Evaluating the CELF‐5 Screening Test and Vineland‐3 for Identifying Language Difficulties in Autism and Attention Deficit Hyperactivity Disorder

ABSTRACT Language screening tools are frequently used to identify children with potential undiagnosed language difficulties. These difficulties are more prevalent in autistic children and those with attention deficit hyperactivity disorder (ADHD) compared to neurotypical peers. Despite the widespread use of tools like the Clinical Evaluation of Language Fundamentals, Fifth Edition Screening Test (CELF‐5 Screener) and the Vineland Adaptive Behavior Scales, Third Edition (Vineland‐3), their sensitivity and specificity for this population have not been empirically validated. This study aimed to evaluate the screening accuracy of the CELF‐5 Screener and Vineland‐3 in children diagnosed with autism and/or ADHD and compare their performance to the gold standard measure. The sample consisted of 132 participants ( n autism = 25; n ADHD = 29, and n autism+ADHD = 78; M age in years = 9.6; % male = 59) from the Monash Autism‐ADHD Genetics and Neurodevelopment Project. The sensitivity, specificity, positive predictive value, and negative predictive value of the CELF‐5 Screener and Vineland‐3 receptive and expressive language subdomain scores were compared against those of the clinician‐administered CELF‐5 receptive and expressive language composite scores. The screening accuracy of each tool was further evaluated through Receiver Operating Characteristic analyses and calculations of Youden's J statistic. The CELF‐5 Screener demonstrated poor sensitivity for receptive language difficulties (35.6%) while demonstrating high specificity (95.3%). Similarly, for expressive language difficulties, the sensitivity was low (37.9%), and the specificity was high (91.1%). The Vineland‐3 showed high sensitivity (93.3%) but low specificity (48%) for expressive language difficulties and inadequate sensitivity (80.9%) and specificity (22.4%) for receptive language difficulties. Both the CELF‐5 Screener and Vineland‐3 may miss a significant number of children with co‐occurring language difficulties related to autism and/or ADHD. Examiners must understand these tools' strengths and limitations, especially when assessing neurodivergent children whose language development might not follow a normative trajectory.