This study conducted in Italy examined whether using the criteria for autism spectrum disorder (ASD) from the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5)1 leads to a different diagnosis than using criteria from the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision (DSM-IV-TR).2
Autism is a neurodevelopmental disorder characterized by restricted interests and repetitive behaviors or activities along with disabilities in social communication and interaction, namely stereotyped and restricted language,3 inability to understand abstract, equivocal, and picturesque language, or figures of speech like metaphor and irony.4
Once considered a rare condition, ASD has gone through a shift in terms of diagnosis, with a prominent increase over time.5 Bent et al.5 reported an increase in the incidence of autism from 2010 to 2013, which plateaued in 2015.
Further evidence in that regard comes from The Autism and Developmental Disabilities Monitoring (ADDM) Network, which provides estimates of the prevalence and characteristics of ASD among children age 8 years whose parents or guardians reside in 11 ADDM Network sites in the United States. In 2012, the combined estimated prevalence of ASD among the 11 ADDM Network sites was 14.6 per 1,000 (1 in 68) in children age 8 years.6
The fluctuation of the diagnosis of autism in these years has sometimes been attributed to the criteria used to define the main characteristics of this syndrome.5 This is because when an international psychiatric classification system such as the DSM or the International Statistical Classification of Diseases and Related Health Problems (ICD) is revised, it is followed by changes in the diagnostic criteria for a disorder (eg, autism), resulting in either an increase or a decrease in incidence of diagnosis.
DSM-I7 had 106 diagnostic categories presented in 130 pages, whereas DSM-51 has more than 300 diagnostic categories presented in 572 pages. Neurodevelopmental disorders, specifically ASD, have undergone considerable diagnostic evolution in the past decade. In the US, the current system in place is the DSM-5,1 whereas in the rest of the world, the ICD-118 serves as a general medical system. Our study is based on the application of DSM-IV and DSM-5 criteria for ASD, as the criteria for ASD in ICD-11 are different. The ICD-11 differs from the DSM-5 in several important ways. Instead of requiring a set number or combination of features for a diagnosis, it lists identifying features and lets clinicians decide whether a person's traits are a match. Because the ICD is intended for global use, it also sets broader, less culturally specific criteria than the DSM-5 does. For instance, it puts less emphasis on what games children play than whether they adhere to or impose strict rules on those games. The ICD-11 also makes a distinction between autism with and without intellectual disability, and it highlights the fact that older people and women sometimes are able to mask their autism traits. However, the instruments defined as “gold standards,” which are used both in research and in the clinic for the diagnosis of autism, are calibrated on the DSM-5 and not on the ICD-11; therefore, our study was based on the criteria in DSM-5.
However, given that the DSM-5 is intended for use in the clinical world (eg, a psychiatrist's practice in a suburban American town, or a pediatric nurse conducting a level-one screener in an urban health clinic on an 18-month-old child), the issue of its impact on eligibility remains to be seen.
The continuous progress of scientific research on autism along with the growing interest about this condition coupled with the numerous criticisms of the previous edition made the authors of the DSM-5 create a new and all- encompassing definition of ASD, as well as a specification of its severity levels. Since then, many autism specialists have been assessing the impact of redefining autism according to DSM-5 classification. Our work aims at investigating the hypothesis of a diagnosis dropout between the DSM-IV-TR and the new classification in DSM-5, which would result in social consequences for the newly excluded patients. We first provide an historical framework of the study focused on autism and how its description has changed during these years. Then, we describe the main differences between DSM-IV-TR and DSM-5 diagnostic criteria. Finally, we investigate whether applying DSM-5 diagnostic criteria resulted in a reduction of cases being diagnosed. We found that the application of DSM-5 diagnostic criteria resulted in fewer ASD diagnoses in a sample of 8-year-old children in Italy who had received a previous diagnosis of ASD with DSM-IV-TR diagnostic criteria.
History of Autism from Kanner to the Present
The original description of autism by Kanner9 in 1943 pointed out two main features: sameness and isolation. In his first publication about autism, Kanner9 described 11 cases of American children who were characterized by the absence of predisposition to sociality. Unlike the current diagnosis, Kanner9 emphasized the “resistance to change” or “insistence on sameness,” rather than the problems of language and communication, as the core features of autism.
