New Beginnings (for children of divorce)

Study 1

Summary

Wolchik et al. (2000 [certified by Blueprints], 2002, 2013) utilized a randomized controlled trial with 240 pairs of divorced mothers and children who were randomly assigned to a control condition or one of two intervention conditions: mother only or dual component (mother-child). The control condition received self-study literature. Data were collected at five time points: baseline, post-intervention, six-month follow-up, six-year follow-up, and 15-year follow-up. Outcome measures included internalizing and externalizing behaviors, as well as mental health problems and disorders, substance use, risky sexual behavior, and GPA.

As summarized by Sandler et al. (2016), the intervention group compared to the control group showed significant improvements in:

• internalizing problems at posttest
• grade-point average at six-year follow-up (combined intervention group)
• number of sexual partners at six-year follow-up (combined intervention group)
• diagnosis of mental health disorder at six-year and 15-year follow-ups (combined intervention group)
• days in jail at 15-year follow-up (combined intervention group)

Evaluation Methodology

Nineteen articles examined the sample, which was assessed at baseline, posttest, six-month follow-up, six-year follow-up, and 15-year follow-up. Three studies (Wolchik et al., 2000, 2002; Dawson-McClure et al., 2004) examined two intervention groups separately, while the remaining studies examined both intervention groups combined or the mother-only intervention group. Three key studies presented the basic findings (Wolchik et al., 2000, 2002, 2013). The others focused on moderation and mediation analyses or summarizing results across multiple studies.

Design: At the start, 240 families were randomly assigned to one of three conditions: (a) mother-only intervention (n = 81), (b) dual-component (mother and child) intervention (n = 83), or (c) literature control condition (n = 76) based on reading books on post-divorce adjustment. At six years, the sample fell to 218, indicating about 9% attrition. At 15 years, 90% of the sample participated in the interviews.

Recruitment: Court records identified 1,816 randomly selected divorce decrees in Maricopa County (Phoenix, AZ) that (a) occurred within two years of the intervention start date and (b) involved a child between ages 9 and 12. Letters and phone calls to these individuals were the primary means of recruitment, but media advertisements were used to recruit 20% of the final sample.

Of all the families contacted (either from court records or through media advertisements), 671 were deemed eligible after an initial screening over the phone. Of these families, 68% (n = 453) completed a recruitment visit; 75% (n = 341) of the recruitment-visit completers agreed to participate in the intervention study; 92% (n = 315) of these families completed the pretest. After the pre-test interview, 16% (n = 49) were deemed ineligible, and an additional 8% (n = 26) withdrew before assignment to treatment or control condition, leaving a final sample of 240 families (38% of the eligible families).

Inclusion Criteria: Inclusion criteria were (a) divorce was granted within the past two years, (b) primary residential parent was female, (c) child was between the ages of 9 and 12, (d) neither the mother nor any residential child was currently in treatment for psychological problems, (e) mother had not remarried, had no live-in boyfriend, and had no plan to remarry during the trial, (f) custody of the child was expected to remain stable over the trial, (g) family resided within a one-hour drive of the site where the program was delivered, (h) mother and child could complete the assessment battery in English, (i) child was not in a special education program for the mentally challenged or learning disabled and (j) if child had a diagnosis of attention deficit disorder for which he or she was taking medication. In families where more than one child met the criteria, one was randomly selected. Families were excluded and referred to treatment if the child scored above 17 or expressed suicide ideation on the Children's Depression Inventory or had a score above the 97th percentile on the Externalizing subscale of the Child Behavior Checklist.

Sample Representativeness: Checks compared 1) a hard-to-locate group with the rest of the sample, 2) intervention acceptors with refusers, and 3) acceptors who were assigned to condition and completed the intervention with withdrawers. First, of the 1,816 families initially identified by court records, 532 were difficult to locate. For those found after special search efforts, members of this hard-to-locate group were more likely to refuse the recruitment visit and intervention than the rest of the sample. Also, children in the hard-to-locate group had significantly higher teacher-reported shy-anxious behavior problems than the rest of the sample. Second, among the full sample, acceptors reported significantly higher incomes and education and fewer children than refusers. Compared to acceptors, children in the refusers group reported more positive coping, and mothers asked more open-ended questions during the mother-child interaction. Third, withdrawers reported less maternal education and lower child support than those assigned to the intervention conditions.

Attrition: All participants assigned to the conditions completed the posttest, and 98% (234 out of 240) completed the six-month follow-up. Of the 83 participants assigned to the dual-component (mother-child intervention), 12% (n = 10) did not complete the program. Of the 81 participants assigned to the mother-only program, 20% (n = 16) did not complete the program. Consistent with the intent-to-treat principle, the 26 subjects who did not complete the program were included in the analysis because follow-up data was available.

At the six-year follow-up, 218 of the 240 randomized (91%) subjects provided data.

At the 15-year follow-up, 89.6% of families provided some data, as 194 young adults and 204 mothers completed the assessment. The rate of completion among young adults for this follow-up was 82% for the intervention group and 79% for the control group.

Sample: Maternal ethnicity was as follows: 88% Caucasian, 8% Hispanic, 2% African American, 1% Asian and 1% Other. The mean age of the mothers was 37.3 years, and 47% reported completing some college courses. Median yearly income was in the range of $20,001 - $25,000. Legal custody arrangements were 63% sole maternal, 35% joint and 2% split. Mothers had been divorced an average of 12 months and physically separated an average of 27 months. The mean number of children was 2.2. The average age of the interviewed children was 10.4 years, and 49% of the children were female.

