The microsegmentation of the autism spectrum: research project

9 Segmenting The Autism Spectrum

9.1 The designation of this research as ‘the Microsegmentation Project’ reflected a fundamental ambition of the study which went beyond the question of providing a foundation for an analysis of the economic consequences of autism in Scotland, namely, to find a meaningful way in which to segment the autism spectrum itself.

9.2 The need for segmentation may be stated clearly in terms of two considerations. First, we cannot plan for research, services or interventions in autism if we treat the whole spectrum as one entity. The cluster of features simultaneously identified as ‘autism’ in the work conducted by Leo Kanner and Hans Asperger in the late 1930s and early 1940s pointed to a group which, on the one hand, had what were viewed as unique similarities in clinical presentation but, on the other hand, nevertheless showed considerable variation (Asperger, 1944/1991; Kanner, 1943). In terms of similarities, Kanner wrote: ‘Since 1938, there have come to our attention a number of children whose condition differs so markedly and uniquely from anything reported so far, that each case merits – and I hope will eventually receive – a detailed consideration of its fascinating peculiarities’ (p.217). Asperger wrote: ‘In what follows, I will describe a particularly interesting and highly recognisable type of child....I have chosen the label “autism” in an effort to define the basic disorder that generates the abnormal personality structure of the children we are concerned with here’ (p.37).

9.3 In terms of variation, in speaking of their ‘essential common characteristics’ (p.242), Kanner stated that the children showed ‘individual differences in the degree of their disturbance, the manifestation of specific features, the family constellation, and the step-by-step development in the course of years’ (pp.241-242). The extent of that variation was demonstrated in sharp relief when he published his follow-up study of the original children 28 years later. Outcomes ranged from largely independent living with full, regular employment to being in institutional care or experiencing early death from epileptic seizures. Asperger, unlike Kanner, identified a feature which was to become of key importance in later research, namely, variation in intellectual status. Although paradoxically (owing to the profile of the children featured in his published case studies) he became the originator of a syndrome defined as being marked by ‘no general delay or retardation in… cognitive development’ (World Health Organization, 1992, p.258), Asperger stated: ‘We have mentioned repeatedly that autism occurs at different levels of ability. The range encompasses all levels of ability from the highly original genius…to the most severe contact-disturbed, automaton-like mentally retarded individual’ (Asperger, 1944/1991, p.74).

9.4 It is recognition of these similarities and differences that has been central to the whole progress of autism research. An abiding commitment to the view that the similarities have clinical validity is the basis on which autism, despite many major reformulations, has endured as a robust clinical syndrome. Equally, it is recognition of the differences that has led to autism becoming recognised over a very long period as a ‘spectrum’ disorder and to the quest for identifying meaningful diagnostic subgroups.

9.5 The second reason for the need for segmentation is the converse of the first, and reflects the quest for meaningful subgroups. While we cannot plan research, services or interventions by viewing autism as one entity it is equally clear that we cannot do so on the basis of treating everyone on the spectrum as being unique. The concept of recognising every person’s unique individuality does not over-ride the need for, and recognition of, meaningful homogeneity in clinical presentation. Identifying the key homogeneous features is a prerequisite for planning research samples, for setting up specialist provision, for providing targeted interventions and for predicting the parameters of future life trajectories.

9.6 Despite the clear need for segmentation it must be recognised that few studies have allowed any form of functional evaluation of the impact of differing presentations of autism either in practical terms for individuals and their families or in economic terms in relation to the national costs of making provision. While there is extensive research relevant to this subject it is not possible to construct any meaningful segmentation framework from the world literature.

9.7 In the attempt to navigate a course between a ‘one size fits all’ approach and establishing some meaningful groupings to allow resource and budget planning, designing service provision, implementing interventions or setting research priorities, various practical schemes have been used. One such approach (commonly used to assist in designing service packages by Scottish Autism) is as follows:

• Group requiring lifelong 24 hour care and support
• Those with autism, or autism plus intellectual disability, or Asperger’s Syndrome needing substantial daily care and support
• Those with autism, autism plus intellectual disability, without serious challenging behaviour, requiring moderate support
• Asperger’s Syndrome, with a measure of independence and structured regular support on a weekly basis
• Asperger’s Syndrome with minimal support requirements
• Asperger’s Syndrome plus challenging, violent or offending behaviour.

9.8 A segmentation of this kind has utility, but it serves to highlight both the strengths and the weaknesses of what can currently be learnt from the literature. As to strengths, it clearly draws from an evidence-based understanding of the place of intellectual disability (‘autism plus intellectual disability’), of assigned diagnosis (‘autism’ or ‘Asperger’s Syndrome’), of having lower symptom severity (‘a measure of independence’) and of having co-occurring conditions or associated features (‘challenging, violent or offending behaviour’). As to weaknesses, it highlights the fact that these features cannot provide a conceptual map of the autism spectrum as they could not be represented either as a continuum or in terms of any coherent overall model. In particular, the first group (those ‘requiring lifelong 24-hour care and support’) are defined only in terms of their service package, but not in terms of any other features, while those is the last group (‘Asperger’s Syndrome plus challenging, violent or offending behaviour’) show a discontinuity in terms of any gradation of need, as their needs, while being greater because of their co-occurring conditions or additional features, are likely to vary significantly within that single group.

