The bar graph shows the proportion of children who met DSM-5 ASD criteria among Autism and Developmental Disabilities Monitoring (ADDM) Network ASD case children (N = 6577). Data are stratified by the number of DSM-IV-TR ASD criteria for surveillance years 2006 and 2008. Percentages denote those who met DSM-5 criteria among children with a given number of DSM-IV-TR criteria.
The graph shows a comparison of previously reported Autism and Developmental Disabilities Monitoring (ADDM) Network autism spectrum disorder (ASD) prevalence estimates (2000-2008 surveillance years) to prevalence using DSM-5 ASD criteria (2006-2008 surveillance years). The population size (or geographic area) for some ADDM Network sites varied from year to year. The data collection methods for the 2002 and 2004 surveillance years are not compatible with the application of DSM-5 criteria. The line with diamonds indicates previously reported ADDM Network ASD prevalence estimates based on DSM-IV-TR criteria. The line with circles shows the prevalence of children meeting DSM-5 ASD criteria. The shaded area shows the prevalence of children who are not currently meeting ADDM Network ASD case status but may meet ASD case status using DSM-5 criteria. The error bars indicate 95% CIs.
eTable 1. Proportion of Children Meeting DSM-5 ASD Criteria, Among Children Meeting Current ADDM Network (DSM-IV-TR–based) ASD Case Status, by Intellectual Disability Status and History of Developmental Regression
eTable 2. Proportion of Children Meeting DSM-5 ASD Criteria, Among Children Meeting Current ADDM Network (DSM-IV-TR–based) ASD Case Status, by Site and Year
eTable 3. Frequency of DSM-IV-TR ASD Behavioral Criteria and Other Characteristics for 3 Groups of Children
eTable 4. Frequency of DSM-5 ASD Criteria for 3 Groups of Children
Maenner MJ, Rice CE, Arneson CL, Cunniff C, Schieve LA, Carpenter LA, Van Naarden Braun K, Kirby RS, Bakian AV, Durkin MS. Potential Impact of DSM-5 Criteria on Autism Spectrum Disorder Prevalence Estimates. JAMA Psychiatry. 2014;71(3):292-300. doi:10.1001/jamapsychiatry.2013.3893
The DSM-5 contains revised diagnostic criteria for autism spectrum disorder (ASD) from the DSM-IV-TR. Potential impacts of the new criteria on ASD prevalence are unclear.
To assess potential effects of the DSM-5 ASD criteria on ASD prevalence estimation by retrospectively applying the new criteria to population-based surveillance data collected for previous ASD prevalence estimation.
Design, Setting, and Participants
Cross-sectional, population-based ASD surveillance based on clinician review of coded behaviors documented in children’s medical and educational evaluations from 14 geographically defined areas in the United States participating in the Autism and Developmental Disabilities Monitoring (ADDM) Network in 2006 and 2008. This study included 8-year-old children living in ADDM Network study areas in 2006 or 2008, including 644 883 children under surveillance, of whom 6577 met surveillance ASD case status based on the DSM-IV-TR.
Main Outcomes and Measures
Proportion of children meeting ADDM Network ASD criteria based on the DSM-IV-TR who also met DSM-5 criteria; overall prevalence of ASD using DSM-5 criteria.
Among the 6577 children classified by the ADDM Network as having ASD based on the DSM-IV-TR, 5339 (81.2%) met DSM-5 ASD criteria. This percentage was similar for boys and girls but higher for those with than without intellectual disability (86.6% and 72.5%, respectively; P < .001). A total of 304 children met DSM-5 ASD criteria but not current ADDM Network ASD case status. Based on these findings, ASD prevalence per 1000 for 2008 would have been 10.0 (95% CI, 9.6-10.3) using DSM-5 criteria compared with the reported prevalence based on DSM-IV-TR criteria of 11.3 (95% CI, 11.0-11.7).