At the six-year follow-up, 25% of the mothers and 63% of the fathers had remarried (but remarriage rates did not differ significantly across conditions).

At the 15-year follow-up, children were 25.6 years old on average, and just over half (51%) were married or living as if married. These young adults most commonly achieved some college, and their median income was in the $30,000 range.

Measures: At all assessments, multimethod, multiagent (mother, child, and teacher) assessment was used. Both mothers and children completed questionnaires, assessing putative mediators and psychological adjustment problems. At posttest and the six-month follow-up, mother-child interactions were videotaped and analyzed for behavioral observation measures (including conversational latitude and attending). Finally, teachers provided reports of children's adjustment problems, including shy-anxious behaviors and acting out.

Psychological Adjustment Problems. The study concentrated on two measures of psychological adjustment. For externalizing behavior problems, mothers completed the 33-item Externalizing subscale of the Child Behavior Checklist. Children completed the Aggression and Delinquency subscales (30 items) of the Youth Self-Report. Teachers completed the six-item Acting-Out subscale of the Teacher-Child Rating Scale. Pretest alphas were .88, .87 and .90, respectively.

For internalizing behavior problems, mothers completed the 31-item Internalizing subscale of the Child Behavior Checklist. Children completed the 27-item Children's Depression Inventory and the 28-item Children's Manifest Anxiety Scale. Teachers completed the six-item Shy-Anxiousness subscale of the Teacher-Child Rating Scale. Pretest alphas were .89, .84, .88 and .82, respectively. Mother and child outcomes were reported as a composite measure.

Putative Mediators. Mediating variables used in the analysis included the following:

• Mother-child relationship quality - Includes items from the Parent-Adolescent Communication Scale and the Child Report of Parenting Behavior Inventory (CRPBI), as well as videotaped behavioral observation measures. Alphas ranged from .71 to .86.
• Discipline - Both mothers and children completed questionnaires on discipline strategies. Mothers and children were asked about the consistency of discipline; mothers were asked about appropriate and inappropriate discipline strategies. Alphas ranged from .59 to .78.
• Interparental conflict - Both mothers and children completed subscales of the Children's Perception of Interparental Conflict Scale. Alphas ranged from .55 to .64.
• Father-child relationship - Mothers completed six questionnaire items that assessed the support of the noncustodial father-child relationship (alpha = .85).
• Appraisals of divorce stressors - Children completed the Threat Appraisal Scale, which includes threats to well-being, including negative self-evaluation, rejection by others and loss of desired objects (alpha = .92).
• Coping - Children completed 36 items from the Children's Coping Strategies Checklist. Items were added to assess active and avoidant coping mechanisms. Alphas ranged from .68 to .87.

Six-Year Follow-up Measures. The six-year follow-up data used some previous measures but also added several new ones.

The follow-up used measures of externalizing problems and internalizing problems similar but not identical to those at earlier assessments. For externalizing problems, the parent rating came from 33 items of the Child Behavior Checklist and the child rating came from 27 items of the Divorce Adjustment Externalizing scale. The two scales were standardized and averaged. For internalizing problems, the parent rating came from 30 items of the Child Behavior Checklist and the child rating came from 28 items of the Revised Children's Manifest Anxiety Scale. The two scales were again standardized and averaged.

Outcome measures new to the six-year follow-up included the following:

• Diagnosis of any mental disorder and any drug abuse or dependence disorder came from the Diagnostic Interview Schedule for children.
• Adolescent symptoms of externalizing and internalizing disorders over the past year were measured using the Diagnostic Interview Schedule for Children. The measure counts the number of symptoms of both externalizing and internalizing disorders.
• Adolescent substance use measured the frequency of marijuana and alcohol use over the past year. In various studies, substance use was measured as the combined frequency of use, the count of drug dependence or abuse symptoms, polydrug use (number of different drugs used), and other drug use.
• Competence was measured with 18 items from the Coatsworth Competence Scale and came from both child and mother ratings (alpha = .83 for child report, .85 for mother report).
• Self-esteem came from the Self-Perception Profile of Children (alpha = .86).
• Academic performance was assessed using school records and cumulative high school GPA.
• Risky sexual behavior was measured as the number of sexual partners over the past year.

The six-year follow-up studies of moderation and mediation examined several additional measures:

• A scale of environmental stressors was based on negative events that occurred to the child, interpersonal conflict, maternal distress, reduced contact with father, and per capita income.
• Negative errors, based on the 50-item Negative Life Events Scale-Child, tallied mean child ratings of the upsettingness of events and negative thinking about the events.
• Adaptive coping, based on the Children's Coping Strategies Checklist, measured problem solving, cognitive decision making, optimism, and control.
• Parental monitoring, based on the 13-item Assessment of Child Monitoring Scale, combined ratings of parent and child.

The most recent studies also used a new measure of baseline risk. Since not all the outcomes at the six-year follow-up were measured at baseline, a generalized measure of risk was constructed. It combined measures of externalizing problems and environmental stressors (negative events that occurred to the child, interparental conflict, maternal distress, reduced contact with father, and per capita income).

Fifteen-Year Follow-Up Outcome Measures (Wolchik et al., 2013).

