Abstract
The coaching service delivery model is often implemented within parent-mediated interventions for infants at an elevated likelihood of autism spectrum disorder. However, less is known about the exact coaching behaviors used within intervention sessions. Therefore, we examined the coaching behaviors implemented within the adaptive responsive teaching intervention. We also investigated the associations between coaching behaviors and parent education levels and the associations between the coaching behaviors and a parent outcome, that is, parent responsiveness. Results indicated that the interventionists demonstrated joint interaction and child-focused behaviors the most frequently. The interventionists much less frequently demonstrated guided practice, caregiver practice, and problem-solving behaviors. The use of joint interaction behaviors was positively associated with parent education levels, whereas the use of child-focused behaviors was negatively associated with parent education level. More information sharing by the interventionists predicted a greater change in parent responsiveness, whereas more child-focused behaviors predicted less change in parent responsiveness.
Keywords
Introduction
Parent-mediated interventions are the most commonly used interventions among children under the age of 2 years who are at an elevated likelihood for a later diagnosis of autism spectrum disorder (EL-ASD) or diagnosed with autism spectrum disorder (ASD) (Bradshaw et al., 2015). These interventions focus on training parents (or other caregivers) to implement strategies with their children since parents typically spend more time with their children than other adults. A common service delivery approach used in parent-mediated interventions is coaching. Rush and Shelden (2011) defined coaching as an adult learning strategy in which the coach promotes the learner’s ability to reflect on his or her actions as a means to determine the effectiveness of an action or practice and develop a plan for refinement and use of the action in immediate and future situations. (p. 8)
Thus, within the coaching model, interventionists collaborate with parents to develop appropriate strategies and practice those strategies within the intervention sessions to promote parents’ use of the strategies during daily interactions with their children (Woods et al., 2011).
There are many benefits of the use of the coaching service delivery model within parent-mediated interventions. This type of service model follows recommended practices of the Division for Early Childhood (DEC) and guidelines of Part C of the Individual with Disabilities Education Act (IDEA), which encourage and require family-centered and family capacity-building practices (DEC, 2004; IDEA, 2004). In addition, within the coaching service delivery model, interventionists are expected to facilitate active participation from the parents throughout the intervention sessions. Active participation is a significant component of adult learning methods because the more opportunities adult learners have to engage within the learning process, the easier it is for them to acquire new knowledge (Trivette et al., 2009). Thus, guidelines governing practices for young children with developmental disabilities and the coaching model itself highlight the importance of interventionists implementing the intervention in a manner that encourages parents to be an active participant within the sessions. Yet, the manner in which the intervention is delivered to the parent is often not discussed in detail in research studies.
Coaching Behaviors
There are many different types of coaching models available, and definitions of coaching vary across studies (Kemp & Turnbull, 2014; Lorio et al., 2020). As a result, the types of coaching behaviors used by interventionists within interventions may differ. Some coaching behaviors examined in prior research are information sharing, observation, joint interaction, direct teaching, demonstration with narration, guided practice, caregiver practice, feedback, problem-solving, and reflection (Brown & Woods, 2015; Friedman et al., 2012). The frequency with which these behaviors are used within intervention sessions varies, with some observed more frequently than others across studies.
One frequently observed interventionist behavior within parent-mediated intervention sessions is the child-focused behavior (McBride & Peterson, 1997; Salisbury & Cushing, 2013). McBride and Peterson (1997) described this behavior as occurring when the interventionist is focused solely on the child. Interventions that primarily use this behavior are following a more traditional approach wherein interventionists work directly with the child and do not explicitly encourage active participation by the parent (Fleming et al., 2011). This behavior does neither follow the guidelines of family-centered practices nor conform to Rush and Shelden’s (2011) definition of coaching; therefore, child-focused behavior often is discouraged within coaching interventions.
In contrast to child-focused behaviors, interventionists use two behaviors less frequently: reflection and problem-solving (Lorio et al., 2020). Lorio et al. reported that problem-solving and reflection were often not included in the intervention descriptions within their review of intervention studies. In addition, they noted that when problem-solving and reflection were included in studies, the authors often failed to provide sufficient descriptions of these coaching behaviors. Lorio et al. review the findings to support the need for further investigation into not only which behaviors are being used within intervention sessions but also factors that impact their use.
Demographic Variables and Coaching Behavior Use
Variability in the use of coaching behaviors may be associated with interventionist factors, such as level of training, area of training, and years of experience, or with characteristics of the family and child such as the culture, motivation, and learning styles of the parent(s) participating in the intervention (Fleming et al., 2011; Friedman et al., 2012; Meadan et al., 2018). One particular demographic variable associated with the coaching behaviors used by interventionists is the socioeconomic status of the families receiving the intervention, specifically the education level of the parent. Sawyer and Campbell (2017) indicated that interventionists are more likely to engage in child-focused behaviors and not describe these behaviors to parents when the parents have attained less education. This may be due to the interventionists’ perceptions of the parents’ needs as it relates to effectively implementing the intervention. For example, Fleming et al. (2011) found that providers indicated that the education level of the parents was one factor associated with parent participation within the intervention sessions. Within this same study, some providers indicated that the parents’ education level was associated with the ease with which they were able to implement the intervention as intended. Therefore, some interventionists may have preconceived notions about a family’s understanding of the intervention and how much they will participate in it based on the family’s education level and select coaching behaviors accordingly.
Coaching interventions and parent outcomes
Finally, research on the impact of the use of specific coaching behaviors on adult outcomes is limited. However, some recent literature has described the relationship between coaching behaviors and parent outcomes. For example, Brown and Woods (2015) indicated that coaching behaviors requiring parent participation (i.e., interventionist observation of the parent, guided caregiver practice, and caregiver practice with feedback) were linked to the parents’ immediate use of intervention strategies within the intervention session. Currently, it is unclear whether this same relationship exists between coaching behaviors used and changes in parent behaviors measured after the intervention ends. Therefore, the relationship between coaching behaviors and parent outcomes over time should also be investigated.
