A scoping review on the psychometric properties of the teacher efficacy for inclusive practices (TEIP) scale

ABSTRACT

Teachers’ self-efficacy for inclusion is emphasized as necessary for enabling inclusive education. One instrument developed for measuring teacher self-efficacy for inclusion is the Teacher Efficacy for Inclusion Practice-scale (TEIP) (Sharma, U., Loreman, T., & Forlin, C. (2012). Measuring teacher efficacy to implement inclusive practices. Journal of Research in Special Educational Needs, 12(1), 12–21). The present study aimed to identify and summarize the empirical literature on structural validation and reliability of the TEIP scale. A scoping review of 15 peer-review articles was conducted. The three subscales found in the original TEIP scale were confirmed. However, there is support for two subscales. The TEIP is also reported to be a construct without multidimensional constructs of teacher-self efficacy. Although the items worked differently in different samples, the internal consistency was generally sufficient or good. Our findings indicate that the TEIP is not yet a scale fit for comparisons between populations and contexts, reflecting the multifaceted nature of the concept of both inclusion and teacher self-efficacy.

KEYWORDS:

• Scoping review
• teacher self-efficacy
• inclusive education
• TEIP
• psychometric

Introduction

Bandura (1997) states that self-efficacy beliefs are a dynamic personal factor crucial to our agency or ability to act. Furthermore, within the framework of social cognitive theory, Bandura defines self-efficacy as an active and learned system of beliefs held in a context. Self-efficacy is thus not considered a stable character trait of a person but changes and varies with a specific task or context (Bandura, 1997). Teacher self-efficacy has been widely researched (Klassen & Chiu, 2010). However, the development of the concept has been described as a history of confusion (Dellinger, 2005; Henson, 2002; Zee & Koomen, 2016). Even if a clear consensus has not been reached about the definition or which particular role teacher self-efficacy plays in the classroom, it is regarded as a critical component for teachers’ classroom functioning (Tschannen-Moran & Woolfolk Hoy, 2001; Zee & Koomen, 2016). The concept has by Tschannen-Moran and colleagues (1998) been defined as “the teacher's belief in his or her capability to organise and execute courses of action required to successfully accomplish a specific teaching task in a particular context” (p. 233). Dellinger and colleagues (2008) formulated teacher self-efficacy “as a teacher's individual beliefs in their capabilities to perform specific teaching tasks at a specified level of quality in a specified situation” (p. 752).

Self-efficacious teachers can adjust strategies and pedagogy (Tschannen-Moran & McMaster, 2009). In a review of 40 years of teacher self-efficacy research, Zee and Koomen (2016) synthesized that self-efficacious teachers are less likely to exclude problematic students from their classes and generally more tolerant toward such students. A more recent study by Woodcock and Jones (2020) indicated, for instance, that teachers with a lower level of teacher self-efficacy may feel personally challenged or under-prepared for the concept of inclusion of all. Teacher self-efficacy for inclusion can be conceptualized as a specific subtype of teacher efficacy (Sharma, 2012). Numerous researchers have investigated the relationships between teacher self-efficacy, inclusion, and professional practice (see, for example, Zee & Koomen, 2016). One crucial factor that has been shown as associated with teachers’ intention to include all children in mainstream classrooms and their attitudes toward inclusive education is teachers’ self-efficacy for inclusive practices (Song et al., 2019).

As most European countries, following the Salamanca Statement and Framework for Action on Special Needs Education (UNESCO, 1994), have acknowledged the premise for educational rights for persons with special educational needs (Haug, 2017), these findings are highly relevant for the evidence-based promotion of inclusive practices. Despite the impact Salamanca Statement has had, as for teacher self-efficacy, the concept of inclusion lacks a general definition of what inclusive education is, which students it addresses, or its pedagogical implications (Haug, 2017; Terzi, 2014). According to the European Agency for Special Needs and Inclusive Education (2014), the key question is no longer about what or the need for inclusion but how it can be achieved.