• Internalizing and externalizing disorders were classified using the Diagnostic Interview Schedule IV. Disorders were assessed since program entry (15 years) and since the last follow-up (nine years).
• Recent internalizing and externalizing problems were determined using subscales of Adult Self Report and Adult Behavior Checklist, completed by the young adult and the mother, respectively. Alphas ranged from .84 to .92.
• Substance-related disorders were assessed using the Diagnostic Interview Schedule. Indicators reflected the presence of any substance-related disorder and the number of substance-related disorders in the last nine and 15 years.
• Use of alcohol, marijuana, and other drugs was captured from the Monitoring the Future Scale, completed by the young adult. Measures represented occasions of alcohol use, marijuana use, other drug use, and polydrug use.
• Substance use problems in the past six months were assessed by standardizing and averaging mother's and young adult's reports from the Adult Behavior Checklist and Adult Self Report, respectively.
• Frequency of binge drinking in the past year was measured using an adaptation of an item from the Quantity and Frequency of Alcohol and Drugs Scale.
• Age of onset of regular drinking was taken from the Diagnostic Interview Schedule.

Analysis: Based on an intent-to-treat approach, the first study (Wolchik et al., 2000) used regression analysis to predict time 2 outcomes from time 1 outcomes plus two dummy variables for the treatment conditions. With the mother-only condition serving as the reference group, one dummy variable measured the literature control condition and a second dummy variable measured the dual-component condition. Thus, main effects were tested for the mother-only condition, and effects for the dual-component condition were examined only in comparison to the mother condition.

In addition, the regression analysis allowed for the interaction of the baseline outcomes with the condition dummy variables. A significant interaction would be expected to show that the intervention had stronger effects for those at high risk (i.e., having worse scores on the outcome at baseline). With the variables centered, the dummy variables alone show the effect of the conditions at the mean of the time 1 predictors. The interactions then show how much the effect of the condition changes at higher levels of the baseline outcomes.

For significant effects, the analysis presented Cohen's d coefficients for effect size, either for the average across all subjects or for high-risk subjects.

Six-Year Follow-Up. The main six-year follow-up (Wolchik et al., 2002) used the same analytic framework but, where appropriate, replaced linear regression with logistic regression and ordinal regression. Except for internalizing and externalizing problems, baseline measures were not available for many outcomes examined in the six-year follow-up. Wolchik et al. (2002) used a measure combining externalizing and internalizing problems to serve as a proxy for baseline risk.

15-Year Follow-Up (Wolchik et al., 2013). The 15-year follow-up used logistic regression, analysis of covariance, and Cox proportional hazards models. It appears that the study did not need to adjust for the groups that intervention participants received the program in, as intra-class correlations indicated little clustering (mean=.02). The analyses combined the dual component and mother programs and compared this group to the control. To adjust for multiple tests, a false discovery rate was applied to effects. The changing nature of the outcomes at the long-term follow-up meant that the baseline outcomes could not be controlled directly. Analyses controlled for the composite measure of baseline risk.

All 240 families were included in analyses, as the article used full information maximum likelihood and multiple imputation to account for missing data (ranging from 0% to 23%). However, measures using Diagnostic Interview Schedule disorders and the measure of onset of regular drinking were exceptions. Participants who exhibited these measures prior to the period analyzed were excluded.

Intent to Treat. The study generally complied with the intent-to-treat principle. Non-completers were included in the analysis when they had follow-up data.

Outcomes

Implementation Fidelity. Mothers attended an average of 77% of the group sessions, and children attended an average of 78%. The proportion of homework completed was .54 for the mother-only condition and .55 for the dual-component condition. Those who participated in the literature control condition reported reading about half of each of the books.

Baseline Equivalence and Differential Attrition. Of the 240 families included in the baseline sample, pretest equivalence of the randomized conditions was assessed and only one significant difference was noted: interparental conflict was lower in the literature control condition than in the dual-component condition.

For the posttest and six-month follow-up, attrition was not a problem. All participants assigned to the conditions completed the posttest, and 98% (234 out of 240) completed the six-month follow-up.

For 218 families with data for the six-year follow-up (91% of the original 240 families), there were no significant baseline differences on demographic variables or baseline levels of externalizing or internalizing problems (Wolchik et al., 2002). Further, there were no significant group-by-attrition effects on externalizing or internalizing problems.

Of the 240 families assigned to condition, 89.6% provided some information at the 15-year follow-up. Rates of attrition did not differ significantly across intervention and control participants (Wolchik et al., 2013). Completers and dropouts differed on baseline self-esteem and internalization, but the analysis controlled for the two measures.

Posttest and Six-Month Follow-Up (Wolchik et al., 2000; Hipke et al., 2002; Tein et al., 2004).

Wolchik et al. (2000) reported average differences for the literature control condition from the mother-only condition, as well as for the dual-component vs. mother-only condition. Significant coefficients for group-by-baseline outcome interaction terms generally mean the effects of the intervention may have special benefits for those at higher risk (i.e., those with worse baseline values on the outcomes).

Psychological Adjustment Problems (Posttest). When compared to the literature control condition, the mother-only intervention resulted in significant improvements in internalizing and externalizing behaviors (for those at higher risk at baseline) but did not improve teacher reports of acting-out problems. In one instance, problems worsened from the intervention: Teacher reports of shy-anxious behaviors significantly increased in the mother-only condition, when compared to the control condition. There were few differences between the dual-component and the mother-only condition, either on average or for those at higher risk.

Psychological Adjustment Problems (Six-Month Follow-up). Results of the six-month follow-up were not as consistent as results at posttest. There were significant improvements in externalizing problems in the mother-only condition when compared to the literature control condition, but there were no significant improvements in internalizing, teacher reports of shy-anxious behaviors or acting out.