Kanner9 called the observed syndrome “precocious infantile autism” and described it as “different, in a unique and substantial way, from all what has been discovered until today.” Kanner did not formulate a diagnosis but listed the following characteristics:
Inability to connect with people, including family members, from the first days of life
Lack of language development or, in children who speak, use of an abnormal, noncommunicative language (Kanner noted that all eight children who could speak had pronominal inversion, echolalia, obsessive questions, and ritual use of language.10
Abnormal response to the environment, including food, loud noises, and moving objects. Kanner described an obsessive need to maintain uniformity, resulting in a limitation of spontaneous activities
Good cognitive potential and excellent memory;
Good muscular coordination but sometimes awkward gait.11
At the same time, Asperger12 described four children between ages 6 and 11 years in Germany. Different from the children Kanner observed, the children described by Asperger acquired language like children without autism, but the content of their speech was repetitive, pedantic, and centered on predefined themes. Additionally, these children showed a lack of interest in other peoples' feelings, and they were not able to predict the consequences of their own actions and their social requests.12
After Kanner and Asperger, Betthelheim13 focused on parental psychopathology causing autism, as the phenomenology of autism and its characteristics had been translate to differentiate it from schizophrenia.14 Notably, Rutter's efforts played an important role for a clearer definition of autism. Before Rutter's work, autism was described in DSM-I7 and DSM-II15 as a subtype of schizophrenia, “childhood type.” Rutter's characterization of autism had a strong impact on DSM-III16 definition of “infantile autism.” The DSM-III-R17 used a polysynthetic definition grouping 16 detailed criteria into a triad of dysfunctional domains: (1) qualitative impairment in reciprocal social interaction, (2) impairment in communication, and (3) restricted interests.18
An important contribution to DSM-IV19 was made by Wing and Gould,20 in which they introduced the concept of an “autism spectrum.” In their research, the Wing and Gould20 advanced the idea that there were no clear boundaries between typical and atypical autism and that this syndrome should be conceived as a spectrum rather an univocal syndrome.
Consequently, when the DSM-IV19 was published in 1994, it had a different definition of autism compared to DSM-I8 and DSM-II.15 This definition of autism in DSM-IV was closer to the one in ICD-1021 and had a good balance of clinical and research utility.22 However, the DSM-IV lacked clarity in defining both Asperger's syndrome and Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS). Specifically, The PDD-NOS diagnosis was problematic because it lacked specific criteria, resulting in abundance of false-positive diagnoses.23–25 Additionally, the main rationale behind these changes in autism definition was the improvement of reliability.26 As a lack of consistency in the definition of Asperger's syndrome22 brought into question the validity of DSM-IV PDD subtypes, some researchers found the Asperger syndrome criteria unclear and too susceptible to the subjective evaluation of the clinician.27,28
At present, ASD has become a more complex syndromic entity in which symptoms and their severity may change from one person to another. Even if there is an increased focus on the diagnostic criteria and treatment of autism worldwide, the biggest parts of the instruments were developed in high-income countries and were not designed to consider cultural and ethnic variables that may influence ASD diagnosis.29 Even in the same country, marked disparity has been shown in diagnosis of ASD in different ethnic groups.30 Cultural factors, such as views about typical bahavior, should be taken into account when examining symptoms in children with ASD, so the scientific community needs to give increasing importance to a cross-cultural diagnosis.
Differences Between Autism Criteria in DSM-IV and DSM-5
DSM-IV diagnostic criteria for ASD require a more severe symptomatology to diagnose autism.31–33 This implies that people previously diagnosed with ASD may no longer be included in this category. In DSM-5, autism is defined as a “spectrum,” and this spectrum replaces the definition in DSM-IV of PDD. Categorical subgroups such as “autistic disorder,” “Asperger's disorder,” “PDDNOS,” and “childhood disintegrative disorder” are not included in DSM-IV, which does not take into account the specificities of each of these distinct entities (Table 1).