Conclusions and Relevance
Autism spectrum disorder prevalence estimates will likely be lower under DSM-5 than under DSM-IV-TR diagnostic criteria, although this effect could be tempered by future adaptation of diagnostic practices and documentation of behaviors to fit the new criteria.
Autism spectrum disorders (ASDs) are a group of neurodevelopmental disorders characterized by impairments in social interaction and communication, as well as repetitive behaviors and restricted interests. The American Psychiatric Association first described infantile autism as a distinct condition in the DSM-III and introduced the category of pervasive developmental disorders (PDDs).1 In 1987, the American Psychiatric Association revised the diagnostic label from infantile autism to autistic disorder and slightly expanded the diagnostic criteria in the DSM-III-R.2 In 1994, the American Psychiatric Association included several subtypes in the DSM-IV: autistic disorder, Asperger disorder, Rett disorder, childhood disintegrative disorder, and pervasive developmental disorder not otherwise specified (PDD-NOS).3 Since the 1994 revision, population prevalence estimates of ASD have regularly included 3 of the 5 PDDs: autistic disorder, Asperger disorder, and PDD-NOS. The criteria described in the DSM-IV-TR published in 2000 were largely unchanged from those presented in the DSM-IV and the International Statistical Classification of Diseases, 10th Revision.4,5 Although the term ASD did not appear in the DSM-IV-TR, it is popularly used in place of PDD (inclusive of subtypes), which is how we use it here.
Population reports from developed countries show consistent, secular increases in ASD prevalence since the mid-1990s.6- 16 Throughout this article, we refer to the estimates obtained from ASD prevalence studies as ASD prevalence. The newly revised ASD criteria17 in the DSM-5 (published in 2013) presents challenges for monitoring ASD prevalence over time. It will be difficult to differentiate changes in prevalence owing to the revised diagnostic criteria from other factors such as enhanced and more widespread screening or changes in risk factors for ASD.
Some experts suggest the DSM-5 criteria require a higher threshold of symptoms.18 Previous studies based on clinical or research samples have reported that a proportion of individuals who meet DSM-IV-TR criteria for ASD fail to meet the DSM-5 criteria.19- 30 We evaluated the potential effects that the revised criteria for diagnosing ASD could have on the population prevalence of ASD among 8-year-old children in a large multisite ASD surveillance system in the United States.
We used data from the Autism and Developmental Disabilities Monitoring (ADDM) Network, an active, population-based surveillance system that implements a multisite, multiple-source, health and education record review methodology. The ADDM Network monitored ASD prevalence among 644 883 8-year-old children who resided in 11 US surveillance sites during the 2006 surveillance year and 14 US sites during 2008.9,10 Data in the ADDM Network are collected every other year. The 14 sites were located in Alabama, Arizona, Arkansas, Colorado, Florida, Georgia, Maryland, Missouri, New Jersey, North Carolina, Pennsylvania, South Carolina, Utah, and Wisconsin (Arkansas, New Jersey, and Utah participated in 2008 only). Sites were funded by the Centers for Disease Control and Prevention through a competitive review process. Although each site represented a local population, the overall sample was not nationally representative. Sites met applicable local institutional review board and privacy/confidentiality requirements under 45 CFR 46; the ADDM Network is a public health surveillance system, which does not require patient consent for record review.
Details of the ADDM Network have been described previously.6- 10 In brief, within each site, records of 8-year-old children were screened at multiple health facilities serving children with developmental disabilities (all 14 sites) and public school special education programs (9 sites). Records for children with a variety of developmental conditions were requested. For children meeting established birth year and residency requirements, records with certain behavioral or diagnostic descriptions contained in developmental evaluations were abstracted. Abstracted information included verbatim behavioral descriptions, psychometric testing and screening results, developmental history or concerns, and developmental diagnoses and classifications. Multiple records from different facilities for the same child were combined into a composite record. Trained clinicians reviewed the information for each child and used a specified protocol to determine (DSM-IV-TR–based) ASD case status.
