Child Maltreatment, Parent-Child Relationship Quality, and Parental Monitoring in Relation to Adolescent Behavior Problems: Disaggregating Between and Within Person Effects

Abstract

Background:

Parent-child relationship quality (PCRQ) and parental monitoring (PM) are associated with adolescent behavior problems following child maltreatment (CM). Whether these associations are best characterized as between (trait) or within-person (state) differences is unknown.

Objective:

Disaggregate between and within-person effects for PCRQ and PM on adolescent behavior problems and test whether these effects vary as a function of prior CM.

Participants and Setting:

Participants (n = 941) are from the Longitudinal Studies on Child Abuse and Neglect (LONGSCAN).

Methods:

Multi-level modeling was employed using PCRQ, PM, and adolescent behaviors assessed at ages 12, 14, and 16 and confirmed CM prior to age 12.

Results:

At the between-person level, adolescents with higher average levels of PCRQ and PM had significantly lower initial levels of externalizing (b = −9.47 and −5.54, respectively, p’s < 0.05; possible range 0–66) and internalizing behaviors (b = −4.45 and −6.41, respectively, p’s < 0.001; possible range 0–62). At the within-person level, greater declines in externalizing and internalizing behaviors were found when individuals reported higher-than-usual levels of PCRQ (b = −4.99 and −2.59, respectively, for externalizing and internalizing, p’s < 0.001) and PM (b = −3.58 and −1.69, respectively, for externalizing and internalizing, p’s < 0.001). There was an interaction between PM and CM on internalizing behaviors over time (b = −1.15, p = 0.026).

Conclusions:

There are between and within-person effects of PCRQ and PM on adolescent behavior problems. Adolescents with CM histories and low levels of PM may be at risk for sustained internalizing behaviors.

Introduction

Internalizing and externalizing behavior problems in adolescence are well-established adverse pediatric health consequences of child maltreatment (Hussey et al., 2006; Trickett et al., 2011). Internalizing behaviors refer to those behaviors that are most often experienced privately, such as worry and depressed mood, whereas externalizing behaviors refer to those behaviors that can most often be observed publicly, such as non-compliance and aggression (American Psychological Association, 2022). Developmental research on normative trajectories of behavior problems during adolescence suggests increases in internalizing behaviors, particularly for females, and decreases in externalizing behaviors for all youth (Bongers et al., 2003; Leve et al., 2006). In contrast, youth who have experienced maltreatment show higher and more consistently elevated levels of both internalizing and externalizing behaviors into and throughout adolescence compared to non-maltreated youth (Godinet et al., 2014; Shenk et al., 2021), and are at risk for increases in particular behavior problems, such as substance use-related consequences, across adolescence (Hagborg et al., 2020). Adolescent externalizing and internalizing behaviors also function as pathways through which child maltreatment leads to later health consequences, such as substance use (Olson et al., 2021) and suicide-related behaviors (Duprey et al., 2020).

However, not all youth who experience maltreatment go on to display the same trajectories of internalizing or externalizing behavior problems during adolescence. That is, there is substantial variability in the average level and growth in internalizing and externalizing behavior problems during adolescence following child maltreatment. For example, some adolescents exposed to or at-risk for maltreatment may experience severe and stable behavior problem trajectories over time compared to other children exposed to or at-risk for maltreatment (Figge et al., 2018). Others exposed to maltreatment may experience more variable trajectories during adolescence due to varying contextual demands or supports (Chen et al., 2017; Meng et al., 2018). Indeed, within child maltreatment samples, there has been evidence of stable trajectories of low or moderate levels of internalizing and externalizing behaviors (Lauterbach & Armour, 2016; Leban, 2021), as well as increasing and decreasing classes of internalizing and externalizing behaviors during adolescence (Figge et al., 2018; Leban, 2021). Similar variability has been found in specific behaviors across adolescence following child maltreatment, such as substance use (Yoon et al., 2020).

Research that elucidates factors that explain such variability in adolescent behavior problems following child maltreatment, across persons and time, has high potential for refining developmental theory and informing intervention. Specifically, identifying factors that explain variability in adolescent behavior problems, and use of specific analytic approaches, such as disaggregation of between and within-person effects of identified factors, is one way to better understand behavior problem variability. This approach can inform theorists or providers on whether increasing average levels of a particular factor, such as parental support, or reducing inconsistencies in the provision of that factor over time (i.e., differentiating between the effect of increasing a caregiver’s “baseline” level of support compared to the effect of when that same caregiver displays more or less support than usual), may be more important in influencing adolescent behavior problems over time following child maltreatment. The current paper addresses the issue of explaining variability in adolescent behavior problems following child maltreatment by disaggregating the between (average level) and within-person (growth or change over time) effects of two established parenting processes known to influence behavior problem trajectories for all adolescents.

Parent-Child Relationship Quality and Parental Monitoring

Parent-child relationship quality plays a critical role in child development, including during adolescence when parents and children navigate shifting social relationships (Laursen & Collins, 2009), and also buffers children and adolescents from the consequences of adversity (Chen et al., 2017). Specifically, parent-child relationship quality is the degree of closeness, understanding, trust, shared decision making, and caring that exists in the parent-child relationship (Runyan et al., 1998) and includes components such as warmth, availability, and proximity (Chen et al., 2017). Parent-child relationship quality has consistently been associated with resilience among youth who have been maltreated (Afifi & MacMillan, 2011; Meng et al., 2018). For example, parent-child relationship quality influences the association between child maltreatment and adolescent aggressive behaviors (Fagan, 2020), substance use (Dubowitz et al., 2019), and internalizing and externalizing behaviors (Steele & McKinney, 2020; Yoon et al., 2021). That is, greater parent-child relationship quality has been shown to attenuate the association between child maltreatment and alcohol and marijuana use (Dubowitz et al., 2019; Shin et al., 2019), and is associated with fewer behavior problems (Fagan, 2020; Steele & McKinney, 2020; Yoon et al., 2021).

