Mindfulness-Based Stress Reduction for Parents Implementing Early Intervention for Autism: An RCT

Participating parents were recruited from a diagnostic clinic from 2015 to 2017. Eligibility criteria included having a child (<36 months of age at consent) with a gold standard ASD diagnosis and parental English fluency. Exclusion criteria included severe child sensorimotor impairment. All procedures were approved by the medical center institutional review board.

We used a random-number generator with a planned maximum enrollment of 70 participants. Allowing ∼10% attrition, this provides 80% power to detect an effect size of 0.71 for 1 primary outcome or an effect size of 0.91 if a multiple-comparison correction was applied for 7 coprimary outcomes. We placed half of the generated numbers into each group. Blinded staff obtained consent then accessed a unique computer file to determine group assignment.

Of 103 eligible families, 63 consented to participate (Fig 1). Two participants did not have data that could be included in analyses: 1 (P-ESDM only) provided inaccurate data and a second (P-ESDM plus MBSR) withdrew without providing any parental data. This yielded a final sample of 61 families included in analyses (Tables 1 and 2). Of these 61 families, 7 withdrew before completing their third P-ESDM session. Reasons for withdrawal included scheduling conflicts (n = 5) and repeatedly missed sessions (n = 2). An additional 8 families were lost to 6-month follow-up (moved away [n = 4] and lost to contact [n = 4]). This attrition rate is similar to that of other studies of parent-implemented, low-intensity ASD interventions^14,23,54,55  and did not differ by group or demographics.

Parent functioning variables were collected at each time point (see the Procedure section).

Parents completed the Parenting Stress Index (PSI), Third Edition, Short Form. This 36-item measure yields 3 subscale scores (Parent-Child Dysfunctional Interaction [PCDI], Parental Distress [PD], and Difficult Child [DC]) used in present analyses. Higher scores indicate more stress.

Parents completed the Center for Epidemiologic Studies Depression Scale (20 items) and the Beck Anxiety Inventory (21 items).⁵⁶  Higher scores on these measures reflect more symptoms.

Parents completed the Satisfaction With Life Scale,⁵⁷  a measure of subjective wellbeing. Its 5 items are scored along a 7-point Likert scale (higher scores indicate more satisfaction).

Parents completed the Five Facet Mindfulness Questionnaire.⁵⁸  The Five Facet Mindfulness Questionnaire consists of 44 items that yield 5 subscales. The total score was used in analyses. Higher scores indicate more mindfulness.

Before random assignment, parents provided the following baseline information: relationship to child, birth date, educational attainment, race and/or ethnicity of self and child, employment status, and annual household income.

The Autism Diagnostic Observation Schedule, Second Edition is a standardized clinical observation system for patients with developmental ages of 12 months and older.⁵⁹  Each module yields Calibrated Severity Scores (range: 1–10) based on age group and language level, which were used in these analyses. Higher scores reflect higher levels of autism symptoms.

The Mullen Scales of Early Learning is a standardized developmental test for children up to age 5 years.⁶⁰  It provides 4 domain scores (visual reception, fine motor, receptive language, and expressive language; mean = 50; SD = 10) and yields an overall ability index (Early Learning Composite; mean = 100; SD = 15). Higher scores reflect higher ability levels.

The Vineland Adaptive Behavior Scales–Second Edition Interview Form is a semistructured interview.⁶¹  It yields 4 domain standard scores (communication, daily living skills, socialization, and motor skills; mean = 100; SD = 15) and an overall Adaptive Behavior Composite (mean = 100; SD = 15). Higher scores reflect better adaptive behavior skills.

The Child Behavior Checklist for ages 1.5 to 5 years⁶²  is a parent-completed checklist of internalizing and externalizing behaviors. It includes 100 symptoms scored from 0 to 2 (higher scores indicate more symptom presence) within the past 2 months. T-scores ≥64 are considered clinically significant for broadband scales. Because of the high correlation between internalizing and externalizing subscales in this work (r = 0.61; P < .001), we used the T-score of the Total Problems scale.