Different Approaches of DSM-IV and DSM-5
DSM-5 groups the first two elements of the diagnosis triad into one social and communication domain, in addition to the “Restricted and Repetitive Interests and Behaviors” (RRIB) domain. A severity scale in accordance with the “spectrum” nature of ASD was also introduced. Some specifiers were also added for a multidimensional approach to autism diagnosis. The specifiers include the presence of an etiological factor, the level of severity (from level 1 to level 3), the presence of an intellectual impairment, the presence of a language impairment, or the presence of catatonia. Furthermore, DSM-5 introduced a new disorder, social communication disorder (SCD), when only the social communication domain is impaired without RRIB (Volkmar & Reichow, 2013).34 SCD is defined by pragmatic difficulties and problems in the use of verbal and nonverbal communication in social contexts. It is presented as a distinct disorder from ASD. This new diagnosis was designed for those children who have problems with social adjustment that does not necessarily reflect the symptoms of autism.
A loss of diagnosis from DSM-IV to DSM-5 is still open for debate. Recent studies report that applying DSM-5's diagnostic criteria may result in a loss of diagnosis. Mattila et al.35 investigated ASD as defined in DSM-IV and compared it with the DSM-5 draft criteria. In this study, the initial screening involved 5,242 children, with 82 children having a diagnosis of ASD. The prevalence of children with autism was 8.4 per 1,000 according to DSM-IV-TR criteria and 65% were high functioning with an IQ greater than 70. Using the DSM-5 criteria, the loss of ASD diagnosis was greater for those with an average IQ or higher. Only 36% of participants with an IQ greater than 70 were identified as having autism using the DSM-5 criteria. These results suggested that DSM-5's diagnostic criteria might be affected by other variables such as IQ.
Other studies have shown that applying DSM-5's diagnostic criteria might result in a loss of diagnosis of ASD. For example, in a study of Matson et al.,32 the authors reported that from a sample of 330 participants 54% received a diagnosis of ASD consistent with DSM-IV diagnostic criteria. The percentage decreased to 34% when DSM-5's diagnostic criteria were used.
Further evidence in that regard comes from Ohashi.36 In this study, the author investigated 68 children between ages 6 and 15 years and found that all children without PDD were not diagnosed with ASD according to DSM-5 criteria. The author concluded that the diagnostic procedure for DSM-5 criteria is ambiguous, especially in children who are high functioning and those diagnosed at an older age.36
Wheeler et al.37 examined a sample of 758 people with Fragile X syndrome. ASD symptoms were classified according to DSM-IV and DSM-5 criteria. In this study, 63.4% of the participants met DSM-IV-TR criteria for ASD, whereas only 39% of the participants met DSM-5 criteria for ASD.
Hartley et al.38 examined data from 1,552 children with a diagnosis of ASD (from Asperger to PDD not otherwise specified), who were seen at the Women and Children's Hospital of Buffalo from 2010 to 2013. Of the 1,552 children, 807 received a diagnosis of PDD according to DSM-IV criteria, whereas 745 received an ASD diagnosis according to DSM-5 criteria.
A consistent finding in the literature is that DSM-5 has a better specificity but lower sensitivity.39 Better specificity implies that fewer patients without ASD would be falsely diagnosed as having autism. On the other hand, lower sensitivity entails the risk of underdiagnosing children with autism.
Other studies also imply that DSM-5 has a better specificity,39 which implies that fewer people would be falsely diagnosed as having autism. In their study, Harstad et al.40 tested the psychometric properties of the DSM-5 and determined how well it performed across different gender, IQ, and DSM-IV-TR subtypes using clinically collected data from 227 participants. The authors found that DSM-5 could be conceived psychometrically better compared to DSM-IV-TR.
Together, these studies provide evidence that there are still some controversies about DSM-5 criteria for the ASD, and it is unclear to what extent the application of these criteria may result in a loss of diagnosis.
Participants. In the present study, 54 children (15 girls and 39 boys; mean age of 86 months [standard deviation = 54 months]) with a previous diagnosis of ASD were recruited from a health center in Italy from October 2016 to September 2017. All children had a first diagnosis of autism according to DSM-IV classification. They attended the health center for periodic follow-up. Parents were asked to consent to repeat the diagnostic procedure and to re-read it simultaneously using the criteria of DSM-IV-TR and DSM-5.