The DSM-IV-TR (see Supplement) included 12 diagnostic criteria for PDD divided into 3 domains of impairment: (1) social interaction; (2) communication; and (3) repetitive behavior or restricted interest. For a diagnosis of autistic disorder, an individual met at least 6 criteria altogether including 2 criteria in the social domain and 1 each in the communication and repetitive/restricted behaviors and interests domains. The onset of impairment must have occurred before age 3 years, and the condition could not be better accounted for by Rett disorder or childhood disintegrative disorder. Diagnoses of PDD-NOS and Asperger disorder required a fewer number or different pattern of criteria than required for autistic disorder.
The ADDM Network ASD case definition was based on the DSM-IV-TR diagnostic criteria for autistic disorder, Asperger disorder, and PDD-NOS in 2006 and 2008. However, the ADDM Network case definitions for Asperger disorder and PDD-NOS were more stringent than the DSM-IV-TR, requiring an autism discriminator in addition to the DSM-IV-TR criteria. Autism discriminators included specific behaviors considered to be common among children with ASD than without or a documented ASD diagnosis.31 This approach is consistent with the notion that using the minimum diagnostic criteria for PDD-NOS leads to more diagnoses than clinically appropriate.32,33 Even when the DSM-IV-TR behavioral criteria were technically met, ADDM Network clinician reviewers could rule out ASD case status if the impairments were likely attributable to another disorder or were otherwise contraindicated. If the clinician was uncertain whether ASD was an appropriate classification, a second clinician independently reviewed the record and the clinicians reached consensus on final case status. In addition to determining case status, ADDM Network clinician reviewers recorded previous diagnoses, history of regression, and other behavioral features consistent with ASD (eg, odd responses to sensory stimuli). Race/ethnicity information was collected from health, educational, and birth certificate records when available and was missing or unknown for 3.6% of the children meeting ADDM Network ASD case status.
The DSM-5 criteria for ASD differ from the DSM-IV-TR criteria in several respects. First, they do not distinguish subtypes of ASD, such as autistic disorder and Asperger disorder, instead classifying a single category of ASD. Second, the DSM-5 recognizes only 2 domains of impairment: social communication and restricted, repetitive patterns of behavior, interests, or activities, and all 3 items in the social communication domain are required. Third, in contrast to the 12 distinct diagnostic criteria of the DSM-IV-TR, the DSM-5 specifies 7 criteria but some of the DSM-5 criteria describe more general principles and behaviors than in the DSM-IV-TR.34 Fourth, the DSM-5 ASD criteria allow for the consideration of historical behaviors in addition to current behaviors, instead of primarily focusing on current behavior as with the DSM-IV-TR ASD criteria.
We operationalized the DSM-5 ASD behavioral criteria using behaviors corresponding to DSM-IV-TR ASD criteria and other behaviors abstracted from a child’s records. Some DSM-5 ASD criteria corresponded directly to 1, or a combination of, specific DSM-IV-TR criteria, making application of DSM-5 criteria to the data recorded by the ADDM Network clinician reviewers straightforward. Other DSM-5 criteria were not directly comparable with DSM-IV-TR criteria but corresponded to other behavioral features already recorded as part of the ADDM Network protocol such as unusual sensory interests. For the purposes of this study, behaviors corresponding to some distinct DSM-IV-TR criteria were allowed to contribute toward meeting more than 1 DSM-5 ASD criteria (Table 1). This operationalization is similar to the coding scheme used in clinical studies of the proposed DSM-5 ASD criteria.20,21 Because the ADDM Network relies on documented descriptions of behaviors through age 8 years, the surveillance system method is consistent with the DSM-5’s inclusion of historical behaviors.