After random assignment, families were scheduled for their first P-ESDM session and, when applicable, paired with an MBSR therapist. MBSR sessions began after the second P-ESDM session. Data were collected at 6 time points: baseline (before treatment), midtreatment (6 weeks), end of treatment (12 weeks), and 1, 3, and 6 months posttreatment.

P-ESDM consisted of 12 consecutive, weekly, clinic-based sessions that were ∼1 hour long. P-ESDM began an average of 1.35 months after diagnosis (SD = 1.45). Because of variable family schedules, it took an average of 15.87 weeks (SD = 3.33) to complete 12 sessions. Six P-ESDM therapists (separate from MBSR therapists and blinded to group assignment) were licensed, board-certified behavior analysts who were trained in fidelity by ESDM-certified trainers. This manualized intervention has a detailed parent-training curriculum and a specific coaching intervention method.⁵  Therapists introduce skills through descriptions, modeling, and coaching with embedded emphasis on skill generalizability. Fidelity was monitored by using the Early Start Denver Model Fidelity Checklist. Therapists covered 94% of intended content across sessions with no significant differences seen between groups (P > .10).

Parents who were randomly assigned to P-ESDM plus MBSR attended 6 additional 1-hour individual sessions. Child care was provided to maximize attendance. The 4 MBSR therapists were clinicians with at least a master’s degree and expertise in behavioral intervention and ASD supervised by a licensed clinical psychologist with formal MBSR training. Fidelity was monitored by therapist-completed content checklists.

The MBSR protocol was based on the work of Dykens et al¹⁴  and modified for individual administration for parents of young children with ASD. It was introduced as a skills-focused stress-reduction program rather than individual therapy. The clinic-based sessions covered topics such as an introduction to mindfulness for managing stress, awareness of the present moment, and cultivating gratitude. Weekly handouts offered written and pictorial practice cues for the home. Suggested homework exercises included formal practice (eg, guided meditation) as well as activities that are easily incorporated into daily routines (eg, breath awareness). Weekly homework logs tracked strategy use.

Multilevel models with discontinuous slopes, using Hierarchical Linear Modeling software,⁶³  were used to test for group differences in outcome changes over time. This allowed for inclusion of all participants with at least 1 time point of data. We modeled discontinuous rates of individual change with 2 time epochs⁶⁴ : (1) linear slope from the first to the third time point (baseline, 6 weeks into treatment, and immediately posttreatment), representing change during active treatment, and (2) change in linear slope from immediately posttreatment through the 6-month follow-up (immediately posttreatment and 1, 3, and 6 months posttreatment), reflecting change posttreatment. We estimated multilevel models for each outcome that included the between-subjects variable of treatment group, with groups centered on 0 (P-ESDM = −0.5; P-ESDM plus MBSR = 0.5). Overall slope estimates in these models reflect change across the entire sample. Including treatment group allowed us to determine if any of the slope estimates (slope representing change during active treatment and change in slope during posttreatment) significantly differed by treatment group. Additional between-persons covariates included child age at treatment initiation, sex of the child, sex of the parent, baseline autism severity, and child behavior problems. Because this is pilot work, no corrections for multiple comparisons were made. All covariates were grand mean centered.

No adverse events were reported. Baseline demographic and diagnostic information about children and participating parents are reported (Tables 1 and 2).

Estimates from the longitudinal models conditioned on treatment group and covariates for each of the outcome variables are presented (Table 3).

Model estimates (including β weights and SEs) for overall change across the sample for each outcome variable are plotted (Fig 2, Table 3). Across the sample, there was statistically significant average improvement during active treatment in all subdomains of parenting stress (PD, PCDI, and DC), depression symptoms, and anxiety symptoms. For the parenting stress subscales and depression symptoms, symptoms worsened slightly during posttreatment data collection (Fig 2). Change in posttreatment slope for anxiety was not statistically significant, indicating that improvement continued. There were no statistically significant average changes during active treatment or posttreatment in life satisfaction or mindfulness.

