Effects of an Adolescent Depression Prevention Program on Maternal Criticisms and Positive Remarks

Abstract

This study examined effects of an adolescent depression prevention program on maternal criticisms and positive remarks, whether the extent of adolescents’ depression accounted for effects, and whether effects of the program on maternal criticisms and positive remarks differed by adolescents’ gender. Participants were 298 adolescent (M_age =14.79, SD=1.36; 59% female) offspring of mothers with histories of depression; youth were randomized to either a cognitive- behavioral prevention (CBP) program or usual care (UC). At baseline and 9-month postintervention evaluations, mothers were administered the Five-Minute Speech Sample to measure number of criticisms and positive remarks made during an open-ended description of their child and their relationship. Adolescents’ depression from pre- through post-intervention was assessed with interviews. A hierarchical generalized linear model showed a significant condition-by-gender interaction, indicating that, controlling for baseline criticism, at post-intervention mothers of girls in CBP made significantly more criticisms than did mothers of girls in UC, whereas mothers of boys in CBP made fewer criticisms than did mothers of boys in UC. The extent of adolescents’ depression from pre- through post-intervention partially mediated the relation between intervention condition and mothers’ criticisms, for boys but not for girls. Second, controlling for preintervention positive remarks, at postintervention, mothers of youth in CBP made significantly more positive remarks about their child than did mothers of youth in UC, regardless of gender; this relation was not mediated by adolescent depression from pre- through post-intervention. We suggest possible explanations for the observed effects of CBP on mothers’ criticisms and positive remarks.

Offspring of parents with depression are at increased risk of psychopathology (e.g., Beardslee, Gladstone, & O’Connor, 2011; Goodman et al., 2011). Some of this risk is likely due to exposure to negative parent behaviors such as withdrawal, disengagement, hostility, irritability, and lack of warmth (e.g., Lovejoy, Graczyk, O’Hare, & Neuman, 2000; Wilson & Durbin, 2010). In particular, mothers with depression tend to be more critical of and less positive with their children as compared to mothers without depression (e.g., Lovejoy et al., 2000). The primary aim of the current study was to evaluate the extent to which an intervention aimed at preventing depression in adolescent offspring of depressed parents affected mothers’ descriptions of their children, as indexed by their number of criticisms and number of positive remarks.

Several randomized controlled trials have shown that interventions can prevent or reduce depression in adolescent offspring of depressed parents (e.g., Clarke et al., 2001; Compas et al., 2009; Garber et al., 2009). Clarke et al. (2001) reported that a cognitive– behavioral (CB) intervention decreased depression in offspring of depressed parents. Using a similar program, Garber et al. (2009) also showed that for youth whose parents were not currently depressed, a group cognitive– behavioral prevention (CBP) program prevented depression in at-risk adolescents. Compas and colleagues (2009, 2011) demonstrated that a family group CB program reduced children’s depressive symptoms and increased observed positive parenting. Thus, these studies indicate that CB interventions significantly prevent depression in at-risk youth.

The purpose of the current study was to explore whether a CBP intervention for preventing depression in adolescent offspring of depressed parents affected mothers’ descriptions of their children, and if so, did the extent of their children’s depressive symptoms from pre- to post-intervention account for changes in mothers’ expressed attitudes about their children. That is, did a depression prevention program that taught adolescents cognitive, behavioral, emotional, communicative, and problem-solving skills affect mothers’ attitudes about their teens? We hypothesized that mothers of youth in CBP would decrease in their criticisms and increase in their positive remarks about their child as compared to mothers of youth in usual care (UC).

A second aim was to explore to what extent the relation between intervention condition and level of mothers’ criticisms or positive remarks was explained (i.e., mediated) by changes in adolescents’ depression. Previous research has shown significant associations between symptoms of adolescent depression and parents’ affect and behavior (Bodner, Kuppens, Allen, Sheeber, & Ceulemans, 2018; Kennard et al., 2008; Shortt et al., 2016). For example, Shortt and colleagues (2016) found that parents of depressed youth engaged in fewer supportive responses and more unsupportive responses as compared to parents of nondepressed youth. Similarly, Bodner et al. (2018) reported that families of depressed adolescents expressed more anger than families of nondepressed adolescents during observed interactions. Kennard et al. (2008) showed that parents’ own depressive symptoms, including negative cognitions, improved after their children’s depressive symptoms remitted. Thus, parents’ behaviors, thoughts, and responses to their children appear to vary along with their children’s depressive symptoms. We hypothesized, therefore, that changes in the levels of adolescents’ depressive symptoms might be a mechanism through which parents’ attitudes about their children change. As such, we tested whether the extent of adolescents’ depressive symptoms from pre- to post-intervention mediated the relation between intervention condition and changes in maternal criticisms and positive remarks.

One key method for assessing parents’ attitudes about their children is the Five-Minute Speech Sample (FMSS; Magaña et al., 1986). The FMSS instructs parents to talk about their child and about their relationship with the child for 5 min. Independent raters code these speech samples for several characteristics including number of criticisms, number of positive remarks, and emotional involvement. These scores yield a classification of the parent as being high or low in overall “expressed emotion” (EE), which is considered to be an important index of the family environment and parents’ attitude about their children (Hooley, 2007).