Parental monitoring refers to “parenting behaviors involving attention to and tracking of the child’s whereabouts, activities, and adaptations” (Dishion & McMahon, 1998, p. 61). Importantly, parental monitoring is a malleable parenting characteristic that has been related to child behavior problems such as deviance and delinquency (Merrin et al., 2019; Pettit et al., 2001), such that more parental monitoring is related to less of these behaviors. Parental monitoring is also a hallmark of the authoritative parenting style which has been extensively associated with improved child outcomes and is characterized by parents who are assertive, set clear standards for child behavior, as well as provide close supervision (Baumrind, 1991). Parental monitoring is especially important during adolescence when youth increasingly spend time away from their parents and garner more independence. In the context of child maltreatment, parental monitoring has moderated the association between maltreatment and emotional distress and initiation of early sexual intercourse (Oberlander et al., 2011). Steketee et al. (2021) also found that child maltreatment is related to lower levels of parental monitoring and that parental monitoring mediated the relationship between the experience of child maltreatment and later adolescent delinquency.

Within and Between-Person Effects

While there is a substantial body of literature on the importance of parent-child relationship quality and parental monitoring in relation to adolescent behavior problems both in normative and child maltreatment samples, no study has simultaneously: 1) examined the impacts of these parenting characteristics on behavior problems over time, 2) investigated both factors simultaneously to determine which may be most important in determining outcomes, or 3) disaggregated the between versus within-person effects of these characteristics on adolescent behavior problems. Disaggregating between and within-person effects of both parent-child relationship quality and parental monitoring on adolescent behavior problems following child maltreatment can uniquely add to our understanding of how these aspects of parenting serve as protective factors on adolescent outcomes. Inferably, adolescents are not only impacted by how positive or negative their relationship is with their caregiver, or how closely their caregiver monitors their behavior on average, but also deviations from these norms. For example, a child may feel especially anxious during a period in which their relationship with their caregiver is highly conflictual, even if it is not so typically. Specifically, disaggregation of between versus within-person effects promotes understanding of not only the average associations between parent-child relationship quality and parental monitoring on adolescent externalizing and internalizing behaviors, but also elucidates the expected, concurrent changes in adolescent behavior problems due to individual deviations from the person-specific averages of these constructs (Curran & Bauer, 2011). Stated differently, between and within-person effects can be thought of as “trait” versus “state” associations respectively, such that the between-person effects represent associations between a caregiver’s average or “trait” level of parent-child relationship quality and parental monitoring, whereas within-person effects represent “state” level deviations and the impact of such deviations on adolescent behaviors. Disaggregation of these effects is an innovative opportunity to enhance specificity in our understanding of how these parenting characteristics explain variability in adolescent behavior problems over time, ultimately promoting risk and resilience with these youth.

The Current Study

The current study leveraged existing data from the Longitudinal Studies on Child Abuse and Neglect (LONGSCAN; Runyan et al., 1998), a multi-wave, prospective cohort study of child maltreatment. Specifically, the present study disaggregated the between and within-person effects of parent-child relationship quality and parental monitoring on adolescent behavior problems to highlight not only which parenting process is most closely related to adolescent behavior problems, but also the dimensions (e.g. level, variability) in which it is related. Such a test has direct implications for developmental theory and for interventions by identifying the most influential parenting process as well as the ways in which to target that process, namely, by improving overall levels or changing variability over time. There were two sets of hypotheses. One, it was hypothesized that there would be both between and within-person effects for parent-child relationship quality and parental monitoring on adolescent internalizing and externalizing behaviors. That is, adolescent behavior problems would be associated with average levels of parent-child relationship quality and parental monitoring as well as within-person deviations in these characteristics. Specifically, it was hypothesized that high levels of parent-child relationship quality and parental monitoring would be related to fewer internalizing and externalizing behavior problems. In addition, variation around an average level of these behaviors (e.g., a reduction in parent-child relationship quality from what is typical for a particular dyad) would also impact adolescent behaviors such that lower-than-average levels of these behaviors would be associated with more behavior problems, and higher-than-average levels of these behaviors would be associated with fewer behavior problems. Two, it was hypothesized that child maltreatment would interact with parent-child relationship quality and parental monitoring to further explain level and change in behaviors over time.

Methods

Sample

LONGSCAN participants (N = 1354) were recruited from five different geographic sites across the United States. Participants at each LONGSCAN site varied in their risk for child maltreatment. Specific information on the unique recruitment targets of each LONGSCAN site can be found in LONGSCAN documentation and publications (Runyan et al., 1998). The present sample is comprised of (n = 941) LONGSCAN participants. Participants from the total LONGSCAN sample were excluded from the present analyses (n = 413) if they did not provide any data on adolescent internalizing and externalizing behavior problems, parent-child relationship quality or parental monitoring at the age 12, 14, or 16 LONGSCAN assessments, or the study covariates at the age 12 LONGSCAN assessment (see Table 1). There were no significant differences between the (n = 941) LONGSCAN participants in the analytic sample and the (n = 413) LONGSCAN participants excluded in the present study according to child maltreatment status birth to age 12, child maltreatment status ages 12 to 16, child gender, child race, child age, or LONGSCAN site (p’s > .185).

Table 1.

Sample Demographics and Covariates (n = 941)

Child Age – Age 12 LONGSCAN Assessment	12.37 (0.44)
Child Age – Age 14 LONGSCAN Assessment	14.33 (0.42)
Child Age – Age 16 LONGSCAN Assessment	16.31 (0.43)
Child Sex
Female	476 (50.6%)
Child Race/Ethnicity
White	240 (25.5%)
Hispanic	62 (6.6%)
Black	514 (54.6%)
Other	125 (13.3%)
Family Income	$20,000 – $24,999
Caregiver Marital Status
Married	359 (38.2%)
Not Married	582 (61.8%)
Confirmed Child Maltreatment Birth to Age 12	473 (50.3%)
Confirmed Child Maltreatment Age 12 to 16	51 (5.4%)

Note. Child age is presented as mean age in years and corresponding standard deviations. Family income is reported as median total family income at age 12 LONGSCAN Assessment. Caregiver marital status is reported status at age 12 LONGSCAN Assessment. Percentages are reported as percentage of total analytic sample (n = 941).