Children at higher risk in the mother-only condition benefited from significant improvements in externalizing problems (mother-child report) and teacher reports of acting out behavior when compared to the control condition.

Putative Mediators (Posttest). Four of the 16 tests of the putative mediators (mother-child relationship quality and effective discipline strategies, as well as the behavioral observation measures validation of content and attending) showed significant improvement in the mother-only condition when compared to the control condition.

For some outcomes, the intervention had more benefits for higher-risk children. High-risk children in the mother-only condition showed significant improvements (when compared to the control condition) in mother-child relationship quality, mother's attitude toward the noncustodial father, parental conflict and supportive coping strategies.

Putative Mediators (Six-Month Follow-Up). Again, the results of the six-month follow-up were not as consistent as for the posttest. There was only one significant improvement (observed attending) in the 11 putative mediators that were assessed at the six-month follow-up in the mother-only condition when compared to the control condition.

Mediation Analysis. Tein et al. (2004) expanded on the analysis of Wolchik et al. (2000) by performing a formal analysis of mediation and moderation. Given previous findings of equivalence in outcomes between the two intervention groups, the authors examined the mother-only intervention group and the control group, without the dual-component intervention group. Using structural equation modeling, the mediation analysis found that improvement in posttest parental methods of discipline and mother-child relationship quality mediated program effects on posttest and six-month follow-up measures of externalizing problems. The mediation effect primarily helped children who began the program with poorer scores on discipline, mother-child relationship quality, and externalizing problems.

Moderation Analysis. Hipke et al. (2002) also used only two of the conditions - the mother-only and control groups (N = 157) - in a moderation study at the posttest and six-month follow-up. The authors noted that the third, dual-component condition had shown no more benefits than the mother-only condition and its moderation effects would likely differ from the mother-only condition. The analysis found no significant main effect (p < .05) of the mother-only condition on externalizing. Consistent with previous findings, a baseline externalizing-by-condition interaction indicated that the program significantly affected only those high on baseline externalizing. The key results came from the between-group regression models that found two significant moderation effects. The program had stronger effects on externalizing when maternal demoralization was low and when children's self-regulatory skills were high. Note, however, that these two significant moderation effects came from the examination of 24 possible combinations of predictors, outcomes, and follow-up periods (see Table 2).

Six-Year Follow-Up (Wolchik et al., 2002, 2007).

In Wolchik et al. (2002), the mother-only program, when compared to the control group, did not significantly influence any of the six outcomes - externalizing, internalizing, mental health or drug abuse disorder, polydrug use, or number of sexual partners.

Two outcomes did show program benefits for subjects at high risk at baseline. For the mother-only program, the program-by-baseline outcome interaction reached significance for externalizing problems and mental disorder symptom count, indicating the program benefit was strongest for those with higher baseline problems.

Wolchik et al. (2007) presented a re-analysis of previous results that combined the dual-component (mother and child) and the mother-only conditions into one intervention group. The authors justified the combination by noting that previous studies found no added benefit of the dual-component version over the mother-only version. The analysis of the combined conditions found significant intervention effects for seven of 15 tests: parent-reported externalizing, teacher-reported externalizing, diagnosis of a mental disorder, marijuana use, number of sexual partners, grade point average, and competence.

Six-Year Follow-Up Moderation Analysis (Dawson-McClure et al., 2004).

Dawson-McClure et al. (2004) adopted a different methodological approach but generally affirmed the results of Wolchik et al. (2002). The key methodological difference was the development of a combined measure of risk to be used as a predictor of diverse outcomes. Rather than using a different baseline measure for each outcome, the models used the same baseline predictor, namely, a scale combining environmental stressors and externalizing problems.

The mother-only program reduced none of the five outcomes, but the baseline risk scale significantly moderated the program effect in eight of 10 tests. In all cases, the program did more to reduce problems of those at high risk. For example, the mother-only program did slightly worse than the control group on percent diagnosed with a mental health disorder among those at low risk (13.9% versus 9.5%) but did much better among those at high risk (25.0% versus 46.1%).

Six-Year Follow-Up Mediation Analyses (McClain et al., 2010; Sandler et al., 2015; Sigal et al., 2012; Soper et al., 2010; Velez et al., 2011; Wolchik et al., 2009; Zhou et al., 2008).

Seven studies performed mediation analyses on the same data set. Note that six studies combined the two intervention conditions, and one dropped the dual-component condition to focus on the mother-only intervention. To justify the combination, studies cited 1) the lack of significant differences in the effects of the two intervention conditions and 2) results of the Box M test demonstrating the equivalence of the variance-covariance matrices across both conditions.

The first study (Zhou et al., 2008) used the combined intervention condition to demonstrate that maternal discipline at posttest (but not mother-child relationship quality) mediated the program effect on GPA at the six-year follow-up. Neither maternal discipline nor mother-child relationship quality mediated the program effect on the number of sexual partners. Relationship quality mediated the program effect on externalizing problems but only for the high-risk group of subjects.

The second study (Wolchik et al., 2009) examined the mother-only condition by excluding those receiving the dual-component condition. Tests demonstrated that the intervention effects on externalizing and internalizing at six years were significantly mediated by posttest relationship quality and six-year child coping skills.

The third study (Soper et al., 2010) used the combined intervention condition to examine both mediators and outcomes at the same time point, the six-year follow-up. The mediators, which differ from those used in the other mediation studies, included parental monitoring, negative errors, and adaptive coping. The outcomes included alcohol/marijuana use, polydrug use, other drug use, and the number of sexual partners. The program increased parental monitoring and parental monitoring decreased alcohol and marijuana use, polydrug use, and other drug use but only for youth with a high risk of developing adjustment problems.