For the 2006 and 2008 surveillance years, ADDM Network clinician reviewers determined 6577 children met ASD criteria based on the DSM-IV-TR. The reviewers identified an additional 1020 children whose behaviors could have met ASD criteria but, after consensus, they were judged to not meet ADDM Network ASD case criteria. We calculated the proportions of these children who met DSM-5 ASD criteria overall and stratified by sex, race/ethnicity, intellectual disability, history of developmental regression, previous ASD classification by a community professional (ASD diagnosis or autism classification in special education), ADDM Network site, and number of DSM-IV-TR ASD criteria documented in evaluations.
To calculate the potential impact on prevalence, we applied DSM-5 ASD criteria to 2 groups of 8-year-old children under surveillance for the years 2006 and 2008: (1) the 6577 children who met both ADDM Network ASD criteria based on the DSM-IV-TR and our operationalized DSM-5 ASD criteria and (2) the 1020 children who did not meet ADDM Network ASD criteria but could plausibly meet DSM-5 criteria. These 1020 children all technically met DSM-IV-TR criteria for PDD-NOS, but the clinician reviewers did not classify them as ASD cases for surveillance purposes; for most of these children, the clinician reviewers concluded that the behaviors were better accounted for by another disorder. For comparability to previously published prevalence estimates, we used the same denominators that were described in the ADDM Network prevalence reports in 2006 and 2008.9,10
We performed χ2 tests to assess differences in proportions, and we calculated 95% CIs using the binomial exact method. All analyses were performed with the R Statistical Computing Package version 2.15.3 (R Foundation for Statistical Computing). Plots were created with ggplot2.35
Among the 6577 children who met the ADDM Network ASD case definition based on the DSM-IV-TR in surveillance years 2006 and 2008, 5339 (81.2%) met the DSM-5 criteria for ASD (Table 1). Of the 3 (required) criteria in the DSM-5 ASD social communication domain, deficits in nonverbal communication was the least frequent, with 86.8% of the 6577 children meeting this criterion. Restricted interests (in the repetitive/restricted behavior domain) was the least frequent overall at 62.8% (Table 1). Nearly all children (96.1%) who met ADDM Network ASD case definition either met, or were within 1 criterion of meeting, DSM-5 ASD criteria (Table 1).
The proportion of children who met DSM-5 ASD criteria among those who met ADDM Network ASD criteria based on the DSM-IV-TR increased from 78.5% in 2006 to 83.1% in 2008 (P < .001; Table 2). Overall, boys and girls were nearly equally likely to meet DSM-5 ASD criteria (81.4% vs 80.0%, P = .28); similar proportions of non-Hispanic black and non-Hispanic white children met DSM-5 ASD criteria (82.2% vs 81.6%, P = .73). Asian children were more likely to meet DSM-5 ASD criteria than Hispanic children (88.0% vs 78.1%, P < .001). Children with a history of developmental regression were more likely to meet DSM-5 ASD criteria than those without a history (89.4% vs 79.0%, P < .001), and children with intellectual disability were more likely to meet DSM-5 ASD criteria than children with an IQ greater than 70 (86.6% vs 82.5%, P < .001). Children with a history of regression remained more likely to meet DSM-5 ASD criteria than children without a history of regression after controlling for intellectual disability (eTable 1 in Supplement). Children identified as having ASD by a community professional (including special education classification of autism) were more likely to meet DSM-5 ASD criteria than those who were not (84.8% vs 69.7%, P < .001). There was variability by ADDM Network site, ranging from 95.6% of children in Utah to 68.8% of children in Florida meeting DSM-5 criteria (P < .001). Some of these pairwise comparisons are not shown in Table 2, which presents overall χ2 statistics for all levels of a factor. In Supplement, eTable 2 shows the proportions of children meeting DSM-5 ASD criteria by site for 2006 and 2008 separately.
We observed a substantial association between the proportion meeting DSM-5 ASD criteria and the number of documented DSM-IV-TR behavioral criteria (Figure 1). None of the 38 children with only 2 DSM-IV-TR behavioral criteria documented met DSM-5 ASD criteria. In contrast, all 899 children with evidence of all 12 DSM-IV-TR criteria met DSM-5 ASD criteria (Figure 1). In Supplement, eTable 3 and eTable 4 show the frequency of specific criteria among children meeting ADDM Network ASD case status only, DSM-5 ASD only, and both sets of criteria.