One specific parent outcome investigated in early childhood development is parent responsiveness. Children diagnosed with ASD demonstrate challenges in their ability to engage with others (Adamson et al., 2009). Adamson and colleagues reported that children with ASD were less likely to demonstrate coordinated joint engagement than children with Down syndrome or children who were typically developing. Thus, within the Adamson et al. study, children with ASD showed fewer instances in which they shifted their attention from an object and acknowledged the other person engaged in the same activity. Parents can perhaps compensate for their children’s impairments in engagement with others by becoming even more responsive than they would naturally be to their child’s cues. One intervention that focuses on building parent responsiveness is Responsive Teaching (Mahoney & Perales, 2005). This intervention aims to increase parent responsiveness to build the child’s pivotal behaviors (e.g., joint attention, imitation) with the long-term goal of promoting better language and social-emotional development (Mahoney & Perales, 2005). Mahoney and Perales documented that, following the intervention, parents demonstrated an increase in responsiveness to the child, and their increased responsiveness was associated with improved child outcomes. Similarly, Watson et al. (2017) used an adapted version of this intervention, adaptive responsive teaching (ART), to train parents of children at EL-ASD on responsive strategies to improve pivotal child behaviors in the domains of social communication and sensory regulation.
The purpose of the current study was to provide new information on the coaching strategies that interventionists who were implementing ART intervention used with parents within intervention sessions, factors that may impact the coaching strategies used, and the relationship between coaching strategies and the changes in adult outcomes. Therefore, the following research questions were investigated in this study:
Method
This study comprised a secondary analysis of extant data collected within a clinical trial that examined the efficacy of ART (Watson et al., 2017). The methods included new coding of coaching videos collected in Watson et al. to address the research questions regarding coaching behaviors.
Overview of ART
Responsive teaching (Mahoney & MacDonald, 2007) was originally developed as a parent-mediated intervention for parents of preschoolers diagnosed with developmental disabilities and was initially implemented in a clinical setting. Responsive Teaching addresses child pivotal behaviors grouped under three developmental domains: cognition, communication/language, and social/emotional. The Responsive Teaching manual provides extensive resources, including collaborative assessment strategies to help parents and interventionists select the pivotal behaviors they will focus on supporting in the child. It also includes “Discussion Points” to use with parents to explain the importance of different pivotal behaviors and the rationale for using responsive strategies, a large number of potential responsive strategies, and guidance for matching responsive strategies to different pivotal behaviors. In adapting Responsive Teaching for infants at EL-ASD and their parents, Watson et al. (2017) modified the intervention to be applicable to a younger population entering intervention without a current diagnosis of a developmental disability (i.e., infants at EL-ASD and their families) and for delivery within the home setting. These modifications included revising some language in discussion points and parent handouts to remove implications that the child had a known disability and adding a supplemental assessment interview of the family’s daily routines and how the child participated in those routines within the home setting. In addition, Watson et al. adapted the intervention to focus on pivotal behaviors within two primary developmental domains particularly relevant to the identification of infants at EL-ASD, that is, social communication and sensory regulatory (Turner-Brown et al., 2013). The pivotal behaviors addressed in the social communication domain (i.e., social play, joint activity, joint attention, vocalization, intentionality, and conversation) were drawn from behaviors already included in Responsive Teaching (Mahoney & MacDonald, 2007) under its cognition and communication/language domains. The pivotal behaviors included under the sensory-regulatory domain included two behaviors from Responsive Teaching’s social/emotional domain (i.e., self-regulation and cooperation) along with the new pivotal behaviors of attention and arousal, exploration, engagement, and adaptability and coping. ART interventionists used the Responsive Teaching (Mahoney & MacDonald, 2007) manual as their primary reference, along with supplemental, project-specific written materials related to the sensory-regulatory pivotal behaviors added for ART. This included added responsive strategies that parents could be coached to use when targeting pivotal behaviors in the sensory-regulatory domain. Watson et al. (2017) also adapted the Responsive Teaching fidelity checklist to align with the modifications made for ART.
In implementing ART, the interventionists identified pivotal behaviors in the sensory-regulatory and/or social communication domain to target based on the interventionist’s and parent’s collaborative assessment of the child’s current behaviors. The interventionist discussed the selected behavior with the parent and then coached the parent to use responsive strategies to encourage that pivotal behavior. The planned intervention schedule for ART was 30 1-hr in-home sessions provided over 6 months (two sessions per week initially, dropping to one session per week after 6 weeks); the intervention could continue up to 8 months if the 30 sessions had not been completed. See supplemental materials from Watson et al. (2017) for more detailed information about the ART intervention.
Participants and Setting
Families
Recruitment for the ART trial occurred through mailings of the First Year Inventory (FYI) (Baranek et al., 2003) to families of 1-year-old infants within six counties in the state of North Carolina (Watson et al., 2017). The families of children whose results met the FYI cut-off criteria for being at an elevated likelihood for later ASD (Turner-Brown et al., 2013) were invited to participate in this research study. There were 87 infants enrolled in this study, with 45 of the infants randomized to the ART intervention (Watson et al., 2017). At least one intervention fidelity video was available for 43 of the 45 families randomized to the ART group; videos available for these 43 families were used for the current study. ART took place in families’ homes, and videos were recorded in that setting. The average age of the children included in this study was 13.8 months (SD = 0.72) at preintervention and 22.6 months (SD = 0.98) at postintervention. The mother was the primary parent who participated in the intervention for 21 of the 43 families (Watson et al., 2017). For the remainder of the families, the father alone (one family), a grandparent (two families), or both parents (19 families) participated in the intervention. For the purposes of this article, the participating caregivers will be called “parents” due to the predominance of parent caregivers in the study. The majority of the parents were White (74%, n = 32). The education level of the primary parent ranged from 9th to 11th grade to a graduate or professional degree. The majority of the parents had obtained a 4-year college degree or greater (72%, n = 31). The age of the majority of parents ranged from 30 to 39 years, and the average age was 33 years old. See Table 1 for specific details about the demographics of the parents. For the purposes of analysis, the different education levels were reduced to three groups (i.e., some college or less, 4-year college degree, graduate/professional degree), as shown in Table 1.