Early studies of EE were conducted with families of patients with schizophrenia, and included ratings of criticisms and emotional overinvolvement (EOI; Magaña et al., 1986). Subsequent studies have shown that “Criticism is the primary component of parental EE impacting depressed youth” (Tompson, Connor, Kemp, Langer, & Asarnow, 2015, p. 1076). Indeed, several studies have not found associations between EOI and youth depression (e.g., Asarnow, Tompson, Hamilton, Goldstein, & Guthrie, 1994; Asarnow, Tompson, Woo, & Cantwell, 2001; McCarty & Weisz, 2002). Therefore, we focused here on maternal criticisms and positive remarks.

Family environments with high levels of criticisms and low levels of positive remarks have been associated with negative treatment outcomes and a more severe clinical course for individuals with schizophrenia, depression, anxiety, bipolar disorder (BD), substance use disorders, attention deficit disorders, and eating disorders (for a review, see Wearden, Tarrier, Barrowclough, Zastowny, & Rahill, 2000). High levels of parental criticism also are linked with more observed negative parent–child interactions, lower parental responsiveness, less child attunement (Nelson, Hammen, Brennan, & Ullman, 2003; Sher Censor, 2015), greater family dysfunction and conflict (Tompson et al., 2015), and elevated externalizing and antisocial symptoms (e.g., Cussen, Sciberras, Ukoumunne, & Efron, 2012; Frye & Garber, 2005). Thus, considerable evidence exists of the deleterious outcomes associated with high levels of parental criticism and low levels of positive remarks in youth, including offspring of depressed parents.

Some interventions exist that alter parents’ EE. For example, in a study of parents of children with anxiety, a CB treatment that taught parents skills for managing their children’s anxiety also reduced parental EE (Gar & Hudson, 2009). Family-based interventions with parents of youth with BD (Kim & Miklowitz, 2004) and with youth with anorexia nervosa also improved parental EE (Moskovich, Timko, Honeycutt, Zucker, & Merwin, 2017). These studies demonstrate that EE is modifiable via interventions with intensive parenting or family components. We do not know, however, whether changes in parents’ expressions of criticism and positive remarks about their children occur in parents whose teen received an intervention aimed at teaching adolescents skills for preventing depression in youth. This study tested the effects of a CBP intervention with at-risk adolescents on mothers’ criticisms and positive remarks about their offspring.

A third aim was to examine whether the effect of the CBP intervention on mothers’ criticisms and positive remarks varied as a function of adolescents’ gender. Results of studies of differences in parents’ EE regarding their sons versus daughters are mixed. In adolescents with BD, parents were more critical of girls than boys, although parents were more critical of boys with childhood-onset BD and of girls with adolescent-onset BD (Coville, Miklowitz, Taylor, & Low, 2008). In contrast, a review of five studies of parental EE in samples of depressed parents or depressed youth did not find gender differences in parents’ EE about their offspring (Peris & Miklowitz, 2015).

Previous studies of gender differences in the effects of prevention programs on youth depression also have yielded inconsistent results (e.g., for reviews, see Brunwasser, Gillham, & Kim, 2009; Garber & Downs, 2011; Merry et al., 2012). In a comprehensive meta-analytic review of the prevention of depression literature in youth, Merry et al. (2012) concluded that, on average, the magnitude of effect of prevention programs on depression was similar for boys and girls. To date, however, no study has examined whether mothers’ expressions of criticisms and positive statements about their adolescent offspring differs for boys versus girls who have had an intervention aimed at preventing depression in at-risk youth.

Finally, elsewhere we (Garber et al., 2009) reported that the CBP program as compared to UC significantly reduced the rate of onsets of adolescents’ depressive episodes for youth whose parents were not currently depressed at the beginning of the intervention, but no differences emerged between CBP and UC for youth whose parents were depressed at intake. This finding is consistent with other intervention studies (Brent et al., 1998; Clarke et al., 2001; Lewinsohn, Rohde, & Seeley, 1998) that have shown that CB approaches are no more efficacious than UC if parents are depressed at the initiation of treatment of youth. We speculated that the observed differences in the effects of the CBP program between adolescents whose parents were versus were not currently depressed might be due to more criticisms and less positivity in parents with current depression. Therefore, the final aim of this study was to test the hypothesis that mothers who were currently depressed would make more criticisms and fewer positive remarks as compared to mothers whose past depression had remitted.

Method

Participants

The current analytic sample included 298 participants (177 females, 59%) ranging in age from 13 to 17 years old (M_age = 14.79, SD = 1.37), from 266 families (within which there were 32 sets of siblings); the 266 parents were mothers. However, the 39 fathers who participated as the target depressed parent in the outcome study did not complete the FMSS. Therefore, only mothers were included in this study. Mothers ranged in age from 29 to 58 (M_age = 43.64, SD = 5.84); 25% of families identified as a racial or ethnic minority (12.5% African American, 1% American Indian, 1% Asian, 5% reporting multiple races; 7% Latino).

Families were eligible if a parent had experienced a depressive disorder (major depressive episode [MDE] or dysthymia) in the past 3 years, three or more MDEs, or three or more cumulative years in an MDE or dysthymic episode within the child’s life. Inclusion criteria for youth were that they currently reported a score of 20 or greater on the Center for Epidemiologic Studies Depression Scale (CESD; Radloff, 1991), had experienced a prior depressive episode that was in full remission for at least the past 2 months; or both. Adolescents were excluded if (a) they or their parent had a diagnosis of bipolar 1 or schizophrenia, (b) they met Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM–IV; American Psychiatric Association, 1994) criteria for a current mood disorder diagnosis, (c) they were currently taking an antidepressant, or (d) they had received more than 8 sessions of CB therapy for depression. Approval for this study was provided by the institutional review boards of each site. Participating parents and adolescents provided written informed consent and assent, respectively. Parents and adolescents were compensated financially for participation in assessments at baseline and follow-up.

