Parenting and the Development of Impulse Control in Youth with Type 1 Diabetes: The Mediating Role of Negative Affect

Abstract

Parents are important for the development and maintenance of regulatory control. The current longitudinal study examined associations between parental coercion and autonomy support and impulse control in 117 youth (ages 8–16; M_age= 12.87, SD=2.53; 44% male) with Type 1 diabetes and explored whether youth negative affect mediated these associations. Parental coercion (but not autonomy support) was concurrently associated with lower impulse control and higher negative affect within individuals. Increases in youth negative affect partially mediated the within-person association between parental coercion and impulse control. These findings suggest that parent-directed interventions to reduce parental coercion may be most beneficial for impulse control if combined with youth-directed interventions to help them regulate negative affect. Replication of the current findings in a larger sample of youth with and without a chronic illness is needed to address the theoretical and empirical importance of negative affect as a potential mechanism through which parental coercion impacts youth impulsivity.

Introduction

Self-regulation refers to the multiple processes that individuals use to act and react in ways that are socially appropriate and in line with their long-term goals (Duckworth, 2011; Rothbart, Ellis, & Posner, 2004). A core component of self-regulation is impulse control, or the ability to resist the urge to act rashly (Nigg, 2017; Steinberg et al., 2008). Impulse control requires the integration of deliberate (‘top-down’) cognitive processes to suppress spontaneous, reward-driven (‘bottom-up’) responses to tempting stimuli (Nigg, 2017). Impulse control translates into important skills, such as raising a hand before speaking in class or reaching for a healthy snack instead of candy. In clinical populations, such as youth living with a chronic illness like type 1 diabetes (T1D), impulse control facilitates adherence to medications and contributes positively to health outcomes (e.g., King et al., 2012; Stupiansky et al., 2013; Silva & Miller, 2019). For example, youth with relatively higher impulse control may have less difficulty stopping an enjoyable activity to perform a health-related task, such as checking blood glucose levels or taking a medication. Impulse control may be particularly beneficial to health-related behaviors and outcomes as youth transition to adolescence and begin taking more responsibility for managing their illness (Silva & Miller, 2019). Despite its importance for illness management, no prior studies have examined developmental patterns of impulse control in youth with a chronic illness or explored how parenting contributes to changes to impulse control.

Numerous studies of youth without a chronic illness demonstrate that impulse control improves across childhood and adolescence (Harden & Tucker-Drob, 2011; Galvan, Hare, Voss, Glover, & Casey, 2007; King, Lengua, & Monahan, 2013; Leshem & Glicksohn, 2007; Steinberg et al., 2017) and that individuals vary both in levels of impulse control and the rate at which impulse control improves (e.g., Blonigen et al., 2009; Harden & Tucker-Drob, 2011; King et al., 2011; Monahan, Steinberg, Cauffman, & Mulvery, 2009, 2013). Belsky & Beaver (2011) found that parents are important contributors to individual differences in the development and maintenance of impulse control across childhood and adolescence. However, less attention has been paid to examining fluctuations in impulse control within individuals and how family context contributes to these intra-individual changes. This is an important gap in the literature because, although impulse control is known to improve with development, a person-centered approach considers how individual experiences at specific time points temporarily diminish or boost impulse control depending on the nature, and importance, of those individual experiences.

The present longitudinal study was exploratory and focused on two aspects of parenting (coercion and autonomy support) as contextual factors that may contribute to changes in impulse control within individuals. Coercive parenting is a type of intrusive, manipulative control that is characterized by hostility toward the child (Silk et al., 2003). It is manifested through overt strategies such as invalidating feelings, inducing guilt, and parental pressure on the child to comply with expectations regardless of the child’s needs and opinions. Parental coercion creates an environment in which acceptance is contingent on the youth’s behavior. In contrast, autonomy supportive parenting is empathic and non-controlling, and reflected in parents’ acceptance and encouragement of the child’s individual expression of opinions, choices, and decisions (Silk et al., 2003; Soenens et al., 2007). We first examined age-related changes in impulse control from late childhood to late adolescence. Second, we used multilevel mediation to explore the effect of parental coercion and autonomy support on impulse control both directly and indirectly through youth negative affect. The present study relied on a sample of youth (ages 8–16 at baseline) with T1D. Although these analyses did not focus on health outcomes, prior research indicates that parenting, youth affect, and impulse control are important predictors of T1D management in adolescence (e.g., Butner et al., 2017; Landers, Friedrich, Jawad, & Miller, 2016; Lansing, Berg, Butner, & Wiebe, 2016).

