--	--	--	0.09 (0.04)†	0.09 (0.05)†	0.08 (0.04)†
PSI-PD, CBCL-Anx	0.41 (0.06)*	0.24 (0.08)*	0.28 (0.08)*	0.16 (0.10)†	--	--	--	0.14 (0.05)*	0.15 (0.05)*	0.12 (0.04)*
PSI-DI, CBCL-Anx	0.36 (0.07)*	0.23 (0.08)*	0.21 (0.09)*	0.09 (0.10)	--	--	--	0.13 (0.04)*	0.14 (0.05)*	0.13 (0.04)*
PSI-DC, CBCL-Anx	0.47 (0.06)*	0.48 (0.07)*	0.37 (0.08)*	0.07 (0.10)	--	--	--	0.12 (0.04)*	0.13 (0.04)*	0.13 (0.04)*
PSI-PD, CBQ-M	−0.50 (0.06)*	−0.48 (0.07)*	−0.12 (0.09)	−0.37 (0.08)*	--	--	--	−0.12 (0.05)*	−0.12 (0.05)*	−0.10 (0.04)*
PSI-DI, CBQ-M	−0.60 (0.05)*	−0.41 (0.07)*	−0.18 (0.09)*	−0.43 (0.08)*	--	--	--	−0.11 (0.04)*	−0.12 (0.05)*	−0.11 (0.04)*
Autoregressive Models
CBQ-M, CDI	−0.14 (0.08)†	−0.14 (0.08)†	0.03 (0.10)	0.02 (0.10)	--	--	--	--	--	--
PSI-PD, CDI	0.04 (0.03)	0.06 (0.05)	0.08 (0.06)	0.05 (0.04)	--	--	--	--	--	--
PSI-DC, CDI	0.02 (0.03)	0.04 (0.04)	0.06 (0.06)	0.05 (0.05)	--	--	--	--	--	--
PCC, CBCL-Agg	−0.10 (0.08)	−0.02 (0.09)	0.01 (0.09)	−0.16 (0.10)	--	--	--	--	--	--
CBQ-C, CBCL-Agg	−0.08 (0.03)*	−0.13 (0.04)*	−0.16 (0.05)*	−0.18 (0.06)*	--	--	--	--	--	--
CBQ-M, CBCL-Agg	−0.66 (0.04)*	−0.51 (0.07)*	−0.51 (0.07)*	−0.49 (0.08)*	--	--	--	--	--	--
PSI-DI, CBCL-Agg	0.16 (0.03)*	0.27 (0.05)*	0.32 (0.05)*	0.32 (0.05)*	--	--	--	--	--	--
PCC, CBCL-Anx	0.00 (0.02)	−0.01 (0.04)	−0.01 (0.06)	−0.01 (0.05)	--	--	--	--	--	--
PSI-PD, PCC	−0.03 (0.03)	−0.05 (0.05)	−0.05 (0.06)	−0.04 (0.04)	--	--	--	--	--	--
PSI-DI, PCC	−0.12 (0.03)*	−0.20 (0.05)*	−0.26 (0.06)*	−0.22 (0.05)*	--	--	--	--	--	--
PSI-DC, PCC	−0.08 (0.03)*	−0.12 (0.04)*	−0.17 (0.06)*	−0.17 (0.05)*	--	--	--	--	--	--
PSI-PD, CBQ-C	−0.07 (0.03)*	−0.11 (0.05)*	−0.13 (0.06)*	−0.11 (0.05)*	--	--	--	--	--	--
CBQ-C, CBCL-Anx	−0.07 (0.03)*	−0.10 (0.04)*	−0.16 (0.06)*	−0.15 (0.06)*	--	--	--	--	--	--
CBQ-M, CBCL-Anx	−0.38 (0.07)*	−0.40 (0.07)*	−0.28 (0.08)*	−0.13 (0.10)	--	--	--	--	--	--
PSI-DC, CBQ-C	−0.06 (0.03)*	−0.09 (0.04)*	−0.13 (0.06)*	−0.14 (0.06)*	--	--	--	--	--	--

Note. -- indicates nonsignificant paths that were not retained in the final model. Standardized parameter estimates displayed with standard errors in parentheses.

