Clinician Confidence, Child Characteristics and Accuracy: Screening for Autism Spectrum Disorder in Toddlers

Background:

Professionals on the front line of identifying ASD make critical decisions that are likely to have long-term consequences for children and families.  While screening instruments are available it has been suggested that experienced clinical judgment is more reliable in detecting ASD.  Clinicians must account for factors such as family history, environment, comorbidity and differential diagnoses that cause overlaps in symptoms which can obscure diagnostic boundaries, and must integrate this information to identify probable diagnostic status which then informs important clinical decisions.  Yet, little is known about the interplay between clinician and child factors that might influence these decisions, which potentially, may have a profound impact on prognosis.

Objectives:

Our aim was to examine the relationship between confidence and accuracy (CA) for clinicians’ judgments and to assess the potential contribution of developmental and behavioral profiles to the CA relationship during developmental screening.  Specifically, we explored a) the relationship between clinician confidence and accuracy of predicted diagnosis and b) the influence of child characteristics on clinicians’ confidence. 

Methods:

Participants were 125 children aged under 14-39 months (M = 28.62, SD =5.41) who presented for screening at a hospital child development center due to developmental concerns.  One-hour long screening interviews were conducted by healthcare professionals.  Following the screening interview professionals completed a questionnaire regarding whether or not they thought the child has ASD.  Clinicians provided a risk estimate and a Likert scale was used to generate a confidence score.  Diagnostic status and assessment results were retrieved from medical record review.

Results:

Experienced healthcare workers exhibited good sensitivity in identifying ASD.  We identified a small but positive correlation between confidence and accuracy, r_τ (119) = .24, p = .003, with diagnosticity highest at confidence levels of 90-100%.  Regression analysis identified parent report of unusual behaviors as the only significant predictor of clinician confidence (t = 2.376, p = .02, β = .395). 

Conclusions:

Despite good sensitivity in clinician prediction of diagnosis, this sensitivity appears to come at the cost of over-classification of children who did not receive an ASD diagnosis.  We identified a small but positive CA relationship in referred children, with diagnosticity for positive identifications highest in the 90% to 100% confidence band.  The CA relationship was less well calibrated for negative identifications.  Parent report of unusual behaviors was found to be the only significant predictor of clinician confidence, however this is not surprising in the light of the fact that a wide range of restricted and repetitive patterns of behaviors has been consistently reported among the earliest infant predictors of a later ASD diagnosis.