Sample Demographics of Primary Parent.
Note. GED = general education development.
Interventionists
The six interventionists who provided ART coaching had previous knowledge about or experience working with children with developmental disorders. All six were White and female. Two of the interventionists had a bachelor’s degree in psychology, two had masters’ degrees in occupational therapy, one had a master’s degree in early childhood special education, and one had a master’s degree in psychology. Their years of related experience working with parents and children, including children with developmental disabilities, ranged from 2 to 24 years.
All of the interventionists were trained on the ART intervention by the lead interventionist, who had been trained in Responsive Teaching by two of the developers, Drs. Gerald Mahoney and Frida Perales. The interventionists’ fidelity to the intervention was measured using the ART Implementation and Fidelity checklist, a 25-item tool adapted from Mahoney and MacDonald (2007) by Watson et al. (2017). Fidelity raters were asked to rate the extent to which expected interventionist behaviors were observed (i.e., minimally, moderately, maximally) using a scale of 1 to 7. The more the interventionists used the specific behaviors, the higher the rating score. As reported by Watson et al., the interventionists obtained a fidelity score of .87 on the ART fidelity tool, which was classified as “good” based on qualitative fidelity levels. Although interventionists were not explicitly trained on the coaching behaviors as defined by Friedman et al. (2012) and applied to the current investigation (see Table 2), all of the coaching behaviors included in the current study were either named or described on the fidelity rating tool. Therefore, interventionists’ use of the coaching behaviors investigated in the current study was expected in the ART intervention as part of the behaviors associated with high fidelity implementation of ART.
Coaching Behavior Descriptions (Friedman et al., 2012).
Measures
Coaching behaviors
Coaching behaviors were coded using definitions from Brown and Woods (2015) and Friedman et al. (2012) to capture the frequency of the coaching behaviors used within the videos (see Table 2 for coaching behaviors and definitions).
Parent responsiveness
A parent responsiveness coding system adapted from Yoder et al. (2015) was used to measure parent responsiveness within the original intervention study. The coding manual was adapted to provide instructions for use with different coding software than used by Yoder et al. Also, the project staff added examples of behaviors seen in the interactions between parents and their infants and toddlers at EL-ASD to clarify coding decisions with the project sample. The codes, definitions, and decision rules used by Yoder et al. were maintained for the purposes of this project. Coders were trained to code child leads and parent responses within 5-s intervals (Watson et al., 2017). Watson et al. obtained parent responsiveness percentages by dividing intervals containing parent responsive behaviors by the total of codable intervals.
Demographic form
The demographic form provided information about the educational level of the parents, and the age and race/ethnicity of the children.
Procedures
Preintervention and postintervention
Watson et al. (2017) collected the demographic information at preintervention. Parent responsiveness was assessed at preintervention and postintervention.
Intervention and coaching behavior coding
The ART intervention occurred within the homes of the participants. A maximum of six intervention sessions were recorded for fidelity purposes for each family. The primary author of the current study randomly selected two fidelity videos per family and used those videos to analyze the coaching strategies used within the intervention sessions. There was only one video available to code for four of the families. A total of 82 videos were coded, with each video coded in its entirety. The sessions ranged from 20 to 70 min long, with the average session length being 47.58 min (SD = 10.40).
The first author trained five research assistants on an adapted version of the Coaching Coding Manual developed by Friedman et al. (2012). The manual was adapted for this study to more reliably and completely capture behaviors demonstrated by the interventionists. We reviewed the definitions and examples provided in the manual and then used it to code intervention videos not included in this study. We maintained all definitions of coaching behaviors from Friedman et al; however, we added descriptions and examples to the coaching behaviors defined in the manual to assist in reliably classifying behaviors demonstrated by the ART interventionists. For example, we added more description to the “no coaching” behavior to include times when the interventionist was interacting with a sibling of the child in the intervention and not with the child who was participating in the intervention or the parent. For the “observation” behavior, the original manual included examples of this behavior occurring before specific behaviors (e.g., problem-solving, reflection, guided practice) to distinguish it from caregiver practice in which the interventionist is observing the parent practice a specific behavior. However, in ART implementation, observation did not always occur before those specific behaviors, but nevertheless met the definition of observation. Therefore, we included more examples in the manual to avoid overly restrictive coding of the observation behavior as applied to ART. We also added more examples of the behaviors that required input from both the parent and the interventionist (i.e., problem-solving, information sharing) to assist in recognizing and differentiating the behaviors. We consulted with one of the developers of the manual to ensure that the adaptions aligned with the established definitions.
A total of 12 possible behaviors were coded. One behavior, no coaching, included three modifiers (i.e., child-focused, no opportunity, and other). For the final analyses, we used the three modifiers for “no coaching” behavior to disaggregate “no coaching” into three distinct behavior categories to examine the frequency with which specific “no coaching” behavior categories occurred within the intervention sessions. Therefore, there were a total of 14 behaviors coded in this study.
Coded behaviors defined and described in the Coaching Coding manual by Friedman et al. (2012) were information sharing, observation, joint interaction, direct teaching, demonstration with narration, guided practice, caregiver practice, feedback, problem-solving, reflection, and no coaching-child-focused, no coaching-other, and no coaching-no opportunity (see Table 2 for a description of each behavior). For the purposes of this study, we added “uncodable” to serve as a code for instances in which the coder was not able to see or hear what the interventionist was doing in relation to the family. All coaching behaviors were based on the actions of the interventionists. Therefore, the interventionist had to demonstrate the behavior for it to be coded as a specific behavior. However, some of the coaching behaviors also required interactions from the parent to be coded as that particular behavior. For example, the problem-solving code required both the interventionist and the parent to contribute to the conversation beyond an affirmation (e.g., nodding, saying “yes”) (Friedman et al., 2012).