Parenting and Impulse Control

Much of what we know about how parenting affects the development of impulse control comes from studies in children and preadolescents (Chang et al., 2003; Colman et al., 2006; Fay-Stammbach, Howes, & Meredith, 2014; Graziano, Keane, & Calkins, 2010; Karreman et al., 2008; Lengua, Honorado, & Bush, 2007; Perry et al., 2018), with some studies suggesting that the impact of parenting on impulse control dissipates by late childhood. For example, Tiberio and colleagues (2016) found contemporaneous negative associations between harsh parenting (characterized by parents’ use of physical coercion, verbal hostility, and punitive discipline) and effortful control¹ (a construct related to impulse control; Nigg, 2017) at ages 3, 5, 7, and 11, but not at age 13. However, studies that focus exclusively on late childhood and adolescence support a negative association between harsh and coercive parenting and different aspects of self-regulation, including impulse control, in both cross-sectional (e.g., Finkenauer et al., 2005) and longitudinal samples (e.g., Brody & Ge, 2001; Eisenberg et al., 2005; King, Lengua, & Monahan, 2013; Moilanen, Rasmussen, & Padilla-Walker, 2015; Moilanen, Shaw, & Fitzpatrick, 2010; Rueth, Otterpohl, & Wild, 2017). In a longitudinal study of youth ages 8 to 12, King and colleagues (2013) found that, although impulse control improved for all youth over a three-year period, youth who reported less parental acceptance had lower starting levels of impulse control, suggesting that youth in negative family environments may transition into adolescence with a propensity to behave impulsively.

A longitudinal study based on self-report data from 9–13 year-olds found that youth who perceived their parents to be more coercive reported higher levels of impulsivity/hyperactivity (the opposite of impulse control) two years later (Rueth, Otterpohl, & Wild, 2017). Another longitudinal study of 11 to 16-year-olds found that youth with more coercive parents at baseline reported lower self-control one year later (Moilanen, Rasmussen, & Padilla-Walker, 2015). Consistent with these findings, a cross-sectional study of 10 to 14-year-olds found that more parental coercion was associated with lower youth self-control, while more parental acceptance was associated with higher self-control (Finkenauer et al., 2005). Similar associations have been reported among 18 to 24-year-olds (Moilanen & Manuel, 2017).

Notwithstanding slight variations in the way in which regulatory control is assessed across these studies, there is evidence to suggest that different ways of parenting can differentially influence youths’ ability to control impulses beyond the childhood years (Moilanen, Rasmussen, & Padilla-Walker, 2015). In fact, a related body of literature focused on adolescent problem behaviors presents evidence that different parenting styles evoke different levels of problem behaviors in adolescents, at least in part, by influencing adolescents’ ability to self-regulate (e.g., Brody & Ge, 2001; Finkanauer, Engels, & Baumeister, 2005; Loukas & Roalson, 2006; King, Lengua, & Monahan, 2013). What remains unclear are the mechanisms via which different aspects of parenting influence regulatory control across adolescence. We speculate that negative affect is one of the potential mechanisms by which parenting might be associated with changes in impulse control. This assumption is informed by social-emotional theories of familial influences on adolescent development, as well as emerging experimental research showing that brief exposure to negative emotional contexts can aggravate lapses in regulatory control among adolescents.

Parenting, Affect, and Impulse Control

The tripartite model of familial influence posits that the development of self-regulation (broadly defined) is influenced by the emotional climate of the family (Morris et al., 2007). According to this theoretical model, different aspects of parenting (e.g., coercion, hostility, physical punishment) influence the emotional climate of the family and may heighten youths’ negative affect. Given adolescents’ sensitivity to emotional information (Casey, Heller, Gee, & Cohen, 2017; Cohen et al., 2016), frequent exposure to coercive parenting may diminish youths’ sense of self, contribute to feelings of sadness and hopelessness, and render youths more likely to resort to impulsive behaviors to mitigate these negative feelings (Brody & Ge, 2001; Brenning, Soenens, van Petegem, & Vansteenkiste, 2015). Parental coercion may also hinder the development of impulse control by creating an environment that leads youth to be emotionally over-aroused, which in turn disrupts their ability to exercise control (Casey, 2015; Cohen-Gilbert & Thomas, 2013).

Indeed, an emerging body of work demonstrates that adolescents have more difficulty than adults controlling impulses in emotionally laden situations (Aite et al., 2018; Cohen et al., 2016; Somerville, Jones, & Casey, 2010). For instance, when individuals are exposed to negative or threatening emotional cues, adolescents’ ability to resist impulsive reactions is relatively compromised (Cohen et al., 2016). Some have argued that in adolescence (but not adulthood), heightened negative emotions accentuate activation of the amygdala and undermine functional activity in the prefrontal cortex (Spear, 2013), thereby increasing youths’ tendency to act impulsively (Casey & Caudle, 2013).

Since the tripartite model (Morris et al., 2017) does not explicitly address autonomy supportive family environments, we refer to self-determination theory to understand the link between autonomy supportive parenting and the development of self-regulation (Grolnick & Pomerantz, 2009). From this perspective, individuals are more likely to self-regulate and behave in ways consistent with their values when they feel cared for/accepted and supported to pursue self-chosen experiences. Autonomy supportive parents may be more likely to provide guidance to their children in challenging situations, which may help children to develop adaptive strategies to self-regulate as they pursue their goals (Brenning et al., 2015). Importantly, when parental guidance and support are conveyed in an empathic and non-imposing manner that acknowledges the child’s perspectives, the child may be more likely to internalize and endorse a parent’s problem-solving suggestion as their own, which is thought to facilitate autonomous self-regulation (Soenens & Vansteenkiste, 2005). In contrast to the negative feelings induced when a child adopts a behavior or idea out of pressure and obligation (as is typical of coercive parenting), autonomy supportive parenting is associated with lower levels of sadness and anxiety (Houtepen, Sijsema, Klimstra, Va der Lem, & Bogaerts, 2019; Liew, Kwok, Chang, Chang, & Yeh, 2014) and better regulation of negative emotions (Liew et al., 2014). Thus, autonomy supportive parenting may create an environment that is less emotionally arousing for children, thereby supporting (or at least not interfering with) the development of impulse control.

