Screening for Learning Difficulty Using Teacher Ratings on the Colorado Learning Difficulties Questionnaire

Abstract

The present study examined clinical utility of teacher ratings on the Colorado Learning Difficulties Questionnaire (CLDQ) learning difficulties screening within a referred, school-aged sample ( N =519, 5-18 years). Of this sample, 419 youth had CLDQ reading scale scores from a reading or general education teacher, and 338 had CLDQ math scale ratings from a math or general education teacher. Sensitivity, specificity and receiver operating characteristic (ROC) curves were examined for specific reading and math difficulties (achievement SS<85). Cut-scores were identified to maximize sensitivity (reading: 60–91%; math: 85–89%), but specificity was low (reading: 60–64%, math: 47–48%); AUCs ranged from .70-.87 for reading and .75-.77 for math. Discrimination was comparable in an elementary subsample. Conditional probabilities suggested CLDQ ratings more accurately predicted children without learning difficulties (i.e., true negatives) than with learning difficulties. Parent and teacher ratings were well correlated ( r =.71), but addition of teacher ratings improved classification accuracy and model fit ( p <.001) across domains. Furthermore, CLDQ teacher ratings were helpful in reducing false positives based upon parent ratings alone. Findings suggest teacher ratings via CLDQ can be used to screen children at risk for learning difficulties, with teacher ratings showing added value over and above parent ratings. Subscribe to LDMJ