The coders used an observational coding system set up in Noldus Observer XT 14.2 to code the coaching behaviors observed within the intervention fidelity videos, using partial interval coding as used by Friedman et al. (2012). Each video was segmented into 30-s intervals. The coders coded all of the behaviors that occurred within each 30-s interval. Thus, our data reflect the frequency of intervals in which each coaching behavior was used within each session. We chose partial interval coding over counting coaching behavior events or continuously coding the time that interventionists engaged in each coaching behavior because we judged this approach to provide the best comparison across diverse coaching behaviors. That is, for some behaviors, one distinct event will take place over a longer period of time (e.g., joint interaction and child-focused behaviors). In contrast, for other behaviors, a unique event is typically completed within brief intervals of time (e.g., feedback, guided practice). Thus, measuring the occurrence of these behaviors using partial interval coding provided us with a common “scale” on which we could capture variation in the use of these different types of coaching behaviors. From the partial interval coding, we obtained a proportion for each coaching behavior by dividing the frequency of intervals in which each coaching behavior was demonstrated by the number of 30-s intervals within the full video. The coders reached a kappa reliability of at least .75 across three training videos before they began coding the videos for this study.
Reliability
The primary author randomly selected and coded 20% of the videos for reliability. Reliability was measured by computing intraclass correlation coefficients (ICCs) for the proportions of all of the behaviors combined, using two-way mixed-effects models for absolute agreement. The ICCs for each reliability video ranged from .82 to .99 (Table 3). Based on the standards listed by Koo and Li (2016) and Portney (2020), the average ICC for all videos was excellent (ICC = .97). If appropriate, we also computed ICCs for proportions of individual behaviors, again using two-way mixed-effects models for absolute agreement. Six of the behaviors (i.e., problem-solving, reflection, guided practice, no coaching-no opportunity, uncodable, caregiver practice) showed limited distributions of the number of intervals in which the behavior occurred. These behaviors were zero-inflated, with not even a single occurrence of each behavior coded in at least one third of the videos. In addition, on average these behaviors occurred in less than 3% of the intervals coded for each session. Therefore, a valid ICC could not be calculated for these behaviors. The individual ICCs for the other behaviors ranged from .82 to .98 (Table 4).
Intraclass Correlation Coefficient (ICC) for Each Coded Video.
Intraclass Correlation Coefficients (ICCs) for Coaching Behaviors.
Statistical Analyses
We performed all analyses using Stata 16 (StataCorp, 2019). To answer RQ1 regarding the relative frequencies of coaching behaviors, we averaged the proportion of each behavior across all the sessions. For RQ2, we extracted the parent education level from the demographic form. For the purposes of this analysis, education was grouped into three levels: (a) some college or less, (b) college: 4-year degree, and (c) graduate/professional degree. See Table 1 for specific demographic details. We then averaged the proportions of each behavior for participants with two observations to ensure the proportion of behaviors per participant were equally weighted within the models. For the four families for whom there was only one video available, we used the frequency numbers available for the one video in the analyses. Next, we conducted a multiple regression analysis, regressing the proportion of coaching behaviors on the parent education levels, types of behaviors, and interaction of education levels and types of behaviors. We then conducted an F-test to determine whether there were significant differences between the education level of the parents and the proportion of each type of behavior and the magnitude of those differences. For RQ3, we conducted multiple regression analyses to determine whether any specific coaching strategies were significantly related to the change in parent responsiveness from preintervention to postintervention after controlling for initial parent responsiveness. The following behaviors were not included in the analyses for RQ2 or 3 due to their zero-inflated distribution and limited variability across coaching sessions: problem-solving, reflection, guided practice, no coaching-no opportunity, uncodable, and caregiver practice. The zero-inflated distributions for these behaviors violate the assumptions for multiple regression analysis, and their overall low frequency and variability make it unlikely they would account for meaningful variance in any dependent variables of interest.
Results
Frequency of Coaching Behaviors
Our RQ1 pertained to the relative frequency of coaching behaviors used by ART interventionists. The behavior that occurred most frequently in the sessions was joint interaction (Figure 1). On average, the parent, child, and interventionist were all engaging in activities together for a third of intervals within the session. The next most frequently occurring behavior was the no coaching-child-focused behavior. These findings indicated that the interventionists interacted directly with the child and not the parent for approximately a quarter of the intervals. Information sharing and direct teaching both occurred in approximately one fifth of the intervals within the sessions. These findings reflect that the interventionists often encouraged the parents to share information about their child’s development. They also spent time providing the parents with information about the pivotal behaviors used in ART and corresponding strategies to support the development of the pivotal behaviors.
Average proportion of intervals of coaching behaviors.
Observation and feedback occurred at more moderate frequency levels. On average, each of these behaviors occurred in less than one tenth of the intervals coded. Demonstration with narration occurred in approximately 3% of the intervals coded within a session, on average. The behaviors that occurred at the lowest frequencies overall (<3% of the intervals, and often not observed at all during a full session) included reflection, no coaching-no opportunity, caregiver practice, guided practice, problem-solving, and uncodable.
Coaching Behaviors and Education Levels
Our RQ2 addressed whether interventionists employed different behaviors in association with differing levels of parental education. The eight behaviors included in this set of analyses were: no coaching-child-focused, information sharing, direct teaching, joint interaction, no coaching-other, observation, demonstration with narration, and feedback.