The Present Study

Using a cohort-sequential study design, the present study examined the longitudinal relationship between dimensions of parenting (e.g., coercion and autonomy support) and impulse control from ages 8 to 18 and explored a potential underlying mechanism (e.g., youth negative affect) in the association between parenting and impulse control. We focused specifically on parental coercion and autonomy support as important social contexts that may differentially influence changes in impulse control across adolescence. According to self-determination theory, autonomy supportive parenting is thought to nurture individuals’ growth in self-regulation, while coercive parenting is thought to impede the development of self-regulation (Matte-Gagne & Bernier, 2011; Grolnick & Pomerantz, 2009).

Our first study goal was to estimate changes in impulse control from late childhood to late adolescence (ages 8 to 18). Based on prior research, we hypothesized that impulse control would increase with age. Our second goal was to specify a multilevel mediation model to examine coercion and autonomy support as predictors of intra-individual variability in impulse control and explore whether within-person associations between parenting and impulse control are mediated by within-person changes in negative affect. Whereas existing research has described age-related improvements in impulse control across adolescence (e.g., Harden & Tucker-Drob, 2011; Quinn & Harden, 2013), temporary deviations from this upward trajectory may occur as a function of time-specific experiences within individuals. Modeling temporary deviations in individuals’ impulse control as a function of changes in parenting and emotional experiences would help us determine, for example, whether youth become temporarily more impulsive on occasions when they perceive their parents as more coercive than usual, and whether such temporary effects on impulse control occur, at least in part, because youth feel more distress (e.g., sad, anxious, guilty) than is typical for them. We expected that increases in coercive parenting would be related to decreases in impulse control, and that this association would be mediated by concurrent increases in youth negative affect. We also hypothesized that increases in autonomy supportive parenting would be related to increases in impulse control, and that this association would be mediated by concurrent decreases in youth negative affect.

Methods

This study was a secondary analysis of data from a study designed to examine predictors and outcomes of decision-making involvement in a sample of youth with a chronic illness (Miller & Jawad, 2019). The study used a cohort sequential design, which involves the examination of different age cohorts over the same time period and allows for the combination of multiple short-term longitudinal data points into a single long-term growth pattern using growth curve modeling (Miyazaki & Raudenbush, 2000). Nine age cohorts (8-to-16-year-olds) were assessed at baseline and followed for two years, with assessments occurring every six months (5 total assessments) from October 2011 to June 2013. With two years of data collection, the nine age cohorts were linked to form a common developmental trajectory spanning ages 8 to 18.

Participants and Procedures

Parents and youth were recruited from a tertiary children’s hospital in a large Northeastern city in the United States. Participants were eligible if they were English-speaking, the parent was the biological or adoptive parent, and youth had been diagnosed with T1D for at least one year and lived with the parent at least 50% of the week. Youth were ineligible if they had developmental delay, past-year psychiatric hospitalization, or another life-threatening medical condition, unrelated to T1D, which required daily treatment for more than 6 months in the last year. Research staff identified potential participants from outpatient clinic lists and schedules, sent their parents a letter describing the study, contacted them by telephone or in person (at clinic), and screened them for eligibility.

We identified and screened 167 families by telephone, among which 90% (n=151) were eligible for the study. Of those, 98% (n=148) agreed to participate in the study, but 7% (n=10) did not schedule a study visit, and 9% (n=14) did not show up to the study visit. One (1%) additional family declined in person at the initial study visit. The remaining 81% (n=123) of eligible dyads consented and enrolled in the study, but four (3%) families did not complete the baseline assessment, and two (1%) were withdrawn from the study because they no longer met eligibility criteria. The final sample was composed of 117 participant dyads. The number of participants in each age group (8–16) can be found on Table 1. Data for the present study were based on self-report questionnaires from youth and parents (youth provided data on parenting, affect, and impulse control, while parents provided demographic data). Research personnel read questionnaires to youth ages 8–10 years to promote comprehension. All study procedures were in accordance with U.S. guidelines for the ethical conduct of human subject research and approved by the institutional review board at the Children’s Hospital of Philadelphia.

Table 1.

Sample Characteristics (N=117)

	n	% or Mean (SD)
Age at baseline		12.87 (2.53)
Diabetes diagnosis duration (years)		5.63 (3.53)
HbA1c		9.01 (1.54)
Gender (Male)	51	43.6%
Race/Ethnicity a
Non-Hispanic White	69	58.9%
Non-Hispanic Black	25	22%
Hispanic	12	10.0%
Non-Hispanic Biracial	7	5%
Asian	1	1%
Single-Parent Household a
No	90	76.9%
Yes	23	19.7%
Household Income a
Less than $20K	19	16.2%
$20–40K	13	11.1%
$41–60K	16	13.7%
$61–80K	14	12%
$81–100K	13	11.1%
More than $100K	33	28.2%
Parent full-time employment a
Not currently employed	28	23.9%
Employed part-time	28	23.9%
Employed full-time	58	49.6%

Note: Numbers may not add up to 100% due to missing data (i.e., parent refused response).