Diagnosis was based on the DSM-IV-TR criteria, DSM-5 criteria, expert clinical judgement, and scores from the Autism Diagnostic Interview, Revised (ADI-R) and the Autism Diagnostic Observation Schedule-2 (ADOS-2). Children younger than age 2 years were excluded from the study because study materials were not suitable for children at this age. For all participants, parents signed a consent form before commencement of the study. This study was approved by the local ethics committee, in agreement with the Tenets of the Declaration of Helsinki.41
Procedure. At least 1 year after their first diagnosis of autism, children underwent a further assessment was based on both DSM-IV-TR and DSM-5 diagnostic criteria. Children were assessed using the same diagnostic criteria as during their initial evaluation, to exclude potential time effect differences between DSM-IV-TR and DSM-5. Symptoms in children with ASD vary, thus changes in diagnosis were likely to occur. Clinical diagnoses were established by using the Lord and Huerta supplementary table,42 which divides ADI43 and ADOS44 items applied to DSM-5 and DSM-IV criteria. To do that, we developed an algorithm using the KNIME software application ( www.knime.com).
The ADOS-2 is a semi-structured, standardized assessment of communication, social interaction, play, and restricted and repetitive behavior. It presents various activities that elicit behavior directly related to a diagnosis of ASD. By observing and coding behavior from 0 (absence of symptoms) to 3 (presence of strong symptom), the examiner can obtain information that informs diagnosis, treatment planning, and educational placement. The ADOS is one of the few standardized measures administered to a child by a trained professional that allows for direct observation of behaviors characteristic of ASD. Of those who use the ADOS as part of their assessments, many consider the administration and the ability to capture behaviors of ASD to be advantages of the measure. The most commonly reported disadvantages of the measure were the tendency to over-classify children who have other clinical disorders, the cost of the test kit, and the time it takes to administer the measure.45
There are no contraindications for the use of ADOS, but the nature of the ADOS requires routine use to remain efficient in administration and coding; therefore, those practitioners who have limited opportunities to assess children with ASD may not necessarily be the best candidates to learn the ADOS and be identified as an ASD expert. Our clinicians, (R.F and E.N.) like the majority of those who use ADOS, attended a clinical training on the measure.
The ADI-R is a standardized, semi-structured clinical review for caregivers of children and adults. The interview contains 93 items and focuses on behaviors in three content areas or domains: (1) quality of social interaction (eg, emotional sharing, offering and seeking comfort, social smiling and responding to other children); (2) communication and language (eg, stereotyped utterances, pronoun reversal, social use of language); and (3) repetitive, restricted, and stereotyped interests and behavior (eg, unusual preoccupations, hand and finger mannerisms, unusual sensory interests). The measure also includes other items relevant for treatment planning, such as self-injury and overactivity. Responses are scored by the clinician based on the caregiver's description of the child's behavior. As with the ADOS, a score from 0 to 3 was provided by a clinician for each item.
In the first step, the number of children who received an ASD diagnosis after the DSM-IV and DSM-5 diagnostic criteria was calculated.
Next, we counted the number of children who (1) received a similar diagnosis according to both DSM-IV-TR and DSM-5 diagnostic criteria (hereafter referred to as “Fit”); (2) received a diagnosis of ASD according to DSM-IV-TR diagnostic criteria but failed to meet the DSM-5's diagnostic criteria (hereafter referred to as “No-Fit DSM-5”); and (3) received a diagnosis of ASD according to DSM-5's diagnostic criteria but failed to meet those of DSM-IV-TR (hereafter referred to as “No-Fit DSM-IV-TR”). These data were evaluated using the chi-square goodness-of-fit test to analyze whether Fit, No-Fit DSM-5, and No-Fit DSM-IV-TR were differently distributed.
Then, the number of children whose symptoms did or did not meet ASD diagnostic criteria for both DSM-IV-TR and DSM-5 were tested using McNemar's test (a statistical test used on paired nominal data). We hypothesized that there would be a reduction of ASD diagnosis when DSM-5 diagnostic criteria were used.
Finally, we investigated if the difference in ASD diagnosis between DSM-IV-TR and DSM-5 varied over two age groups. A study by Harstad et al.40 brought some evidence forward in that regard. In this study, Harstad et al.40 reported that younger children tended to meet fewer of DSM-5's diagnostic criteria. Furthermore, these children were more likely to meet social communication and repetitive behavior criteria. In the present study, we addressed this point by dividing our sample in two groups based on the median of their age. For each group we counted the number of cases that met or did not meet the diagnostic criteria of either DSM-IV-TR or DSM-5. Then, these data were submitted to two separate chi-square tests.
