Exploring the Proposed DSM-5 Criteria in a Clinical Sample
- First Online:
- Cite this article as:
- Taheri, A. & Perry, A. J Autism Dev Disord (2012) 42: 1810. doi:10.1007/s10803-012-1599-4
- 1.4k Views
The proposed DSM-5 criteria for Autism Spectrum Disorder (ASD) depart substantially from the previous DSM-IV criteria. In this file review study of 131 children aged 2–12, previously diagnosed with either Autistic Disorder or Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS), 63 % met the new DSM-5 ASD criteria, including 81 % previously diagnosed with Autistic Disorder and only 17 % of those with PDD-NOS. The proportion of children meeting DSM-5 differed by IQ grouping as well, with higher rates in lower IQ groups. Children who did meet criteria for ASD had significantly lower levels of cognitive and adaptive skills and greater autism severity but were similar in age. These findings raise concerns that the new DSM-5 criteria may miss a number of children who would currently receive a diagnosis.
The American Psychiatric Association’s (APA) Diagnostic and Statistical Manual (DSM) is commonly used to define autism, beginning with DSM-III (APA 1980), then DSM-III-R (APA 1987), then DSM-IV (APA 1994), and then DSM-IV-TR (APA 2000), with each successive edition attempting to enhance diagnostic precision based on current understandings of the disorder derived from research and clinical practice. The DSM is currently undergoing another of its periodic revisions and DSM-5 is scheduled to be published in 2013 (APA 2010). Along with other disorders, the criteria for autism will be substantially revised. The DSM-5 draft criteria have been circulated for comment and have met with some alarming media headlines and mixed reviews among established professionals (e.g., Happé 2011; Tanguay 2011; Wing et al. 2011). The present study provides some data of relevance to this debate.
In the DSM-5, it is proposed that the specific diagnoses of Autistic Disorder (AD), Asperger Disorder, Childhood Disintegrative Disorder, and Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS) be merged into one diagnosis called Autism Spectrum Disorder (ASD), and that Rett Disorder be dropped. Along with the changes made to the criteria for ASD, the DSM-5 introduces a new diagnostic category called Social Communication Disorder, which includes all three social communication deficits without the presence of any repetitive and restricted interests or behaviours.
In DSM-IV there were 12 criteria for AD, four each in the “triad of impairments”, with many possible combinations of symptoms meeting the diagnostic algorithm. In DSM-5, the criteria consist of three broad Social Communication criteria (i.e., social-emotional reciprocity, nonverbal communication, and social relationships) and four Repetitive Behaviour/interests criteria (i.e., repetitive behavior, routines and rituals, restricted interests, and sensory). Thus, there are fewer individual criteria (seven vs. 12) and a greater proportion of the individual criteria must be met in DSM-5. All three Social Communication criteria must be met (vs. at least two of four social and at least one of four communication criteria in DSM-IV) and at least two of four Repetitive Behaviour/interests criteria (vs. one of four in DSM-IV).
This restructuring is based on quite robust factor analytic findings (see Shuster et al. 2012 for a review) suggesting the social and communication items (derived from various measures) load together and are distinct from repetitive actions, repetitive speech, use of objects, and sensory difficulties (Frazier et al. 2008, 2012; Gotham et al. 2007, 2008; Mandy et al. 2012). Thus, in the DSM-5, there is no communication domain, per se, and stereotyped verbal behaviours are included as part of the restricted, repetitive behaviours. Furthermore, the repetitive behaviour domain includes unusual sensory behaviours explicitly. Finally, the specific age of onset is removed; the disorder must be present in infancy or early childhood, but may be revealed only in later years due to increasing social demands.
Although the factor analytic studies (e.g., Mandy et al. 2012) support the construct validity of the DSM-5 model, suggesting it provides a better fit than the DSM-IV triad model, at least in higher functioning individuals, important questions remain about the results of direct comparisons of the two sets of criteria in different samples and about how DSM-5 will perform in routine clinical settings, as interpreted by community clinicians. Such studies are beginning to emerge and have resulted in some controversy in the field and concern by professionals and stakeholders alike.