For practical, inclusive education, the challenge thus is implementation (Haug, 2017). Implementation of inclusive education concerns all levels of the educational system. It is embedded in a layer of contexts, from the cultural and societal level to the next level, including rules and regulations, teacher education content, support and resources allocated to schools, local authorities, and school leadership affecting the educational practice in the classroom. Each of these layers can potentially facilitate or hinder the inclusion of a particular student (Ginner Hau et al., 2020; Mitchell, 2005). The complexity of this multifaceted practice offers many challenges in making inclusive education happen in the classroom, where teachers naturally play a fundamental role (Haug, 2017). Transforming the idea of inclusion into inclusive education in the classroom requires highly qualified and competent practitioners who accept commitment and ownership of the process of all children in the classroom (Florian, 2012). As previously mentioned, one crucial factor that has been shown as associated with teachers’ intention and attitudes to include all children in mainstream classrooms is teachers’ self-efficacy for inclusive practices (Song et al., 2019).

Consequently, measurements of teachers’ self-efficacy for inclusive practices could play an essential role in implementing inclusive practices. Researchers and practitioners need valid instruments to be able to measure and document teacher self-efficacy for working in inclusive settings. One widely applied scale developed to measure teachers’ perceived self-efficacy for inclusion is the Teacher Efficacy for Inclusion Practice scale (TEIP; Sharma et al., 2012). Sharma and colleagues conceptualized teacher self-efficacy for inclusion as comprising three dimensions; Efficacy to use Inclusive Instructions, Efficacy in Collaboration, and Efficacy in Managing Behavior. These three dimensions refer to teachers’ belief in their ability to have strategies that promote the inclusion of all learners, work with parents and other professionals, and deal with disruptive behavior. Sharma et al. did not choose a specific definition of inclusion, but they explained that inclusion is a philosophy that all students are different. The schools must adapt to meet all student’s needs regardless of disability or not. Based on a literature review and a group of experts in the field, Sharma et al. developed 30 items investigated among pre-service teachers in four countries (Australia, Canada, Hong Kong, and India). The responses were analyzed with an explorative factor analysis suggesting three subscales, including 6 items each, and a total scale based on 18 items. High scores on the TEIP total scale and subscales indicate higher levels of teacher self-efficacy for working in inclusive practices. The internal consistency varied between the subsamples, but when collapsing the sub-samples, the internal consistency was reported by Sharma et al. to be strong (Cronbach's alpha ranged from .85 to .93).

Recently, literature reviews have shown that the teacher reports on the TEIP scale are related to teacher education and teaching experience, attitudes to and experiences of students with special needs, as well as educational systems (Tumkaya & Miller, 2020; Wray et al., 2022). However, the psychometric properties of the TEIP have not been evaluated. Therefore, the present study aimed to identify and summarize the empirical literature on structural validation and reliability of the TEIP scale. Specifically, we investigated 1) How stable is the factor structure in the TEIP? and 2) How coherent are the items in the TEIP?

Methods

The present study was conducted as a scoping review, which is an appropriate research method for summarizing research findings and identifying research gaps (Arksey & O'Malley, 2005). Initially, we searched Academic Search Premier, Ebsco, Education Collection, ERIC, PsycINFO, PsycARTICLES, Scopus, Social Sciences Database, and Web of Science to identify studies on the TEIP scale (Sharma et al., 2012). We searched for peer-review articles published in English between January 2012 and July 2022. The search terms used were TEIP, teacher-efficacy for inclusive practices, teacher-efficacy or self-efficacy, and inclusive or inclusion. Initially, the abstracts and titles were screened to identify studies on the TEIP, and we also screened articles in full text for eligibility. In addition, we checked reference lists to ensure that all relevant studies were identified.

The inclusion criteria used were studies that had used the TEIP, written in English, peer-reviewed, and contained an aim regarding the psychometric properties of the TEIP scale. We excluded 114 peer-review articles, and 15 were included in the current study.

After searching and selecting articles for the present study, we identified data on structural validation and reliability of the TEIP in each of the included studies. Relevant data on structural validation and reliability were extracted with information on response rate and the samples in each article.