^*p < .05

^†p < .10.

Parenting quality and child outcomes (Hypothesis 1).

The results revealed significant bidirectional cross-lagged effects between parent-child communication and child depressive symptoms. Specifically, more communication at each timepoint predicted fewer depressive symptoms at the subsequent timepoint (β = −.11, p = .04). Similarly, more depressive symptoms at each timepoint were related to less parent-child communication at the next timepoint (β = −.09, p = .04). Covariates were not significantly related to either communication or depressive symptoms at any timepoint. A model including child-reported relationship quality and depressive symptoms revealed significant unidirectional cross-lagged effects from relationship quality to depressive symptoms. Better child-reported relationship quality predicted fewer subsequent depressive symptoms (β = −.22, p < .001). Depressive symptoms did not predict subsequent relationship quality. Covariates were not significantly related to either variable across time. There were no significant cross-lagged effects in bivariate CLPA models examining parenting quality with aggression or anxiety; parenting was not reciprocally related to either aggression or anxiety.

Parenting stress and child outcomes (Hypothesis 2).

Several cross-lagged and bidirectional effects were observed between parenting stress and child functioning. First, the model examining PSI Dysfunctional Interaction and child depressive symptoms demonstrated bidirectional effects. Children whose parents endorsed higher levels of Dysfunctional Interaction at baseline reported more depressive symptoms at 3-month follow up (β = .20, p = .004). In turn, child depressive symptoms at 3-month follow-up were associated with higher Dysfunctional Interaction scores at 9-month follow-up (β = .26, p < .001), and Dysfunctional Interaction at 9-month follow-up predicted more child depressive symptoms at 15-month follow-up (β = .24, p = .001). Covariates were not significantly related to either variable at any timepoint.

Two parenting stress subscales had significant or marginally significant cross-lagged effects with child aggression. First, models examining PSI Parental Distress (PSI-PD) with aggression revealed marginally significant unidirectional cross-lagged effects from aggression to Parental Distress (β = .09, p = .05); more aggressive behavior at one timepoint predicted marginally more Parental Distress at the next timepoint. Parental Distress did not predict subsequent aggression. The bidirectional cross-lagged effects between PSI Difficult Child (PSI-DC) and aggression were significant; higher Difficult Child scores at one timepoint predicted more aggression at the following timepoint (β = .14, p = .002), and vice versa (β = .11, p = .01) Mothers at the Georgia site reported higher Difficult Child perceptions (β = .07, p = .02), but no other covariates were associated with Difficult Child scores or child aggression.

CLPA models evaluating child anxiety with each of the three parenting stress subscales revealed significant cross-lagged effects for all three bivariate pairs. For the Parental Distress (β = .14, p = .046) and Parent-Child Dysfunctional Interaction subscales (β = .13, p = .002), child anxiety symptoms were related to significantly higher subsequent parenting stress on both subscales, but Parental Distress and Dysfunctional Interaction did not predict subsequent child anxiety. For the Difficult Child scale, the effect was bidirectional. Mothers who reported more anxiety symptoms for their children at one timepoint endorsed higher Difficult Child scores at the following timepoint (β = .12, p = .002); and those who endorsed higher Difficult Child scores also reported higher subsequent levels of child anxiety (β = .16, p < .001). In all models, child age was significantly (for Parental Distress and Dysfunctional Interactions) or marginally (for Difficult Child) related to anxiety, such that younger children exhibited more anxiety symptoms than older children. Group assignment and child gender were unrelated to all subscales of parenting stress and anxiety.

Parenting stress and parenting quality (Hypothesis 3).

