Test-retest Reliability and Measurement Invariance across time of the Quantitative-CHecklist for Autism in Toddler

Abstract

The study evaluated the stability over time of the Q-CHAT (i.e., Quantitative-CHecklist for Autism in Toddler), a screening tool for the early detection of Autism. The Q-CHAT is filled by parents when the child is from 18-36 months old, and whether a risk score is found, the clinician refers the child to a more in-depth diagnostic evaluation. Previous studies demonstrated the measure test-retest reliability on a small time interval (1-6 months). Therefore, the present study tested the stability of the scores considering a larger time interval of 18 months. No previous studies have tested its measurement invariance over time. Since the tool is filled by parents during a large time span (18-36 months of life), it is imperative to understand whether there is an invariance over time in their evaluations. The Italian version of the Q-CHAT was completed by 282 parents of children with no pre-existing signs of risk of Autism. The Q-CHAT was administered when children were 18 months (T1) and then 18 months later (T2). The intraclass correlation coefficients for the test-retest reliability ranged from sufficient to moderate. The measurement invariance across time revealed a tolerable configural and metric invariance. Contrary, the scalar invariance was not met meaning that the means of the constructs are not invariant over time. The findings give a further demonstration of the reliability of the Q-CHAT. They give evidence that changes in the scores would reflect real changes in the construct itself, and not in the way individuals interpret the measure items.