^aGiven the accelerated cohort design, the number of participants at each age cohort varied at baseline: 8 years (n=6); 9 years (n=11); 10 years (n=18); 11 years (n=13); 12 years (n=12); 13 years (n=10); 14 years (n=16); 15 years (n=18); 16 years (n=13).

Measures

Reliability was estimated separately for each variable at each time point. We report the range of reliability coefficients and intraclass correlations (ICC) across measurement waves, which measure within-subject correlations.

Impulse Control.

At all visits, youth completed the eight-item impulse control subscale from the Weinberger Adjustment Inventory (Weinberger, 1997). Participants indicated how accurately a series of eight statements (e.g., “I stop and think things through before I act;” “I say the first thing that comes into my mind without thinking enough about it” (reverse coded); “I do things without giving them enough thought” (reverse coded)) described them on a 5-point Likert-type scale (ranging from 1=almost never to 5=almost always). Scores were averaged such that high scores indicated greater impulse control. Cronbach’s alphas across visits ranged from 0.76 to 0.84. The intraclass correlation was moderately high, ICC=0.70.

Parenting.

At all visits, youth completed the Parents as Social Context Questionnaire (PASCQ; Skinner, Johnson, & Snyder, 2009), which measures youths’ perceptions of their parents across six parenting subscales: autonomy support, coercion, structure, chaos, warmth, and rejection. Although all subscales were administered, we limit our description of this measure to the Autonomy Supportive and Coercive Parenting subscales (the focal predictors in the present study). For each subscale, participants indicated the accuracy of a series of four statements on a 4-point Likert-type scale (ranging from 1=not at all true to 4=very true). Examples of items measuring parental autonomy support include: “My mother tries to understand my point of view,” and “My mother lets me do the things I think are important.” Examples of items measuring parental coercion include: “My mother thinks there is only one right way to do things—her way;” and “My mother says ‘no’ to everything.” Youths’ responses were averaged across items to create separate scores for autonomy supportive and coercive parenting. Higher scores indicated greater perceived autonomy support or coercion from parents. Cronbach’s alphas across visits ranged from 0.61 to 0.79 for autonomy supportive parenting, and from 0.75 to 0.83 for coercive parenting. Intraclass correlations were moderately high for both dimensions of parenting, ICC=0.52 for autonomy support and ICC=0.57 for coercion.

Negative Affect.

At visits 1, 3, & 5, youth completed the child/adolescent version of the Positive and Negative Affect Schedule (PANAS; Laurent et al., 1999), a measure of positive and negative affect. The negative affect subscale consists of 15 emotions words (e.g., sad, ashamed, guilty, scared) that participants rated on a 5-point Likert-type scale (ranging from 1=not at all true to 5=extremely true) to indicate the extent to which they “felt this way during the past few weeks” for each emotion. Responses were averaged across items to create a single score, such that higher scores indicated higher negative affect. Cronbach’s alphas across visits ranged from 0.89 to 0.92. The intraclass correlation was moderately high, ICC=0.55.

Glycemic Control.

Research staff completed chart reviews of the most recent clinic visit (after study visits 1, 3, and 5 only) to obtain glycated hemoglobin (HbA1c) values for each participant. HbA1c reflects three-month average blood glucose levels and is used to assess how well the blood glucose level has been controlled over the prior 3 months. HbA1c values were measured on the day of the clinic visit by point-of-care HbA1c tests or at a lab within two months before or after the visit. Eleven participants (9%) had HbA1c testing done on a different day than the most recent clinic visit with a range of 1 to 58 days from the visit. In healthy individuals, HbA1c levels are typically below 5.7%; HbA1c values between 5.7% and 6.4% are indicative of prediabetes, and values greater than 6.5% are one of the diagnostic criterion for diabetes.

Attrition and Missing Data

Of the 117 evaluable cases at visit 1, 83% (n=97) completed surveys at visit 2, 86% (n=101) at visit 3, 83% (n=97) at visit 4, and 82% (n=96) at visit 5. Overall, 78 dyads completed all follow-up visits, 33 completed between 1 and 3 follow-up visits, and 6 did not complete any follow-up visits. There were no significant differences between participants who completed no follow-ups, 1–3 follow-ups, and all 5 follow-ups with respect to baseline age, demographics, or main variables analyzed in the present study (all ps>0.05).

The amount of missing data was low and ranged from 1% to <2.6% for each study visit. Across study visits, two youth had missing data on impulse control, and five youth had missing data on dimensions of parenting and negative affect. Statistical tests indicated that youth with missing data on the parenting measure were all males (p<0.01) and reported significantly higher parental coercion (p<0.05) and lower autonomy support (p<0.05) at baseline compared to youth without missing data. Youth with missing data on the impulse control measure reported significantly lower autonomy support compared to youth without missing data (p<0.01). No differences were found between youth with vs. without missing data on negative affect (all ps>0.05). Missing data were addressed using Full Information Maximum Likelihood estimation.