Mother and child reports of relationship quality both demonstrated significant cross-lagged effects with parenting stress. For two parenting stress subscales (Parental Distress and Dysfunctional Interaction), better mother-reported relationship quality predicted less parenting stress at subsequent timepoints (Parental Distress: β = −.12, p = .011; Dysfunctional Interaction: β = −.11, p = .011). However, these two aspects of parenting stress did not predict subsequent relationship quality. For the Difficult Child subscale, mother-reported relationship quality and Difficult Child scores were mutually related to one another across time, with relationship quality predicting less Difficult Child perceptions (β = −.11, p = .013) and vice versa (β = −.10, p = .034). One aspect of parenting stress, the Dysfunctional Interaction subscale, significantly inversely predicted subsequent child-reported relationship quality (β = −.11, p = .011); however, children’s perceptions of relationship quality did not predict subsequent Dysfunctional Interaction scores. Child-reported relationship quality was not reciprocally related with other subscales of parenting stress. There were no significant cross-lagged effects between all subscales of parenting stress and parent-child communication.

Multivariate Cross-Lagged Panel Model (Exploratory)

Based on the results of bivariate CLPA models and pattern of bivariate associations, a multivariate CLPA model was tested examining bidirectional cross-lagged effects among PSI Parent-Child Dysfunctional Interaction (PSI-DI), child-reported relationship quality, and child depressive symptoms (CDI). These specific aspects of each construct were selected because they are theoretically related to one another, empirically significantly related at the bivariate level, and allow for inclusion of both mother- and child-reported variables. To reduce the likelihood of Type I error, only one multivariate model was examined as an exemplar of the transactional associations among parenting, parenting stress, and child functioning, rather than examining all possible multivariate combinations.

To test this model, an autoregressive model was first specified with no cross-lagged paths. In order to improve model fit, both first- and second-order autoregressive paths were included. No covariates (group assignment, study site, child age, or child gender) had significant associations with any primary variable at any timepoint. Next, a fully saturated transactional model was specified with all possible cross-lagged paths included between variables. Model fit for the fully saturated cross-lagged model was significantly improved compared to the autoregressive model. However, some cross-lagged paths were nonsignificant and did not contribute significantly to model fit. Nonsignificant cross-lagged paths were then removed from the model, in order to estimate the best-fitting and most parsimonious cross-lagged model. In addition to all autoregressive paths, the final model included significant cross-lagged effects from child depressive symptoms to Dysfunctional Interaction, from relationship quality to depressive symptoms, and marginally significant paths from Dysfunctional Interaction to depressive symptoms. Better child-reported relationship quality predicted less subsequent depressive symptoms (β = −.22, p = .044). In turn, lower levels of child depressive symptoms predicted lower subsequent Dysfunctional Interaction scores (β = .09, p = .041). Lower Dysfunctional Interaction scores marginally predicted subsequently lower child depressive symptoms (β = .07, p = .074). The final multivariate model is presented in Figure 2.

Figure 2.

Multivariate cross-lagged panel model displaying longitudinal relations among parenting, child depression, and parenting stress.

Note. Relationship quality measured using CBQ – Child Report. Child depression measured using CDI. Parenting Stress – Dysfunctional Interactions measured using PSI-DI.

Model fit indices: χ² = 110.36, df = 76, p = 0.01; CFI = .96; RMSEA = .05.

Standardized parameter estimates are displayed.

Four timepoints displayed: Baseline; 3-month follow-up; 9-month follow-up; 15-month follow-up.

── Solid lines indicate significant paths (*p < 0.05).

-·-·- Dashed lines indicate marginally significant paths (^†p < 0.10).

┈┈ Dotted lines indicate nonsignificant paths (p > 0.10).

Covariates, including group assignment, study site, child age, and child gender, are not depicted and had no significant associations with primary variables. For each construct, autoregressive paths with a lag of two timepoints (e.g., time 3 regressed on time 1, time 4 regressed on time 2) were estimated in the model but are not displayed due to space limitations.