Results

We identified 15 original peer-review studies with a psychometric aim (see Table 1). The studies were conducted in 12 different countries, with samples of 105 to1231 participants. In the included studies, the samples consisted of self-ratings by pre-service teachers (k = 7, 46.7%), in-service teachers (k = 7, 46.7%), as well as combinations of pre-service and in-service teachers (k = 1, 6.7%). The pre-service and in-service teachers were preschool, primary, secondary, vocational, and/or special education teachers. The response rate was only reported in 6 of the 15 studies, varying between 32% and 96%.

Structural validation of the TEIP

Structural validity refers to which degree the scores of an instrument are an adequate reflection of the dimensionality of the construct teacher self-efficacy for inclusive practices (cf., Mokkink et al., 2010). According to the international consensus on measurement properties, Mokkink et al. explained that the “structural validity should be assessed to determine or confirm existing subscales” (p. 741). Hence, the included articles in the current study should show replicability regarding the distribution of the 18 items in the three-factor structure of the TEIP by the original study by Sharma et al. (2012). Therefore, each subscale item is expected to load on the specific subscale to be a pure measure of the subscale (Tabachnik & Fidell, 2013).

All 15 studies included data on the structural validity of the TEIP. In these studies, the researchers have applied different types of analyses investigating the coherence of the items and the stability of the factor structure of the TEIP (cf., Tables 1 and 2). For example, the subscales of the TEIP have been explored with principal component analysis (PCA) in 2 studies, explorative factor analysis (EFA) in 3 studies, and confirmatory factor analysis (CFA) in 11 studies (cf., Table 2). The PCA is recommended to be applied in studies where data should be empirically summarized without a theoretical foundation (Tabachnik & Fidell, 2013). The EFA is associated with theory development, whereas CFA should be used when the aim is to study underlying constructs in the data and test a theory. Four studies using PCA or EFA showed an overall explained variance of 56-65%. However, two studies revealed better support for two dimensions than three. The explained variance for the first dimension was 37-47% in all five studies using PCA or EFA. Two or four items had to be omitted to reach an acceptable fit.

Table 1. Reviewed articles on the TEIP scale with 18 items (cf., Sharma et al., 2012).

Study	N	Country	Teachers	Response rate	Psychometric aim	Structural validity	Reliability
Alnahdi (2019a)	312	Saudi Arabia	In-service	-	Construct validity	Rasch analysis	Person separation index
Alnahdi (2019b)	432	Saudi Arabia	In- and pre-service	-	Psychometric characteristics	PCA, CFA, Rasch analysis	Cronbach's alpha
Alnahdi and Yada (2020)	250	Japan	In-service (grade 1–12)	-	Construct validity	Rasch analysis	Person separation index
Cardona-Molto et al. (2020)	475	Spain	Pre-service (early childhood, elementary)	71%	Factor structure and reliability	EFA, CFA	Cronbach's alpha
Malinen et al. (2012)	437	China	In-service (primary, middle)	>90%	Factor structure	CFA	Cronbach's alpha
Malinen et al. (2013)	552	China	Pre-service (early childhood, general, special education)	-	Factor structure	CFA	Cronbach's alpha
Martins and Chacon (2020)	308	Brazil	In-service (early childhood, elementary)	-	Psychometric properties	EFA	Cronbach's alpha
Miesera et al. (2019)	887	Germany	Pre-service (elementary, lower secondary, upper secondary, vocational, special education)	-	Factor structure	CFA	-
Emam and Al-Mahdy (2019)	287	Sultanate of Oman	In-service (basic school)	96%	Factor structure	CFA	Cronbach's alpha
Narkun and Smogorzewska (2019)	291	Poland	In-service (primary)	≈50%	Psychometric properties and factor structure	PCA	Cronbach's alpha
Park et al. (2016)	134	USA	Pre-service (early childhood)	-	Factor structure	CFA	Cronbach's alpha
Tanrıverdi and Özokçu (2018)	567	Turkey	Pre-service (early childhood, primary, special education)	-	Psychometric properties	CFA	Cronbach's alpha
Vogiatzi et al. (2022)	465	Greece	In-service (secondary)	-	Psychometric properties	EFA, CFA	Cronbach's alpha
Yada et al. (2021)	105	Finland	Pre-service	32%	Factor structure	CFA	Cronbach's alpha
Yada et al. (2018)	1231	Finland, Japan	In-service (primary, secondary)	≈60% Finland, 49% Japan	Construct validity	MGCFA	-

Notes: EFA = Explorative factor analysis, PCA = Principal component analysis, CFA = Confirmatory factor analysis, MGCFA = Multigroup confirmatory factor analysis.