Measures

Confirmed Child Maltreatment.

Confirmed child maltreatment was measured through coding of official case records via the Modified Maltreatment Classification System (MMCS; English et al., 2005). A team of independent raters reviewed official case records for all LONGSCAN participants and rated the presence of child maltreatment according to pre-defined criteria for child maltreatment. LONGSCAN MMCS coders achieved 90% inter-rater reliability and the MMCS has demonstrated improved child maltreatment detection and identification rates above and beyond official case records alone (Runyan et al., 2005). For the present investigation, two separate dummy coded variables were created, one representing confirmed child maltreatment occurring between birth and age 12 and the second representing confirmed child maltreatment occurring between ages 12 and 16. For both variables, ‘1’ represented one or more instances of any confirmed child maltreatment occurring during the respective period, while ‘0’ represented no confirmed cases of any child maltreatment during that same period. Two separate maltreatment variables were used in the models to isolate the effects of confirmed child maltreatment occurring between birth and age 12, a key moderating variable of interest, while controlling for subsequently confirmed child maltreatment on adolescent behavior problems.

Child Internalizing and Externalizing Behaviors.

Internalizing and externalizing behavior problems at the age 12, 14, and 16 LONGSCAN assessments were measured via caregiver report on The Child Behavior Checklist (Achenbach, 1991). The CBCL is comprised of 113 items where respondents report the frequency of a specific problem over the last six months using a 3-point Likert scale (0 = Not true, 1 = Somewhat or sometimes true, and 2 = Very true or often true). Raw scores were utilized for modeling in the present study given their sensitivity to change over time. The possible range of raw internalizing and externalizing scores on the CBCL is 0 – 62 and 0 – 66 respectively. In the present sample, the internal consistency of items within the externalizing problems scale ranged from α = 0.92 to α = 0.94 across the age 12, age 14, and age 16 LONGSCAN assessments, and the internal consistency of items within the internalizing problems scale ranged from α = 0.87 to α = 0.89 (Table 2).

Table 2.

Descriptive Statistics of Main Study Variables

	Maltreatment (n = 473)	Comparison (n = 468)	Total Sample (n = 941)
Age 12 LONGSCAN Assessment
Child Externalizing Behaviors	13.33 (10.07)	10.70 (9.02)	12.02 (9.65)***
Child Internalizing Behaviors	8.09 (6.85)	7.01 (6.70)	7.55 (6.79)*
Parent-Child Relationship Quality	4.34 (0.46)	4.43 (0.46)	4.38 (0.47)**
Parental Monitoring	1.80 (0.27)	1.85 (0.23)	1.82 (0.25)**
Age 14 LONGSCAN Assessment
Child Externalizing Behaviors	13.44 (10.80)	10.18 (9.21)	11.79 (10.16)***
Child Internalizing Behaviors	8.43 (7.53)	6.55 (6.76)	7.48 (7.21)***
Parent-Child Relationship Quality	4.28 (0.51)	4.39 (0.49)	4.34 (0.50)***
Parental Monitoring	1.74 (0.35)	1.81 (0.29)	1.77 (0.32)***
Age 16 LONGSCAN Assessment
Child Externalizing Behaviors	11.83 (10.86)	9.66 (9.72)	10.77 (10.37)**
Child Internalizing Behaviors	7.50 (7.66)	5.83 (6.14)	6.69 (7.00)***
Parent-Child Relationship Quality	4.16 (0.59)	4.28 (0.60)	4.22 (0.60)**
Parental Monitoring	1.66 (0.44)	1.76 (0.35)	1.70 (0.40)***

Note.

^*= p < 0.05

^**p = < 0.01

^***= p< 0.001. Child maltreatment and comparison conditions refer to those with confirmed child maltreatment between birth to age 12 and those without. Statistical tests are based on independent samples t-tests comparing child maltreatment and comparison conditions. Child behavior problems are reported as raw scores and corresponding standard deviations. Possible ranges for externalizing behaviors are 0 – 66. Possible ranges for internalizing behaviors are 0 – 62. Parent-child relationship quality and parental monitoring are presented as mean scores. Parent-Child Relationship Quality Scores ranged from 1 to 5 and parental monitoring ranged from 0 to 2 at each LONGSCAN assessment.

Parent-Child Relationship Quality.

Parent-child relationship quality at the age 12, 14, and 16 LONGSCAN assessments was measured via a caregiver-report measure adapted from the Add Health Study (Resnick, 1997). The measure was comprised of six items, including “How close do you feel to your [child]?” and “How often do you and s/he get along well?” Caregivers reported on a 5-point Likert scale from 1 = Not at all to 5 = Very Much and 1 = Never to 5 = Always, respectively. Per LONGSCAN, a quality of relationship mean score was created by taking the mean of the 6 items, which was subsequently used in statistical models. A greater mean represents greater parent-child relationship quality. The internal consistency of items constituting the Parent-Child Relationship Quality scale ranged from α = 0.69 to α = 0.80 across the age 12, age 14, and age 16 LONGSCAN assessments in the analytic sample (Table 2). Interested readers can request access on complete item content for this measure from NDACAN.

Parental Monitoring.

Parental monitoring at the age 12, 14, and 16 LONGSCAN assessments was measured via caregiver-report on 5 items where caregivers reported on the extent to which they know about their child’s friends and activities on a 3-point scale (0 = Don’t Know, 1 = Know a little, 2 = Know a lot) (Patterson & Stouthamer-Loeber, 1984). Sample items include “How much do you really know about who [child’s] friends are?” and “How much do you really know about what [child] does with his or her free time?” Per LONGSCAN, an overall parental monitoring score was computed as the mean of the five items, which was included in subsequent modeling. A greater mean represents greater levels of parental monitoring. The internal consistency of items making up the Parental Monitoring scale ranged from α = 0.61 to α = 0.81 across the age 12, age 14, and age 16 LONGSCAN assessments in the analytic sample (Table 2). Interested readers can request access on complete item content for this measure from NDACAN.