McPartland et al. (2012) reported on a reanalysis of DSM-IV field trial data. They rescored the items available in order to rate the DSM-5 algorithm and reported that only 60 % of children with previous diagnoses of AD, Asperger, or PDD-NOS met the DSM-5 criteria for ASD. Specificity was high but sensitivity was very poor in groups with Asperger or PDD-NOS, and in children with IQ over 70. Likewise, in a recent Australian study, 50 % of children with a DSM-IV diagnosis of PDD-NOS were found not to meet criteria for ASD according to DSM-5 (Gibbs et al. 2012). Similarly, in a study from Finland of 8-year-olds screened for autism, agreement between DSM-IV and DSM-5 criteria was lower in groups with IQ over 70 and in children with PDD-NOS and, especially, Asperger Syndrome (Mattila et al. 2011). Finally, a set of studies by Matson and colleagues on adults (Matson et al. 2012), children and adolescents (Worley and Matson 2012), and toddlers (Matson et al. in press) all suggest that the proportion of individuals meeting DSM-5 criteria is lower than for the DSM-IV criteria. Thus, there is some significant and growing concern about some unintended consequences of the new criteria.
The Neurodevelopmental Disorders Workgroup, a group of highly respected and well intentioned professionals, has raised a number of logical and methodological issues with these studies, however (Swedo et al. 2012). The use of datasets and measures designed to address criteria for DSM-IV may not include all the information needed to evaluate properly the DSM-5 criteria which are intended to be broader (so as to include milder symptoms) and studies may differ in terms of using historical versus current ratings. The Workgroup report preliminary data from the DSM-5 field trial suggesting similar rates of ASD in DSM-IV and DSM-5, but details of the method and sample are not yet published. Swedo et al. (2012) also suggest that the DSM-5 criteria should be evaluated against expert clinical diagnosis, not just against DSM-IV, and that more research is needed in community and clinical samples to address the validity and utility of the new criteria.
The purpose of this study was, therefore, to report on a retrospective file review study of the proposed DSM-5 ASD criteria in relation to established methods of diagnosis, using DSM-IV and expert clinical judgement, in a clinical sample of children who had been previously diagnosed with AD or PDD-NOS. First, the DSM-5 diagnosis algorithm (all A criteria and at least two B criteria) was applied in the total sample and subgroups of different IQ levels and compared to previous diagnosis based on the DSM-IV. Based on the intended inclusivity of the spectrum under the new proposed criteria, it was hypothesized that all children previously diagnosed with AD or PDD-NOS should meet the ASD criteria (this hypothesis was developed prior to the publication of studies raising concerns about DSM-5). Second, we examined the seven individual criteria (three Social Communication criteria and four Repetitive Behavior/interests criteria) for ASD to determine the number of children in our sample who met or did not meet each criterion and set of criteria (the three A criteria and two or more B criteria). Third, the subgroups of children who met and did not meet the proposed DSM-5 criteria for ASD were compared in terms of age, autism severity, cognitive level, and adaptive behaviour skills.
The data for this study were collected through a retrospective file review of all available files (i.e., not selected in any way) from several studies of behavioural intervention done in our lab in the past 5 years. Eligibility for the behavioural intervention program required a “diagnosis of autism or a disorder that would be considered at the severe end of the autism spectrum”, that is AD and some PDD-NOS but not Asperger. Children included some who were just beginning intervention, some in the comparison groups, and some who had been discharged for some time and were being followed up (including some who had made large gains and were functioning in the average range). Assessments were conducted by a team of graduate students and an experienced psychologist, based on clinical observations, developmental assessments and parent interviews. Files contained a demographic information form, an early history form, one of several IQ tests, the Vineland-II parent interview protocol, a DSM-IV checklist, the CARS protocol, a CARS supplementary form, and sometimes other measures. Specific information gleaned from files for the present study is described below.