Model estimates (including β weights and SEs) for each outcome variable by treatment condition are plotted (Fig 3, Table 3). Groups were similar at baseline on all outcome variables except for PD. During active treatment, P-ESDM plus MBSR had greater improvements than P-ESDM in PD and PCDI. There was also a statistically significant treatment effect on mindfulness, with the P-ESDM plus MBSR parents’ effect increasing during active treatment and the P-ESDM group’s effect declining (Table 3, Fig 3). For PD, treatment group significantly impacted the change in slope posttreatment: P-ESDM plus MBSR had greater slowing of improvement posttreatment, resulting in posttreatment slopes similar to those of the P-ESDM group (Fig 3). Treatment group did not significantly predict slopes (during active treatment or posttreatment) for DC, depressive symptoms, anxiety symptoms, or life satisfaction.

A second aim was to examine how key covariates impacted response to treatment. As seen in Table 3, the covariate most consistently related to parental functioning was child behavior problems. Higher levels of behavior problems related to poorer parental functioning on all baseline variables. Relative to parents of children with low levels of behavior problems, parents of those with high low levels of behavior problems were higher on PD, PCDI, and depression at baseline but experienced significantly greater response to intervention during active treatment (Supplemental Fig 4).

Parent sex significantly predicted all estimates of PD (Table 3, Supplemental Fig 5). Relative to mothers, fathers started out with higher PD scores at baseline, improved more rapidly during active treatment, but then had greater slowing of improvement during posttreatment with scores that began rising during this time (whereas mothers stayed more constant during posttreatment). Parent sex also predicted slopes for mindfulness. Only mothers in the P-ESDM plus MBSR group increased in mindfulness during active treatment. Fathers declined in mindfulness regardless of treatment group, although that decline was less pronounced if they were in P-ESDM plus MBSR. Change was attenuated in all groups during posttreatment.

Child age predicted initial life satisfaction, change in slope during active treatment, and change in slope posttreatment. Parents of younger children at baseline had life satisfaction scores that improved during treatment but then declined posttreatment. Parents of older children showed the opposite pattern, with scores that declined during treatment but improved posttreatment (Supplemental Fig 5).

Other covariates had minimal or no impact on treatment response. The sex of the child predicted posttreatment response for life satisfaction but not response during active treatment. Child autism severity did not relate to initial status or slopes for any parental outcome variables.

In addition to our primary analyses, we ran multilevel models to examine whether the point estimates of parent functioning at the end of active treatment (3 months) and the end of follow-up (9 months) differed by group. To do this, we reran the multilevel models for each outcome that included treatment group and covariates but changed the intercept from baseline to the 3-month follow-up and from baseline to the 9-month follow-up. Estimates representing the difference between groups at each of these time points are presented (Table 4). At the end of active treatment, P-ESDM plus MBSR had marginally lower scores on PD and PCDI. These marginally significant differences persisted through the end of follow-up for PCDI. Note that none of the group differences in point estimates reached statistical significance at P < .05.

This study provided MBSR to the parents of newly diagnosed young children with ASD undergoing low-intensity, parent-mediated early behavioral intervention. This combination treatment approach addressed the well-established risk for increased parenting stress and psychopathology while also teaching parents foundational skills for interacting with their children.

Relative to parents who received P-ESDM only, parents who received P-ESDM plus MBSR showed significantly greater reductions in PD and perceptions of dysfunctional child interactions. Other work has documented that parent training may stabilize⁶⁵  or reduce⁶⁶  parenting stress. Our results suggest that giving parents stress-reduction strategies may enhance that potential. Although there were differences in the rate of change, 3 months was not long enough for that to translate into statistically significant group differences in parent functioning at the end of treatment. This suggests that detecting specific point-in-time group-level differences may require a larger sample size or prolonged treatment exposure.

Self-rated mindfulness improved for P-ESDM plus MBSR but decreased for P-ESDM only. Gains in mindfulness made by the P-ESDM plus MBSR group during active treatment were maintained during follow-up, suggesting that the impact of brief, targeted MBSR training may persist past the intervention period. Both groups showed reduced depression and anxiety symptoms during active treatment. However, with the exception of anxiety, most improvements that were made during active treatment were slowly lost over time with scores that approached baseline levels by 6-month follow-up. Thus, families may benefit from ongoing treatment to maintain initial gains. This contrasts with the findings of Dykens et al,¹⁴  who noted sustained gains in most outcomes other than anxiety, especially for mothers of children with ASD.