Procedure

Participants were from communities surrounding Boston, MA; Nashville, TN; Pittsburgh, PA; and Portland, OR. We recruited families through a computerized database, a university medical center listserv, letters to community physicians, letters to parents at local schools, and local advertisements (see Garber et al., 2009 for details). Adolescents were randomized to either the CBP program (n 148) or UC (n 150); we used yoked randomization to ensure that families with multiple eligible adolescents were in the same condition. Randomization was stratified on age, gender, race/ethnicity, and inclusion criteria (i.e., adolescents’ depression history or baseline CESD scores) resulting in a balanced distribution on these variables across conditions. Participants did not differ significantly by study intervention condition on any demographic (e.g., family socioeconomic status; parent education; adolescents’ age) or clinical characteristics (e.g., adolescents’ CESD scores) at study entry (Garber et al., 2009).

We conducted evaluations at baseline prior to randomization and at the completion of the intervention at 9 months. Participants (n 298) who had FMSS data at both baseline and at the postintervention evaluation were included in the current study. No variables differed significantly at baseline between those who had both baseline and postintervention evaluation data from those with only baseline data. The overall retention rate at the 9-month follow-up evaluation for all study participants was 90.5%; rates did not differ significantly between conditions (CBP 89.3% vs. UC: 91.7%; Garber et al., 2009).

Measures

Adolescent depression.

We used the CESD (Radloff, 1991) to assess adolescents’ depressive symptoms over the past week, a 20-item, self-report measure of behavioral and emotional symptoms of depression, rated on a 4-point scale. Youth completed the CESD at baseline and at the postintervention evaluation. Coefficient alphas for the CESD in the current sample were .87 at baseline and .82 at follow up. In adolescent samples, mean CESD scores typically are above 15, which is higher than mean CESD scores generally reported in adult samples (Roberts, Andrews, Lewinsohn, & Hops, 1990).

Adolescent depression also was assessed at the postintervention evaluation using the Longitudinal Interval Follow-up Evaluation (LIFE; Keller et al., 1987), which is an interview of mothers and adolescents about the teen’s depressive symptoms during the follow-up interval. The LIFE provides a continuous assessment of symptoms, onset, and offset of disorders since the last evaluation. Mothers and adolescents were interviewed separately with the LIFE; examiners derived consensus scores from this multi-informant report using guidelines similar to those used to score other diagnostic interviews such as the Kiddie-Schedule for Affective Disorders and Schizophrenia (Kaufman et al., 1997). A score (1– 6) on the Depression Symptom Rating (DSR) scale is given for each week of the follow-up period. Scores of 1 or 2 indicate zero, one, or two symptoms with no or mild impairment. A score of 3 reflects at least three symptoms with mild to moderate impairment; scores of 4 indicate at least four symptoms with mild to moderate impairment; and scores of 5 or 6 indicate that the person meets definite criteria for an MDE, with scores of 6 being more severe (e.g., hospitalized).

We used DSR scores to calculate the metric of adolescents’ “depression-free days” (DFDs), which has been used in several depression intervention trials (e.g., Brent et al., 2015; Vannoy, Arean, & Unützer, 2010). DSRs were converted to disorder weights of 0.00 (DSR=1), 0.33 (DSR=2), 0.66 (DSR=3), or 1.00 (DSR ≥ 4). Weekly disorder weights were multiplied by 7 and summed to derive the number of days with depression. Adolescents’ DFDs were calculated by subtracting the total number of days with depression from the total number of days of the follow-up for each week of the 9-month follow-up period. Scores were adjusted with a reflection procedure to create nonzero data and then log transformed to adjust for skewed data.

Maternal depressive episodes.

The Structured Clinical Interview for DSM–IV Axis I Disorders—Patient Edition (SCID-I/P; Version 2.0; First, Spitzer, Gibbon, & Williams, 1997) was administered at baseline to assess parents’ prior and current mood disorder diagnoses. All participating mothers had had a depressive episode during the child’s life; 45% of mothers were in a current depressive episode at baseline (see Garber et al., 2009 for more details about the parent assessment).

Maternal criticisms and positive remarks.

The FMSS (Magaña et al., 1986) was used to assess mothers’ criticisms and positive remarks about their children. The FMSS instructs parents to talk about their child and about their relationship with the child for 5 minutes during which the experimenter does not interact or interrupt. The instruction was to describe: “What kind of person [teen’s name] is and how you get along.” The FMSS was administered at baseline and at the 9-month evaluation. Speech samples were audio-recorded, transcribed, and coded. All samples were 5 minutes in length. Examiners stopped recording speech samples at 5 minutes; no recordings were stopped prior to 5 min. If a parent stopped talking prior to the end of 5 minutes, the experimenter repeated the instructions and continued to record.

Independent coders were unaware of intervention condition or time point rated. Statements were coded as criticisms if mothers referred to or described their adolescent with phrases such as, “I don’t like it when ____[teen’s name] … ,” “It annoys me when,” or “I resent it when … ” Examples would be “Susan is very lazy,” or “I hate it when he leaves his clothes around the house,” (Magaña et al., 1986). Positive remarks were statements that described positive behaviors or characteristics of the adolescent, such as “Sam is a kind person,” or “Rachel helps take care of her younger sister” (Magaña et al., 1986). In the current study, the primary variables derived from the FMSS were the total number of criticisms and the total number of positive remarks. Reliabilities were calculated with intraclass correlations (ICCs) on 20% of the sample rated by a second, independent coder, and yielded an ICC=0.82 for number of criticisms and an ICC=0.79 for positive remarks.

