Using student performance during a reading intervention to predict student outcomes and performance on accountability measures of reading.

Abstract

An existing data set for a sample of 3rd grade students was used to determine the relationship between performance during a reading intervention and short-term achievement test outcomes, and long-term risk status. Students participated in a reading intervention, one-on-one practice with a trained adult, during which weekly curriculum based measurement-reading data and dosage of intervention service time were collected. The sample was comprised of reading curriculum based measurement progress monitoring data for 99 students from 8 different schools. The recommendation was for students to receive 60 minutes per week of fluency sessions, the average student received a dosage of 51 minutes per week. Student scores on district and state reading tests from the spring of third grade, and student participation in pre-referral or special education programming 1 and 2 years post participation were also examined. Reading test data were provided by the schools including student performance on the spring 3rd grade No Child Left Behind state reading test, and for 6 of the schools, a district reading test by the Northwest Evaluation Association called the Measures of Academic Progress. Schools also provided existing information about free/reduced lunch eligibility and risk-status, for each of the students. The following research questions guided the study: (a) how does performance during a third-grade reading intervention predict future performance on district and state reading tests; and (b) how does performance during a third-grade reading intervention predict future risk-status in 4th and 5th grades? Multiple linear regression models were used to determine the amount of variance accounted for in state and district reading test performance, using the predictor variables related to participation in the 3rd grade reading fluency intervention. Twenty-three percent of the variance in state test scores, and 44% of the variance in district test scores could be explained using multiple linear regression models with the predictor variables: lunch status, slope, baseline and end levels, and dosage. There were different significant coefficients in each model, and 2 of the schools did not administer the MAP. Risk-status 1 and 2 years following participation in the 3rd grade reading intervention also was determined, using the predictors: performance during the intervention, dosage, and lunch status. Logistic regression models used to predict placement in special education in 4th or 5th grades were not significant. The logistic regression model used to predict student participation in pre-referral programming in 4th grade was not significant, but the model was significant in 5th grade. Variables were backwards eliminated to further determine which variable(s) in the model was/were significant in predicting pre-referral participation in 5th grade, and baseline level at the onset of the 3rd grade reading intervention was the significant predictor. An estimated probability plot illustrated that students with a higher baseline level were less likely to participate in pre-referral programming in 5th grade, versus students with lower baseline level performance. The findings from the first part of this study further support the existing research that reading curriculum based measurement does predict performance on other reading tests (Baker, et al., 2008; Chard, Vaughn & Tyler, 2002; Lennon & Slesinski, 1999) and state reading tests (Stage & Jacobsen, 2001; Good, Simmons & Kame’enui, 2001; Crawford, Tindal, Stieber, 2001; McGlinchey & Hixson, 2004; Keller-Margios, Shaprio & Hintz, 2008). The current study is unique in its application of progress monitoring data for students that participated in a reading intervention, versus seasonal benchmark scores. The second part of the study examined risk-status in 4th and 5th grades for the students who participated in the reading intervention service in 3rd grade. Results supported the findings of other studies, that some intervention participants struggle to maintain gains in subsequent years, (Byrne & Fielding-Barnsley, 1993; Bus & Van Ijzendoorn, 1999; O’Connor, Notari-Syverson, & Vadasy, 1996, 1998) and that curriculum-based measurement baseline level, can be a strong predictor of growth (Silberglitt and Hintz, 2007). The current study was unique in that much of prior research has examined results for students that participate in interventions higher in dosage, diverse in skill areas and prior to 3rd grade. The findings from this study suggest that progress-monitoring data at the on-set, and during an intervention, do relate to performance on other reading assessments, and may predict future risk-status.