There were significant differences between the use of the no coaching-child-focused behavior when comparing the parents with lower levels of education to the parents with higher levels of education (Figure 2). Specifically, interventionists working with parents whose education level were classified as some college or less used child-focused behaviors in more intervals than they did with parents with a 4-year college degree (contrast = −0.17, F= 33.24, p < .01) and parents with a graduate or professional degree (contrast = −0.19, F = 44.51, p < .01). In contrast, the interventionists used joint interaction in fewer intervals with parents whose education was classified as some college or less than parents who had a 4-year college degree (contrast = 0.13, F = 18.45, p < .01) or a graduate/professional degree (contrast = 0.14, F = 25.29, p < .01). There were no significant differences in the other interventionist behaviors based on the education level of the parents.
Average proportion of intervals of coaching behaviors based on education level.
Coaching Behaviors and Change in Parent Responsiveness
To answer our RQ3, we examined whether any of the eight interventionist behaviors observed consistently across sessions was associated with changes in parent responsiveness over the course of the ART intervention. Parent responsiveness at preintervention was significantly related to the change in parent responsiveness within the families that participated in the ART intervention (β = −0.63, t =−4.53, p < .01), such that parents with higher responsiveness at preintervention showed less change in responsiveness over time (Table 5, Model 1). In addition, two coaching behaviors were significantly associated with change in parent responsiveness, after controlling for preintervention parent responsiveness. First, the proportion of intervals with information-sharing behaviors was associated with change in parent responsiveness (β = 47.98, t = 2.40, p = .02) (Table 5, Model 2). Specifically, the more intervals that included information sharing, the larger the change in parent responsiveness from pre- to postintervention. Second, the association between the proportion of intervals that contained child-focused behaviors and the change in parent responsiveness after controlling for initial parent responsiveness was also significant (β = −48.09, t = −3.39, p = .02) (Table 5, Model 8), with the negative direction of effect indicating that the more intervals in which interventionists participated in child-focused behaviors, the less change occurred in parent responsiveness. In addition, the association between the observation behavior and the change in parent responsiveness approached significance (β = 63.55, t = 1.77, p = .09) (Table 5, Model 3), such that more use of observation behavior by the coach was associated with a greater change in parent responsiveness. The remaining five interventionist behaviors included in this set of analyses (direct teaching, joint interaction, demonstration with narration, feedback, and no coaching-other) showed no association with changes in parent responsiveness. Finally, we conducted a regression model with both coaching behaviors that were significant independently (i.e., information sharing and no coaching); within this model, both preintervention parent responsiveness (β = −0.77, t = −4.55, p < .01) and no coaching (β = −40.04, t = −2.44, p = .02) remained significant (Table 5, Model 10), but information sharing did not account for significant unique variance.
Results of Models of Multiple Regression Analyses Predicting Change in Parent Responsiveness From Initial Parent Responsiveness and Coaching Behaviors.
Note. Model 1 only includes initial parent responsiveness at Time 1. Models 2–9 include initial parent responsiveness and individual coaching behaviors, and Model 10 includes initial parent responsiveness, information sharing, and no coaching-child-focused.
p < .05, two-tailed significant.
Discussion
The purpose of this study was to investigate the frequency of coaching behaviors used within the ART intervention and explore the relationship between those behaviors and parent education levels and parent responsiveness. When examining the average use of all 14 behaviors within all coded videos, we saw that interventionists engaged in facilitating joint interactions in more intervals than any other coaching behaviors. This finding contrasts with previous research in which interventionists were reported to be engaged in child-focused behaviors most frequently (Peterson et al., 2007). This may imply that the ART interventionists were encouraging parents to play a more active role in the intervention session and facilitating opportunities for parents to be involved to a greater extent than occurred with previously examined parent-mediated interventions. This behavior, however, has its limitations within the context of coaching behaviors. As joint interaction captures a variety of interactions in which the interventionist is not explicit about the intended focus or purpose, we know less about if or how the interventionists are training parents on the child’s skills or recommended strategies within this behavior (Marturana & Woods, 2012). Therefore, it is important to recognize that although this coaching behavior involves participation of the parent, the use of joint interactions may not necessarily facilitate adult learning to the same degree as some other coaching behaviors that encourage active participation from the parent, such as caregiver practice or problem-solving.
Although, on average, joint interaction behaviors occurred the most frequently across all of the sessions, the no coaching behaviors, particularly the child-focused behavior, also occurred frequently within the ART intervention sessions. The frequent use of this set of behaviors showed that a large amount of time was spent implementing behaviors that are not designed to directly address parent–child interactions and build the parents’ capacity to aid in their child’s development (Woods et al., 2011). Therefore, it is essential for us to develop a better understanding of why these behaviors continue to be implemented so often within intervention sessions.
Child-focused behavior, the second most frequently occurring type of behavior, is often discussed in the intervention literature as a behavior that should occur less frequently within parent-mediated intervention sessions (McBride & Peterson, 1997; Peterson et al., 2007; Woods et al., 2011). However, as noted by Salisbury and Cushing (2013), there are some instances in which the child-focused behavior is warranted. For example, interventions that focus on a child’s motor abilities might require interventionists to work directly with the child during a portion of the session (Salisbury & Cushing, 2013). In this case, the interventionists may need to focus on the child to assess the child’s skills accurately and to recommend appropriate interventions for the family. Therefore, it is important to note that this behavior is useful in some contexts and should not be completely removed from all intervention services. With this notion, Salisbury and Cushing (2013) also stressed the importance of keeping this behavior to a minimum, because time spent in child-focused behavior suggests that there is less time spent in interactions between the parent and interventionist and the parent and child. These ideas encourage more consideration about the function of the child-focused behavior and whether or not it can be shaped into other behaviors more conducive to active parent participation. For instance, the interventionist could describe the skills that they are assessing to the parent while they are working directly with the child or try to incorporate the parent in the interaction with the child by prompting the parent to implement specific strategies to assess the child’s skills.