Interventions

CBP program.

The intervention was a CBP program for adolescents, comprised of eight weekly 90-min sessions and six monthly 90-min continuation sessions. Adolescents met in small groups (M=6.6 adolescents per group, SD=1.6); on average, youth attended 6.5 of the eight weekly sessions and 3.8 of the six, monthly continuation sessions. One clinician with a master’s or PhD degree in a mental health field led each group and was trained and supervised by experienced clinicians. No therapist effects were found. The CBP program focused on cognitive restructuring techniques and problem-solving skills in the first eight sessions. In continuation sessions, these skills were reviewed and another skill (e.g., relaxation, assertiveness, behavioral activation) was taught. In Weeks 1 and 8 of weekly sessions, parents attended meetings during which another therapist provided the rationale and description of skills being taught to their children; 82% of the target parents attended at least one parent session. No other direct intervention was provided to parents.

Usual Care.

All families, regardless of condition, were permitted to receive other mental health services after randomization. Although youth who currently were receiving antidepressant medications for depression or had cognitive-behavioral therapy (8+ sessions) prior to randomization were excluded, youth in both conditions could seek any treatment after randomization. Elsewhere, we (Garber et al., 2009; Lynch et al., 2019) reported that conditions did not differ significantly in the extent or type of mental health services received prior to randomization (9.6% UC vs. 16.4% CBP, p=.13) or during the 9 months of the intervention period (29.3% UC vs. 30.8% CBP, p=.74).

Data Analysis Plan

We conducted separate models to examine differences between CBP and UC in the number of maternal criticisms and number of positive remarks made at the 9-month evaluation, controlling for baseline criticisms and positive remarks, respectively. To examine gender differences, we included a Gender X Condition interaction. Analyses controlled for either mothers’ criticisms or positive remarks at baseline, adolescents’ level of depressive symptoms at baseline and at follow up, adolescents’ age, and gender; predictors were grand mean centered. Given the nested structure of the data (i.e., youth nested within families), multilevel modeling was used to account for interdependency. Because the outcomes were count variables that follow a Poisson distribution, we ran hierarchical generalized linear models, which use the log link function. (See the online supplemental materials for a further discussion of the model.)

Results

Descriptive Statistics

Table 1 presents the means, standard deviations, and correlations for all study variables. On average, mothers made 0.47 criticisms (SD=1.05, range 0 –7) and 3.53 positive remarks (SD=3.07, range 0 –17) at baseline. There were no gender or condition differences and no Gender X Condition interactions at baseline for either maternal criticisms or positive remarks. At postintervention, on average, mothers made 0.23 criticisms (SD = 0.71, range 0 – 6) and 3.01 positive remarks (SD=2.92, range 0 –7). The number of positive remarks and criticisms were negatively correlated at baseline, r(297)= −0.18, p <.001, and at follow-up, r(297) −0.15, p .021. The correlation between number of maternal criticisms and adolescents’ CESD was not significant, r(297) =0.09, p=.083. The correlation between number of positive remarks and adolescents’ CESD scores was significant and negative at baseline, r(297) = −0.11, p =.049, such that a higher number of maternal positive remarks was associated with a lower level of adolescent depressive symptoms at baseline. Intervention conditions did not differ significantly in maternal depressive symptoms or disorder at baseline, t(297) = 0.68, p=.494.

Table 1.

Means, Standard Deviations, and Correlations Among Study Variables

Study variables	M (SD)	Age	Gender	Parent depressed T1	Condition	T1 CESD	T2 CESD	T1 criticisms	T2 criticisms	T1 Positive remarks	T2 Positive remarks	Adolescent DFD
Age
Gender	14.80 (1.36)		.073	.085	−.005	−.087	−.044	−.086	−.088	.027	.016	.143*
Mother dep at T1				−.004	.002	.057	.141*	− .003	.023	−.038	.035	−.072
Condition					.021	.025	.080	.060	.069	−.083	−.042	−.126*
T1 CESD	15.67 (9.71)					−.041	−.119*	− .070	.019	−.044	.113	.123*
T2 CESD	12.02 (8.46)						.465**	.091	−.006	−.113*	−.091	−.266***
T1 criticisms	0.47 (1.05)							.051	.094	−.097	−.085	−.279***
T2 criticisms	0.23 (0.71)								.417**	−.183**	−.136*	−.277***
T1 positive remarks	3.53 (3.07)									−.101	−.147*	−.200**
T2 positive remarks	3.01 (2.92)										−.308**	.171**
Adolescent depression-free days (DFD)	237.12 (55.03)											.151*

Note. T1 =baseline; T2 = post-intervention; Condition: usual care = 0, cognitive behavior prevention program = 1; CESD = Center for Epidemiological Studies Depression Scale; DFD = depression-free days

^*p < .05.

^**p < .01.

^***p < .001.

Effect of CBP versus UC on Mothers’ Criticisms at Post-intervention

Results of the model examining the effect of the intervention on number of criticisms mothers made about their adolescents at the postintervention evaluation, controlling for their baseline criticisms, revealed a significant main effect for intervention condition, and a significant Gender X Condition interaction (Table 2). Mothers of girls in the CBP program made significantly more criticisms as compared to mothers of girls in UC, t=2.20, p=.03. Mothers of boys in CBP made fewer criticisms as compared to mothers of boys in UC; this difference was not significant (t =−1.69, p<.09; Table 3 and Figure 1).

