End-of-Kindergarten Screening Measures
Table 3 provides classification and criterion validity results for several end-of-kindergarten screening measures used to predict end-of-grade 1 reading. As before, researchers employed different performance levels to divide unsatisfactory from satisfactory reading levels on the criterion reading measure, which in turn affected the proportion of the school population designated at-risk by the screen. Also as before, some researcher identified risk-groups in a post-hoc fashion by linking performance on the criterion test to cut-scores on the screening test. In contrast, Speece et al. (2003) arbitrarily designated the bottom 25% of their sample as at risk.
| Measure/Study | Sample | Type of Evidence | Result | |||
| Combination of: (1) Letter Naming Fluency (LSF) (2) Phoneme Segmentation (3) Sound Repetition O'Connor and Jenkins (1999) |
129 April K | Classification | At Risk 19% |
Sensitivity 100% |
Specificity 87% |
Criterion Below 8th percentile WRMT Basic Reading - Spring Grade 1 |
| Same combination with revised cut-scores O'Connor and Jenkins (1999) |
101 April K | At Risk 17% |
Sensitivity 100% |
Specificity 91% |
Criterion Same |
|
| TRPI (1) Letter Sound Identification (LSI) (2) Blending Phonemes Foorman et al. (1998) |
421 April K | Classification | At Risk 50% |
Sensitivity 90% |
Specificity 62% |
Criterion Below 23rd percentile WJ-Broad Reading - Grade 1 |
| Composite of CTOPP-Blending and Elision Speece et al. (2003) |
39 Spring K | Classification | At Risk 25% |
Sensitivity 66.7 42.9 |
Specificity 81.8 78.1 |
Criterion Below 26th percentile WJ-R Word Attack CBM-ORF |
| Concurent Validity | .68 WJ-R Letter-Word ID | |||||
| Predictive Validity Spring Grade 1 |
.73 WJ-R Letter-Word ID .73 WJ-R Word Attack .62 CBM-ORF |
|||||
| DIBELS-PSF Good et al (2001) |
353 Spring K | Concurent Validity | .62 NWF | |||
| Predictive Validity | .68 WJ-R Reading Cluster | |||||
| Spring K-Winter Grade 1 | .62 CBM-ORF | |||||
| Spring K-Winter Grade 1 | .54 WJ-R Reading Readiness Cluster | |||||
| Letter Name Fluency (LNF) Speece et al. (2003) |
39 Spring K | Classification | At Risk 25% |
Sensitivity 50.0 87.5 |
Specificity 78.8 87.5 |
Criterion Measure WJ-R Word Attack CBM-ORF |
| Concurent Validity | .55 WJ-R Letter-Word ID | |||||
| Predictive Validity Spring Grade 1 |
.55 WJ-R Letter-Word ID .44 WJ-R Word Attack .69 CBM-ORF |
|||||
| Other Predictions Spring Grade 1 |
(1) LNF accounted for no unique variances in Letter Word ID, Word Attack, or CBM-ORF, after controlling for Phonological Awareness and NWF. | |||||
| DIBELS-NWF Speece et al. (2003) |
39 Spring K | Classification | At Risk (Lowest 25% of sample on NWF in Spring K) |
Sensitivity 50.0 85.7 |
Specificity 72.7 81.3 |
Criterion Below 26th percentile WJ-R Word Attack CBM-ORF |
| Concurent Validity | .91 WJ-R Letter-Word ID | |||||
| Predictive Validity Spring K-Spring Grade 1 |
.59 WJ-R Letter-Word ID .59 WJ-R Word Attack .71 CBM-ORF |
|||||
| Other Prediction Spring Grade 1 |
(1) NWF did not accout for
significant variances in WJ-R Letter-Word ID once phonology awareness was
controlled.
(2) NWF accounted for significant variance in WJ-R Word Attack and CBM-ORF, after Phonological Awareness was controlled. (3) NWF accounted for significantly more variance than LNF in Word Attack and CBM-ORF. |
|||||
Sensitivity levels attained by candidate screening measures varied depending on the criterion measure used to measure reading outcomes. Letter Naming Fluency (LNF) and DIBELS NWF showed poor (50%) sensitivity when WJ-R Word Attack was the criterion measure, but good sensitivity (88%) when CBM-ORF was the criterion measure (Speece et al., 2003).
The highest sensitivity levels (93% and 100%) were reported by Foorman et al. (1998) and O'Connor and Jenkins (1999), respectively. The former used a combination of Letter Name/Sound Knowledge (LN/S) and blending onset-rimes in screening, and the latter used a combination of phoneme segmentation, LNF, and sound repetition. However in evaluating the sensitivity achieved by the various screens, it is important to note that both Foorman et al. and O'Connor and Jenkins established screening cut-scores by working backward from the criterion measure--a way to guarantee reasonable sensitivity. Their specificity ranged from 63% to 87%.
Previous Page | Next Page
(Promising Measures) | (Beginning First Grade)