When ASD prevalence estimates are adjusted to include only children meeting DSM-5 ASD criteria, the prevalence of ASD is lower than previously reported (DSM-IV-TR) ASD prevalence for both 2006 and 2008 (Figure 2). The adjusted prevalence estimates included 5339 (of the 6577) children meeting both DSM-IV-TR ADDM Network criteria for ASD and DSM-5 ASD criteria and 304 (of the 1020) children who did not meet ADDM Network ASD case status after clinician review but did have documented behaviors consistent with DSM-5 ASD criteria. For 2006, the prevalence estimate using DSM-5 criteria was 7.4 per 1000 (95% CI, 7.1-7.7) compared with 9.0 (95% CI, 8.6-9.3) based on the ADDM Network application of the DSM-IV-TR. For 2008, the DSM-5–based estimate was 10.0 per 1000 (95% CI, 9.6-10.3) compared with 11.3 (95% CI, 11.0-11.7) based on the DSM-IV-TR (Figure 2).
These results suggest that fewer children would have been classified as having ASD using the DSM-5–based criteria than the DSM-IV-TR–based ADDM Network surveillance approach. The net reduction on prevalence estimates for surveillance years 2006 and 2008 would have been approximately 18% and 12%, respectively. While a number of recent clinic-based studies examined the percentage of patients in specific clinics or research samples meeting DSM-IV-TR criteria for autism (or PDD) who also meet DSM-5 ASD criteria, this population-based study evaluated the potential impact of the DSM-5 criteria on the population prevalence of ASD. The percentages found in clinic-based studies,19- 30 summarized in Table 3, ranged from 46% to 93%. The analogous percentage from our study was 81%.
The potential reduction in ASD prevalence under the DSM-5 criteria was smaller in 2008 than 2006 both in terms of absolute prevalence (a reduction of 1.3 per 1000 in 2008 compared with 1.6 per 1000 in 2006) and the proportion of ADDM Network ASD cases who did not meet DSM-5 criteria (16.9% in 2008 compared with 21.5% in 2006). This suggests that the adoption of the DSM-5 criteria is unlikely to reverse the trend of increasing ASD prevalence over time (Figure 2). Continually increasing awareness of ASD—leading to more detailed descriptions of behaviors in evaluations—could potentially explain why the ADDM Network ASD case children in 2008 were more likely to meet DSM-5 criteria than in 2006. These same factors could also contribute to the considerable variability across sites in ASD prevalence and the proportion of children meeting DSM-5 ASD criteria. A previous evaluation of the ADDM Network methods cited these factors among the most difficult to assess.36
An advantage of the population-based design was that it is representative of all children in defined populations who meet ASD criteria and are evaluated in typical community settings rather than selected samples attending a particular clinic or enrolled in specific research projects. Another important strength of this study was that we considered children who may meet DSM-5 ASD criteria but did not meet ADDM Network DSM-IV-TR–based criteria. The inclusion of this group allowed us to estimate the potential net impact of the new criteria on prevalence accounting for cases lost and gained. However, it is likely that many of the 304 ASD cases gained could be classified as having another disorder and not ASD, which is how most were classified using the current ADDM Network methods.
Previous introductions of new criteria for ASD suggest that the process by which professionals become trained in and familiar with the new diagnostic or eligibility criteria is gradual.37- 40 With the advent of the DSM-5, it is likely that policies for service eligibility and reimbursement—as well as diagnostic tools—will be adapted in response to the revised criteria. It is also possible that as clinicians become cognizant of the revised criteria, they will assess and document additional symptoms to substantiate an ASD diagnosis. If community professionals will be more inclined to document symptoms that receive greater emphasis in the DSM-5, the differences between ASD prevalence based on the DSM-IV-TR and DSM-5 could decline over time.