The fourth study (McClain et al., 2010), which used the combined intervention condition, was the most detailed. This study laid out a full path model: The program improved the mother-child relationship and discipline at posttest, which improved internalizing and externalizing at the posttest and six-month follow-ups, which in turn lead to better outcomes relating to academics, self-esteem, internalizing problems, externalizing problems, drug use, and risky sexual behavior at the six-year follow-up.

Program correlations with the six-year follow-up measures were significant for only one of the six outcomes at the last follow-up - risky sexual behavior (r = -.16). The indirect effects were small, but in a few cases significant. The effect of the program via the quality of the mother-child relationship and internalizing reached significance at the six-year follow-up for two of six outcomes: internalizing (CI = -.01 to -.15) and self-esteem (CI = .01 to .20). The effect of the program via discipline and externalizing reached significance at the six-year follow-up for three of six outcomes: externalizing (CI = -.03 to -.41), substance use (CI = -.01 to -.17), and academic performance (CI = .01 to .04).

The fifth study (Velez et al., 2011) used the combined intervention condition to examine how the impact of the intervention on child coping at six months and six years was mediated by parent discipline and relationship quality at posttest. The models tested for 12 possible mediated effects, with three proving statistically significant. Significant mediated effects emerged for posttest relationship quality on coping efficiency at six months and for active coping and coping efficiency at six years.

The sixth study (Sigal et al., 2012) used the combined intervention condition in a moderation-mediation study of two non-behavioral outcomes - education expectations and job aspirations. The analysis first showed the program significantly reduced educational expectations and job aspirations but only for those at high baseline risk. The analysis of the posttest mediators for the six-year outcome found that none of the mediated effects reached significance. The analysis of the six-year mediators for the six-year outcomes found significant mediated effects but only for those at high risk at baseline. When inserting all mediation effects into models simultaneously, three remained significant. For those at high risk at baseline, self-esteem and academic competence mediated the program effect on educational expectations, and academic competence mediated the program effect on job aspirations.

The seventh study (Sandler et al., 2015) presented a mediation model for the combined condition that differed modestly from the model in McClain et al. (2004) but overall offered similar results (see Figure 1). Based on the model, the program improved positive parenting at posttest, which reduced child internalizing and externalizing problems at six months and thereby improved multiple child outcomes at six years.

15-Year Follow-Up (Wolchik et al., 2013; Luecken et al., 2015; Christopher et al., 2017).

The key 15-year analyses came from Wolchik et al. (2013). For both the 15-year study period and the 9-year period since the previous follow-up, the combined intervention group (mother and dual component programs) showed a significantly lower percentage of young adults that had developed an internalizing disorder compared to the control group. Hazard analyses showing a reduced rate of symptom onset further demonstrated the program effect on internalizing disorders. Comparing percentages of adults with specific disorders indicated that the program effect on internalizing disorders was primarily due to a reduction in major depression. Developing an internalizing or externalizing disorder showed marginal significance with the adjustment for multiple tests for both durations. Externalizing disorders, recent externalizing and internalizing problems, and nine substance use outcomes showed no significant program effects.

Moderation analyses looking at program effects across females and males did not show any significant interactions at the .05 level with the adjustment for multiple tests. Marginally significant effects indicated that the program was strengthened for males on polydrug use, other drug use, substance use problems, and substance use disorders. Females showed marginally significant increased drinking for those in the program, though the study noted that this outcome might not have clinical significance.

Luecken et al. (2015) studied changing cortisol levels in response to a socially stressful task. The mixed-model analysis of 161 (67%) participants with data focused on the combined intervention condition at 15 years. It found no main effects of the intervention on cortisol reactivity but identified moderation by age. For older participants, the intervention reduced cortisol reactivity relative to the control group, but for younger participants, the control group exhibited a greater decline after the task than the intervention group.

Christopher et al. (2017) studied the combined intervention condition at 15 years and five outcomes involving painful feelings about divorce. The analysis showed a significant main effect of the combined conditions on maternal blame and filter of divorce and significant moderation effects such that the intervention did more to reduce painful feelings among those at high baseline risk. In 15 moderation tests, the study found two significant results indicating that the intervention moderated the harmful effect of maternal blame on externalizing and the harmful effect of the filter of divorce on substance use.

15-Year Follow-Up Mediation Analyses (Mahrer et al., 2014; Wolchik et al., 2016).

Mahrer et al. (2014) examined two non-behavioral outcomes, parenting warmth attitudes and harsh discipline attitudes, for the sample children at the 15-year follow-up, when they had reached young adulthood. The study found no main effects of the combined intervention on the outcomes, but it did find moderation effects. The intervention proved more effective for young adult parenting attitudes when the mothers had low warmth and harsh discipline at baseline. The mediation analysis then focused on the subgroups of young adults with mothers showing these characteristics. For the subgroups, warmth of the mother at posttest mediated the intervention effect on warm parenting attitudes of the young adults. However, no similar mediation effect emerged for harsh discipline. The authors argued that the results demonstrated that the intervention can disrupt the intergenerational transmission of negative parenting attitudes.