Analysis Plan

Since the current study was based on secondary data analysis, it was important to ensure we had a sufficiently large sample size to achieve 80% power in the estimation of longitudinal mediation effects. With an intraclass correlation of approximately 0.60 for the measures of interest, and a study design of five observations per subject, the estimated number of subjects required to achieve a medium-to-large mediation effect size is 100 (Pan, Liu, Miao, & Yuan, 2018). If we assume three observations per subject (since data on youth negative affect were collected at three time points only), the estimated number of subjects to achieve a similar mediation effect size is 115 (Pan, Liu, Miao, & Yuan, 2018). These sample size estimates assume a longitudinal lower-level mediation analysis that treats all path coefficients in the model (Figure 1) as fixed-effects. A study sample size of 117 youth gives us confidence in our analytic approach and ability to address our research questions using the present data.

Figure 1.

Adapted conceptual representation of a lower level mediation model (with permission from Pan. Liu, Miao, & Yuan, 2018), depicting the direct and indirect effects of parenting (X_ij) on impulse control (Y_ij) through youth negative affect (M_ij)

Note: Path a represents the fixed effect of parenting on negative affect; path b represents the fixed effect of negative affect on impulse control; path c represents the direct of parenting on impulse control, adjusted the negative affect.

Model building.

Growth curve models were first estimated to identify the best-fitting trajectory of change in impulse control from ages 8 to 18, using age as the marker of time, centered at the grand mean (13.77 years). Linear and quadratic fixed effects models were sequentially tested, as was the inclusion of random intercepts and slopes. The best-fitting fixed-effects model was determined by statistical significance of the growth parameters (e.g., intercept, linear slope, quadratic term), and fit of random effects was determined by comparing log likelihood values using a chi-square difference test. Given the cohort sequential design, models included youths’ baseline age (centered at the grand mean) to reflect cohort effects and adjust for variance in impulse control due to being younger or older at the start of the study. To account for any age-related differences in impulse control at baseline, we also regressed the growth parameters on individuals’ baseline levels of impulse control (centered at the grand mean).

After finding the best-fitting growth trajectory of impulse control, we conducted longitudinal mediation analysis using multilevel modeling in Mplus 7.1 (Muthén & Muthén, 1998–2012). All of our main variables of interest (parenting, youth negative affect, and impulse control) were time-varying (i.e., measured as repeated measures within individuals), which allowed us to specify a multilevel mediation model (Kenny, Korchmaros, & Bolger, 2003; Krull & ManKinnon, 2001; Preacher, Zyphur, & Zhang, 2010). The use of time-varying variables permits the disaggregation of within-person (level 1) from between-person (level 2) effects. Following recommendations from Zhang, Zyphur, & Preacher (2009), the predictor and mediator were person-mean centered (i.e., each individual acts as his or her own control) at level 1 so that within-person estimates are not conflated with between-person characteristics or relations. At level 2, each person’s mean score on the predictor and the mediator were grand-mean centered. Separate mediation models were run for coercive and autonomy supportive parenting. In each mediation model, we specified random intercepts and fixed slopes (net the random effect of age on impulse control). In these models, path a estimated the effect of parenting on negative affect; path b estimated the effect of negative affect on impulse control; and path c’ estimated the effect of parenting on impulse control, controlling for negative affect. The indirect effect estimated the effect of parenting on impulse control via negative affect. The total effect of parenting on impulse control (without adjusting for negative affect) was also estimated.

Results

Sample Description

Table 1 displays participant demographic data. The mean age at baseline was 12.87 (SD=2.53). Approximately 55% of youth in the sample were female, and 39% were racial or ethnic minorities. The majority of youth lived in two-parent households and had been living with T1D for approximately 5.63 years (SD=3.53). The average HbA1c for the sample at baseline was 9%, which is higher than the recommended <7.5% for youth in this age group.

Trajectory of Impulse Control

A quadratic model was the best overall fit to the data. Model statistics and parameters are summarized in Table 2. The means of the linear and quadratic factors were both positive, implying that impulse control increased gradually from late childhood through mid-adolescence with slight acceleration in the late teens (Figure 2). The estimated level of impulse control for an adolescent that is 13.77 years old was 3.72 (SE=0.08, 95% CI=3.61– 3.83) and increased 0.13 units (SE=0.03, 95% CI=0.06–0.19) with every 6-month increase in age. While the rate of change and growth curvature of impulse control did not vary between individuals, there were significant individual differences around the intercept, indicating that youth varied in the estimated starting levels of impulse control. In addition, significant variability in impulse control within individuals indicated that individuals deviated from the predicted underlying trajectory of change at specific points in time. The presence of within-person variability allowed us to examine whether fluctuations in parenting and youth affect were associated with time-specific deviations in impulse control within individuals.

Table 2.