Table 2.

Hierarchical Generalized Model Results Predicting Positive Remarks and Parents’ Criticisms at the Post-intervention Evaluation

Outcome variables	Coefficient	SE	T ratio	p value
Outcome: Positive remarks
Age	0.04	0.04	0.99	0.34
Gender	0.09	0.13	0.71	0.49
Parents’ baseline positive remarks	0.06	0.02	3.27	0.005
Adolescent CESD at baseline	−0.005	0.01	−0.60	0.56
Adolescent CESD at 9 months	0.003	0.01	0.27	0.79
Condition (UC vs. CBP)	0.27	0.12	2.23	0.03
Outcome: Criticism
Age	−0.18	0.08	−2.15	0.04
Gender	−0.01	0.29	−0.04	0.97
Parents’ baseline criticism	0.54	0.06	9.43	0.001
Adolescent CESD at baseline	−0.05	0.02	−3.05	0.008
Adolescent CESD at 9 months	0.06	0.02	3.57	0.003
Condition (UC vs. CBP)	0.68	0.28	2.41	0.017
Gender X Condition	1.69	0.58	2.93	0.003

Note. Coefficients are log link of original scale.

CBP = cognitive behavior prevention program; UC=usual care; CESD =Center for Epidemiological Studies Depression

Table 3.

Expected Number of Maternal Criticisms at the Post-intervention Evaluation by Condition and Gender

Group	Expected log rate	Expected Number of Criticisms	95% Confidence Interval
Girls in UC group	−2.36	0.09	[.06, .16]
Girls in CBP group	−1.32	0.27	[.17, .41]
Boys in UC group	−1.81	0.16	[.09, .31]
Boys in CBP group	−3.23	0.04	[.01, .28]

Note. CBP = cognitive behavior prevention program; UC = usual care

Figure 1.

Number of criticisms expressed by parents at the pre- and post-intervention evaluations, as a function of intervention condition and adolescents’ gender. CBP = cognitive behavior prevention program; UC = usual care. See the online article for the color version of this figure.

We also examined gender differences within each condition. At the postintervention evaluation, there was a nonsignificant trend for mothers of youth in UC to make more criticisms of boys than of girls, t=1.95, p=.0525, whereas for youth in the CBP program, the nonsignificant trend was for mothers to make more criticisms of girls than of boys (t= −1.94, p=.054). Overall, the significant Condition X Gender interaction for maternal criticism primarily was due to differences in maternal criticisms of girls in the CBP program versus UC. In the model with the significant Gender X Condition interaction term (see Table 3), adolescent age, baseline and concurrent adolescent depressive symptoms, baseline maternal criticisms, and condition also significantly predicted maternal criticisms at post intervention.

Effect of CBP versus UC on Mothers’ Positive Remarks at Post-intervention

Results of the model examining the effects of the CBP intervention on mothers’ positive remarks about their adolescent at the postintervention evaluation revealed a significant main effect for condition (see Table 2). At postintervention, mothers of adolescents in the CBP program made significantly more positive remarks about their adolescents (3.34 remarks, 95% CI [2.83, 3.85]) as compared to mothers of adolescents in UC (2.68 remarks, 95% CI [2.14, 3.22]), controlling for number of mothers’ positive remarks at baseline. Adolescents’ age, gender, and baseline and concurrent depressive symptoms were not significant predictors of mothers’ positive remarks. The Condition X Gender interaction was not significant.

Does the Amount of Depression in the Adolescents from Pre- to Post-intervention Mediate the Effect of Condition on Maternal Criticisms or Positive Remarks?

Maternal criticism.

To test the extent to which the amount of adolescent depression, measured in terms of the number of adolescent DFDs between baseline and postintervention (9 months), accounted for the relation between intervention condition and maternal criticism, and whether this effect varied by gender, we conducted a moderated-mediation model, evaluated as a structural equation model in Mplus (Version 7.4; Muthén & Muthén, 1998– 2012). Confidence intervals for indirect effects were calculated using the distribution of the product method (MacKinnon, Lockwood, & Williams, 2004) using the RMediation package (Version 1.1.4; Tofighi & MacKinnon, 2011) in R (Version 3.5.1; R Core Team, 2013). As there was evidence of overdispersion in the maternal criticism variable, we used a negative binomial model for all regressions in which maternal criticism was the outcome variable. To accommodate the nonindependence of data points due to the clustering of adolescents within families, we used the robust maximum likelihood estimator (MLR in Mplus), which calculates standard errors with a sandwich estimator (Muthén & Muthén, 1998–2012). We relied on the missing at random assumption, which presumes that missingness can be accounted for by non-missing, observed data.

The final model was based on data from 298 adolescents who had complete data for the exogenous variables (i.e., model covariates). The traditional chi-square test of model fit and approximate fit statistics (e.g., root-mean-square error of approximation) are not available in models with count outcomes. The regression of DFDs on CBP (Path a) was positive and significant when adjusting for adolescent age, sex, and baseline CESD score: b 15.43, 95% CI [4.037, 26.83]. For Path b, the regression of maternal criticism on adolescent DFDs (adjusting for adolescent age, and baseline CESD and maternal criticism), there was evidence of moderation by adolescent sex: b=0.03, 95% CI [0.01, 0.04]. Among male adolescents, the estimated effect of DFDs on maternal criticism was negative and significant: b=−0.03, 95% CI [−0.04, −0.01]. Among female adolescents, there was no evidence of an effect of DFDs on maternal criticism: b=−0.001, 95% CI [−0.008, 0.005]. Finally, there was evidence of a negative indirect effect of CBP on maternal criticism though DFDs among male (a × b=−0.43, 95% CI [−0.87, −0.10]) but not female adolescents (a × b=0.02, 95% CI [−0.12, 0.08]). Overall, the moderated-mediation analyses revealed that youth in the CBP program had more DFDs than youth in UC; for boys, a higher number of DFDs then predicted lower levels of maternal criticism at postintervention (see Figure 2).