Wolchik et al. (2016) presented multiple models of cascading effects that included the following causal paths: 1) from the combined intervention to parenting at posttest, 2) from parenting at posttest to child internalizing and externalizing at six months, 3) from child internalizing and externalizing at six months to mental health and substance use at six years, and 4) from mental health and substance use at six years to externalizing, internalizing, marijuana use, and alcohol use at 15 years. The complex mediation and moderation results cannot be summarized simply, but the authors concluded that the models demonstrated the enduring benefits of improved parenting. The benefits emerged in a cascade such that improvements in functioning during adolescence fostered improved functioning during young adulthood. For example, higher academic competence in adolescence due to improved parenting had spillover effects on internalizing problems and externalizing problems in young adulthood.

All-Wave Summary (Sandler et al., 2016).

Sandler et al. (2016) offered a comprehensive overview of the findings from the multiple assessments. Table 5.3 lists main effects, moderated/mediated effects, and null effects (either not significant or not analyzed) for 21 outcomes.

The posttest results identified one main effect on internalizing, one moderated effect on externalizing (by baseline externalizing), and two non-significant effects. The six-year results combined the mother-only and dual component modalities into one intervention group and identified three main effects (sexual partners, mental disorder diagnosis, and grade point average) and 10 moderated/mediated effects (by baseline risk). The 15-year results also used a combined intervention group and identified three main effects (mental disorder diagnosis, days in jail, and mental health services used) and 10 moderated effects (by gender, age, and baseline risk). Overall, then, the study found some main effects with many moderated effects of the program for subgroups of the sample.

Study 2

Summary

Wolchik et al. (1993), a randomized controlled trial, had a small sample (94 families) in which participants were assigned to the mother-only intervention or a wait-list control group. Data were collected at baseline and post-intervention, and outcomes included parenting skills and mother-child relationship quality, as well as child depression, manifest anxiety, aggression, conduct problems and total problems.

Wolchik et al. (1993) found that, compared to the control group, the program showed significant improvements in:

• child-reported aggression
• parent-reported communication, positive routines, control, and willingness to change visitation

Evaluation Methodology

Design: The sample came from a narrowly defined population. Lists of potential participants were gathered from court records of filings for divorce and responses to media articles or school presentations. As in Wolchik et al. (2000), investigators prescreened the potential participants in a telephone interview to ensure they met seven criteria: a) the divorce decree was granted during the last two years, b) there was at least one child between 8 and 15 years of age, c) the custodial or primary residential parent was female, d) neither the custodial parent nor child was currently in treatment for psychological problems, e) the custodial parent had not remarried and had no plans to remarry during the study period, f) the custodial arrangement was expected to remain stable during the study period; and g) English was the primary language of the mother and children. Without listing the number of families screened, the study identified 177 families to proceed to the next step.

A second stage of screening eliminated children who did not appear to be at risk and mothers or children with clinical levels of depression. Of the 177 families examined, 94 were selected for randomization. The sample consisted of seven separate cohorts, each of which had 6-8 families who participated in a single group. Treatment mothers attended 10 weekly group sessions and two individual sessions. The control group waited six months to begin the program.

Assessments were done before the intervention and at six months, immediately after the intervention.

Of the 94 families selected for randomization, 70 completed pretest and posttest interviews. About 40% of the attrition occurred prior to the first session. The attrition rate was 22% for the control group and 29% for the intervention group.

Sample: Ninety percent of the mothers were Caucasian. The mean age of the mothers was 36.8 years, and 74% had attended college. The average yearly income was $20,001 to $25,000. The parents had been divorced an average of 11.0 months and physically separated an average of 23.1 months. The mother had sole custody in 74% of the families. The average age of the child was 10.6 years, and 61% of the children were male.

Measures:

The 25 measures, obtained by interviewers blind to the condition, were based on self-reports of children and mothers. The measures can be divided into four groups:

1. Eight scales of modifiable risk and protective factors (or mediators) from children's reports. These relate to communication with parents, discipline, time with parents, and perceived parental conflict.

2. Eleven scales of modifiable risk and protective factors (or mediators) from mother's reports. These relate to family communication, relationship quality, discipline, time with children, parental conflict, and attitude toward child's relationship with father.

3. Four scales of psychological adjustment from children's reports, including scales of depression, manifest anxiety, aggression, and conduct problems.

4. One scale of children's psychological adjustment from mother reports that focused on total problem behaviors.

The scales have strong reliability. Other measures cover basic demographic information.

Analysis. A series of ANCOVAs used pretest scores as covariates and group membership as the key predictor of the outcomes. The models tested for the interaction of the pretest scores with group membership. For significant interactions, calculations determined whether the program most benefited those with low or high pretest scores.

Outcomes

Implementation Fidelity. Group leaders received training and ongoing supervision. A process evaluation measured time spent on each component of the program, attendance of participants, self-ratings by leaders of their knowledge and mastery of the program material, ratings of leaders by participants, and participants' reports on use of the skills taught in the program. According to the results of the process evaluation, participants attended 9.6 of the 10 scheduled group sessions. The leaders closely followed the session outlines, nearly all leaders (more than 90%) rated their knowledge of materials as high, and participants gave leaders positive ratings (4.75 on a 5-point scale). Participants further reported favorable changes in usage of 22 skills taught in the program (a rated mean of 2.76 on a 3-point scale).

Baseline Equivalence. The study did not examine baseline equivalence.

Attrition. The rate of attrition did not differ significantly across the intervention and control groups (p = .68). Tests of interaction for condition-by-attrition status were done for the pretest variables. No differences existed for mothers' reports, while two of 12 baseline measures from children's reports - aggression and depression - were significant. The results indicated that completers in the control group tended to be less aggressive and less depressed than completers in the intervention group. The authors summarize: "Given that the control group completers tended to be healthier, any effects of attrition would be expected to be biased against showing positive effects of the treatment."