Descriptive information on sample (N = 131)
M (SD) or n (%)
Previous DSM-IV diagnosis (n = 129)
93 (72.1 %)
36 (28.1 %)
Autism severity (CARS)
Cognitive level (n = 128)
Full scale IQ
59 (46.1 %)
39 (30.5 %)
30 (23.4 %)
Adaptive behavior (Vineland-II standard scores) (n = 129)
Daily living skills
One of several tests was administered to measure children’s cognitive levels depending on their age and level of functioning. These included the Mullen Scales of Early Learning (MSEL; Mullen 1995), the Stanford-Binet Intelligence Scale: 4th or 5th Editions (Roid 2003; Thorndike et al. 1986), the Wechsler Preschool and Primary Scale of Intelligence (3rd ed.; WPPSI-III; Wechsler 2002), or the Wechsler Intelligence Scale for Children (4th ed.; WISC-IV; Wechsler 1999). When children’s level of ability and age precluded standard scores being available, Ratio IQs were computed.
The Vineland Adaptive Behavior Scales-II (Vineland-II; Sparrow et al. 2005), based on parent interview, was used to assess adaptive functioning in: Communication, Daily Living Skills, and Socialization, as well as the Adaptive Behavior Composite (ABC).
Diagnostic observations were made throughout the assessment by several trained observers, using the 15 items of the Childhood Autism Rating Scales (CARS; Schopler et al. 1988), supplemented by parent report information.
Checklist used for autism spectrum disorder (DSM-5)
A. Persistent deficits in social communication and social interaction across contexts, not accounted for by general developmental delays, and manifest by all threeof the following:
1. Deficits in social-emotional reciprocity; ranging from abnormal social approach and failure of normal back and forth conversation through reduced sharing of interests, emotions, and affect and response to total lack of initiation of social interaction
2. Deficits in nonverbal communicative behaviors used for social interaction; ranging from poorly integrated verbal and nonverbal communication, through abnormalities in eye contact and body-language, or deficits in understanding and use of nonverbal communication, to total lack of facial expression or gestures
3. Deficits in developing and maintaining relationships, appropriate to developmental level (beyond those with caregivers); ranging from difficulties adjusting behavior to suit different social contexts through difficulties in sharing imaginative play and in making friends to an apparent absence of interest in people
B. Restricted, repetitive patterns of behavior, interests, or activities as manifested byat least twoof the following:
1. Stereotyped or repetitive speech, motor movements, or use of objects; (such as simple motor stereotypies, echolalia, repetitive use of objects, or idiosyncratic phrases)
2. Excessive adherence to routines, ritualized patterns of verbal or nonverbal behavior, or excessive resistance to change; (such as motoric rituals, insistence on same route or food, repetitive questioning or extreme distress at small changes)
3. Highly restricted, fixated interests that are abnormal in intensity or focus; (such as strong attachment to or preoccupation with unusual objects, excessively circumscribed or perseverative interests)
4. Hyper- or hypo-reactivity to sensory input or unusual interest in sensory aspects of environment; (such as apparent indifference to pain/heat/cold, adverse response to specific sounds or textures, excessive smelling or touching of objects, fascination with lights or spinning objects)
Specific file review notes used for DSM-5 ratings
DSM-IV A4, A3, A1
CARS I, XI, III
V-II expressive, interpersonal
A2 (nonverbal communication)
CARS XII, I
DSM-IV A2, B4
CARS I, V
V-II play & leisure
B1 (repetitive behavior)
DSM-IV B3, C3
CARS IV, V, XI
B2 (routines & rituals)
CARS VI, III
B3 (restricted interests)
DSM-IV C1, C4
CARS VII VIII, IX
Inter-rater agreement for each DSM-5 criterion (n = 30 cases)
A1. Social-emotional reciprocity
A2. Nonverbal communication
B1. Stereotyped behav/speech
B3. Restricted interests
M = 88.43
Agreement on overall diagnosis
DSM-5 Diagnosis in Relation to the DSM-IV
Percentage of sample and subgroups meeting DSM-5 ASD diagnosis
n (%) meeting ASD on DSM-5
Total sample (N = 131)
Sample broken down by DSM-IV Diagnosis (n = 130)
Autistic disorder (n = 93)
PDD-NOS (n = 36)
Sample broken down by IQ (n = 124)
IQ <40 (n = 58)
IQ 40–69 (n = 39)
IQ 70+ (n = 27)
The proportion of children meeting DSM-5 diagnostic criteria also varied by IQ group, with the vast majority (89.7 %) of the severe/profound group meeting the criteria, a lower proportion (56.4 %) in the intermediate IQ group (mild/moderate) and the lowest proportion (22.2 %) in the group with IQ of 70 and above (χ² = 39.99, p < .001). These data are also shown in Table 5.