Data Analysis

The study hypotheses were tested via multi-level modeling in R, Version 4.1.0, using the nlme package (V3.1–155; Pinheiro et al., 2022). As aforementioned, (n = 413) participants from the total LONGSCAN sample (N = 1354) were excluded from the analyses if they did not provide any data on adolescent internalizing and externalizing behavior problems, parent-child relationship quality, or parental monitoring at the age 12, 14, or 16 LONGSCAN assessments, or the study covariates at the age 12 LONGSCAN assessment, leaving an analytic sample of (n = 941) for model fitting. Remaining missingness in the dependent variables (1% - 22.42%) was handled via full-information maximum likelihood estimation. Two separate models were run—one for externalizing behavior problems and one for internalizing behavior problems (see Equation 1). In each model, repeated measurements of child behavior problems (Level 1) were nested within-person (Level 2). Child sex (0 = Male, 1 = Female), child race (Hispanic, Black, and Other with White representing the reference category), and LONGSCAN site (East, Midwest, Northwest, and South with Southwest representing the reference category) were included as covariates in the models. Total family income (1=<$5,000; 2=$5,000-$9,999; 3=$10,000-$14,999; 4=$15,000-$19,999; 5=$20,000-$24,999; 6=$25,000-$29,999; 7=$30,000-$34,999; 8= $35,000-$39,999; 9=$40,000-$44,999; 10=$45,000-$49,999; 11=>$50,000) and caregiver marital status (0 = Married, 1 = Not Married) at the age 12 LONGSCAN assessment were also included as covariates. Confirmed child maltreatment experienced between ages 12 and 16 (0 = No, 1 = Yes) was included as a covariate to control for confirmed maltreatment occurring after age 12 based on research that maltreatment occurring during adolescence may be influential on adolescent outcomes (Thornberry et al., 2001). All covariates were included as time-invariant (Level 2) predictors of the intercept to control for hypothesized relations with adolescent internalizing and externalizing behavior problems. The time variable in the present models, child age, was centered such that the intercept is representative of the youngest child in the analytic sample at the age 12 LONGSCAN assessment (11.01 years). Of note, while three assessment time points were used (age 12, age 14, and age 16 LONGSCAN assessments), child age across these assessment points varied in the analytic sample between 11 to 18 years.

Centering was used to facilitate disaggregation and interpretations of the between versus within-person effects of both parent-child relationship quality and parental monitoring on adolescent behavior problems (Curran & Bauer, 2011). One, between-person effects were determined using grand-mean centering, where a single, sample average score was derived from the three assessments of parent-child relationship quality and then subtracted from each participant’s own average across these same assessments. The same procedure was then repeated for parental monitoring. These grand-mean centered scores for relationship quality and monitoring were simultaneously entered as between-person (Level 2) predictors (i.e., “trait parent-child relationship quality, “trait” parental monitoring) in the model. Two, within-person effects were determined using person-mean centering, where a single person’s average score was derived from their own three assessments of parent-child relationship quality and then this person-specific average was subtracted from their own individual scores obtained from these three assessments. The same procedure was repeated for parental monitoring. These person-centered mean scores were entered simultaneously as within-person predictors (Level 1) (i.e., “state” parent-child relationship quality, “state” parental monitoring) in the model. The proposed model is expressed as:

Level 1:

$$\begin{gathered} \mathit{A}\mathit{d}\mathit{o}\mathit{l}\mathit{e}\mathit{s}\mathit{c}\mathit{e}\mathit{n}\mathit{t}\mathit{B}\mathit{e}\mathit{h}\mathit{a}\mathit{v}\mathit{i}\mathit{o}\mathit{r}\mathit{P}\mathit{r}\mathit{o}\mathit{b}\mathit{l}\mathit{e}\mathit{m}{\mathit{s}}_{ti}={\beta }_{0i}+{\beta }_{1i}ChildAge-{11.01}_{ti}+{\gamma }_{20}StatePCR{Q}_{ti}+ \\ {\gamma }_{30}StateParentalMonitorin{g}_{ti}+e_{ti} \end{gathered}$$

Level 2:

$${\mathit{\beta }}_{\mathbf{0}\mathit{i}}={\gamma }_{00}+{\gamma }_{01}ChildGende{r}_i+{\gamma }_{02}TotalFamilyIncom{e}_i+{\gamma }_{03}CaregiverMaritalStatu{s}_i+{\gamma }_{04}Hispani{c}_i+{\gamma }_{05}Blac{k}_i+{\gamma }_{06}Othe{r}_i+{\gamma }_{07}LONGSCANEASit{e}_i\text{+}{\gamma }_{08}LONGSCANMWSit{e}_i\text{+}{\gamma }_{09}LONGSCANNWSit{e}_i+{\gamma }_{010}LONGSCANSOSit{e}_i+{\gamma }_{011}ConfirmedCM0-{12}_i+{\gamma }_{012}ConfirmedCM12-{16}_i+{\gamma }_{013}TraitPCR{Q}_i+{\gamma }_{014}TraitParentalMonitorin{g}_i+{\gamma }_{015}TraitPCRQ\ast ConfirmedCM0-{12}_i+{\gamma }_{016}TraitParentalMonitoring\ast ConfirmedCM0-{12}_i+U_{0i}$$ (1)

$${\mathit{\beta }}_{\mathbf{1}\mathit{i}}={\gamma }_{10}+{\gamma }_{11}ConfirmedCM0-{12}_i+{\gamma }_{12}TraitPCR{Q}_i+{\gamma }_{13}TraitParentalMonitorin{g}_i+{\gamma }_{14}TraitPCRQ\ast ConfirmedCM0-{12}_i+{\gamma }_{15}TraitParentalMonitoring\ast ConfirmedCM0-{12}_i+U_{1i}$$

