In-Home Training for Fathers of Children with Autism: A Follow up Study and Evaluation of Four Individual Training Components

The ADI-R (Lord et al. 1994) is a standardized parent interview for assessing presence and severity of symptoms. A diagnosis of autism is established if an individual scores at or above the cutoff score in the three ICD-10 symptom domains. A trained certified member of the research team administered the ADI-R to the primary caregiver (parent who reported spending the most time with the child).

The ADOS (Lord et al. 2002) is a semi-structured observational assessment administered directly to the child. Like the ADI-R, instrument validity is based on diagnostic criteria for autism in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-R 1994) and ICD-10 symptom domains. ADOS modules were administered by team members trained and certified in its use.

The Vineland (Sparrow et al. 1984) is a widely used measure of adaptive behaviors from birth to age 18. This semi-structured interview with a parent or a caregiver assesses adaptive (real life) skills in communication, daily living, maladaptive behavior, and socialization. It is commonly used in autism research because it offers overall areas of deficits in individuals with autism and dimensional metrification of a hallmark deficit, socialization (social competence; Volkmar et al. 1997). All instruments described above have well-established psychometric properties.

Research Assistants (RAs), who successfully completed extensive behavioral coding training, videotaped 15 min each of father–child and mother–child in-home play sessions. The first 5 min were “settling in” and not coded; in the following 10 min, coders videotaped occurrences of parent and child reciprocity behaviors. Coders worked independently and behavioral frequency counts were determined by number of occurrences counted during the 10-min session. Coding and analysis of target behaviors were facilitated with the widely used computerized Observer Program, which allowed data to be entered into computers, labeled and organized, and stored in a larger desktop computer with network access. Throughout the research protocol the PI, who taught the interventions, was not involved in the coding process, and coders were not told which interventions were being taught. To evaluate for possible rater drift, a second rater coded 20% of randomly selected videotaped sessions. Interrater scores ranged from 0.71 to 1.0, with a mean of 0.83. First, the team coded frequencies of the four father skill components: imitating with animation, expectant waiting, following the child’s lead, and commenting on the child. The team also coded frequency of the same skill components for the mothers. Child initiating, child responding, child non-speech vocalizations, and intelligible words were also coded for both the father–child and mother–child session.

In earlier research and clinical work, many parents verbalized they did not know how to play with their children during play sessions. Parents either sat passively or aggressively tried to direct interactions, not allowing children sufficient time to respond (Elder 1995). The first intervention, “imitating/animating,” was designed to address these concerns by promoting basic turn-taking play interactions. Enhanced with specific videotaped examples and written directions about following the child’s lead in play, fathers were instructed to attend to and imitate their children’s vocalizations and/or actions in an animated manner. As in our earlier work, parents reported that imitating the children was “fun” and helped them relax during play sessions. “Expectant waiting” continued to be more difficult to employ, as it requires waiting for child responses. Thus, we continued to teach “imitating/animating” first, followed by “expectant waiting.” In our prior study we also observed that even with instruction emphasizing waiting for the child to initiate, father and mother initiating did not decrease significantly across conditions and that many parents relied on directive questioning. To address this in the current study, we added another component to specifically teach fathers to comment on the child’s verbalizations and actions, rather than question.

Figure 1 provides an overview of PI-father training sessions, father-conducted training sessions, and videotaping data collection sequence. As noted, there were three PI-father training sessions: introducing and teaching “imitating/animating” and “following the child’s lead” in play interventions (Session 1), introducing and teaching “expectant waiting” and “commenting on the child” (Session 2), and combining the four strategies (Session 3). Following PI-father training, fathers implemented what they learned. Each time an intervention was introduced by fathers, a new condition was established (I-III).

Once pre-evaluation data were analyzed for father–child dyads, the first author (PI) used father–child videotaped baseline sessions to teach fathers to recognize child initiations, follow child’s lead in play, and apply the intervention “imitating/animating.” Mothers were not present during father training. After training, fathers were instructed to immediately begin using Intervention I in everyday interactions with their autistic children. Skill acquisition was assessed via videotape during father–child play sessions two times a week for two consecutive weeks. Fathers were instructed to use the intervention component with their child for 3–4 days before being videotaped again. The research team used data from these sessions to determine number of opportunities for using the intervention and how often fathers responded to these opportunities. The PI discussed these findings with fathers and determined if they were ready to advance to the next condition or if remediation was indicated. Mothers were videotaped during each home visit as well. Father–child and mother–child sessions remained separate, and order of videotaping was counterbalanced across sessions. As in the baseline phase, children were given breaks between all sessions.

Prior to collecting data in Condition II, fathers were asked to train mothers in Intervention I and to begin using Intervention II, themselves. Data collection methods of child/parent behaviors were identical to Condition I. PI-father training of Intervention II involved two strategies: “expectant waiting” (signaling child for a desired action or word and waiting expectantly a minimum of 3-s for a response) and “commenting on the child,” instead of asking questions, a more common parental behavior. From our previous work we noted that even when instructed to discontinue Intervention I, some fathers had already incorporated it into their daily father–child interactions, and it was unrealistic to expect them to “unlearn” the skills set. Therefore, in the current study fathers decided whether to continue using Intervention I without overt instruction, and Intervention I use was documented throughout all conditions. As in Condition I, the research team used data obtained from the first two videotaped sessions in Condition II to determine number of opportunities for using the intervention and how often fathers responded to these opportunities. The parent trainer (PI) discussed findings with fathers to determine if they were ready to advance to the next condition or if booster training was indicated. If videotapes showed a father was not using the strategies as instructed, a team member made one subsequent home visit to provide additional instruction and encouragement, and to answer questions. At least two more video sessions were then conducted to ensure the father was using the strategies before asking him to instruct the mother and move to the next intervention.

In this training session fathers viewed father–child videotapes and data, and received additional instruction regarding combining both previously learned strategies. Training was individualized, based on assessment data, the child’s progress, and expressed needs of fathers. Data collection was identical to Conditions I and II.

In summary, each family was videotaped twice a week for 8–12 weeks. The range in weeks reflects each family’s uniqueness regarding need for training, family illness, and/or last-minute filming session cancellations. Variation was accounted for in statistical analysis of grouped data.