Children identified as having ASD by a community professional were more likely to meet DSM-5 ASD criteria than those who were not (84.8% vs 69.7%). Furthermore, most of the ADDM Network ASD case children who did not meet the DSM-5 ASD criteria were only lacking by 1 criterion. Similar to the findings of 3 other studies,23,26,41 relaxing the DSM-5 criteria to require 1 fewer (in either domain) included almost all children (96.1%) meeting the current ADDM Network (DSM-IV-TR) ASD case definition. This suggests that the DSM-5 ASD criteria could have a smaller effect on eligibility for services than their effect on prevalence estimates (which include children never diagnosed as having ASD).
A limitation of this study was its reliance on symptoms documented in records by professionals in the community during a time when the DSM-IV-TR held sway over the ASD diagnostic process; the children were born in 1998 or 2000. Because most DSM-5 ASD criteria refer to behavioral features documented in developmental assessments performed to evaluate DSM-IV-TR criteria and the ADDM Network included documentation of other clinical features specified under DSM-5 ASD criteria (eg, unusual responses to sensory inputs), the retrospective analysis presented in this study was possible and provides a reasonable estimate of the anticipated impact of the DSM-5 criteria on ASD prevalence.
Some children meeting PDD criteria under the DSM-IV-TR will not meet DSM-5 ASD criteria but might meet the DSM-5 criteria for social communication disorder.34,42 We were unable to assess social communication disorder prevalence because the criteria for it did not readily correspond to existing measures in the ADDM Network data.
The results of this population-based study suggest ASD prevalence estimates may be lower under the DSM-5 ASD criteria than under the criteria based on the DSM-IV-TR. Autism spectrum disorder prevalence estimates for 2014 and beyond should be interpreted in context. Future changes in evaluation and reporting practices, as well as refinements to standardized diagnostic instruments, will also affect future trends in ASD prevalence estimation and may run counter to the potential effects of the DSM-5 criteria suggested by this study.
Corresponding Author: Matthew J. Maenner, PhD, National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, 1600 Clifton Rd, MS E-86, Atlanta, GA 30333 (firstname.lastname@example.org).
Submitted for Publication: April 7, 2013; final revision received September 20, 2013; accepted September 20, 2013.
Published Online: January 22, 2014. doi:10.1001/jamapsychiatry.2013.3893.
Author Contributions: Dr Maenner had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Maenner, Rice, Schieve, Carpenter, Kirby, Durkin.
Acquisition of data: Maenner, Rice, Arneson, Cunniff, Schieve, Carpenter, Van Naarden Braun, Bakian, Durkin.
Analysis and interpretation of data: Maenner, Rice, Cunniff, Schieve, Van Naarden Braun, Kirby, Durkin.
Drafting of the manuscript: Maenner, Rice, Schieve, Durkin.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Maenner, Schieve, Kirby.
Obtained funding: Maenner, Cunniff, Carpenter, Bakian, Durkin.
Administrative, technical, and material support: Rice, Arneson, Carpenter, Van Naarden Braun, Bakian, Durkin.
Study supervision: Rice, Cunniff, Durkin.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by a grant from the Autism Science Foundation and by the Centers for Disease Control and Prevention Cooperative Agreements UR3/CCU523235, UR3/DD000078, and UR3/DD000677. Additional support was from National Institute of Child Health and Human Development grant T32 HD007489-04 and the Waisman Center (grant P30 HD03352).
Role of the Sponsor: The Centers for Disease Control and Prevention funds and administers the Autism and Developmental Disabilities Monitoring Network and co-authors working at the Centers for Disease Control and Prevention made contributions to this article. The other funders had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The findings and conclusions in this article are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Additional Contributions: We are grateful to the many staff, scientists, and clinicians who have contributed to the Autism and Developmental Disabilities Monitoring Network project.
Correction: This article was corrected online February 5, 2014, for a typographical error in Table 2.