.png){kind=icon}
.png){kind=icon}
Return
How Accurate Are Computer-Adaptive Reading Screeners? A Systematic Review and Meta-Analysis of Validity and Classification Accuracy
.png){kind=icon}
PRE
AI
L
Y
Z
M
C
N
DOI：10.1080/2372966X.2026.2633597.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This meta-analysis evaluated the criterion-related validity and classification accuracy of computer-adaptive tests (CATs) used as universal screeners for reading, with statewide achievement tests or norm-referenced assessments serving as criterion measures. Across 30 included studies, CATs demonstrated strong criterion-related validity (r = 0.69, 95% CI [0.66, 0.72]). Pooled classification accuracy estimates indicated satisfactory specificity (0.83, 95% CI [0.80, 0.86]) and moderate sensitivity (0.77, 95% CI [0.73, 0.80]), with sensitivity falling below commonly recommended benchmarks for universal screening (e.g., ≥0.80). Moderator analyses revealed systematic variation in CATs’ performance. Compared with early elementary grades, sensitivity was lower, but criterion-related validity was higher in upper elementary and secondary grades, and specificity was higher in secondary grades. Data-driven threshold selection methods were associated with higher sensitivity but lower specificity relative to publisher-recommended cutoffs. Sensitivity decreased, and specificity increased as base rates increased. In addition, specificity was higher for CATs based on three-parameter IRT models than for those based on one-parameter models. Implications for school-based universal screening practice, CAT development, and future research are discussed.
Impact Statement
Schools increasingly adopt computer adaptive tests (CATs) to screen students for reading risk. This meta-analysis shows that while CATs demonstrate strong criterion-related validity, their classification accuracy, especially sensitivity, is often below recommended benchmarks. To support effective data-based decision-making, schools should use locally validated thresholds and consider supplementing CATs with additional tools to avoid missing students in need of early reading intervention.
Keywords:
Reading
computer adaptive test
universal screening
meta-analysis
Benjamin G. Solomon
Journal
IF:
4.2
Papers:
515
Citations:
3.3K
Organization
Cited Papers