Selecting at-risk readers in first grade for early intervention: A two-year longitudinal study of decision rules and procedures.

TLDR

RTI models for identifying learning disabilities depend on accurately spotting children who would develop reading disability without Tier 2 tutoring. The study investigated whether adding first‑grade word identification fluency and its progress‑monitoring data improves RD prediction, and whether classification tree analysis outperforms logistic regression. Four classification models were evaluated using data from 206 first‑grade children tracked through the end of second grade. Combining initial WIF, WIF‑Level, and WIF‑Slope with classification tree analysis significantly improved prediction, supporting their use in RTI.

Abstract

Response to intervention (RTI) models for identifying learning disabilities rely on the accurate identification of children who, without Tier 2 tutoring, would develop reading disability (RD). This study examined 2 questions concerning the use of 1st-grade data to predict future RD: (a) Does adding initial word identification fluency (WIF) and 5 weeks of WIF progress-monitoring data (WIF-Level and WIF-Slope) to a typical 1st-grade prediction battery improve RD prediction? and (b) Can classification tree analysis improve the prediction accuracy compared to logistic regression? Four classification models based on 206 1st-grade children followed through the end of 2nd grade were evaluated. A combination of initial WIF, WIF-Level, and WIF-Slope and classification tree analysis improved prediction sufficiently to recommend their use with RTI.