Data-Based Decision Making in Reading Interventions: A Synthesis and Meta-Analysis of the Effects for Struggling Readers

15 studies of reading interventions incorporating data-based decision making, published 1975-2017

Struggling readers in grades K-12

The modern replacement for the 1986 figure, and independent of the developers of curriculum-based measurement. It reports two contrasts, and the second is the one that answers the actual question: comparing DBDM interventions to business-as-usual confounds DBDM with getting an intervention at all, whereas the six-study subset comparing the SAME intervention with and without DBDM isolates what the data-based decision making itself adds. Graded C: pooled mixed designs, k = 6 for the clean contrast, and the main-effect confidence interval nearly touches zero.

Reading intervention with DBDM against business-as-usual gives g = 0.24 (95% CI 0.01 to 0.46). The cleaner contrast - the same intervention with versus without DBDM - gives g = 0.27 (95% CI 0.07 to 0.47) across six studies. So the decision-making layer adds roughly a quarter of a standard deviation, and that is the honest expected value for "use the data to decide what to teach next", against the 0.70 the 1986 meta-analysis reported. The authors are explicit that experimental investigation is still needed to establish DBDM as an evidence-based practice at all, which for a practice this widely mandated is a striking admission.

Low. The clean contrast holds the intervention constant and varies only whether progress data drove instructional changes.

Consistent in magnitude with Jung et al. (2018) on data-based individualisation (g = 0.37) and with the broader formative-assessment literature at scale; deflates Fuchs & Fuchs (1986) by roughly two-thirds, in the direction this archive's efficacy-to-effectiveness rule predicts.