Improving Reading Screening for Third Grade Students from Culturally & Linguistically Diverse Backgrounds

Improving Reading Screening for Third Grade Students from Culturally & Linguistically Diverse Backgrounds

First Author: Shawn Kent -- University of Houston
Keywords: Reading, Screening, Technology
Abstract / Summary: 

Purpose: Hispanic students, many of whom are ELLs, are the fastest growing population of students in public schools; nearly 80% of ELLs speak Spanish. Hispanic students, including ELLs, are at increased risk for disparities in reading achievement and possible placement in special education. Early identification has the potential to ameliorate reading problems for these students, yet limited research exists on screening with students from culturally and linguistically diverse (CLD) backgrounds. Research does suggest measures of language to improve screening for ELLs. Also, computer adaptive tests (CATs) represent a potential avenue for screening, but few studies have directly examined their diagnostic accuracy for screening. We sought to evaluate the diagnostic accuracy of individual measures, including CATs, and multivariate screening approaches, in predicting Grade 3 state reading assessment outcomes for students from CLD backgrounds. Method: Data from 220 Grade 3 students from a public charter academy in a large, urban area was included. Of sample, 96% were Hispanic and 40% designated ELLs. In fall, students were administered two CATs, MAP Reading and ISIP Vocabulary, along with ORF and a running record. In spring, students took the state reading assessment; 59% met or mastered expectations. Logistic regression and ROC curve analyses were conducted for individual and combined measures. Results: Single measures provided inadequate diagnostic accuracy, with very low sensitivity (< .62), though adequate specificity. Improved accuracy noted with multivariate approach (SE/SP> .80), though using more than two measures does not appear warranted. A combination of MAP and ORF was most accurate and given these measures, represents an efficient screening procedure. When examining CATs, together they accounted for 60% of variance in outcome. Future research utilizing locally-derived cut scores appears warranted in order to increase diagnostic accuracy of screening process.