Patterns of processing strengths and weaknesses for LD identification : identification rates, agreement, and group characteristics
Two models for learning disabilities (LD) identification are explicitly allowed in federal regulations: (a) ability-achievement discrepancy and (b) response to intervention. Dissatisfaction with both models has led to calls for a third model, which identifies a pattern of cognitive processing strengths and weaknesses (PSW model) as a marker of LD. However, little empirical research has investigated this proposed model. This study investigated two proposed approaches for implementing a PSW model: (a) the concordance/discordance model (C/DM) and (b) the cross battery assessment (XBA) approach. All 139 participants demonstrated inadequate response to a Tier 2 intervention in sixth or seventh grade. Following Tier 2 intervention, participants completed a comprehensive battery of cognitive and academic tests. I utilized results to empirically categorize each participant as either meeting or not meeting LD criteria according to the two PSW approaches at different academic deficit cut points. Resulting group status was utilized to determine: (a) LD identification rates, (b) agreement between approaches, and (c) the relative academic performance and sociodemographic characteristics of resulting groups. The number of participants that met LD criteria varied widely, dependent upon the approach and deficit cut point (range: 10.8% - 47.5%). More participants met criteria for both approaches at higher deficit cut points. More participants met C/DM criteria than XBA criteria at similar cut points. Agreement between the two approaches was generally low. Kappa ranged from -.04 - .56 when comparing classification decisions across different iterations of the two approaches. Comparisons of groups that met and did not meet C/DM and XBA criteria on external academic and sociodemographic variables were largely null. The results highlight several potential challenges to widespread implementation of a PSW model. Both approaches identified a low percentage of students, raising questions of efficiency. Low agreement is an inevitable result of measurement error and implementation differences between the two approaches. Such variability in classification decisions suggests the models may be incompatible and should be independently validated. Further, the failure to find qualitative differences in academic needs between groups that met and did not meet LD criteria for either approach raises questions about the utility of the identification model.