Of relevance to this point, McBride and Peterson (1997) noted that interventionists may show child-focused behaviors within intervention sessions due to their belief that they are modeling the strategies to the parents. In fact, there were instances within the recorded ART intervention videos in which interventionists appeared to be modeling a strategy. However, they did not describe the strategy they were attempting to model, and therefore, these intervals were coded as child-focused. The description of the strategy is significant because it provides parents with cues that the interventionist is performing the behavior for the explicit purpose of demonstrating (Woods et al., 2011), rather than trying to implement a child-focused intervention. Also, an explanation of the strategy creates opportunities for parents to better understand the interventionist’s actions as the parent will be able to see the strategy and hear a corresponding description of the strategy. Based on this explanation, it appears that one way to facilitate the decrease in child-focused behaviors is to encourage interventionists to explain strategies that they are demonstrating to the parent. Although there continues to be a need for a decrease in child-focused behaviors, the results for this behavior show that, on average, these behaviors did not occur the most frequently of all behaviors in the current study. These results contrast the earlier findings of Peterson et al. (2007) for a different parent-mediated intervention. Perhaps this reflects a general shift toward a practice in which the parents play a more active role in intervention sessions, or the results may have been specific to the ART intervention itself.
Some aspects of the other behaviors not considered to be coaching may be important within interventions, although they are not necessarily perceived as being essential to parent–child interactions. For example, the time the interventionists spent writing notes about the session and scheduling the next session with the parent was coded as no coaching. Writing notes is often necessary within community-based early intervention sessions to document child progress, provide information relevant to revising the intervention plan, and bill for the intervention session. Scheduling the next session is also important to encourage the continuity of the intervention. Although these tasks are necessary, having efficient and effective ways to perform them may allow more time for productive coaching activities.
Within some intervention sessions, the interventionist and parent engaged in conversations that were not directly related to the child. The conversations or chit-chat not related to the child’s development were also coded as no coaching-other, but they may have served a purpose in the session. These conversations often encouraged communication between the parent and interventionist and, perhaps, helped to establish rapport between them. According to Ebert and Kohnert (2010), speech-language pathologists identified these components as being important for the clinician-client relationship. Thus, the interventionists may have recognized those conversations as an opportunity to create a bond with the families. Therefore, there may also be benefits to the use of no coaching-other behavior as it relates to conversations with the parents for a short duration during the intervention session.
Information sharing between parents and interventionists was another behavior that frequently occurred during ART sessions. Peterson et al. (2007) noted that parents were often highly engaged during information sharing behaviors within their study. Therefore, the fact that information sharing occurred frequently is a promising sign that there were moments in which parents were highly engaged during the ART sessions. The coders selected the information-sharing behavior within intervals in which the parent and interventionists engaged in conversation about the child’s development, access to additional services, and his or her progress with the intervention strategies (Friedman et al., 2012). The frequency with which this behavior occurred across sessions suggests that the interventionists were consistently encouraging this behavior. As a result, parents were actively engaged in the session for at least a portion of the session.
Just as important as the behaviors frequently demonstrated are the behaviors that were not demonstrated as often within intervention sessions. Of particular significance are those behaviors through which the interventionist could have provided opportunities for parents to be more active. These behaviors are observation, caregiver practice, guided practice, problem-solving, and reflection. Trivette et al. (2010) noted that to encourage adult learning, adults must have the opportunity to apply the information that they are being taught. As a result, we would expect that more instances of the interventionist guiding caregiver practice and urging more independent caregiver practice would facilitate the parents’ understanding of the strategies by providing them with opportunities to practice the strategy during the intervention session. Specifically, on the ART implementation fidelity rating form, there is an item explicitly related to this expectation: “Coach parents while they interact with their child.” Yet this behavior was observed at a very low frequency.
In addition, observing the parent and child provides the interventionist with an opportunity to see the natural interactions between parent and child in hopes of helping to build on these interactions (Woods et al., 2011). Unless interventionists implement these observation coaching behaviors, they may be missing out on opportunities to learn more about the parent and child interactions and provide the optimal type of individualized support to build target skills.
There is a similar case for the low-frequency behaviors of problem-solving and reflection behaviors. Both behaviors encourage parents to engage in conversations with the interventionist and not just state what is happening, but why they believe it might be occurring or possible solutions to difficulties that may arise. As with the other low-frequency coaching behaviors, it is important to understand why interventionists may not be using these behaviors. The infrequent use of these behaviors may be because the interventionists have limited knowledge about these behaviors or training on how to implement them, or perhaps because these behaviors are harder to implement. For example, when examining the conversations used within coaching sessions, Jayaraman et al. (2015) noted that reflection occured less frequently. Similarly, Meadan et al. (2018) indicated that interventionists used reflection and feedback the least. Yet, in both studies, the interventionists noted that all of the behaviors, including reflection, were important. These findings indicate that interventionists were aware of reflection as a coaching behavior but may have found it to be difficult to implement the reflection strategy within the session. Similarly, in the ART intervention, interventionists appeared to know about the reflection behavior, as it was listed on the fidelity tool. Still, they had difficulty implementing it in the context of the intervention. As stated by Romano and Schnurr (2020), additional professional support may be useful to facilitate the implementation of these behaviors within intervention sessions. More frequent use of behaviors supporting active parent engagement in learning intervention strategies arguably may increase the efficacy of parent-mediated interventions in impacting the desired outcomes for both parents and their children.
Coaching Strategies and Education Level
The frequency of two coaching behaviors varied based on education level. Those two behaviors were joint interaction and no coaching-child-focused. Interventionists working with parents with a higher education level (i.e., college degree or graduate/professional degree) were more likely to use joint interaction more when compared with interventionists working with parents with lower education levels (i.e., high school diploma and some college). In contrast, interventionists were more likely to demonstrate more no coaching-child-focused behaviors with parents with lower education levels. These two behaviors, no coaching-child-focused and joint interaction, reflect two different types of parent engagement. Within joint interaction, parents are participating in the action with the child and interventionist. In contrast, during child-focused behaviors, parents are more likely to observe the interventionist and child interacting with one another.