Discussion

Families affected by maternal HIV face unique stressors that impact child functioning, parenting quality, and parenting stress; yet the longitudinal, bidirectional interplay among these processes has not been examined in this population. Guided by the process model of parenting (Belsky, 1984), the current study sought to understand the complex longitudinal interplay among parenting quality, parenting stress, and child functioning among a sample of 174 MLH and their 6- to 14-year-old children. MLH-child dyads completed self-report measures at four waves spanning 15 months, and data were analyzed using cross-lagged panel analysis. The findings were largely consistent with hypotheses, revealing significant cross-lagged unidirectional and bidirectional effects among parenting quality, parenting stress, and child functioning. Child depressive symptoms emerged as uniquely reciprocally associated with parenting quality, but all three aspects of child functioning (depressive symptoms, anxiety, and aggression) had bidirectional relations with aspects of parenting stress. Parenting stress and parenting quality also reciprocally predicted one another across time. Notably, in bivariate pairs with significant bidirectional relations, parent- and child-driven effect sizes were comparable. This study extends previous literature through reliance on a racially, ethnically, and geographically diverse sample of HIV-affected families. This group is considered an extremely vulnerable population in that the families are dealing not only with a chronic illness (HIV/AIDS), but also with myriad contextual challenges (e.g., poverty, stigma, stressors related to mental health outcomes, trauma and exposure to violence, etc.). Given the stressors faced by this population, understanding reciprocal parent-child interactions is necessary to identify appropriate malleable prevention and intervention targets.

The results revealed linkages between parent-child communication and child depressive symptoms across time. Children who perceived that they had strong communication with their parents were less likely to report depressive symptoms after a period of several months; and conversely, children’s depressive symptoms impeded subsequent parent-child communication. This finding is consistent with hypotheses and with prior literature suggesting associations between parent-child interactions and children’s depressive symptoms (Sander & McCarty, 2005; Yap et al., 2014). This bidirectional effect suggests that clinicians should work with parents to bolster communication skills (e.g., reflective listening) in order to enhance children’s coping and thereby prevent and reduce internalizing symptoms; and simultaneously, treatment should target children’s emotion regulation and coping skills to facilitate more effective family communication. Although not directly measured in this study, maternal depression also likely plays a role in contributing to both child depression and parent-child communication. In the presence of maternal depression or other psychopathology, aspects of parenting quality such as communication and warmth may buffer youth against the negative impact of parental psychopathology (Brennan et al., 2003). Therefore, given high rates of mental health concerns among MLH (Rochat et al., 2018), treatment approaches with these families should offer coping skills to parents in order to increase use of effective communication skills. It is important to note that in this sample, the reciprocal findings between communication and depressive symptoms both rely on child report, suggesting the possibility that the finding may be due in part to common reporter variance. Future research should replicate these findings with multi-reporter approaches.

Overall, parenting stress and child functioning were found to have significant longitudinal associations. Specific parenting stress subscales had somewhat unique associations with aspects of child functioning. The Difficult Child (DC) subscale, which taps into parents’ perceptions of the child’s difficult temperament, was reciprocally related with both anxiety and aggression; this aspect of parenting stress both contributed to and was exacerbated by children’s internalizing and externalizing difficulties. Child anxiety and aggression were both reported by mothers; thus, this finding may reflect a general negative perception of the child’s functioning and temperament. Parenting stress may also negatively influence mothers’ views of child behavior, as they perceive that they cannot keep up with the demands of meeting their child’s needs. Consistent with a call in the literature to examine parenting stress in older stages of child development (Crnic & Ross, 2017), these findings extend previous work conducted among younger children (e.g., Neece et al., 2012) by examining these reciprocal processes in middle childhood and early adolescence, suggesting that across development, parenting stress and child difficulties are mutually exacerbating.