Age-Related Changes in Impulse Control

Estimates	Model 1 Unconditional growth curve B (SE)	Model 2 +Baseline age B (SE)	Model 3 +Baseline impulse control B (SE)
Fixed Effects
Intercept	3.72 (.08)***	3.72 (.08)***	3.73 (.06)***
Baseline age	-	−.11 (.04)**	−.09 (.04)**
Baseline impulse control	-	-	.65 (.07)***
Linear slope	.04 (.02)*	.12 (.04)***	.12 (.03)***
Quadratic slope	.01 (.01)	.01 (.01)	.01 (.01)
Random Effects
Level 1: Within-person residual	.20 (.02)***	.20 (.02)***	.21 (.02)***
Level 2: Intercept	.58 (.10)***	.57 (.10)***	.22 (.05)***
Linear slope	.03 (.02)	.02 (.02)	.004 (.02)
Quadratic slope	.002 (.001)+	.002 (.001)+	.002 (.001)+
Covariance
Intercept with linear slope	.03 (.02)	.03 (.02)	−.02 (.01)
Intercept with quadratic slope	−.03 (.01)***	−.03 (.01)**	−.02 (.01)**
Linear slope with quadratic slope	−.004 (.002)*	−.003 (.002)+	.002 (.001)+
Model Fit
−2LL	948.04	941.45	795.38
BIC	978.49	974.94	831.81
df	10	11	12

Note:

⁺p<0.10;

^*p<0.05;

^**p<0.01;

^***p<0.001.

SE=standard error; −2LL=−2log-likelihood; BIC= Bayesian information criterion; df=degrees of freedom. Analyses were repeated with HbA1c (an indicator of glycemic control) as a covariate, and the pattern of results did not change.

Figure 2.

Estimated growth trajectory of impulse control from ages 8 to 18

Multilevel Mediation Models

Total, direct, and indirect effects, as well as the coefficients for each path in the mediation models are summarized in Table 3. With regard to the average within-person total effect (not accounting for negative affect) of coercive parenting on impulse control, we found that more parental coercion was associated with lower impulse control intra-individually (Level 1: total effect=−.30, SE=0.05, 95% CI: −.39, −.20, p<0.001). That is, youth impulse control was lower on occasions when youth perceived their parents to be more coercive than usual. Notably, the total effect of coercive parenting on impulse control was not observed between individuals (Level 2: total effect=−0.011, SE=0.07, p=0.09), suggesting that, on average, youth who reported more parental coercion were not necessarily more impulsive. Youth who perceived more parental coercion did, however, report higher negative affect compared to youth who perceived their parents to be less coercive (Level 2: B_a=0.27, SE=0.09, p=0.04), but this was unrelated to individual differences in impulse control.

Table 3.

Results from Multilevel Longitudinal Mediation Analyses

Parenting Predictor	B (SE)			Average effect [95% CI]
	a path	b path	c’ path	Total effect	Indirect effect
Coercion
Level 1	.27 (.04)***	−.27 (.08)***	−.22 (.05)***	−.30 [−.39, −.20]	−.07 [−.12, −.02]
Level 2	.26 (.09)**	−.07 (.07)	−.10 (.07)	−.11 [−.24, .02]	−.02 [−.06, .02]
Autonomy support
Level 1	−.11 (.06)+	−.34 (.08)***	.05 (.07)	.09 [−.05, .24]	.04 [−.01, .08]
Level 2	.15 (.11)	−.09 (.07)	.15 (.11)	.17 [−.04, .38]	−.01 [−.04, .01]

Note: Path a represents the fixed effect of parenting on negative affect (B_a); path b represents the fixed effect of negative affect on impulse control (B_b); path c’ represents the direct of parenting on impulse control, adjusted the negative affect (B_c’).

⁺p<0.10;

^**p<0.01;

^***p<0.001.

CI=confidence interval; SE=standard error. Analyses were repeated with HbA1c (an indicator of glycemic control) as a covariate, and the pattern of results did not change.

With regards to within-person mediated effects, higher parental coercion at a given time point for an individual was associated with more negative affect (Level 1: B_a =.27, SE=0.04, p<0.001; Figure 3), which in turn was linked to lower impulse control at that time point for the same individual (Level 1: B_b=−.27, SE=0.08, p=0.001). The average direct effect of coercive parenting on impulse control (after controlling for youth negative affect) remained significant (B_c’=−.22, SE=0.05, p<0.001; Figure 3), indicating that negative affect partially mediated the association between coercive parenting and impulse control within individuals (Level 1: indirect effect=−.07, SE=0.03, 95% CI: −.12, −.02; Table 3).

Figure 3.

Negative affect partially mediated the within-person effect of coercive parenting on impulse control

Autonomy supportive parenting, on the other hand, was not associated with intra-individual changes in impulse control (Level 1: total effect=.09, SE=0.07, 95% CI: −.05, .24, p=0.21; Table 3). Changes in youths’ perceptions of parental autonomy support were marginally associated with changes in negative affect within individuals (Level 1: B_a= −.11, SE=0.06, p=0.09). However, the indirect effect of autonomy support on impulse control through negative affect did not reach statistical significance (Level 1: indirect effect=0.04, SE=0.02, 95% CI: −.01, .08). Therefore, we did not find evidence that negative affect mediated the relationship between autonomy supportive parenting and impulse control.

We ran exploratory analyses with glycemic control (HbA1c) as a covariate and the pattern of results did not change. Although increases in HbA1c were related to increases in self-reported negative affect within-individuals, HbA1c was not related to impulse control and its inclusion as a covariate did not change the relationships between parenting, affect, and impulse control.

