The Influence of Clinical Judgment in a Record-Review Surveillance System on Autism Spectrum Disorder Prevalence Estimates

Background: Symptoms of autism spectrum disorder (ASD) overlap with other conditions in early childhood (e.g., attention deficit hyperactivity disorder). Differential diagnosis by a clinician is important to distinguish ASD from other disorders. One way to estimate ASD prevalence while mitigating resources needed for in-person evaluations is to have clinicians apply professional judgment to abstracted service (educational and health) records to determine case status. Clinical judgment applied to service records may exclude children who meet a standardized definition of ASD, but have symptoms that are better accounted for by another disorder.

Objectives: The objectives of this analysis were to (1) identify the number of children in a population-based surveillance system who met a standardized definition of ASD, but were disqualified as having ASD by a clinician who reviewed service records, (2) assess associations between ASD status determined by a clinician who reviewed the service records and the clinician who examined the child and authored the service record, and (3) estimate the change in ASD prevalence if children who were disqualified by a clinician who reviewed service records were included in ASD prevalence estimates.

Methods: Participants were children who met the Autism and Developmental Disabilities Monitoring Network (ADDM) case definition for ASD in surveillance year 2012. Education and health records were abstracted by project personnel and sent to clinicians for comprehensive review. Clinicians applied a standardized coding scheme based on DSM-IV-TR criteria to abstracted information to determine ASD case status. A child was “disqualified” if he or she met standardized coding criteria for ASD, but the clinician felt that symptoms were better accounted for by another disorder or any other reason. Children who were disqualified were not included in ASD prevalence estimates.

Results: 6,112 children met ADDM coding criteria for ASD; 5,063 (83%) were not disqualified and 1,049 (17%) were disqualified after record review. Logistic regression found that children who were disqualified were less likely than children who were not disqualified to have 1) a diagnosis of autistic disorder, pervasive developmental disorder – not otherwise specified, or Asperger disorder (9.1% versus 73.8%); 2) a special education classification of autism (1.0% versus 23.9%); and ASD characteristics noted in service records (16% versus 67.1%). However, children who were disqualified were more likely than children who were not disqualified to have ASD excluded by the clinician who evaluated the child and authored the service record (28.4% versus 10.0%). ASD prevalence increased from 14.6 per 1,000 to 17.6 per 1,000 when children who were disqualified were included in prevalence estimates.

Conclusions: The estimated prevalence of ASD is higher when clinical judgment is not applied to surveillance case definitions. More children who do not have ASD noted in service records – and who do have ASD excluded by a clinician who evaluated the child – are omitted from surveillance counts when clinicians disqualify children based on professional judgment. More research is needed to determine the influence of clinical judgment on ASD prevalence estimates, and the clinical characteristics that are associated with disqualification of ASD surveillance case status.