Further, child anxiety and aggression predicted (but were not predicted by) later parenting stress within the Parental Distress (PD) and Dysfunctional Interaction (DI) subscales, suggesting over time there may be effects of children’s emotional and behavioral functioning on parents’ perceptions of themselves in the parental role, and perceptions of unpleasant interactions with their child. Examples of unpleasant interactions include perceptions that the child rarely does things that make the parent feel good, that the child does not want to be close to the parent, and that the child does not often smile at the parent. For this population, HIV-related challenges can negatively impact children’s functioning (e.g., children worrying about mothers’ well-being or perceiving stress in the home), which in turn contributes to mothers’ sense of being overwhelmed by their parenting responsibilities while managing a chronic illness. Child depressive symptoms were transactionally related with only one aspect of parenting stress: the Dysfunctional Interaction (DI) subscale, perhaps reinforcing the previously described connections between parent-child communication quality and child depressive symptoms. Children’s depressive symptoms and mothers’ negative perceptions of their interactions intensified one another across time. Depressive symptoms were reported by children; perhaps the DI subscale was the aspect of parenting stress that was most strongly related to children’s perceptions of their own functioning because it taps into dyadic interactions.

Overall, consistent with hypotheses and previous literature (e.g., Liu & Wang, 2015; Neece et al., 2012), parenting stress is worsened by caregivers’ and children’s perceptions that the child is not functioning well. In turn, parenting stress contributes to more child symptoms (e.g., depression, anxiety, aggression) across time. In this sample, parenting stress—beyond parenting quality—appears to play a central role in influencing child functioning. Parenting stress is likely compounded by the stress of managing HIV as well as contextual challenges. This pattern of findings supports the body of literature highlighting the influential role of parenting stress on child and family functioning, including among studies using similar methods (e.g., CLPA) within the general population (Neece et al., 2012), and extends these findings to a vulnerable understudied population.

Parenting stress and parent-child relationship quality evidenced significant cross-lagged associations. Mothers’ perceptions of positive relationship quality predicted lower subsequent parenting stress in all subscales; and one aspect of parenting stress—the perception that the child has a difficult temperament—reciprocally predicted worse parenting quality. For mother report of the relationship, parenting appears to drive the association with parenting stress. In contrast, mothers’ parenting stress regarding difficult interactions with the child predicted worse subsequent relationship quality from the child’s perceptions. This pattern suggests the possibility of a self-fulfilling prophesy: when mothers perceive a weaker relationship with their child, their parenting stress is subsequently higher; and higher parenting stress in turn shapes children’s perceptions that the relationship is weaker.

Descriptive analyses indicated some differences in parenting variables by site, despite demographic similarities (e.g., average age of the children, percentages of girls and boys, family socioeconomic status). Mothers in Georgia reported greater parenting stress at multiple timepoints than mothers in California, and children in California reported better parent-child relationships compared to those in Georgia. One possible explanation for these differences is the limited availability of child-focused mental health resources in Georgia, particularly relative to California, as discussed in LeCroix, Goodrum, Hufstetler, and Armistead (2017). For example, qualitative work with this sample revealed that children of MLH at the California site participated in extra-curricular activities such as after-school programs specifically designed for families affected by HIV (Goodrum et al., 2021). Expansion of these services, tailored appropriately to each community and geographic setting, may help to alleviate parenting stress and HIV-related stress for MLH who do not currently have access to these specialized services.