Otherwise, attrition occurred similarly across the intervention and control groups. Mothers with higher levels of interparental conflict were more likely to drop out, and children with higher levels of contact with nonparental adults were more likely to drop out. However, the similarity in attrition across conditions limits the potential for bias.

Posttest. The study organized the results by the four groups of outcome measures. In general, the program had stronger effects on measures obtained from mothers than from children.

Mediators from Child Reports. Of the eight outcome measures, the program significantly improved one (acceptance-rejection), but only for those with worse pretest scores. Six others showed no significant differences at the .05 level, and one other (support from non-parental adult) improved more for the control group. Rather than a true iatrogenic effect, this latter finding may have resulted simply from increased contact time with the mother.

Mediators from Parent Reports. Of the 11 outcomes, the program significantly improved four (communication, positive routines, control, and willingness to change visitation) across all pretest levels and significantly improved two others (consistency of discipline and negative divorce events) among those with worse pretest scores.

Outcomes from Child Reports. Of the four outcomes measuring psychological adjustment, the program significantly improved one (aggression). For depression and anxiety, there were no statistically significant differences between the treatment and control groups.

Outcomes from Mother Reports. The one measure of psychological adjustment showed significantly greater improvement among the intervention group, but only among those with worse pretest scores.

Mediation. A full mediation analysis treated the mother's report of the quality of the mother-child relationship (a combination of scores on acceptance/rejection, communication, and positive routines) as mediating the influence of the intervention on mother's report of child problem behaviors. The indirect effect of the intervention on the outcome via the quality of relationship was significant (beta = .35 * -.22 = - .077, p < .01).

Study 3

This study is an effectiveness randomized trial that evaluated an adaptation of the New Beginnings Program (NBP) for divorced families tested in previous efficacy trials (Wolchik et al., 2007) for larger-scale delivery (Sandler et al., 2016). The authors stated that program modifications were surface level (e.g., more culturally relevant examples/content), and NBP's core components and fidelity were maintained. However, shortening the intervention from 11 sessions to 10 sessions and using a single local leader, rather than two leaders, may have changed the program substantially. Additionally, this study included fathers and ran separate groups for mothers and fathers.

Summary

Sandler et al. (2020) included mothers and fathers in a randomized controlled trial of 830 families who were assigned to the intervention or a low-dosage control condition. The intervention condition included separate groups for mothers and fathers. Measures of parenting skills, parent-child relationship quality, child internalizing, externalizing, and total problems were collected at pretest, posttest and 10-month follow-up.

Sandler et al. (2020) found that, compared to the control group, the program showed significant improvements in parent-reported:

• internalizing
• externalizing
• total problems
• relationship quality
• effective discipline

Evaluation Methodology

Design:

Recruitment: The study was conducted in partnership with family courts in two urban and two small-town, rural counties in Arizona. Parents were primarily recruited (92.4%) by viewing a 12-min invitational DVD shown during a four-hour parent information program mandated for all parents seeking a divorce or separation. Parents were also recruited through judge-, lawyer-, and self-referrals and through media announcements.

Parents who expressed interest were screened by phone based on the following eligibility criteria: filing for divorce or modification of a divorce decree within the past two years or, if never married, being in court in the past two years to establish or change a parenting time agreement following separation; having at least one child age 3 to 18 with whom the parent spends three hours or more each week or one overnight every other week; being able to complete the program and assessments in English; and not being mandated to a parenting class by the Juvenile Court or Child Protective Services.

Assignment: Using computer-generated random numbers, a total of 830 families (including 886 total parents with 56 pairs) were randomized to either the New Beginnings Program for divorced families treatment condition (N = 445) or the low dosage control condition that involved two group sessions (N = 385). The authors stated, "We excluded data from the second parent enrolled in each of these 56 pairs, resulting in a sample of 830 parents for our analyses."

Assessments/Attrition: The study included pretest, posttest, and 10-month follow-up assessments. The trial was conducted in four cohorts between 2012 and 2014. Attrition rates were 18% at posttest (N = 152 families) and 27% at the 10-month follow-up (N = 227 families) assessments.

Sample: The sample consisted of families in which either the mother (n = 474; 57.1%) or father (n = 356; 42.9%) enrolled in the study. At the pretest assessment, 262 (31.6%) of the 830 parents were divorced, 442 (53.3%) were legally married but divorcing, and 126 (15.2%) were never legally married but were in court to establish or change a parenting time agreement following separation. Parents were ethnically diverse (59.4% non-Hispanic White, 31.4% Hispanic, and 9.2% other race or ethnicity). Parents' education levels were: 3.7% less than a GED or high school diploma, 15.7% GED or high school diploma, 36.5% some college or vocational training, 14.5% associate degree, and 29.6% bachelor's degree or higher. Parents ranged from 18 to 63 years old (M = 37.46) and children ranged from 3 to 18 years old (M = 8.43); 47.8% of the children were female.

Measures: Data were collected using telephone interviews with parents and children ages 9-18 years and using online or paper-and-pencil questionnaires for teachers of children ages 6-18 years. Because parents were learning new parenting skills from the program, applying the program principles to their children, and thus helping deliver the program, the parent-rated measures of children may not be independently obtained. The timeframe for all measures was within the "past month." The authors reported good reliabilities for the well-known measures.

