Do Curriculum-Based Measures of Comprehension (CBMs-C) contribute beyond oral reading fluency differentially for student subgroups?

Do Curriculum-Based Measures of Comprehension (CBMs-C) contribute beyond oral reading fluency differentially for student subgroups?

First Author: Steve Amendum -- University of Delaware
Additional authors/chairs: 
Kristin Conradi Smith; Meghan D. Liebfreund
Keywords: Demographic subgroups, Reading fluency, Predictors of reading skills, Grade 3, Regression models
Abstract / Summary: 

The purpose was to examine how well CBM assessment data typically used by schools related to end-of-grade performance on a high-stakes state test of 3rd grade reading comprehension and whether relationships varied by students’ demographics. We asked: (RQ1) Does oral reading fluency (ORF) predict end-of-year reading achievement and do CBMs-C predict end-of-year achievement above and beyond ORF? (RQ2) Do ORF and CBMs-C predict end-of-year reading achievement differently based on student demographic factors? The sample included 9,967 grade 3 students from one U.S. school district, including around 50% students of color, 33% with subsidized lunch, 12% with limited English proficiency, and 4% who received special education. The data sources were two school-based assessments. The first was DIBELS Next (Dynamic Measurement Group, 2014), specifically DIBELS ORF (DORF), RETELL, and DAZE (the DIBELS comprehension maze test). The second was the state reading comprehension achievement test (RCAT) administered at the end of grade 3. Demographic data included subsidized lunch status (LUNCH), Limited English Proficiency status (LEP), special education status (SPED), and student race/ethnicity (RACE). A series of hierarchical regression models were used with RCAT as the outcome. Main conclusions were: (RQ1) DORF explained meaningful variation in RCAT. While the CBMs-C also accounted for unique variance, the additional amount accounted for may not have been meaningful. (RQ2) DORF explained more variation in RCAT for students who received subsidized lunch, had limited English proficiency, received special education services, and identified as Black. Notably, the addition of the CBMs-C explained similar, but small, amounts of unique variation in RCAT across demographic subgroups (ranging from 1%-4%). Findings raise questions about potentially unequal benefits of ORF for different student subgroups, as well as questions about the utility of CBMs-C for predicting later achievement.