One important factor that might help us understand differences in teachers’ responses to COMPASS is teacher self-efficacy. For teachers, self-efficacy refers to beliefs in one’s ability to produce desired educational outcomes. Self-efficacy originates from Bandura’s work (1986) that showed that when people hold perceptions that their actions can result in chosen outcomes, they are more likely to have higher motivation, effort, and persistence for achieving those outcomes in the face of obstacles and barriers (Soto and Goetz 1998). Much like goal setting theory we described earlier (Ryan 1970) that suggests the importance of clear, time-limited, and well-specified goals, self-efficacy may also play a significant mediating role in goal accomplishment. If we could document what influences self-efficacy of teachers of students with ASD, then we would be able to use this information in future research for improving educational outcomes.

Bandura identified four factors that predict self-efficacy: (a) mastery experience (perception of past performance); (b) vicarious experience (observation of others); (c) social persuasion (persuasive messages received by others); and (d) physiological and emotional states (somatic and affective reactions regarding performance). In other words, teachers who perceive themselves as performing well in the past, have the opportunity to observe competent peers, receive encouragement from others, and experience positive feelings regarding their work have high self-efficacy. Our first study on self-efficacy (Ruble et al. 2011) explored three of these four factors—mastery, social persuasion, and physiological/emotional states. We measured mastery by number of years teaching students with ASD, social persuasion by leadership support, and emotional state by burnout. Burnout was broken down further into one of three areas: emotional exhaustion, depersonalization (treating students as if they were objects), and reduced personal accomplishments (low sense of reward from teaching). We administered a generic measure of teacher self-efficacy called the Teacher Interpersonal Self-Efficacy Scale (TISES) and examined teacher responses against the three sources. The TISES taps into self-efficacy for three general educational domains: (a) classroom management; (b) obtaining colleagues’ support; and (c) obtaining principal’s support.

When we correlated each of the three types of self-efficacy against the sources, we found that self-efficacy for classroom management significantly correlated with all three emotional states of burnout. In other words, teachers who reported higher ability to manage the classroom reported reduced emotional exhaustion and depersonalization, and higher sense of personal accomplishment. However, self-efficacy subscales measuring the perceived ability to obtain colleague and principal support did not correlate with any of the proposed sources of self-efficacy. Moreover, two of the sources of self-efficacy, mastery, as measured by number of years teaching students with autism, and social persuasion, as measured by administrator support, failed to correlate with any self-efficacy subscale. Perhaps when judging their level of mastery, teachers of students with autism are unable to rely on prior experience due to the heterogeneity in symptom presentation of students. The diverse set of symptoms may create difficulties generalizing experiences with one child with ASD to a different child.

Another possible explanation is that teachers may simply feel underprepared, regardless of prior levels of experience. Most current teacher training in autism is unsuccessful and is not based on methods that result in changes in classroom practice. Effective training in autism-specific interventions (Jennett et al. 2003), such as those that are hands-on and classroom-based is limited. However, the failure to find an association between self-efficacy and social persuasion also could be an artifact of low power due to low sample size (n = 35). Specifically, the correlation coefficient for classroom management was of moderate size (r = 0.28) and would have been significant with a slightly larger sample. Alternately, special education teachers may not rely on the same sources of support as general education teachers. Instead they may derive support from individuals with direct involvement in supporting their teaching (e.g., autism specialist or special education director) rather than general administrators such as principals, because persuasive messages are most influential when the persuader is intimately familiar with the task and the teacher holds confidence in their opinions (Bandura 1977).

Overall, our findings failed to find strong support for the relationship between self-efficacy measured by the TISES and putative sources of self-efficacy other than the role burnout might play in child educational outcomes. One possible explanation for these generally negative results is that our measure of self-efficacy was too broad, since it was designed to be used with general education teachers. Consistent with this concern, Bandura has emphasized that self-efficacy is not a general or broad construct, but narrowly focuses on very specific tasks and behaviors. Thus, we decided to re-examine these hypotheses using a specific measure that related to the teaching tasks associated with learners with ASD.

Accordingly, we created a new measure to assess teacher self-efficacy directly related to the tasks required to teach young students with ASD and tested it in our second RCT (Ruble et al. 2013). The new measure, the Autism Self-Efficacy Scale for Teachers (ASSET), consists of 30 questions that represent various tasks associated with effective teaching (e.g., confidence assessing social interaction, using visual structure to increase independence) and was completed by 44 teachers of students with ASD. Items were rated using a 100-point scale (1 = cannot do at all to 100 = highly certain can do). Teachers reported an overall mean score across items of 74.5. After initial inspection of responses indicated a general failure to use the lower part of the scale, we converted the 100-point scale to a 6-point scale using the following metric: 0–50 to 0, 51–60 to 1, 61–70 to 2, 71–80 to 3, 81–90 to 4, and 91–100 to 5. The mean score on the new 6-point scale was 2.58. The high internal consistency of the total scale (0.96) indicated that the ASSET was unidimensional and represented by a single construct.

