Discordance Across Time in Caregiver Report during the Autism Diagnostic Interview-Revised (ADI-R): Findings from a Canadian Inception Cohort of Children with Autism Spectrum Disorder (ASD)

Background: The ADI-R algorithms rely on accurate timing of the first manifestations of ASD symptoms. Caregiver recall of developmental milestones and the emergence of symptoms are thus key to an accurate diagnosis, especially when the evaluation occurs later in life. Previous studies in ASD found “telescoping” in caregiver report of language milestones, such that older ages of attaining milestones were reported as time passed (Hus et al. 2011). Several proposed factors influencing recall (Barsky, 2002; Hus et al., 2011; Jones et al., 2015) include characteristics of the informant (e.g., state at the time of recall, beliefs about etiology), the child (e.g., delays in development), and the interview (e.g., order of questions, time since the recalled event).

Objectives: In the present study we examined telescoping in caregivers’ longitudinal reports of their children’s early development in a prospective cohort study of children diagnosed with ASD. Specifically, we focused on ADI-R reports of age of first concerns, age of first words, and age of first phrases. We also examined possible moderators of telescoping including caregiver education level and child’s age and cognitive level at the time of diagnosis.

Methods: Data were drawn from Pathways in ASD, a study of 400 preschoolers followed across five Canadian sites. We included in the analysis cases for whom three longitudinal ADI-R interviews were available (n = 136 participants; 117 Males) and excluded cases for which caregivers either reported that their child had no functional language at study entry and/or possible/definite language regression at any point. ADI-R interviews were completed when the children were 2-4 years old (around time of diagnosis; mean = 46.6 months, SD = 7.9), 6-7 years old (mean = 79.8 months, SD = 4.0), and 10-11 years old (mean = 128.4 months, SD = 2.3). We explored as covariates: maternal education, child’s age of diagnosis (mean = 44.7 months, SD = 7.7), and the child’s cognitive ability measured with the Merril-Palmer-Revised (mean = 77.9 months, SD = 24.5) around the time of diagnosis.

Results:  We found no telescoping effects in report of symptom onset; caregivers reported similar ages of onset across the three interviews. In contrast, we observed significant telescoping effects in reported age of language milestones. Over time, caregivers tended to report older ages of acquiring first words (p=.006) and first phrases (p<.001). We also examined potential moderators of telescoping. Caregiver reports of symptom onset over time did not interact with maternal education or with child age and cognitive ability around diagnosis. However, more telescoping of language milestones was evident in caregivers of children with lower cognitive ability (first words p=.01; first phrases p<.001). There were no significant interactions between report of language milestones over time and child’s age of diagnosis or maternal education.

Conclusions: Overall, our findings in a Canadian inception cohort replicate previous results of telescoping in caregiver-reported ages of language acquisition but not of first concerns. We will discuss implications of reliance on retrospective recall of information for clinical diagnosis and research, especially later in life.