Figure 2.

The test of moderated mediation was significant. The relation between condition and maternal criticism at the postintervention evaluation (9 months) was partially mediated by adolescent depression-free days for boys but not for girls, controlling for maternal criticism at baseline. CBP = cognitive behavior prevention program; UC = usual care. **p < .01.

Maternal positive remarks.

We conducted a second analysis to determine whether adolescents’ DFDs mediated the relation between condition and maternal positive remarks at the post-intervention evaluation. Analyses controlled for adolescents’ depressive symptoms at baseline, maternal positive remarks at baseline, adolescent age, and gender. The regression of DFDs on CBP (Path a) was positive and significant when adjusting for adolescent age, sex, and baseline CESD score: b=15.43, 95% CI [4.037, 26.83]. For Path b, the regression of maternal positive remarks on adolescent DFDs (adjusting for adolescent age, and baseline CESD and maternal positive remarks), there was no evidence of moderation by adolescent sex: b=0.007, 95% CI [−0.02, 0.01]. Among male adolescents, the estimated effect of DFDs on maternal positive remarks was not significant: b=0.007, 95% CI [−0.02,0.01]. Among female adolescents, there was no evidence of an effect of DFDs on maternal positive remarks: b= 0.001, 95% CI [−0.003, 0.004]. There was no evidence of an indirect effect of CBP on maternal positive remarks though DFDs among male (a × b=0.12, 95% CI [−0.06, −2.27]) or female adolescents (a × b= 0.11, 95% CI [−0.06, 0.27]). There was no evidence of effects of adolescents’ DFDs on maternal positive remarks at 9 months. Also, there was no evidence of indirect effects for either male or female adolescents. Thus, adolescents’ DFDs did not significantly account for the relation between condition and maternal positive remarks at 9 months.

Current Maternal Depression and Maternal Criticisms and Positive Remarks

Current maternal depression was not significantly associated with mothers’ criticisms or positive remarks at either baseline (criticisms: r=.060, p=.31; positive remarks: r= −.083, p= .16) or at the post-intervention evaluation (criticisms: r=.069, p=.32; positive remarks: r=−.042, p=.54). Moreover, current maternal depression did not significantly moderate the effect of the CBP program on maternal criticisms or positive remarks at the postintervention evaluation (Table 4). The three-way interaction of Condition X Gender X Maternal Depression was not significant for either maternal criticisms (p=.46) or positive remarks (p=.84).

Table 4.

Hierarchical Generalized Model Results Predicting Mothers’ Criticisms at the Post-intervention Evaluation (i.e., 9-Month Follow Up; Left Panel) and Mothers’ Positive Remarks at the Postintervention Evaluation (i.e., 9-Month Follow Up; Right Panel)

	Outcome: Maternal criticism				Outcome: Maternal Positive Remarks
Study Variables	Coefficient	SE	T ratio	p	Study Variables	Coefficient	SE	T ratio	p
Age	−0.15	0.09	−1.66	0.12	Age	0.03	0.04	0.82	0.43
Gender	0.12	0.33	0.35	0.73	Gender	0.14	0.13	1.07	0.31
Parent depression T1	0.32	0.29	1.10	0.29	Parent depression T1	0.01	0.12	0.05	0.96
Parents’ criticism T1	0.51	0.06	8.57	0.001	Parents’ positive remarks T1	0.06	0.02	3.13	0.008
Adolescent CESD at T1	−0.05	0.02	−3.24	0.006	Adolescent CESD at T1	0.001	0.01	0.12	0.91
Adolescent CESD at T2	0.06	0.02	3.82	0.002	Adolescent CESD at T2	−0.01	0.01	−1.01	0.33
Condition (UC vs. CBP)	0.86	0.30	2.86	0.005	Condition (UC vs. CBP)	0.24	0.13	1.84	0.07
Parent Dep T1 X Condition	−0.33	0.58	−0.56	0.59	Parent Dep T1 X Condition	0.04	0.25	0.16	0.88

Note. Coefficients are log link of original scale. CESD = Center for Epidemiologic Studies Depression; Dep = Depressions; CBP = cognitive behavior prevention program; UC = usual care; T1 = Time 1 (baseline); T2 = Time 2 (9-month evaluation)

Discussion

Several interesting findings emerged from this study. First, we found a significant main effect of the intervention on maternal positive remarks and a significant intervention condition by adolescent gender interaction on maternal criticisms. Second, there was a significant moderated mediation effect of the intervention on mothers’ criticisms through adolescents’ level of depressive symptoms for boys. The significant condition by gender interaction indicated that, controlling for baseline criticism, at the post-intervention assessment, mothers of boys in the CBP program made significantly fewer critical comments about their sons than did mothers of boys in UC. In contrast, mothers of girls in the CBP program made more criticisms about their daughters than did mothers of girls in UC.