Discussion

The present longitudinal study tested the mediating role of youth negative affect in the link between two aspects of parenting (e.g., coercion and autonomy support) and impulse control in a sample of youth ages 8 to 16 with T1D. Our goals were to estimate a growth trajectory of impulse control and specify multilevel mediation models to examine longitudinal associations between coercive and autonomy supportive parenting, youth negative affect, and impulse control. This is the first study to explore age-related change in impulse control in a sample of youth with T1D and, more broadly, to examine whether youths’ experiences of negative affect help explain the effects of parental coercion and autonomy support on impulse control in this population.

With respect to the first study aim, we found that the developmental changes in impulse control followed a quadratic growth trajectory. This pattern of change is consistent with other longitudinal studies investigating changes in impulse control in non-clinical samples (e.g., Harden & Tucker-Drob, 2011; Monahan et al., 2013), with one notable difference. While prior studies have reported a gradual linear increase in impulse control that is followed by a growth deceleration or stabilization in late adolescence/young adulthood (e.g., Harden & Tucker-Drob, 2011; Quinn & Harden, 2013), the present study showed a gradual linear increase followed by a slight growth acceleration in late adolescence. This discrepancy could be attributed to differences in methodology, including study design and the age span of samples across studies. However, it is also possible that for youth with T1D, much of the maturational gain in impulse control occurs later in adolescence. Youth living with diabetes need constant medical management and may be closely monitored by adults to a greater degree (and for a longer period of time) than youth without a chronic illness (Suris, Michaud, & Viner, 2004). It is possible that the daily involvement of parents in the management of a chronic illness—especially during late childhood and early adolescence—provides fewer opportunities for youth to make independent decisions and for potential lapses or improvements in impulse control to occur (Christin, Akre, Berchtold, & Suris, 2015). As youth age and begin to manage their illness independently, the demand for impulse control may increase (Silva & Miller, 2019). An important implication of this finding is that youth with T1D may continue to need support from sources outside the family as they enter young adulthood.

With respect to the second study aim, results indicated that youth negative affect partially mediated the link between coercive parenting and impulse control within individuals. Specifically, within a given child, more perceived parental coercion at a given time point was associated with higher negative affect, which in turn was linked to lower impulse control. This finding suggests one mechanistic pathway through which coercive parenting impacts fluctuations in impulse control within individuals. Contrary to expectations, autonomy supportive parenting was not associated with impulse control, nor was it related to youth affect. This finding is partly consistent with a prior study suggesting that youth depression and anxiety may be more related to the presence of parental coercion than to the presence of parental autonomy support (Silk et al., 2003). Although some prior research indicates that autonomy supportive parenting is negatively associated with depression/anxiety symptoms in adolescents (Houtepen et al., 2019; Liew et al., 2014), the strength of these negative associations is comparatively much weaker than the positive associations between coercive parenting and depression (Houtepen et al., 2019; Rueth, Otterphl, & Wild, 2017). Moreover, while youth with more autonomy supportive parents exhibit higher levels of emotional regulation (Liew et al., 2014; Rueth, Otterphl, & Wild, 2017), some research indicates that autonomy support is not related to impulse control per se (Rueth, Otterphl, & Wild, 2017). Thus, it is possible that, in contrast with coercive parenting, autonomy support creates an environment that is not negatively arousing for youth, and thereby does not interfere with the development of impulse control.

Consistent with this thinking, one potential reason for the lack of association between autonomy supportive parenting and impulse control in the present study may be that our measure captures a person’s ability to suppress an unwanted emotion- or reward-driven response. In this respect, impulse control represents not only the cognitive processes required to behave in a goal-directed manner, but also the ability to do so in the context of social or emotional arousal. Therefore, one speculation is that autonomy support may promote deliberate thinking that helps youth behave in goal-directed ways, but these underlying cognitive processes may not be sufficient to regulate spontaneous, reward-driven reactions (Deci & Ryan, 2000; Meuwissen & Carlson, 2019). That is, parental autonomy support may enhance aspects of self-regulated behaviors that are typically carried out in neutral or non-arousing contexts, such as planning (e.g., Won & Lu, 2018), but the engagement of these processes may not be sufficient to support regulated decision-making when youth are under social or emotional arousal (an aspect of self-regulation that impulse control is thought to capture). Future research is needed to understand the ways in which autonomy support may be related to different aspects of self-regulation. Nonetheless, we acknowledge that youth in the present study with missing data on impulse control reported significantly lower autonomy support compared to youth without missing data. Therefore, our analyses could be biased to youth with higher autonomy support, and we may have been able to detect a significant association between autonomy support and impulse control if our sample had included more youth with lower autonomy support.

The present study provides initial evidence that parental coercion is related to concurrent changes in impulse control within individuals, at least in part, through its impact on youth negative affect. Our findings reinforce the contention of Darling & Steinberg (1993) that aspects of parenting set the emotional climate in which the development of self-regulation unfolds. More important, our findings raise the potential need for intervention efforts designed to help youth become aware of, and modify, their own affective responses to parental coercion as a means to mitigate the propensity to act impulsively. The ability to regulate affective responses to interpersonal stressors is a fundamental skill that supports positive development and well-being and that may precede adolescents’ ability to behave prudently. Knowing how to modify, or cope with, negative emotional responses to the surrounding environment may facilitate youths’ ability to resist acting impulsively (Rueth, Otterpohl, & Wild, 2017). In addition, interventions targeted to youth with poor diabetes adherence or glycemic control could incorporate the identification and amelioration of problematic parenting behaviors related to T1D management. Such interventions could be family-based because parents of youth with T1D may need guidance on how to promote healthy diabetes behaviors without the use of coercion.