Covariates emerged as significant predictors of some outcome variables. Fathers improved more during active treatment but had greater slowing of improvement during follow-up. Fathers may warrant specific attention within the literature as an understudied group with potentially different pathways of stress and coping.^{39–41,67–69}  Parents who initiated treatment at a younger age reported higher levels of stress and distress at baseline, improved quickly during treatment, but had slowed progress during follow-up relative to slightly older children. Of note, children within our study were on the cusp of transitioning from Part C to Part B services within the Department of Education. The impact of changes in concurrent intervention use on parent outcomes will be examined in future publications.

Baseline child behavior problems were related to higher baseline levels of parenting stress and depression as well as more rapid improvements in these outcomes during intervention. It is unclear to what degree this may be a function of all 3 variables being self-reported or the phenomenon of regression to the mean. The relation between child behavior problems and parenting stress in ASD is well established^70–73  but complex, with recent work suggesting an iterative, interactive process.³⁷  Future work should consider how all of these variables may interact over time given the potential bidirectional nature of these relationships.³⁷ 

This initial diagnostic period for young children represents a critical point worthy of attention. Higher levels of parent stress may impact a family’s ability to obtain and implement intervention.^28,46,65  This period also represents an opportunity for increased understanding and relationship building between parents and young children, whose core social and communication symptoms may be especially challenging for parents to interpret. Importantly, however, baseline autism severity did not significantly impact initial scores on parenting outcomes nor change in outcomes over time. Rather, it was parent-reported levels of behavioral challenges that were significantly related to parental stress and depression.

Although this work was longitudinal, its follow-up time frame was relatively brief. Our sample size was underpowered to conduct subgroup analyses and test mechanisms of treatment response. Several participants were lost to attrition, which was similar across groups and split between families that were lost to contact and families that moved away. This suggests that adding an extra weekly MBSR visit was not too great a scheduling burden relative to P-ESDM only. Parents were allowed to reschedule missed sessions, which added variability to treatment length. Additionally, analyses included all participants who provided valid data regardless of level of participation; this provided a more conservative estimate of treatment effects. Although we collected information on fidelity of clinician P-ESDM implementation, fidelity of parent implementation was not tracked, preventing us from drawing conclusions about the impact of MBSR on parents’ treatment implementation. Most participants self-identified as white women, limiting generalizability of results to other groups, and information on parenting self-efficacy was not collected. Most participating families had limited access to high-intensity treatments, and it is unclear how MBSR would pair with other intervention models. Also, although random assignment of children to a no-treatment group would be unethical, without an untreated group, questions remain about which longitudinal effects are because of treatment versus maturation or another time-related correlate.

Finally, by nature of the study design, the P-ESDM plus MBSR group received more intervention sessions than the P-ESDM group. Thus, it is unclear whether additional parent functioning gains made by the P-ESDM plus MBSR group are due to greater attention or the MBSR program itself. Future work should include attention-matched groups to further evaluate standalone intervention impact. Examining mindfulness as a mediator of treatment on parenting stress will also be important when determining whether MBSR is the mechanism of action, as will investigating the impact of child response to intervention on parent outcomes.

This study suggests that high-quality, low-intensity early intervention was associated with improvements in PD and parenting stress. Adding MBSR resulted in additional improvements in parenting stress. At no point did parents in P-ESDM plus MBSR show more symptoms of stress than parents in P-ESDM only, suggesting that the additional requirements on parents’ time did not seem to be harmful within this time-limited intervention study. Findings may have implications for systems that partner with parents to care for young children as well as practitioners who work directly with parents independent of child intervention. Future analyses will explore whether particular profiles of children and families are more amenable to this combination.

ASD

autism spectrum disorder

DC

Difficult Child

MBSR

mindfulness-based stress reduction

PCDI

Parent-Child Dysfunctional Interaction

PD

Parental Distress

PSI

Parenting Stress Index

P-ESDM

parent-implemented Early Start Denver Model