The fact that there were associations of these two behaviors with parent education level raises more questions within this area of research. While interventionists may be adapting themselves to the needs of the family, it is vital to understand the reason why adaptations may occur specifically for these two behaviors. Potentially, parents with higher education levels were more assertive and did not perceive interventionists to be there solely to provide information or direct services to the child but also to aid the parent in using strategies with their child. Interventionists may have facilitated more joint interactions with these parents and perceived less of a need for these parents to observe. In contrast, the parents with less education may have been less sure of their role and the interventionist’s role within the intervention session. As a result, they may have decided to play a more passive role in the intervention, such as observing the interventionist work with the child. Another possible explanation for the more frequent use of the child-focused behavior with the families of lower education levels is that interventionists may have thought that the child-focused behavior gave the parents more opportunities to observe the interaction between the interventionist and the child. Therefore, while it is vital to know in which way interventionists alter their therapy sessions based on the education level of the parent, it is also essential to understand why they are altering their sessions. This is particularly important as it pertains to using parent-mediated interventions in a variety of communities, such as under-resourced communities. Interventionists should be aware of their own assumptions as they work with families. If interventionists are adjusting the strategies that they implement based on assumptions related to a parent’s education level, ultimately, that could impact the parents’ understanding of the type of intervention being offered and their role in the intervention. Therefore, throughout the intervention, interventionists should strive to actively engage all parents in the process.
Relationship Between Change in Parent Responsiveness and Coaching Behaviors
After controlling for the initial parent responsiveness, there were two interventionist behaviors associated with change in parent responsiveness. The first behavior was information sharing. As mentioned previously, information sharing is a coaching behavior in which a parent and interventionist converse about aspects of child development (Friedman et al., 2012). There are many advantages to this coaching behavior. First, parents are actively engaged in the intervention strategy and are encouraged to share information about their child or events related to their child. In addition, this coaching behavior provides interventionists with the opportunity to learn more about the child’s recent progress or concerns that parents may have about their child’s development. From this information, it is possible that the interventionists selected the most appropriate strategy to meet the family’s needs. As Woods et al. (2011) stated, it is important for parents to understand why a specific strategy is being encouraged as it helps them to understand the relevance of the strategies recommended. Thus, both the parent and the interventionist are likely to see the value in dialogue about the child’s development. Therefore, it is not a surprise that this coaching behavior was associated with a change in parent responsiveness as it appears to help establish a relationship between the parent and interventionists and potentially leads to more individualized intervention strategies.
Information sharing is also one of the three coaching behaviors that encourage parents to engage in conversation, as a consequence of the interventionist facilitating input from the parent. The other two similar coaching behaviors are problem-solving and reflection (Friedman et al., 2012). As problem-solving and reflection were not implemented frequently within these intervention sessions, the frequency of their use was less likely to be correlated with the change in parent responsiveness. However, the results about information sharing raise the possibility that increases in problem-solving and reflection would also encourage active participation of the parents and provide parents with examples of how strategies may relate to their child’s development and could, therefore, have a positive relationship with parent responsiveness.
Another coaching strategy that related to the change in parent responsiveness was the no coaching-child-focused behavior, with more child-focused behaviors associated with less of a change in parent responsiveness from preintervention to postintervention. Perhaps, the passive role that parents take within this coaching behavior does not build the parent’s capacity to implement the strategies with their child that equates to an increase in parent responsiveness. Parents do not have an opportunity to practice interaction strategies while this behavior is being used by the interventionist. In addition, parents may not know how to identify the strategy that the interventionist is implementing or understand its importance for their child’s development. So, the purpose of the interventionist’s child-focused behaviors may not be apparent to the parents nor easily replicated by them.
Of even more concern is that the interventionists used the child-focused behavior more with parents of lower education levels when compared with parents with higher education levels. While interventionists may be assuming that specific families will benefit from observing these child-focused behaviors, there is no evidence that this behavior is effective in improving the outcomes of parents at any educational level. Therefore, interventionists should continue to monitor the use of this behavior. If they often use the child-focused behavior within therapy sessions, they should determine why this behavior is occurring and what alternative coaching behaviors they could use to promote more active parent participation. For example, instead of using child-focused behaviors throughout the session, they can narrate the behavior that they are trying to encourage with the child (i.e., demonstration with narration) and then prompt the parent to practice this same behavior (i.e., guided practice). Based on the parents’ needs, interventionists may need to provide more or fewer prompts for parents to practice the strategy. Following guided practice, they can allow the parent to practice independently (i.e., caregiver practice) and offer feedback as needed. This change in the initial behavior can lead to opportunities to employ other coaching behaviors. For example, having the parent reflect on what happened with their child during the practice is another active participation strategy.
One final coaching behavior that approached significance in accounting for change in parent responsiveness was observation. An increase in the use of observation by the interventionists was associated with a larger change in parent responsiveness. Not only does observation allow the interventionist to see the parent and child interact together but it also ensures that the parent plays an active role and has the potential to boost the parent’s confidence (Woods et al., 2011). This further supports the notion that within this ART intervention study, behaviors that encouraged active participation from the parents often related to a greater change in parent responsiveness when compared with behaviors that did not encourage active participation (e.g., no coaching-child-focused).