where ${\beta }_{0i}$ represents individual i’s intercept, or initial level of problem behaviors at the age 12 LONGSCAN assessment, and was hypothesized to be explained by individual differences in gender, total family income, caregiver marital status, child race and ethnicity, LONGSCAN site, confirmed child maltreatment between birth and age 12, confirmed child maltreatment between ages 12 and 16, “trait” levels of parental monitoring and parent-child relationship quality, and the interaction of “trait” parental monitoring and parent-child relationship quality with confirmed child maltreatment between birth and age 12. In turn, ${\beta }_{1i}$represents the age-related slope in adolescent behavior problems, that is, the change in behavior problems for each additional year of age across adolescence, and was hypothesized to relate to individual differences in CM status, “trait” parent-child relationship quality and parental monitoring, and the interaction of “trait” parental monitoring and parent-child relationship quality with confirmed child maltreatment between birth and age 12. Both ${\beta }_{0i}$ and ${\beta }_{1i}$ were set to be random. That is, the intercept and slope were allowed to vary across participants; a statistically significant random effect for slope and intercept suggest that these parameters significantly differ across individuals, or that the level of behavior problems at the age 12 LONGSCAN assessment and the rate of change in behavior problems across adolescence, varies across participants. In addition, nlme (V3.1–155; Pinheiro et al., 2022) provides t-values and corresponding p-values for all fixed effects. An alpha threshold of p = 0.05 and two-tailed tests were used to determine statistical significance of model parameters. The parameter estimate, standard error, t-value, and corresponding 95% confidence interval are provided for all fixed effects (Tables 3 and 4). Normality in residuals and random effects (Schielzeth et al., 2020) was assessed using a combination of Q-Q plots, histograms, and plots of residuals versus predicted values. Post-hoc comparisons of the relative strength of parent-child relationship quality and parental monitoring on adolescent behaviors were examined by re-fitting nested variations of each model and examining associated changes in model fit (Tables 3 and 4).

Table 3.

Multi-Level Model for Externalizing Behavior Problems

Fixed Effects	b	SE	t-value	95% CI
Intercept (β0i)
Intercept (γ00)	16.64***	1.12	14.81	14.45, 18.84
Female (γ01)	−1.22**	0.45	−2.70	−2.11, −0.34
Total Family Income (γ02)	−0.20*	0.08	−2.57	−0.36, −0.05
Caregiver Marital Status (γ03)	0.44	0.52	0.84	−0.58, 1.46
Hispanic (γ04)	−1.97	1.03	−1.91	−3.98, 0.05
Black (γ05)	−0.85	0.64	−1.33	−2.10, 0.40
Other (γ06)	−0.02	0.78	−0.03	−1.55, 1.51
LONGSCAN Site: East (γ07)	−1.46	0.82	−1.78	−3.06, 0.14
LONGSCAN Site: Midwest (γ08)	−3.80***	0.77	−4.95	−5.30, −2.30
LONGSCAN Site: Northwest (γ09)	−0.99	0.71	−1.39	−2.39, 0.40
LONGSCAN Site: South (γ010)	−1.87*	0.81	−2.30	−3.46, −0.28
Confirmed CM Birth to age 12 (γ011)	0.57	0.67	0.85	−0.74, 1.88
Confirmed CM ages 12 to 16 (γ012)	2.49*	1.04	2.40	0.47, 4.52
Average PCRQ (γ013)	−9.47***	1.10	−8.62	−11.61, −7.33
Average PM (γ014)	−5.54*	2.16	−2.57	−9.75, −1.32
Average PCRQ*Confirmed CM Birth to age 12 (γ015)	−2.85	1.55	−1.84	−5.88, 0.17
Average PM*Confirmed CM Birth to age 12 (γ016)	1.73	2.72	0.64	−3.58, 7.04
Age-Based Slope (β1i)
Intercept (γ10)	−0.51***	0.10	−5.06	−0.71, −0.32
Confirmed CM Birth to age 12 (γ11)	−0.10	0.14	−0.68	−0.37, 0.18
Average PCRQ (γ12)	−0.29	0.25	−1.16	−0.78, 0.20
Average PM (γ13)	0.35	0.50	0.71	−0.62, 1.33
Average PCRQ*Confirmed CM Birth to age 12 (γ14)	0.14	0.35	0.41	−0.54, 0.83
Average PM*Confirmed CM Birth to age 12 (γ15)	−0.24	0.62	−0.39	−1.45, 0.97
Person and Time-Varying Covariates
Time-Varying PCRQ (γ20)	−4.99***	0.38	−13.12	−5.73, −4.24
Time-Varying PM (γ30)	−3.58***	0.55	−6.52	−4.65, −2.51
Random Effects
Intercept (σu0)	7.02 [6.47, 7.62]
Correlation (Intercept, Slope)	−0.47 [−0.58, −0.35]
Slope (σu1)	1.02 [0.88, 1.18]
Level 1 Residual (σe)	4.50 [4.30, 4.71]
Model Fit (Log Likelihood; LL)	−8061.48

Note.

^*= p < 0.05

^**= p < 0.01

^***= p < 0.001. PCRQ = Parent Child Relationship Quality. PM = Parental Monitoring. CM = Child Maltreatment. Total family income and caregiver marital status assessed at Age 12 LONGSCAN Assessment. LONGSCAN Site: Southwest was the reference category. Random effects are reported as standard deviations and corresponding 95% confidence intervals. To compare the relative strength of parent-child relationship quality and parental monitoring in relation to adolescent externalizing behaviors, we re-fitted nested variations of the externalizing model with subsets of the parenting characteristics omitted and examined associated changes in model fit. The associated changes in LL value for parent-child relationship quality with omission of between-person (γ₀₁₃ γ₀₁₅, γ₁₂, γ₁₄), within-person (γ₂₀), and all parameters were ΔLL = 166.21, 129.80, and 296.62 respectively. The respective changes in LL value for parental monitoring were ΔLL = 7.81, 21.15, and 28.82.

Table 4.

