Witmer, S. E., & Roschmann, S. (2020). An examination of measurement comparability for a school accountability test among accommodated and non-accommodated students with autism . Education and Training in Autism and Developmental Disabilities , 55 (2), 173–184. https://www.jstor.org/journal/eductraiautideve

Several accommodations were identified as part of the analyses of the mathematics assessment dataset. In part, they included live/in-person oral delivery, recorded human voice oral delivery, test administered across multiple days, and other accommodations—each in lower frequencies.

An extant data set from Michigan (U.S.) consisting of students in grade 4 and grade 7 was analyzed. Three groups of students were identified: (a) students with autism who received assessment accommodations, (b) students with autism who did not receive assessment accommodations, and (c) a reference group of students without disabilities who did not receive assessment accommodations. Students with autism included 567 grade 4 students and 593 grade 7 students. The grade 4 students with autism included 240 students receiving accommodations and 327 students not receiving accommodations. The grade 7 students with autism consisted of 217 students receiving accommodations and 376 not receiving accommodations. For each grade level (grades 4 and 7), a reference group of 1,000 students without disabilities not receiving accommodations was randomly selected. Participant demographics including sex, ethnicity/race, and socioeconomic information were reported.

An extant data set from Michigan consisting of grade 4 and grade 7 mathematics scores from 2012 for students with and without disabilities was obtained and analyzed. Specifically, student responses to multiple-choice test items were analyzed; constructed-response test item scores were not examined. The grade 4 mathematics construct included: Geometry, Measurement and Data, Number and Operations in Base Ten, Number and Operations for Fractions, and Operations and Algebraic Thinking. The grade 7 math construct included: Expressions and Equations, Geometry, The Number System, Ratios and Proportional Relationships, and Statistics and Probability. Two aspects of differential item functioning (DIF) presence were investigated to determine the comparability of measurement across the participant groups: (a) presence of DIF on state math assessments when comparing students with autism who received accommodations to non-accommodated students without disabilities, and (b) presence of DIF on state math assessments when comparing students with autism taking the assessment without accommodations to non-accommodated students without disabilities. These analyses were completed separately for grade 4 math tests and grade 7 math tests.

Differential Item Functioning (DIF) analysis results indicated no systematic performance differences  on math test content or item features for any group of students, whether students with or without disabilities, in either grade 4 or grade 7. Put another way, despite some items (5–8%) functioning differently to a small degree on average for some student groups, the overall effects, such as on test scores, were not substantial. The implications are that the math test, with few and small exceptions, primarily measured the same academic content at the same difficulty level for students with autism and students without disabilities. Item-level comparisons for students with autism receiving or not receiving accommodations found that slightly more test items had small yet significant differences in item functioning for students with autism not using accommodations in comparison to fewer test items functioning differently for students with autism who used accommodations. The implications are that the accommodations did not problematically affect the academic content being assessed in a systematic manner for accommodated and non-accommodated students with autism. The researchers offered additional discussion about the need for including students with autism in state accountability assessments. Limitations of the study were reported, and future research directions were suggested.