Limitations and Directions for Future Research

The findings of this study should be interpreted in the context of the study’s limitations. Due to the eligibility criteria of the parent study, at baseline all MLH in this sample had not disclosed their HIV status to their children. Therefore, the study findings may not be generalizable to MLH who choose to disclose without the aid of a disclosure intervention, as there may be notable differences in the characteristics and challenges faced by both groups (Goodrum et al., 2021). It is estimated that between 20 to 67% of MLH in the U.S. disclose their status to their children (Qiao et al., 2013). To address the variability in disclosure decisions in this population, future research should be conducted to replicate these findings across a broader sample of MLH who have and have not disclosed their HIV status. More information about disclosure rates in the current sample across time can be found in the primary RCT outcome paper (Schulte et al., 2021). Second, statistical power limitations precluded the ability to examine a comprehensive multivariate model including all aspects of parenting, parenting stress, and child functioning simultaneously; future research with larger samples of HIV-affected families should examine these processes simultaneously. Future research with larger samples may also be able to disentangle the role of covariates such as child age and gender, including potential moderating effects, as these effects may have been washed out to do power limitations in the multivariate model. Additionally, given evidence that financial and contextual stressors are an important contributor to family functioning for this population (Goodrum et al., 2021), future research should incorporate contextual variables (e.g., socioeconomic status, financial/employment concerns, HIV-related stigma, healthcare challenges) in addition to the family variables examined in this study. Families in the current study were predominately low-income, which in itself may be a risk factor for increased stress and negative family outcomes, highlighting the importance of future research to examine linkages among economic disadvantage, parenting stress, and child and family functioning in this population. The role of physical health and functioning of MLH on these processes also warrants further investigation, particularly given the potential impact of HIV-related health functioning on families. Further, future research should examine shared genetic influences on these family processes within this population (Paschall & Mastergeorge, 2016). Finally, given advances in using fixed effects models to examine bidirectional family interactions (Besemer et al., 2016), an important next step will be to examine within-person change in these family processes.

Implications

Despite its limitations, this study represents an important step in illuminating family processes among families affected by maternal HIV. This study is the first to examine longitudinal bidirectional associations of family processes in a sample of families affected by HIV. The findings have important implications for clinical practice. These results support the use of evidence-based behavioral parenting interventions among families affected by HIV to improve children’s emotional and behavioral functioning. For example, stronger parent-child communication skills—as perceived by children—may reduce depressive symptoms. Given significant access barriers for mental healthcare among this population (Remien et al., 2019), these interventions could be integrated into standard care in HIV clinics. Additionally, equipping mothers with skills to promote a close, interactive, and supportive parent-child relationship may prevent elevated parenting stress, which in turn can be protective for children’s perceptions of the relationship and their emotional and behavioral functioning.

Beyond promoting positive parenting quality among MLH, the study findings point to the central importance of targeting parenting stress in prevention and intervention efforts with these families. Parenting behaviors may be less effective in the context of high maternal stress (Kashdan et al., 2004), particularly for this vulnerable population. Given the direct linkages between parenting stress and child functioning in this sample, it is imperative for clinicians to routinely assess and target parenting stress to support MLH and, in turn, their children. Evidence-based interventions aimed at reducing parenting stress are needed for this population. However, particularly at the Georgia site, the availability of such interventions is very limited (LeCroix et al., 2017).

Conclusions

Families affected by HIV face a number of unique challenges which may negatively impact youth adjustment; yet, the bidirectional associations among parent and child processes have not been examined in this population. This study filled this gap by examining bidirectional, longitudinal relations among parenting quality, parenting stress, and child functioning in a sample of 174 mothers living with HIV and their children. The results revealed significant cross-lagged unidirectional and bidirectional effects among parenting quality, parenting stress, and child functioning, and were largely consistent with hypotheses. Overall, the findings highlight bidirectionality in parent-child interactions and family functioning across time. The longitudinal interplay among parenting quality, parenting stress, and child functioning is complex and multifaceted, with both child factors and parent factors driving the effects. This study highlights the ways in which these bidirectional effects can be viewed and utilized in ways to improve parent-child interactions, with a focus on reducing parenting stress and improving parent-child communication and the parent-child bond that in turn will reduce parenting stress. These findings can directly inform intervention, prevention, and policy efforts to promote the health and well-being of MLH and their children.