Multi-Level Model for Internalizing Behavior Problems

Fixed Effects	b	SE	t-value	95% CI
Intercept (β0i)
Intercept (γ00)	10.72***	0.86	12.44	9.04, 12.40
Female (γ01)	1.12**	0.34	3.27	0.45, 1.79
Total Family Income (γ02)	−0.13*	0.06	−2.16	−0.25, −0.01
Caregiver Marital Status (γ03)	0.67	0.40	1.69	−0.11, 1.44
Hispanic (γ04)	−1.84*	0.78	−2.35	−3.36, −0.31
Black (γ05)	−2.08***	0.48	−4.29	−3.03, −1.14
Other (γ06)	−1.23*	0.59	−2.07	−2.39, −0.07
LONGSCAN Site: East (γ07)	−1.17	0.62	−1.88	−2.39, 0.04
LONGSCAN Site: Midwest (γ08)	−1.92**	0.58	−3.30	−3.06, −0.78
LONGSCAN Site: Northwest (γ09)	−0.45	0.54	−0.83	−1.51, 0.61
LONGSCAN Site: South (γ010)	−0.85	0.62	−1.38	−2.06, 0.36
Confirmed CM Birth to 12 (γ011)	−0.49	0.54	−0.91	−1.55, 0.57
Confirmed CM 12 to 16 (γ012)	0.71	0.79	0.90	−0.83, 2.25
Average PCRQ (γ013)	−4.45***	0.89	−4.98	−6.20, −2.71
Average PM (γ014)	−6.41***	1.76	−3.65	−9.85, −2.98
Average PCRQ*Confirmed CM Birth to 12 (γ015)	−1.13	1.26	−0.89	−3.59, 1.33
Average PM*Confirmed CM Birth to 12 (γ016)	4.44*	2.22	2.01	0.12, 8.77
Age-Based Slope (β1i)
Intercept (γ10)	−0.47***	0.09	−5.54	−0.64, −0.31
Confirmed CM Birth to 12 (γ11)	0.16	0.12	1.36	−0.07, 0.39
Average PCRQ (γ12)	−0.09	0.21	−0.42	−0.50, 0.32
Average PM γ13)	1.16**	0.42	2.78	0.34, 1.97
Average PCRQ*Confirmed CM Birth to 12 γ14)	−0.22	0.29	−0.75	−0.79, 0.35
Average PM*Confirmed CM Birth to 12 (γ15)	−1.15*	0.52	−2.23	−2.17, −0.14
Person and Time-Varying Covariates
Time-Varying PCRQ (γ20)	−2.59***	0.32	−8.07	−3.21, −1.96
Time-Varying PM (γ30)	−1.69***	0.46	−3.66	−2.59, −0.79
Random Effects
Intercept (σu0)	5.55 [5.03, 6.13]
Correlation (Intercept, Slope)	−0.58 [−0.68, −0.46]
Slope (σu1)	0.85 [0.68, 1.07]
Level 1 Residual (σe)	3.81 [3.61, 4.01]
Model Fit (Log Likelihood; LL)	−7549.96

Note.

^*= p < 0.05

^**= p < 0.01

^***= p < 0.001. PCRQ = Parent Child Relationship Quality. PM = Parental Monitoring. CM = Child Maltreatment. Total family income and caregiver marital status assessed at Age 12 LONGSCAN Assessment. LONGSCAN Site: Southwest was the reference category. Random effects are reported as standard deviations and corresponding 95% confidence intervals. To compare the relative strength of parent-child relationship quality and parental monitoring in relation to adolescent internalizing behaviors, we re-fitted nested variations of the internalizing model with subsets of the parenting characteristics omitted and examined associated changes in model fit. The associated changes in LL value for parent-child relationship quality with omission of between-person (γ₀₁₃, γ₀₁₅, γ₁₂, γ₁₄), within-person (γ₂₀), and all parameters were ΔLL = 77.95, 80.49, and 159.86 respectively. The respective changes in LL value for parental monitoring were ΔLL = 8.65, 6.72, and 14.29.

Results

Externalizing Behavior Problems

Hypothesis 1: Within-Person Effects.

Results of the externalizing model are shown in Table 3. At the within-person level, there was a significant effect of both parent-child relationship quality (${\gamma }_{20}$ = −4.99, p < .001) and parental monitoring (${\gamma }_{30}$= −3.58, p < .001) on adolescent externalizing behavior problems (Table 3; possible range of raw externalizing behaviors 0 – 66). These results indicate that for every 1-unit increase in an individual’s parent-child relationship quality mean score, adolescent externalizing behavior problems are 4.99 points lower on average. Further, for every 1-unit increase in an individual’s parental monitoring mean score, adolescent externalizing behavior problems are 3.58 points lower.

Hypothesis 1: Between-Person Effects.

At the between-person level, there was a significant effect of parent-child relationship quality on the intercept of adolescent externalizing behavior problems (${\gamma }_{013}$ = −9.47, p < .001) (Table 3; possible range of raw externalizing behaviors 0 – 66). This between-person effect indicates that on average, those adolescents with greater parent-child relationship quality had fewer externalizing behavior problems. Specifically, across individuals, a 1-point increase in parent-child relationship quality mean score equates to a reduction of 9.47 points in adolescent externalizing behavior problems. There was also a significant between-person effect of parental monitoring on adolescent externalizing behavior problems (${\gamma }_{014}$ = −5.54, p = .011) on adolescent externalizing behavior problems (Table 3; possible range of raw externalizing behaviors 0 – 66). This between-person effect indicates that on average, those adolescents with greater parental monitoring had fewer externalizing behavior problems. Specifically, across individuals, a 1-point increase in parental monitoring mean score equates to a reduction of 5.54 points in adolescent externalizing behavior problems.

Hypothesis 2: Confirmed Child Maltreatment Birth to 12 and Interactions with Parent-Child Relationship Quality and Parental Monitoring.