Individual Criteria Met and Other Permutations
Percentage of children rated as meeting individual DSM-5 criteria and combinations
n (%) Met
Social communication criteria
A2 (nonverbal communication)
Meeting all 3 A criteria
Repetitive behavior/interests criteria
B1 (repetitive behavior)
B2 (routines & rituals)
B3 (restricted interests)
Meeting at least 2 B criteria
Meeting DSM-5 Algorithm (all 3 As and at least 2 Bs)
Meet 3 As and at least 1 B
Meet at least 2 As and at least 1 B
Meeting SCD definition (all 3 As; no B)
As suggested by other investigators who have experimented with “relaxing” the criteria, we examined three other permutations. Reducing the B requirement to one instead of two as in DSM-IV (but still requiring three A criteria) resulted in a slight increase to 73.3 % of children meeting criteria. Relaxing both sets of criteria, to at least two A criteria and at least one B, resulted in 83.2 % of children meeting criteria. Finally, we explored whether any of the children not meeting the ASD criteria would be likely to meet criteria for Social Communication Disorder (all three A criteria and no B criteria) but only two children showed this pattern.
DSM-5 ASD Criteria in Relation to Other Measures
Developmental and autism severity scores in two subgroups with DSM-5 criteria Met or Not Met
Met (n = 82)
Not Met (n = 49)
M (SD )
Autism severity (CARS)
Full scale IQ
Adaptive level (Vineland)
Daily living skills
We report here on a file review study of the proposed DSM-5 criteria for ASD. We had initially hypothesized that all the children in the sample, whether previously diagnosed with AD or PDD-NOS, should meet the new ASD criteria. However, only 63 % of children met the criteria for ASD, including most of the children with a previous clinical diagnosis of AD, but very few of the children with PDD-NOS. Furthermore, children with lower IQs were more likely to be rated as meeting the criteria.
We examined the seven individual criteria for ASD in the proposed DSM-5 to determine the number of children who met or did not meet each criterion. The Social Communication criteria were met by the majority but not all children. In the Restricted, Repetitive Behavior, Interests or Activities domain, two criteria were frequently met and two were not met by the majority of the present sample.
Some authors (e.g., Frazier et al. 2012; Mattila et al. 2011) have suggested or experimented with relaxing or adjusting the DSM-5 criteria. We found that relaxing either or both of the sets of criteria slightly increased the number of children who would receive a diagnosis but, even still, only 83 % would meet these looser criteria. We also attempted to determine whether the children who did not meet ASD criteria might obtain a diagnosis of Social Communication Disorder but we found that only two children in our sample would qualify, since most did have some repetitive behavior.
Children who met the DSM-5 criteria for ASD and those who did not were compared on several diagnostic and developmental variables. Children who met the DSM-5 criteria had significantly higher scores on the CARS, i.e., greater autism severity, consistent with the Matson et al. (2012) study of adults (using a different autism severity measure). Thus, the new diagnostic criteria seem to detect autism in children with more severe symptoms, but may be missing children diagnosed on the basis of milder symptom severity (even within the range included in the present study).