Table 2. Studies on the structural validation of the TEIP self-rated by teachers.

Study	N	Teachers	Stability
Alnahdi (2019a)	312	in-service	The initial Rasch analysis on 18 items did not support the unidimensionality of the TEIP scale (unidimensionality test = 14.4%). However, when removing 27 participants regarded as outliers, omitting 5 items, and rescoring 4 items, the unidimensionality of the TEIP scale was improved to 4.9% on the lower limit of 95% CI.
Alnahdi (2019b)	432	in-service and pre-service	In the initial Rasch analysis, 2 of the 18 items and 21 participants were reported to misfit the unidimensionality of the TEIP. Nextly, analyses with PCA revealed that 4 items did not function as expected according to a three-factor solution of the TEIP. There was support for two components, and 45% of the variance was explained with the first components. CFA on the three-factor structure of the TEIP was reported to have an acceptable fit (CFI = .910, TLI = .895, RMSEA = .075, SRMR = .0466). A model with omitted items and removed outliers based on finding from the Rasch analysis did not fit the data better. Thus, the authors report a slightly better fit when errors from 8 items were covaried.
Alnahdi and Yada (2020)	250	in-service	The initial Rasch analysis on 18 items did not support the unidimensionality of the TEIP scale. When omitting 4 items and 18 participants and rescoring 1 item, the unidimensionality test of the TEIP scale was 4.46% on the lower limit of 95% CI. Rasch analyses on the three subscales did not support multidimensionality within the TEIP scale.
Cardona-Molto et al. (2020)	475	pre-service	In the EFA, on 18 items, the explained overall variance was 61.52% (KMO = 0.932, Bartlett's test p < .001). The first factor explained 44.43% of the variance, the second 9.07%, and the third 7.02%. With promax oblique rotation the items in three factors had loadings between .476 and .865. Further, the CFA on the three-factor structure of the TEIP and 18 items was reported to not have an acceptable fit to the data (CFI = .845, TLI = .842, RMSEA = .111). The CFA model was reported to improve when 3 items were omitted (CFI = .864, TLI = .861, RMSEA = .090). Correlations between the subscales were reported to be high (r = .74–.90).
Malinen et al. (2012)	437	in-service	CFA on the three-factor structure of the TEIP was reported to have an acceptable fit (CFI = .92, TLI = .91, RMSEA = .06, SRMR = .06). Two items loaded on all factors and were removed from the model.
Malinen et al. (2013)	552	pre-service	CFA on the three-factors structure of the TEIP was reported to have a good fit (CFI = .96, TLI = .95, RMSEA = .05, SRMR = .04). Correlations between the subscales were between .53 and .60. One item loaded on all factors and was left out.
Martins and Chacon (2020)	308	in-service	An initial EFA explained 62.27% of the overall variance for three dimensions (KMO = 0.90, Bartlett's test p < .001). The first factor explained 46.96% of the variance, the second 8.82%, and the third 6.50%. Two items were observed to load on two factors. Also, only two items had loadings on the third factor, and the researchers argued for an additional model consisting of 16 items divided into two factors. The total explained variance was 56.95%, and the first factor explained 47.07% and the second 9.88%.
Study	N	Teachers	Stability
Miesera et al. (2019)	887	pre-service	CFA on the three-factors structure of the TEIP including 16 items proposed by Park et al. (2016) were reported to have an acceptable fit (CFI = .913, Gamma-hat = .956, RMSEA = .066, SRMR = .069).
Emam and Al-Mahdy (2019)	287	in-service	Initial CFA on the three-factor structure of the TEIP and 18 items had not an acceptable fit to the data (CFI = .864, NFI = .844, TLI = .841, RMSEA = .138). Three items had low loadings (< .31) and 2 items had cross-loadings. Further, a CFA on three-factor structure with 15 items and 4 covaried items errors was reported to have good fit (CFI = .961, NFI = .041, TLI = .952, RMSEA = .079).
Narkun and Smogorzewska (2019)	291	in-service	The PCA explained 65.05% of the overall variance (KMO = 0.923, Bartlett's test p < .0001). The first component explained 46.61% of the variance, the second 10.57% and the third 7.86%. One item loaded on component that was not expected in relation to the TEIP scale. Orthogonal rotation solution with varimax rotation resulted in loadings with values between .429 and .873 for items in the three components. Correlations between the total scale and the subscales were report to high (r = .855–.860). The correlations between the subscales (r = .573–.631) were also high.
Park et al. (2016)	134	pre-service	CFA on three-factor structure was reported to have an acceptable fit (CFI = .994, TLI = .993, WRMR = .886, RMSEA = .085, LL = .070, UL = .100). Factor loadings between .01-.47 for the factor Instructions, .17-.41 for the factor Collaboration, and .01-.46 for the factor Managing behavior. Additionally, unidimensionality of the TEIP was also confirmed (CFI = .997, TLI = .996, WRMR = .651, RMSEA = .064, LL = .045, UL = .082).
Tanrıverdi and Özokçu (2018)	567	pre-service	CFA on the three-factor structure of the TEIP was reported to have an acceptable fit of the data (CFI = .96, NFI = .95, SRMR = .054, RMSEA = .101). Factor loadings were between .44 and .81. Two items in the factor Collaboration were assumed to measure the same behavior. Also, two items in Managing behavior were reported to measure the same behavior.
Vogiatzi et al. (2022)	465	in-service	EFA explained 55.77% of the overall variance (KMO = .882, Bartlett's test p < .005). The first factor explained 37.449% of the variance, the second 9.741% and the third 8.578%. Loadings between .38-.85 for the Instructions, .53-.76 for the Collaboration, and .51-.78 for the Managing behavior. Different factor structures were explored with CFA, and the three-factor structure was reported to fit the data best (CFI = .90, TLI = .96, RMSEA = .013, SRMR = .003). Correlations between factors were reported to be between .55 to .65.
Yada et al. (2021)	105	pre-service	CFA on three-factor structure was reported to sufficiently fit the data after covariation of items (SRMR = .061, RMSEA = .060). The factors were highly correlated with each other (r = .73-.83), and factor loadings were reported to be between .47-.84 for the factor Instructions, .69-.84 for the factor Collaboration, and .57-.81 for the factor Managing behavior.
Yada et al. (2018)	872 359	in-service	An initial CFA model on the three-factor structure of the TEIP was reported to have an adequate fit (RMSEA = .052, SRMR = .061). When scalar invariance was explored, an insufficient fit was reported and explained by lack of invariance in some item intercepts (RMSEA = .067, SRMR = .093).