Parents and children completed the following parenting, parent conflict measures: parenting skills (parent-child relationship quality and discipline); family routines, using the adapted Family Routines Inventory; involvement; communication, using the Communication subscale of the Parent-Adolescent Communication Scale; closeness; child monitoring; three subscales of the Child Report of Parental Behavior Inventory (acceptance, rejection, and consistency of discipline); and, interparental conflict from the Children's Perception of Interparental Conflict Scale. Children also reported on their perception of being caught in the middle of their parents using the Caught in the Middle Scale, and on their exposure to badmouthing.

Parents also completed three subscales of the Oregon Discipline Scale (follow-through, appropriate use of discipline, and inappropriate use of discipline), the Child Behavior Checklist for children aged 6-18, and the Preschool CBCL for children aged 3-5 to measure internalizing, externalizing, and total problems. Children aged 9-18 and teachers completed the Brief Problem Monitor to measure internalizing, externalizing, and total problems. Teachers also completed the Teacher-Child Rating Scale to assess the following developmental competencies: learning problems, task orientation, frustration tolerance, assertive social skills, and social competence.

Analysis: Using the Mplus program, the main effects analyses in Sandler et al. (2020) included baseline covariates for the outcome, risk, county of residence, parent gender, and an indicator of whether both parents were enrolled in the effectiveness trial. The study reported clustering by family (mean ICCs for the parent-reported monitoring measure = .58; child-reported measures = .26; and teacher-reported measures = .16 across pretest, posttest, and 10-month follow-up). Outcome analyses also included a sandwich estimator for the standard error computations to adjust for clustering by family for parent reports of monitoring, child reports, and teacher reports.

This study conducted moderation analyses of program effects by baseline status on the outcome, parent gender, parent ethnicity, child age, and interparental conflict. For all main effect and moderator analyses, the authors applied a false discovery rate correction (FDR) to control the Type I error rate. For outcome domains with multiple measures (parent reports of parenting; child-report of interparental conflict; and parent-, child-, and teacher-reports of child mental health problems and developmental competencies), the authors reported the main and moderated effects with FDR p-values ≤ .10. For domains of outcomes with a single indicator, main and moderated effects were reported with ps ≤ .05

Intent-to-Treat: Sandler et al. (2020) conducted intent-to-treat analyses by using FIML estimation that includes participants with missing data.

Tien et al. (2018) supplemented the ITT approach by examining program non-attenders separately from program attenders. The approach was justified by the low attendance at program sessions and the use of an active control group, both of which limited the ability to detect intervention benefits. A total of 26 baseline measures were used to obtain propensity scores for three groups: the active control group, the inactive intervention group of non-attenders, and the active intervention group of those attending at least one session. The study used multiple-treatment generalized boosted models to generate the propensity scores. The analysis also imputed missing data and, to equalize differences across the non-randomized conditions, the models used inverse probability weights based on the propensity scores. With the weighting, all of the baseline standardized mean differences across the three conditions fell below .20.

Outcomes

Implementation Fidelity: Of the 445 parents randomized to the intervention, the mean number of sessions attended (either regular or make-up) was 5.59 (55.9% of the 10 sessions). A significantly greater proportion of intervention parents than comparison parents never attended, χ2 (1) = 6.44, p = .01. The mean fidelity of implementation score across activities indicated a high level of fidelity to the manual (2.86 out of 3).

Baseline Equivalence: The treatment and control groups differed on three of 33 demographic and baseline outcomes at pretest. Compared to the control condition, parents in the treatment condition were higher on child reports of positive parenting and lower on interparental conflict and badmouthing. Overall, the groups were equivalent on demographic and baseline outcomes.

Differential Attrition: There were no statistically significant differences in attrition rates across conditions at posttest or 10-month follow-up. Only four of the 60 attrition status main effects were significant and only two of the 60 Condition × Attrition Status interaction effects were significant. Overall, differential attrition was minimal given the large number of tests.

Posttest: The ITT results in Sandler et al. (2020) focused on moderation rather than main effects, with few significant main effects at posttest and 10-month follow-up. All significant main effects were parent-reported. Main effects results showed that five of 23 outcomes were statistically significant (p < .05) at posttest. Compared to control group parents, intervention group parents reported significantly more effective relationship quality and discipline (d = 0.22), while also reporting fewer child internalizing problems, child externalizing problems, and total child problems at posttest. Additionally, treatment group parents reported greater monitoring than control group parents, but only at the trend level (p = .07).

At the 10-month follow-up, there were no significant main effects at p < .05. However, there was one trend-level effect with parent reports of greater monitoring (p = .09) for the treatment group, compared to the control group.

Overall, the moderation results indicated positive effects for the intervention group compared to the control group to strengthen parenting at posttest and reduce child mental health problems at posttest and the 10-month follow-up. Many moderated effects showed greater benefits for non-Hispanic White families than Hispanic families (e.g., greater parent-child relationship quality, lower rejection, lower parent-reported child internalizing, externalizing, and total problems).

The non-ITT results presented in Tein et al. (2018) included 18 tests for main effects. They showed that the active intervention group reported significantly (p < .05) better outcomes than the inactive intervention group for five of the tests. At posttest, there were significant intervention effects for parent-child relationship quality, discipline, parental distress, and a child risk index. None remained significant at the 10-month follow-up, however. The only significant outcome at 10 months, interparental conflict, was not significant after application of the false discovery rate adjustment. Otherwise, the inactive intervention group did not differ significantly from the active control group on any of the outcomes.

Long-Term: Not examined.