We tested one possible mediator of the relation between intervention condition and mothers’ criticisms of their children — number of adolescent DFDs between pre- to post-intervention. We hypothesized that the CBP program would produce fewer days with adolescent depressive symptoms, which then would predict lower levels of maternal criticism. Results partially confirmed this expectation. The mediation model revealed that the number of adolescent DFDs experienced between baseline and the post-intervention evaluation accounted for a significant amount of variance in the effect of the intervention condition on mothers’ criticisms of boys at postintervention, controlling for baseline maternal criticisms. Thus, the CBP program produced significantly more days without depression in adolescents, which, in turn, predicted lower levels of maternal criticism for boys, although not for girls. This result is consistent with a finding by Shortt et al. (2016) that parents of depressed youth respond more negatively to expressions of sadness in boys than in girls. Thus, in the current study, youth in CBP had significantly fewer days with depressive symptoms, which then predicted fewer criticisms by mothers about their sons, but not their daughters. Without continuous measures of adolescents’ depressive symptoms and mothers’ descriptions of their offspring, however, we do not know the precise temporal course of these constructs in relation to each other. That is, it is possible that mothers of boys in CBP showed an early decrease in their criticisms of their sons, which, in turn, then decreased boys’ depressive symptoms.

Several other explanations for the significant condition by gender interaction for maternal criticism are possible. For the whole sample, maternal criticisms decreased from baseline to post-intervention, but the extent of this decrease varied by condition and gender. Mothers of girls in CBP made significantly more critical comments than did mothers of girls in UC, whereas mothers of boys in CBP made fewer criticisms than did mothers of boys in UC. We speculate that the greater level of maternal criticisms of girls in CBP might be due, in part, to these girls implementing some of the new skills they learned in CBP, such as assertiveness or questioning the accuracy of their own and others’ beliefs. The mothers of girls in CBP might have disapproved of their daughters’ expression of such behaviors. Without a careful content analysis of mothers’ critical statements, however, we cannot know if this was the case.

When girls and boys engage in similar behaviors such as direct, assertive expressions of their thoughts and feelings, others might respond differently to them. Indeed, a study of children as young as 4 years old found that parents reacted more positively to assertiveness in boys than in girls, with boys receiving more praise and girls receiving more negative responses (Kerig, Cowan, & Cowan, 1993). For adolescents, adults maintain prescriptive gender norms for girls that emphasize being nondominant and communal compared to norms for boys that value agency and independence (Koenig, 2018). As such, it is possible that parents’ expectancies were violated by their daughters’ expressions of assertiveness and questioning the accuracy of beliefs. Differential responses to assertiveness by gender also are well-documented in the negotiation literature where women are expected to be less direct to avoid backlash from peers and to align with gender stereotypical expectations (Bossuyt & Van Kenhove, 2018). Based on the data from the current study, however, we are unable to determine which behaviors or characteristics of girls contributed to observed levels of mothers’ criticisms of daughters in CBP. We need further evaluation of which of the skills taught in the CBP program the girls actually used, and how the expression of these new skills affected the mother-adolescent relationship. Future studies also should examine the range of parents’ reactions to their adolescents’ new behaviors and explore why parents respond differently to sons versus daughters.

Regarding mothers’ positive remarks about their adolescent offspring, there was a significant main effect of condition, but neither a significant mediation nor moderation effect by adolescent gender. Controlling for maternal positive remarks at baseline, mothers of youth in the CBP program made significantly more positive remarks about their teen at the postintervention assessment as compared to mothers of youth in UC, regardless of youths’ gender. If youth in the CBP intervention used some of the cognitive, behavioral, emotional, communication, and problem-solving skills taught in the program, then this may have contributed to mothers making a greater number of positive statements about their teens. This is consistent with literature showing bidirectional relations between offspring’s behaviors and parents’ EE (e.g., Peris & Miklowitz, 2015). Thus, it is possible that the CBP program “worked” by helping adolescents to improve their regulation of their affect and behaviors, and their mothers’ descriptions of them reflected this. Although this is a plausible speculation, the design of the current study does not allow us to determine the direction of such effects. Just the very act of the youth attending the CBP program might have led mothers to describe their teens more positively. That is, mothers might have appreciated that their child participated in the intervention and therefore described them more positively. If mothers expressed such positivity toward their adolescent, then it is possible that the teens’ behaviors improved resulting in a positive cycle of interactions.

Another possible explanation for the mothers of youth in CBP making more positive remarks than mothers of youth in UC may be because mothers knew to which intervention condition (i.e., CBP or UC) their adolescents was assigned. Perhaps mothers of youth in the CBP program were more positive about their teens due to cognitive dissonance. That is, because they had invested in getting their adolescent to the sessions, mothers might have talked more positively about their teen to justify the time and effort expended.

A related explanation is social desirability or demand characteristics. Mothers of youth in CBP might have thought that they should be positive about their teens because of their participation in the program. That is, mothers may have believed that their teens must have been behaving more positively because they were in the program. One argument against this explanation, however, was the finding of gender differences in maternal criticisms. It is not clear why mothers of girls in the CBP program would be less likely to have such cognitive dissonance or respond less to demand characteristics as compared to mothers of boys. Thus, given that the study design was a randomized controlled trial, we can conclude that there was an overall effect of the CBP intervention on mothers’ positive statements about their adolescents, but we do not know why. Further research is needed to determine the mechanisms through which these mothers were more positive in their descriptions of their adolescent offspring at the post-intervention evaluation as compared to mothers of youth in UC.