The present study contributes to a growing body of evidence suggesting that parents—and specific aspects of parenting—continue to play an important role in the development and maintenance of regulatory control beyond childhood (e.g., Moilanen & Rambo-Hernandez, 2017). Given that different facets of parenting (e.g., support, criticism) may facilitate or hinder a wide array of youth self-regulatory characteristics relevant to health outcomes, especially in youth with chronic illness (Lansing & Berg, 2014), it may be important for healthcare providers to continue offering parenting guidance beyond infancy and childhood. Such guidance may be particularly relevant for parents of youth with T1D since one commonly reported concern is parents’ tendency to engage in parenting practices that both parents and youth perceive as intrusive and overbearing (e.g., Spencer, Cooper, & Milton, 2010; Tsiouli, Alexopoulos, Stefanaki, Darviri, & Chrousos, 2013). Given the implications of such parenting strategies for the development of impulse control, it may be critical for providers to offer parent-directed guidance as youth transition to adolescence. For instance, existing family-based intervention programs could emphasize the array of negative outcomes associated with coercive parenting, including its impact on youths’ emotional well-being and development of regulatory abilities, and offer strategies to reduce the use of coercion. It may be equally important for family-based programs to help parents identify goals for their developing youth and highlight strategies to help parents achieve those goals without the use of coercion.

Strengths and Limitations

Although the present study is strengthened by its use of longitudinal assessment and advanced statistical methodology, it is limited in several respects. First, the present study relied exclusively on self-report measures from a single reporter. Thus, we can only conclude that parenting is related to the development of impulse control as perceived and felt by the youth. Without data from other informants, we cannot assess the extent to which youth characteristics bias their own perceptions of parental behaviors, or address the extent to which youth characteristics impact parenting practices (from the parents’ perspective). For instance, adults may opt for less effective ways of parenting in response to their child’s maturational status or behavior. It would be valuable to replicate these findings using parent reports and/or observational and behavioral measures.

A second limitation of the present study is that it is based on secondary data analysis from a study designed to address different research questions in a sample of youth with chronic illness. Consequently, the current results may only be generalizable to youth living with T1D, who may have different developmental trajectories of impulse control, and for whom negative emotions and coercive and autonomy supportive parenting may have different effects, due to the demands of managing the illness on a daily basis. It will be important for future studies on this topic to include a healthy comparison group.

A third limitation is that we cannot infer causality from the concurrent associations that we observed, and it will be important for future research to consider the bi-directionality of the relationships explored in this study. Importantly, given research that depressive symptoms increase more rapidly from early to late adolescence in youth with chronic illness compared to those without (e.g., Ferro, Gorter, & Boyle, 2015), it will be relevant for future work to consider the role of depressive symptoms on youths’ perception of parenting practices. An additional limitation is that the study was not designed for multilevel mediation and was therefore not sufficiently powered for a more nuanced investigation. For instance, it is possible that youth differ in the extent to which changes in parental coercion impact their own emotional state, and youth may also differ in the extent to which shifts in negative affect impact their capacity for impulse control at a given time point. Some youth may withdraw and retreat in response to coercive parenting, while others may become highly emotional and resort to impulsive behaviors to mitigate negative affective states. Replicating the current findings in a larger sample would be a natural next step, and exploring whether the mediated effects reported here vary across individuals would be a valuable contribution to this line of research.

It will be important for future research to replicate the present findings using additional measures that tap into other aspects of self-regulation, such as emotional regulation. It is possible that the combined experience of higher negative affect and lower impulse control at specific time points translates into youths’ compromised ability to regulate emotions. Alternatively, increases in negative affect that result, at least in part, from changes in parental coercion may also provide opportunities for youth to manage or cope with such emotional arousal. As that occurs, youths’ capacity for emotional regulation may improve, and the effect of parental coercion on impulse control (at least indirectly through negative affect) may diminish. Given these potential scenarios, it is possible that coercive parenting impacts the development of different aspects of self-regulation in different ways. Using measures that capture various dimensions of self-regulation within the same individual may help elucidate how associations between parenting and youth emotionality impact maturational changes in impulsivity, emotional regulation, and other aspects of self-regulation, such as attention and delay of gratification, that are easily disrupted in emotionally-laden contexts.

Conclusion

In the current sample, impulse control increased gradually from ages 8 to 18, with a slight acceleration in late adolescence. We found significant associations between coercive (but not autonomy supportive) parenting and impulse control within individuals, such that youth reported lower impulse control on occasions when they perceived parents as more coercive than usual. Coercive parenting was also associated with higher negative affect between- and within- individuals. The present study demonstrated that changes in youth-perceived parental coercion hinders impulse control within individuals, and that this may occur, at least in part, through increases in youth negative affect. Replication of the current findings in a large sample of youth with and without a chronic illness is needed to address the theoretical and empirical importance of youth affect as a potential mechanism through which parental coercion impacts youths’ ability to control impulsive urges.