Similarly, one or several items were omitted in the CFA studies to reach an acceptable or good fit (see Table 2). The omitted items vary between the studies, indicating that more than a specific item should be removed to reach a stable instrument. Additionally, there were problems with cross-loadings and item errors.

Furthermore, Rasch analyses were applied in three studies to investigate whether the items measure a homogeneous latent construct, such as the teacher self-efficacy for working in inclusive practices. With this statistical analysis, the participants’ response patterns are investigated (cf., Müller, 2020). In the three studies using Rasch analyses, modifications were necessary, and some participants had to be omitted, and items had to be rescored. In addition, various numbers of items misfitted the TEIP scale. They were not regarded as a measurement unit on the continuum of the construct of teacher self-efficacy for inclusive practices. Between two and five items were suggested to be omitted to reach a more consistent scale. The results did not support a multidimensional construct of teacher-self efficacy.

In all, the results on the structural validity of the TEIP varied between the studies (cf., Table 2). Mostly, there was support for a scale with three subscales. Thus, a couple of studies demonstrated a better fit with only two subscales, and there were also findings indicating no support for dividing the TEIP into different subscales.

Reliability of the TEIP

Reliability can be explained as “the degree to which the measurement is free from measurement error” (Mokkink et al., 2010, p. 741). In the original study by Sharma et al. (2012), the 18 items in the TEIP scale were reported to reach good reliability. Among the 15 included articles in the current study, reliability was explored in 13 (see Table 1). The reliability was presented with Cronbach’s alpha in 11 studies, and the Person separation index was applied in 2 studies.