Our results showed fewer diagnosed cases of ASD with DSM-5's diagnostic criteria compared to DSM-IV-TR. Although we should be cautious in interpreting these results, they were consistent with those of Harstad et al.40 In a recent study, Sung et al.46 examined different pros and cons of DSM-IV-TR and DSM-5 in regard to ASD. DSM-IV-TR criteria compel clinicians to be specific and detailed about the presentation of each case so as to correctly identify the diagnostic sub-group classification, but this simultaneously yielded more disagreements and discrepancies between the categories of sub-groups. DSM-5 criteria make the clinical decision a binary process but with a loss of details, which in turn may result in a loss of ASD diagnosis. The goal is for the DSM-5 and the ICD-11 to serve as complementary systems to improve diagnosis, prognosis, and treatment for children and adults diagnosed with neurodevelopmental disorders.
A recent systematic review by Smith et al.18 and a meta-analysis carried out by Bennett and Goodall47 examined the sensitivity and specificity of DSM-IVTR and DSM-5 criteria for ASD, and the number of people diagnosed using these criteria. These studies reported an overall reduction in the number of people eligible for a DSM-5 diagnosis of ASD as compared with those of DSMIV and DSM-IV-TR. To be diagnosed with autism based on DSM-5, people must meet two of four “restricted and repetitive interest, behavior, and activities” criteria. Previous studies showed that the main reason for the diagnostic difference was the necessity to fulfill two of four criteria in the “restricted and repetitive interest, behavior, and activities” field.31,48
Our results showed a loss of diagnosis when DSM-5's diagnostic criteria were used. Furthermore, the analysis showed this loss was higher in the younger group. Although we should be cautious in interpreting preliminary data, our results are consistent with those of Matson et al.,33 (see also Mattila et al.,35 Ohashi et al.36) in which the percentage of ASD diagnosis decreased when children were classified according to the DSM-5's criteria.
One such loss could rely on different characteristics of DSM-5. For example, Posar et al.49 showed that DSM-5 tried to unify heterogeneous conditions in ASD. These include (1) the severity and complexity of the characteristic ASD symptoms; (2) the conditions that may occur in comorbidity (primarily intellectual disability); and (3) the underlying etiologies. These definitions of ASD had an important consequence, namely that the ASD diagnosis lost its specificity and became a diagnostic category whose border with other different types of pathologies may become blurred.
Many studies have suggested that the RRBI domain may account for some of the people who meet DSM-IV-TR criteria but not DSM-5 criteria for ASD.50 Other studies5 pointed out the three impairments in the social communication area that underlie the decrease of those diagnoses based on DSM-5.
Other authors5,18,31,51 advanced the hypothesis that DSM-5's diagnostic criteria could not be sensitive to Asperger's syndrome. In fact, the incidence of Asperger's syndrome and PDD-NOS significantly decreased from 20135 after the introduction of DSM-5. Although previous research has reported lower rates of diagnostic stability for people with a diagnosis of Asperger's syndrome and PDD-NOS.18,31,51 The reduction in incidence seen in these data suggest they may reflect the removal of these diagnostic categories in DSM-5.
These studies highlight some of the limits of DSM-5, which in turn can affect the qualitative description of autism (ie, the removal of sub-diagnoses eliminates the possibility to validate them, with an important impact for researchers and clinicians).
As far as we are concerned, these results suggest that the attempt of DSM-5 to insert objective but dimensional metrics of severity has not been achieved and that the diagnosis remains categorical, even with the addition of a single, comprehensive scale of severity.
Importantly, a loss of diagnosis along with a decrease of sensitivity in DSM-5 may have dramatic consequences for children with ASD and their families. For example, children who previously received a diagnosis of ASD with DSM-IV-TR criteria could now be classified as not having ASD with DSM-5 criteria, meaning they would lose qualified support and health cares offered by the National Health Service of Italy.
Our findings, along with the evidence from other studies, suggest that the diagnosis of ASD decreased when DSM-5 criteria, rather than DSM-IV-TR diagnostic criteria, began to be used, making further studies investigating the basis for critical aspects of DSM-5 diagnostic criteria unnecessary.