The interactions between parent-child relationship quality and child maltreatment between birth and age 12 (${\gamma }_{14}$ = 0.14, p = .684), or between parental monitoring and child maltreatment birth and age 12 (${\gamma }_{15}$ = −0.24, p = .698), did not further explain variability in adolescent externalizing behavior problems over time (Table 3; possible range of raw externalizing behaviors 0 – 66). Child maltreatment status between birth and age 12 ${(\gamma }_{11}=$ −0.10, p = .494), average parent-child relationship quality ${(\gamma }_{12}=$ −0.29, p = .247), and average parental monitoring ${(\gamma }_{13}=$ 0.35, p = .477), also did not separately explain adolescent externalizing behavior problems over time (Table 3; possible range of raw externalizing behaviors 0 – 66).

Internalizing Behavior Problems

Hypothesis 1: Within-Person Effects.

Results of the internalizing model are shown in Table 4 (possible range of raw internalizing behaviors 0 – 62). At the within-person level, there was a significant effect of both parent-child relationship quality (${\gamma }_{20}$ = −2.59, p < .001) and parental monitoring (${\gamma }_{30}$ = −1.69, p < .001) on adolescent internalizing behavior problems. These results indicate that for every 1-unit increase in an individual’s parent-child relationship quality mean score, adolescent internalizing behavior problems are 2.59 points lower on average. Further, for every 1-unit increase in an individual’s parental monitoring mean score, adolescent internalizing behavior problems are 1.69 points lower.

Hypothesis 1: Between-Person Effects.

At the between-person level, there was a significant effect of parent-child relationship quality (${\gamma }_{013}$ = −4.45, p < .001) and parental monitoring (${\gamma }_{014}$= −6.41, p < .001) on the intercept of adolescent internalizing behavior problems (Table 4; possible range of raw internalizing behaviors 0 – 62). These between-person effects indicate that on average, those adolescents with greater parent-child relationship quality will have fewer internalizing behavior problems. Specifically, across individuals, a 1-point increase in parent-child relationship quality mean score equates to a reduction of 4.45 points in adolescent internalizing behavior problems. Similarly, a 1-point increase in parental monitoring equates to a reduction in 6.41 points in adolescent internalizing behavior problems.

Hypothesis 2: Confirmed Child Maltreatment Birth to 12 and Interaction with Parent-Child Relationship Quality and Parental Monitoring.

There was a significant interaction between parental monitoring and confirmed child maltreatment between birth and age 12 in predicting adolescent internalizing behavior problems over time (${\gamma }_{15}$= −1.15, p = .026) (Table 4; possible range of raw internalizing behaviors 0 – 62). This suggests that the influence of parental monitoring on adolescent internalizing behavior problems over time varies as a function of child maltreatment status between birth and age 12 (see Figure 1). Examination of Figure 1 highlights that for those adolescents who experienced child maltreatment between birth and age 12, a high level of parental monitoring was associated with consistently lower levels of internalizing behavior problems across adolescence. A similar pattern was evident for those adolescents who did not experience child maltreatment between birth and age 12, such that a low level of parental monitoring was associated with more internalizing behavior problems throughout much of adolescence. There was no significant effect of parent-child relationship quality on adolescent internalizing behavior problems over time (${\gamma }_{12}$ = −0.09, p = .663), nor a significant interaction between average parent-child relationship quality and confirmed child maltreatment in predicting internalizing behavior problems over time (${\gamma }_{14}$ = −0.22, p = .451).

Figure 1. Interaction between Adolescent Age, Confirmed Child Maltreatment and Parental Monitoring.

Child Maltreatment is confirmed child maltreatment between birth and age 12. Child age range extends beyond 12 to 16 due to varying child ages at the Age 12, Age 14, and Age 16 LONGSCAN assessments. Internalizing behavior problems are CBCL raw scores. Low and high parental monitoring refer to 1 standard deviation below and above the analytic sample mean, respectively.

Discussion

Adolescence is a particularly important time to study behavior problem trajectories following child maltreatment as pre-existing, childhood behavior problems may continue and new behavior problems such as delinquency, substance use, and problems with intimate relationships may emerge and persist well into adulthood (Trickett et al., 2011). Adolescence also represents one of the last opportunities to address behavioral concerns and to prevent the emergence of deleterious outcomes prior to adulthood. Research on adolescent behavior problems after child maltreatment is important for explaining variability in risk and resilience trajectories that may pinpoint promising intervention targets to prevent adverse health outcomes.

Literature on the associations between parent-child relationship quality, parental monitoring, and adolescent behavior problems suggests the importance of these parenting characteristics for adolescent development. For example, parental monitoring mediates the relationship between child maltreatment and later violent delinquency (Steketee et al., 2021) and greater parental monitoring is associated with less emotional distress following child maltreatment for females and less adolescent sexual intercourse (Oberlander et al., 2011). Parental monitoring is also a target in interventions seeking to prevent behavior problems and child maltreatment for those at risk (Chu et al., 2015). Parent-child relationship quality has similarly been implicated in the association between child maltreatment and adolescent internalizing and externalizing behaviors including aggression (Fagan, 2020), depression (Song et al., 2022), and psychological symptoms (Jiang et al., 2021).

The present study contributes to this growing literature in three specific ways. One, it examined parent-child relationship quality and parental monitoring simultaneously in models to examine the unique effects of each parenting factor while controlling for the other. Two, the present study disaggregated the between and within-person effects of both parenting characteristics on adolescent behavior problems. This is an especially innovative contribution to the child maltreatment literature and, to our knowledge, has not yet been done. Three, it examined the impact of confirmed child maltreatment between birth and age 12, parent-child relationship quality, and parental monitoring on adolescent behavior problems over time. All of these contributions were achieved using a large, geographically and ethnically diverse sample well-known in the child maltreatment field (LONGSCAN; Runyan et al., 1998).