Cronbach's alpha is a measure of internal consistency and describes the coherence between the items in each subscale and the total scale (Pallant, 2010). Items reflecting the same phenomena should associate strongly (Mokkink et al., 2010). The internal consistency is reported to be satisfactory or strong for the subscales and the total scale of the TEIP (see Table 3), with translations from the original English version into other languages (Cardona-Molto et al., 2020; Martins & Chacon, 2020; Narkun & Smogorzewska, 2019; Tanrıverdi & Özokçu, 2018). However, the Cronbach's alpha value is sensitive to the number of items in the (sub)scale, and therefore, (sub)scales with less than ten items and a Cronbach's alpha below the threshold are recommended to be presented with average inter-item correlations (Pallant, 2010). Cronbach's alpha should not be interpreted as a measure of a subscale's or total scale's unidimensionality (Cortina, 1993). The inter-item correlation might be low with ten or more items even though the Cronbach alpha value is sufficient (Cortina, 1993). Inter-item correlations are reported in two of the studies (Tanrıverdi & Özokçu, 2018; Vogiatzi et al., 2022). In several studies, Cronbach’s alpha was reported for all 18 items, also when all items had not been included in the PCA or CFA.

Table 3. Studies on the internal consistency, evaluated with Cronbach's alpha, for the TEIP.

Study	N	Country	Teachers	Efficacy in inclusive instructions	Efficacy in collaboration	Efficacy in managing behavior	Total scale
Alnahdi (2019b)	432	Saudi Arabia	pre-service and in-service	.82	.82	.83	.92
Cardona-Molto et al. (2020) ^a	475	Spain	pre-service	.84	.85	.89	.92
Malinen et al. (2012)	437	China	in-service	.75	.87	.88	.91
Malinen et al. (2013) ^b	552	China	pre-service		.90
Martins and Chacon (2020) ^c	308	Brazil	in-service		.92
Emam and Al-Mahdy (2019)	287	Sultanate of Oman	in-service	.83	.93	.93	.93
Narkun and Smogorzewska (2019)	291	Poland	in-service	.83	.88	.92	.93
Park et al. (2016)	134	USA	pre-service	.93	.95	.94	.98
Tanrıverdi and Özokçu (2018)	567	Turkey	pre-service	.77	.79	.68	.89
Vogiatzi et al. (2022) ^d	465	Greece	in-service				.89
Yada et al. (2021)	105	Finland	pre-service				.93

Note: ^a Reliability based on 15 items due to the omission of 3 items in the structural validation of the TEIP. ^b Cronbach’s alpha was not specified for each scale but was reported to be between .75-.85 for the subscales. ^c Cronbach’s alpha is reported on 16 items, divided into 2 subscales (alphas were .88 and .89). ^d Cronbach’s alpha was reported to be between .88 and .89 for all items.

The Person separation index (PSI) is a measure similar to Cronbach’s alpha (Robinson et al., 2019). The PSI is calculated on logit values, while raw values are used in Cronbach’s alpha calculations. In the two studies using PSI, the internal consistency was reported to be good for the total scale of the TEIP. In one of the studies, 5 items were omitted, and the reliability analysis was performed on 13 items (PSI = 0.88) (Alnahdi, 2019a). In the other study, 4 items were omitted, and the reliability was presented for 14 items (PSI = 0.91) (Alnahdi & Yada, 2020).

In general, the internal consistency of the TEIP is reported to be satisfactory or strong for the subscales and the total scale (cf., Table 3).

Discussion

To our knowledge, the psychometric properties of the TEIP have not previously been summarized. Our search procedure revealed that the TEIP scale is widely used as it has been applied in at least 129 peer-reviewed articles. Still, we identified only 15 peer review articles that investigated the TEIP scale's psychometric properties.

The structural validation of the TEIP was performed with various statistical methods, and the researchers have been more or less conservative in their choice of fit indices and statistical methods (cf., Table 2). Several studies confirmed the TEIP scale's three-factor structure, but there were studies also suggested a one- or two-factor solution. Hence, the subscales in the TEIP scale should be interpreted with caution. There are items with cross-loadings and item errors that covary, indicating that more research is needed before the TEIP scale can be considered to have a stable factor structure.