Furthermore, the subgroup with DSM-5 criteria met were significantly lower functioning on all cognitive and adaptive scores, suggesting the new criteria identify autism in lower functioning children, but may be missing autism symptomatology as manifest in relatively more capable children (again within the range studied which included mainly lower functioning children). These findings were consistent with those of others who have studied samples of children with different clinical profiles and used different measures and methods (Gibbs et al. 2012; Mattila et al. 2011; McPartland et al. 2012).
These results were initially somewhat surprising and concerning to us, given the apparent goal (and the clear intention of the DSM-5 Workgroup on Neurodevelopmental Disorders) of the ASD category to be more broad and inclusive of different types of children said to be “on the spectrum”. The findings are, however, quite consistent with those of other studies and imply that the proposed DSM-5 criteria, at least as operationalized in this study, could result in many children who are currently diagnosed being missed under the DSM-5.
It is important to acknowledge a number of significant limitations in the present study and to emphasize that the results may be limited to the particular sample and methodology we used. Only children with AD and PDD-NOS were examined; the study did not include other PDD categories (especially, Asperger syndrome) that are proposed to be merged, nor did we have children not on the spectrum. Therefore, we could not calculate sensitivity and specificity. Also, the sample size was modest and many of the children had quite low levels of cognitive and adaptive skills. The sample may have differed in these respects from other samples that have been examined using the DSM-5 to date and it is possible that symptom structure is different in children of different ages and IQ levels (Frazier et al. 2012; Shuster et al. 2012). Guidelines for rating each criterion for different age and ability groups could perhaps be developed and might be helpful in improving DSM-5 in practice.
In terms of methodological limitations, the DSM-5 diagnoses were based on file review using a DSM-5 checklist, not a clinical re-evaluation, and the information available to the clinician did not map perfectly onto the DSM-5 criteria (cf. Swedo et al. 2012), although there were several sources of information from clinical notes on the DSM-IV, CARS, and Vineland-II pertaining to the DSM-5 criteria. The clinician completing the checklist was highly experienced and was the same one who had diagnosed the children previously (based on direct observation and multiple measures), which minimizes any variance associated with different diagnostic practices across clinicians. Nonetheless, she was not, of course, blind to the DSM-IV diagnoses and this could have resulted in some bias. However, if anything, her bias would have been in the direction of the study’s original hypothesis that all children should fall under the ASD umbrella (recall that the ratings were done prior to publications suggesting issues with DSM-5 missing milder cases). Finally, it is unclear whether the procedure of using a checklist and evaluating each criterion individually will yield similar results to those obtained by clinicians in future, who might use the criteria more globally. The psychologist doing the ratings reported anecdotally that the four repetitive behavior criteria were straightforward to judge but the three social-communication criteria were difficult to evaluate because they were so very broad in the way they were stated. It is possible that her threshold for considering a criterion ‘met’ was different than another clinician might set. However, the inter-rater reliability for these items was very high.
It is crucial for the DSM to classify autism with good reliability and validity and it is to be expected that DSM-5 should be an improvement on DSM-IV. Failure to recognize or correctly diagnose ASD could have major implications in both research and clinical practice. Children who display symptoms but do not receive a diagnosis may not have access to appropriate resources and this will also have a major impact on primary caregivers. In addition, research conducted at different times, using different definitions of autism, may not be comparing the same types of children. Even the apparent prevalence of the disorder may be affected. It appears that Wing et al.’s (2011) question: “Better or worse than the DSM-IV?” awaits further empirical evidence before a clear answer is forthcoming.
We are most grateful to Dr. Nancy Freeman, Surrey Place Centre, Toronto, ON, who performed the ratings for diagnostic reliability.