Results from the first hypothesis test highlighted that there were both between and within-person effects for parent-child relationship quality and parental monitoring on adolescent internalizing and externalizing behavior problems. Parent-child relationship quality and parental monitoring each explained unique variability in adolescent behavior problems when simultaneously estimating the effects of the other parenting process. Without implying a causal role, this suggests that each parenting process is important in shaping current and future adolescent behavior problems. Results specifically demonstrated that adolescents who experience overall higher levels of parent-child relationship quality and parental monitoring have fewer internalizing and externalizing behavior problems, adding to the plethora of research on parenting styles highlighting that appropriate levels of warmth, monitoring, and parental control are predictive of positive youth outcomes (Baumrind, 1991; Dishion & McMahon, 1998). This study expands this literature by demonstrating, via within-person effects, that concurrent decreases in these problem behaviors can also be expected when parents increase their level of parent-child relationship quality or parental monitoring from their own trait levels of these characteristics. While both parenting characteristics demonstrated statistically significant between and within-person effects on adolescent internalizing and externalizing behaviors, change in model fit statistics when either parenting characteristic is or is not estimated suggest that parent-child relationship quality may play a stronger role on behavior problems when directly compared to parental monitoring. In sum, these results bode well for parenting interventions designed to address adolescent behavior problems, where improvements in both of these aspects of parenting are often key intervention targets.

The second hypothesis—that confirmed child maltreatment would interact with both parent-child relationship quality and parental monitoring to predict levels of adolescent internalizing and externalizing behaviors problems and their change over time—was partially supported. Based on our statistical models, there was evidence of a significant interaction between adolescent age, parental monitoring, and confirmed child maltreatment between birth and age 12 in the prediction of adolescent internalizing behavior trajectories. Probing of this interaction suggested that for youth who experienced confirmed child maltreatment between birth and age 12, high levels of parental monitoring served as a buffer and was associated with lower levels of internalizing behavior problems across adolescence in comparison to those youth with a child maltreatment history and low levels of parenting monitoring. Stated differently, for those with a child maltreatment history, low parental monitoring was associated with elevated trajectories of internalizing behaviors during adolescence. Notably, Figure 1 demonstrates declines in internalizing behavior problems for all adolescents. While seemingly at odds with literature in normative populations that suggests slight increases in internalizing problems during adolescence (Bongers et al., 2003; Leve et al., 2006), the range in internalizing behaviors within these samples still remains lower than that found in this child maltreatment sample, despite the observed increase and decrease in the samples respectively. That is, adolescents who have been maltreated, or who are in at-risk samples, may have consistently elevated levels of internalizing behavior problems, even with apparent declines. This significant interaction demonstrates that parental monitoring, while simultaneously accounting for the role of parent-child relationship quality, explained variability in adolescent internalizing behaviors over time following child maltreatment. Thus, enhancing parental monitoring may be useful for adolescents after maltreatment when there are concerns of internalizing behaviors.

Our statistical models showed no evidence that confirmed child maltreatment birth to 12, parental monitoring, parent-child relationship quality, nor their interactions predicted change in adolescent externalizing behavior problems over time. There was also no evidence for the impact of confirmed child maltreatment between ages birth and 12, parent-child relationship quality, nor their interaction predicting change in internalizing behavior problems. There are several potential explanations for the absence of these significant effects. Regarding child maltreatment, parental monitoring, and externalizing behaviors, it is possible that parental monitoring is more influential on level of externalizing behaviors compared to their change over time. The same could be true for parent-child relationship quality in association with both internalizing and externalizing behaviors. In addition, for internalizing behavior problems, it is possible that when considered simultaneously, parental monitoring has a stronger predictive role than parent-child relationship quality on change trajectories. Despite these non-significant interactions, it warrants repeating that there were significant between and within-person effects of parental monitoring and parent-child relationship quality on internalizing and externalizing behavior problems. Thus, these parenting characteristics were still shown to have an influence on the level of adolescent behavior problems and therefore may be promising intervention targets.

Limitations and Future Directions

Several limitations of the present study should be noted. One, the present study assessed trajectories of behavioral problems, and the time-varying predictors of parent-child relationship quality and parental monitoring, across three time-points in adolescence. Future studies should utilize more assessments, as three occasions is a fairly limited number of time-points to assess within-person change, and may limit accurate estimates of within-person variation. Two, caregiver reports of parent-child relationship quality and parental monitoring at each of the age 12, 14, and 16 assessments were relatively high (see Table 2). The potential lack of variability in these scores may have made it difficult to assess the impact of low levels of these parenting characteristics on adolescent behavior problems for adolescents with and without experiences of child maltreatment. Three, not all aspects of parenting or the parent-child relationship associated with adolescent internalizing and externalizing behavior problems were examined in the present study. Future studies may wish to explore additional aspects of parenting and the parent-child relationship that may be associated with adolescent behavior problems such as parent-child conflict (Moed et al., 2015), parenting stress (Marçal, 2022), or more discrete aspects of the parent-child relationship, such as characteristics of parent-child communication (Shenk & Fruzzetti, 2014). Four, while approximately half of the analytic sample had confirmed child maltreatment between birth and age 12, LONGSCAN participants were all initially recruited due various levels of being at-risk of child maltreatment. Therefore, both the maltreatment and comparison groups in the present study were at-risk for high levels of behavior problems, potentially making it difficult to detect differences between child maltreatment and comparison conditions. Five, data were aggregated across LONGSCAN sites, and the effect of study site was incorporated into models to account for site variability, consistent with LONGSCAN recommendations. LONGSCAN sites varied in the specific characteristics of children recruited and we refer interested readers to a detailed description of the variability in child maltreatment recruitment methods and characteristics in this cohort (Runyan et al., 1998). While not a primary research question of the present study, future studies may investigate how the results of the present study vary when considering LONGSCAN sites separately, or when considering other subsamples of children at varying risk of child maltreatment. Six, approximately 30% of LONGSCAN sample was removed from the present analytic sample due to missing data. Future studies may use other missing data handling techniques (e.g., multiple imputation) to preserve a larger sample size. Finally, future studies may wish to replicate the results of this study using a multi-informant approach (caregiver and child-report) of adolescent behavior problems, parent-child relationship quality, and parental monitoring.