The internal consistency on the 18 items seemed to be sufficient or good (cf., Table 3). However, inter-item correlations were only reported in two of the studies. None of the included studies investigated re-test reliability; consequently, we cannot conclude how consistent the teachers’ responses will be over time.

Besides the limited number of studies focusing on the psychometric properties of the TEIP, there are some additional limitations to be aware of in interpreting the results. The response rates are not reported in about half of the studies. Also, studies have with low response rates (cf., Table 1). This means that there may be a selection bias in the samples. Teachers who are more interested in questions related to inclusion may have responded to a greater extent than those with more limited interest. Alternatively, the responding teachers might be more pessimistic about teaching students with special needs in their classrooms. Teachers’ attitudes to inclusion are shown to be related to their teacher-self efficacy for inclusion (Yada et al., 2022). Besides, the teachers’ response pattern in the TEIP is shown to vary due to their experiences of working with students with special needs, and those with more experience tend to report higher teacher self-efficacy for inclusion (Wray et al., 2022). Also, about half of the studies are conducted among pre-service teachers, who might lack teaching experience. Their belief in their ability to work inclusively might be based on hypothetical reasoning about giving instructions and collaborating with colleagues within inclusive practices.

A possible explanation for the somewhat incoherent results might be the lack of a general agreement on a specific definition of inclusive education (Haug, 2017; Terzi, 2014). Differences between countries in how inclusion has been implemented (Haug, 2017) could have influenced the teachers’ reports on self-efficacy for inclusion. The current scoping review includes studies from twelve countries on three continents. Thus have data been collected in countries that are likely to vary greatly in terms of policies and practices of inclusive education (see, for example, Emam & Al-Mahdy, 2020; Malinen et al., 2013; Martins & Chacon, 2020; Narkun & Smogorzewska, 2019), as how inclusive education manifests in the classroom varies within countries and between schools and individual classrooms (Haug, 2017). Of relevance is also how participants have been recruited from various populations, ranging from pre-service teachers for early childhood inclusive education (Park et al., 2016) to in-service secondary teachers (Alnahdi & Yada, 2020). It is reasonable to assume that these factors contribute to the inconsistent results reported in the included studies. Therefore, the TEIP scale has limitations regarding comparing different populations and contexts.

The inconsistent results might also be connected to the fundamental problem of measuring concepts without distinct definitions, such as teacher self-efficacy (Dellinger, 2005; Henson, 2002; Zee & Koomen, 2016) and inclusion (Haug, 2017; Terzi, 2014). Therefore, operationalizing these two concepts is complex; consequently, measuring teacher self-efficacy for inclusion is challenging. Due to the lack of general definitions of inclusion and teacher self-efficacy, the validation process of the TEIP could benefit from clearly articulating the definitions the scale is based on.

Instead of continuing the validation process of the TEIP, another approach could be to develop scales that target specific special needs and teachers’ belief in their ability to teach a student with these particular special needs in inclusive settings (cf., Mintz, 2019). According to Mintz (2019), approaching inclusion as a general concept will not be successful when teacher self-efficacy is based on the variation in special needs among students. Another approach could be the direction suggested by Zee and Koomen (2016), namely to focus on the further developing the concept and measurement of teacher self-efficacy rather than developing several scales for different forms of teacher self-efficacy.

Conclusions

The widespread use of the TEIP might reflect the need for reliable and valid measurements to evaluate various aspects of inclusion. Based on the present scoping review, it can be concluded that there is a limited number of studies validating the TEIP. The results are somewhat inconsistent, but a three-factor solution is confirmed in several samples, and that the internal consistency is reported to be sufficient or good. Our results suggest that the TEIP needs further validation before it can be used to compare populations and different contexts. Therefore, considering the multifaceted nature of the concept of inclusion and the concept of teacher self-efficacy, the validation process of the TEIP could benefit from clearly articulating the specific definitions the scale is based on.

Disclosure statement

No potential conflict of interest was reported by the authors.