How Has DSM-5 Affected Autism Diagnosis? a Five-Year Follow-up Systematic Literature Review and Meta-Analysis

Background: Autism spectrum disorder (ASD) is the fastest growing development disability, and its increasing prevalence during the last two decades has prompted public health concerns. Despite the marked increase in ASD research, its diagnosis remains behaviorally based on clinical diagnostic criteria. The 2013 publication of the Fifth Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) significantly revised autism criteria with vastly fewer options to obtain a diagnosis. This has caused concern that DSM-5 criteria may fail to capture some individuals who would have previously met ASD diagnostic criteria under DSM-IV, Text-Revised (DSM-IV-TR), but may still benefit from services. Most recent data estimate ASD prevalence to be 1 in 59 children; however, these data primarily included those evaluated under DSM-IV-TR, leaving the effect of DSM-5 unknown.

Objectives: The purpose of this systematic literature review and meta-analysis were to: (1) determine the changes in frequency of ASD diagnosis in the first five years after publication of the revised DSM-5 ASD criteria, and (2) identify the DSM-IV-TR autism subtypes most affected by these new diagnostic criteria.

Methods: Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we searched the literature for studies published between April 2013 and July 2018 that applied both DSM-IV-TR and DSM-5 ASD diagnostic clinical criteria to study samples. Scientific rigor was rated using the Quality Appraisal of Reliability Studies. Data on sample size, individuals meeting DSM-IV-TR ASD criteria, and those no longer meeting an ASD diagnosis under DSM-5 criteria were extracted. Pooled effects were estimated for ASD and DSM-IV-TR subtypes of autistic disorder (AD), Asperger’s Disorder, and pervasive developmental disorder-not otherwise specified (PDD-NOS) using random effects meta-analysis models. Heterogeneity of each model, subtype analyses to explore reasons for heterogeneity if present, and publication bias was assessed.

Results: Of 898 studies identified, 33 met inclusion criteria for the review and meta-analysis; of these, 19 studies specifically examined DSM-IV-TR subtypes (AD n=17; Asperger’s Disorder n=14; PDD-NOS n=18). Overall risk of bias across studies was unclear. Most studies utilized the Autism Diagnostic Interview–Revised (ADI-R) and the Autism Diagnosis Observation Schedule (ADOS) as evaluation tools. There were statistically significant pooled decreases in ASD diagnoses under DSM-5 criteria [20.8% (95%CI 16-27), p<0.001] and for the DSM-IV-TR subtypes of AD [10.1% (95%CI 6.2-16.0), p<0.001] and Asperger’s [23.3% (95%CI 12.9-38.5), p=0.001]; however, the pooled decrease for PDD-NOS was not significant [46.1% (95%CI 34.6-58.0), p=0.52]. Two variables contributed to heterogeneity across ASD and subtype models: age group and type of clinician who made the diagnosis. Notably, when the diagnosis was made by a team of a physician and a psychologist, the decrease in diagnosis rates between DSM-IV-TR and DSM-5 was lowest.

Conclusions: While all previous literature reviews that examined this topic found ASD rates could decrease by at least one-third, findings from this five-year follow-up demonstrated smaller pooled decreases for ASD and all DSM-IV-TR subtypes. Nevertheless, future research is needed as concerns remain for impaired individuals without a specific diagnosis as well as to stratify for future clinical intervention studies.