Surprisingly, the correlation between the use of joint interaction and the change in parent responsiveness was not significant. This suggests that parents and interventionists engaging in activities with the child may not have as much of an impact as the specific coaching behaviors mentioned previously, although parents are actively involved during joint interaction. Similarly, Brown and Woods (2015) reported that joint interaction was not related to the parent’s use of targeted strategies within the intervention session. A more in-depth investigation of what is occurring within joint interaction would be informative, specifically as it relates to intervention outcomes. One explanation for its non-significant relationship with final outcomes is that the behaviors of the interventionist, parent, and child often vary within joint interactions. For example, the interventionist, parent, and child may all be playing together with limited connection to the strategy introduced during the intervention session, or the interventionist and parent, instead, may be practicing the strategy with the child, although without any explicit narration or feedback from the interventionist, facilitation of problem-solving, etc. Thus, the goal of the joint interaction may be different based on the parent, interventionist, child, or context in which it is implemented. Therefore, to determine its effectiveness in building parent and child outcomes, we would need to explore the actions occurring during joint interaction, the purpose of this activity, and whether or not parents play a more active or passive role within those particular interactions. In addition, it would be interesting to test which of the active behaviors combined with joint interaction may have the largest effect on outcomes.
Limitations
One limitation of this study was that it was focused specifically on the ART intervention. Therefore, findings from this study may not be generalizable to studies of other interventions employing coaching as a service delivery model. However, ART mimicked many other parent-mediated programs in attempting to teach parents improved strategies for interacting with their child. Thus, there may be more similarities across interventions than differences.
In addition, we did not measure the quality of the interventionists’ use of each of these coaching behaviors. Such data may have provided additional insight into why some behaviors were used more frequently than other behaviors. Also, it is possible that the quality of coaching behaviors may be more strongly associated with parent outcomes than the frequency of their use or that there may be additive or interaction effects related to quantity and quality.
A limitation of the sample is that the generally high levels of parent education provided us with less opportunity to observe whether there may have been additional differences in strategy use related to parent education than would have been possible if the parent sample was more diverse educationally. For example, we combined all of the parents who had indicated that their education level was some college or less for the purpose of analysis to create relatively equally-sized parent education groups, but important differences in coaching strategies might be seen in interventionists working with parents who have less than a high school education compared with parents who have completed coursework toward a bachelor’s degree.
A limitation of the procedure is that we specifically coded the occurrence of coaching behaviors within 30-s intervals, with a specific behavior only being coded once for any given interval if it was observed during any part of the interval. Therefore, the data in this study neither provide the precise frequency of occurrence data (e.g., in some cases, a behavior may have occurred more than once in a 30-s interval but would only have been coded one time) nor the exact duration of each of these behaviors across the intervention sessions. Nevertheless, given the variable nature of the coaching behaviors, we considered partial interval coding to offer the best approach for us to compare the relative use of these behaviors in this study.
Another limitation of the study was that the distributional properties of some of the interventionist behaviors resulted in their exclusion from the analyses addressing our RQ2 and RQ3. Within this study, due to their overall low-frequencies and limited variability, it is unlikely these excluded behaviors would have been significantly associated with parent education levels or would have accounted for significant variance in changes in parent responsiveness, even if we had employed a nonparametric analytic approach. Nevertheless, the theoretical importance of coaching behaviors that promote parent active participation in the coaching process argues for a focus on increasing the use of problem-solving, reflection, guided practice, and caregiver practice in future parent-mediated interventions and evaluating whether, with greater overall usage, variability in these behaviors would demonstrate an association with parent outcomes.
Future Directions and Conclusions
Interventionists should have opportunities to receive professional development regarding evidence-based coaching behaviors and should be trained on how to implement these coaching behaviors with diverse populations. In addition, preservice students should also receive training on how to implement coaching interventions, as it has been noted that preservice students often have limited knowledge about coaching before entering early intervention (Stewart & Applequist, 2019). Future research should also continue to study the behaviors demonstrated by interventionists within intervention sessions and determine whether these coaching behaviors improve parent outcomes and potentially mediate the relationship between parent behaviors and child outcomes. If these coaching behaviors are deemed to be important for parent-mediated interventions, investigations should move toward addressing questions not only regarding whether these behaviors are being implemented within intervention sessions but also for how long they are being implemented, whether professional development can successfully increase the use of strategies observed to occur with low frequency, and if increases in the use of specific strategies impact long term outcomes.
In addition to understanding why interventionists often implement certain coaching behaviors, we should also explore the attitudes and beliefs of the parents about these intervention strategies. It is difficult to determine if coaches were more likely to implement specific strategies due to their own preferences or because the families were less comfortable with specific coaching strategies. Therefore, it is not only essential to investigate the beliefs about these coaching strategies from the perspective of the interventionists but also from parents who are primary stakeholders in the intervention. Potentially such input from parents in parent-mediated interventions could support the development of a tool to assess parent preferences or other variables predicting which coaching strategies are more likely to meet a parent’s needs, align with their preferences, and ultimately more effective in supporting parent implementation of intervention strategies with their child.
Footnotes
Acknowledgements
The authors acknowledge the staff and volunteers who assisted in the data collection and the coding of the videos and the families who participated in this study. The authors thank Dr. Mollie Romano for providing consultation on the definitions of the coaching behaviors. We acknowledge Dr. Cathy Zimmer and Dr. Chris Weisen for their assistance with the statistical analyses.
Authors’ Note
Jonet Artis is now at New York University in the Department of Communication Sciences and Disorders. This article reports a novel study using video data collected for a previously published randomized controlled trial (Watson et al., 2017); it is based on a dissertation completed by 
.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work was supported by a grant from the Institute of Education Sciences, U.S. Department of Education (Grant R324A100305) to the University of North Carolina at Chapel Hill (PIs: Watson and Crais). The opinions expressed represent those of the authors and do not represent views of the Institute or the U.S. Department of Education. This research was also supported by the James J. Gallagher Dissertation Award through the Frank Porter Graham Child Development Institute and the H325D160060: Bridging the Gap Between Research and Practice Leadership Grant from the Office of Special Education Programs, U.S. Department of Education.