Escala de triagem para identificação de sinais de autismo para professores (TEA-prof) : evidências de validade, precisão e parâmetros individuais dos itens

Abstract

An important way to track Autism Spectrum Disorder (ASD) in school-age children is through information from education professionals who live with them daily. Thus, this current study aimed to verify the evidence of validity, precision and individual parameters of the items of the “Screening Scale for Identification Autism Signs for Teachers (ASDTeacher)”. Data collection was carried out remotely, being the ASD-Teacher Scale available via GoogleForms® and applied to 1237 elementary school teachers belonging to two different groups: Group A: Teachers of autistic children, with at least six months of work experience; and applied to 29 parents and teachers; B: Teachers of typical children, with at least six months of work experience; and 29 parents and teachers corresponding to Group C: Parents of autistic children, according to their own report; and Group D: Teachers with at least six months of experience with children whose parents belonged to Group C, according to their own report. For the data comparison analysis, between the ASD-Teacher Scale responses of Groups A and B, the Multivariate Analysis of Variance (MANOVA) was used, using the Statistical Package for the Social Sciences, version 22.0. The results indicated that the scale differentiated typical children from children with ASD, as well as identified which clinical grades the latter fell into (Mean Square = 396483.195; F = 253.288; Eta partial square = .388; p>0.005). Structural validity verification was performed using Confirmatory Factor Analysis in R studio software. As a result, adequate fit indices were found for a hierarchical model. In this case, two factors: (a) persistent deficits in social communication and social interaction in multiple contexts; (b) restricted and repetitive patterns of behavior, interests or activities; are justified by the superior factor called “ASD”. To assess the instrument's accuracy, the hierarchical McDonald's Omega (ωh) was used for the total accepted model and Cronbach's alpha (α) for each independent factor identified. The partial results present, in relation to the precision of the test, total consistency of ωh = 0.99 (Confidence Interval [CI] 0.99 – 0.99). The precision of each existing factor was also found: for Persistent Deficits in Social Communication and Social Interaction (CSI): α = 0.97 (CI 0.97 – 0.97); Restricted and repetitive patterns of behavior, interests or activities (PRR): 0.98 (CI 0.98 – 0.98). It was also performed another analysis of the correlation between the instruments applied to the parents, Childhood Autism Rating Scale (CARS) and Autism Behavior Checklist (ABC), with the ASD-Teacher Scale, applied to the teachers (Groups C and D). To correlate the scores of the instruments, the Spearman Rank Correlation Coefficient (rs) was used and to measure the effect size, the Determination Coefficient (r) was used. As a result, the correlation obtained between ABC and CARS was the one with the best indices: rs = 0.79; p = < 0.01; r² = 0.62. In other words, there was a strong and significant correlation with a great effect. The results obtained for the correlation between ASDTeacher scale and ABC were: rs = 0.30; p = 0.11; r² = 0.09. For ASD-Teacher Scale and CARS: rs = 0.35; p = 0.06; r² = 0.12. That is, they presented a moderate level in both cases, but were not significant. Despite this, the effect sizes were medium. For the Item Response Theory (IRT) analysis, the R studio software was used. Graded response (GRADED), graded scale (RSM) and generalized partial credit (GPCM) models were tested. As a result, GRADED proved to be the most adequate model. Furthermore, the most significant items were also chosen for the CSI and PRR factors, in addition to the scale's skill level being also adequate, presenting discriminative capacity for children with low levels of symptoms/skills. Therefore, the ASD-Teacher Scale presented adequate levels of precision, in the same way that its scores presented adequate indices for the investigated validities.