Also noteworthy was the fact that the correlation between maternal criticisms and positive remarks, although significant, was not strong. This weak correlation may reflect the inherent differences in these two indices. That is, the absence or presence of positive remarks does not necessarily indicate the absence or presence of criticisms. Positive remarks and criticisms may be only weakly associated with each other. What are the clinical implications of understanding the distinctness of these two constructs? Increasing parental positive remarks may have little effect on decreasing parental criticisms, and vice versa. As such, it may be particularly beneficial for clinicians to train parents not to simply reduce their criticisms of their teens, but also to increase their positive remarks about their offspring, simultaneously and explicitly.

One other possible clinical implication of these results is that it might be beneficial to add a more direct parenting component to the CBP program. A few interventions with more intensive parenting elements have been found to improve parents’ expressions of criticisms and positive remarks (e.g., Miklowitz et al., 2013). Some youth depression prevention programs have not found that including parents in the intervention improved offspring’s outcomes as compared to the same intervention without a parent component (Clarke et al., 2001; Gillham et al., 2006). Clarke et al. (2001) noted, however, that parent attendance was low and therefore their trial might not have been an adequate test of their parent component. Other youth depression prevention programs that work directly with parents have shown improvements in both parenting behaviors and youth’s outcomes (e.g., Compas et al., 2009; Forehand et al., 2012). Indeed, Compas and colleagues (2009) tested the efficacy of a family group cognitive– behavioral (FGCB) prevention program that explicitly targeted improving parenting behaviors in families with a parent who had a history of depression and found that parents in the FGCB program showed significantly more positive parenting relative to parents in a control condition. Thus, changing parents’ thoughts, feelings, and behaviors about their teens may occur more efficiently through programs with a primary focus on parenting. Direct comparisons of these different programs regarding changes in parents’ criticisms and positive remarks about their offspring in relation to the children’s adaptation would be interesting and potentially worthwhile.

A final aim of the current study was to explore whether mothers with current depression expressed more criticisms and fewer positive remarks compared to mothers whose depression was not current. Results indicated that the valence of mothers’ descriptions of their adolescent offspring was not associated with their being currently depressed. Previous studies of EE and parental depression have reported differences between depressed and nondepressed parents, but have not indicated whether parents’ depression was current or past (e.g., Nelson et al., 2003). Thus, our previously reported finding of differential effects of CBP on youths’ depression as a function of parents’ depression diagnoses at baseline (Garber et al., 2009) likely was not simply due to more criticisms or fewer positive remarks by the currently depressed parents. Strengths, Limitations, and Future Directions

Strengths of the current study include its longitudinal design and use of a measure of parents’ views about their children that was not a self-report questionnaire. In the FMSS, parents do not know what is coded from their speech sample, making it a more indirect measure of parents’ overall “attitudes” about their children (Hooley, 2007). It is still possible, however, that parental responses on the FMSS are subject to some level of social desirability bias insofar as parents might feel pressure to portray a description of their child and their relationship that would put them and their child in a good light.

Additionally, the FMSS ratings do not provide information about the actual types of criticisms and positive remarks (i.e., global vs. specific). Importantly, although the FMSS is considered an index of the family environment, it is not an explicit measure of parental behaviors toward their children. Medium effect sizes have been reported in studies of behavioral correlates of criticisms expressed during the FMSS (Cruise, Sheeber, & Tompson, 2011). Future studies should examine the relations of behavioral correlates of parents’ positive remarks as well.

Other limitations of this study highlight directions for future research. First, although the sample was representative of the communities from which participants lived, only 25% of the sample identified as members of a racial or ethnic minority. In the current sample, we did not find any significant differences as a function of race or ethnicity, but a larger, more diverse sample is needed to generalize these findings more broadly. Second, we only collected the FMSS at two time points. To examine the maintenance and stability of the effects of CBP on maternal criticisms and positive remarks, we should have assessed these indices at later time points as well. It also would be interesting to explore the extent to which maternal criticisms and positive remarks predict subsequent depression in the adolescents.

Third, this sample was comprised of adolescents at risk for depression. Results may not generalize to a normative sample. Fourth, although there were fathers in the sample, we were not able to obtain FMSS data from them, thus limiting the application of these finding to mothers only. Engaging fathers in clinical research is an important goal for future research on parental depression. Finally, the overall number of maternal criticisms was low, and therefore these findings may lack meaningful clinical utility. Even at low rates, however, criticisms have been shown to be powerful predictors of relapse, with evidence that as few as two criticisms predicted worse outcomes in individuals with unipolar depression (Hooley, 2007; Vaughn & Leff, 1976).

Conclusion

In sum, a group CBP intervention for at-risk adolescents resulted in mothers of youth in the CBP program being more positive about their adolescents at the postintervention evaluation than were mothers of youth in UC. The extent of maternal criticism of youth in CBP versus UC varied by adolescents’ gender. Mothers of boys in CBP made fewer criticisms than did mothers of boys in UC; this relation was partially mediated by the greater number of adolescent DFDs between baseline and postintervention in boys in CBP as compared to boys in UC. In contrast, at the post-intervention follow up, mothers of girls in CBP made significantly more criticisms about their daughters than did mothers of girls in UC.

We speculated that perhaps mothers reacted negatively to girls’ use of some of the skills taught in CBP (e.g., questioning the accuracy of their own and others’ beliefs). Nevertheless, an adolescent-focused CBP program with even minimal parent involvement (e.g., 2 sessions) resulted in changes in mothers’ criticisms and positive remarks about their adolescents, although somewhat differently for boys than for girls. Supplementing youth intervention programs with more direct efforts to reduce parental criticisms and increase positive remarks may further improve parents’ attitudes about their at-risk offspring.