As before, we analyzed the associations between self-efficacy as measured with the ASSET and the three potential sources of self-efficacy—administrative support, mastery, and emotional responses. Similar to our prior study, self-efficacy was unrelated to administrator support and mastery. Moreover, unlike our prior study using the TISES, the ASSET failed to correlate to burnout. However, ASSET scores were related to teacher stress instead. Specifically, low self-efficacy on the ASSET was related to increased teacher stress as measured by subscales indexing higher self-doubt and need for support and higher teaching disruption. We also observed significant correlations between all subscales related to teacher stress (self-doubt, loss of satisfaction, disruption of teaching, frustration with parents) and all subscales related to burnout (exhaustion, depersonalization, lack of personal accomplishment). This is a potentially important finding. The difference between stress and burnout is burnout is the long-term consequence of chronic stress. Perhaps the ASSET may be helpful as an early indicator of stress that as it accumulates, potentially leads to teacher burnout. If identified early, the long-term consequences of stress may be able to be reduced through teacher stress/burnout interventions.

A critical question was whether self-efficacy, an internal personal characteristic of teachers, also impacted teaching quality and child outcomes (Fig. 7.1). When we correlated ASSET scores collected at the beginning of the year with child goal attainment change scores and IEP quality, no significant associations were found. But when we examined ASSET scores obtained at the end of the school year, findings were different. Teacher self-efficacy collected at the end of the school year was associated with child goal attainment change scores (r = 0.31, p = 0.04) and with IEP quality (r = 0.44, p = 0.00). That is, as predicted by Self-efficacy Theory (Bandura 1977), teacher self-efficacy both impacted teacher behavior and the outcome of that behavior, child goal improvement. Thus, these findings suggest that self-efficacy as measured with the ASSET appears to be related to outcomes through its impact on IEP quality and perhaps indirectly through its impact on stress and burnout. Moreover, the findings are consistent with our overall model positing the potential impact of teacher internal factors on student outcomes. Finally, the fact that self-efficacy impacted outcomes, even within a well conducted clinical trial, suggests that more research is needed on understanding ways to enhance and support teacher self-efficacy.

The most critical teacher variables are those that directly impact child educational success. So far, we have discussed three important factors—observable teacher behaviors associated with IEP quality, teaching adherence/quality, and engagement, and internal teacher reports of self-efficacy. But are there other internal teacher factors that impact child educational outcomes? We alluded to the importance of stress and burnout above because there is now a considerable body of research identifying stress and burnout as major contributors to teacher turnover (Awa et al. 2010; Carlson and Thompson 1995). For example, the annual attrition rate for special educators (13 %) is twice that of general educators and the 3 year attrition rate is approximately 25 %, i.e., one in four teachers is lost every 3 years (Boe et al. 2008; Cook and Boe 2007; McLeskey et al. 2004; McLeskey and Billingsley 2008; Nichols et al. 2008). Furthermore, stress and dissatisfaction with their teaching positions lead an additional 20 % of special educators each year to transfer to general education or seek another position in special education (Boe et al. 2008; Boe 2014). These high turnover rates are further exacerbating the national shortage in special educators, estimated at 11.2 % (U.S. Department of Education 2008), with the unfortunate result that many special education teachers are hired without adequate preparation (Boe 2014; McLeskey and Billingsley 2008). Pairing our most challenging learners, especially students with complex needs such as those with ASD, with our least prepared educators is clearly undesirable.

Moreover, instability in our teacher workforce has negative consequences for schools and most importantly for students. Although there is a large amount of attention given to teacher burnout, surprisingly, to our knowledge, no study has directly examined or linked burnout to child educational outcomes. We wanted to understand what impact, if any, burnout has on student outcomes. However, to provide context, we also wanted to understand the potential predictive strength of burnout relative to other factors from our model. Accordingly, in this analysis we focused on several measures of the internal factors (stress, burnout, years teaching, experience with ASD), a single measure of external factors (administrator support), and teacher and student engagement impacting teacher behavior and teaching quality. In Table 7.1 we show the correlations between teacher burnout and stress and teacher internal and external factors for our sample of teachers who participated in both RCTs—the control and experimental group participants. All the measures were obtained at Time 2, the end of the school year, with the exception of administrator support. It should be noted that we were unable to include self-efficacy in our analysis because we used different measures in the two RCTs.