9.9 Attempts to formulate segments at a conceptual level within the spectrum have followed three main lines of enquiry. The first relates to diagnostic subgroups. Do the separate subgroups, as described in terms of the main historical classifications of childhood autism, Asperger’s Syndrome and atypical autism, together with other proposed variants, offer a meaningful basis for segmentation in ways that would inform service needs, economic impact or prediction of outcomes? The second relates to identifying different ASD profiles, with or without these mapping on to specific diagnostic subgroups. These have focussed mainly on such features as intellectual functioning, verbal language usage or behavioural presentation. The third relates to co-occurring conditions. Does the presence of additional conditions such as ADHD or mental health difficulties provide a consistent basis for segmentation? All of these lines of enquiry have been helpful and each has made some relevant contribution towards meaningful segmentation. However, none has offered a sufficient evidence base to allow any form of robust framework to be constructed.

9.10 In relation to diagnostic subgroups, the two international classification systems, while not quite merging, became very closely aligned both in terms of what the subgroups are and how they should be operationally defined from the time of ICD-10 (World Health Organization, 1992, 1993) and DSM-IV (American Psychiatric Association, 1994) until the publication of DSM 5 (American Psychiatric Association, 2013). For both classifications the two key categories were autism ( ICD ‘childhood autism’, DSM ‘autistic disorder’) and Asperger’s Syndrome ( DSM ‘Asperger’s disorder’). In addition there was the ICD subgroup ‘atypical autism’, corresponding most nearly to DSM ‘pervasive developmental disorder not otherwise specified’ ( PDD-NOS). However, any perusal of the wording of these classifications will indicate why it was unlikely that they could have any real utility in relation to segmentation. Basically, atypical autism covered almost everything that might look like autism but did not meet one or more of the key criteria, whether in terms of age of onset or of symptomatology or of both of these. This led to further sub-classifications of atypical autism to cover all the main possibilities. The matter was confused further in ICD by the presence of an additional ‘catch-all’ classification of ‘pervasive developmental disorder, unspecified’, to cover anything that seemed to be pervasive developmental disorder but could not be fitted into the diverse range of options already available.

9.11 The three diagnostic segments of childhood autism, Asperger’s Syndrome and atypical autism, or their DSM equivalents, became the basis on which the autism spectrum was defined. Thus, within a Scottish context, the Public Health Institute of Scotland’s Needs Assessment Report on ASD (the PHIS Report, Public Health Institute of Scotland, 2001) defined the spectrum on this basis, and the SIGN guideline on ASD likewise stated, ‘The term autism spectrum disorders has been used throughout this guideline to cover conditions termed autism, atypical autism and Asperger’s syndrome’ (Scottish Intercollegiate Guidelines Network, 2007, p.3).

9.12 However, diagnostic subgroups, and the various attempts to reformulate these or add further variants, have not only been unable to support a useful segmentation framework but have also failed in themselves to have an enduring basis in terms of their clinical validity. This may be most clearly illustrated in relation to the subgroups Asperger’s Syndrome and atypical autism. Despite the general popularity of Asperger’s Syndrome as a classification and a vast literature specific to it, its status as a diagnostic category was viewed from the beginning as tentative. ICD-10 noted that it was ‘of uncertain nosological validity’ (World Health Organization, 1992, p.258) while DSM-IV stated that the diagnostic validity of the disorder was unknown (American Psychiatric Association, 1994). This remained the case, despite efforts to distinguish Asperger’s Syndrome from ‘high functioning autism’ (Chiang, Tsai, Cheung, Brown, & Li, 2014; Cuccaro et al., 2007; Macintosh & Dissanayake, 2006; Mukaddes, Herguner, & Tanidir, 2010; Nayate et al., 2012; Thede & Coolidge, 2007), and latterly there was not an evidence base to support its continued recognition as a separate diagnostic entity in DSM 5 or in the Beta Draft of ICD-11.

9.13 The diagnoses of ‘atypical autism’ and ‘pervasive developmental disorders – not otherwise specified’ ( PDD-NOS) have been defined as ‘a large depository for complex or atypical cases, tremendously heterogeneous and poorly defined… a kind of terra incognita’ (Klin, Volkmar & Sparrow, 2000, p.7). The ICD-10 definition of atypical autism, as something which is subthreshold in age of onset, in symptomatology or in both, is essentially a negative definition – ‘not, or not quite, autism’ (Klin et al., 2000, p.331). While this would suggest a less severe or less full manifestation of ASD symptoms than in autism (and indeed this is how it is often used in diagnostic practice), ICD-10 in fact intends the opposite, noting that is ‘arises most often in profoundly retarded individuals whose very low level of functioning provides little scope for exhibition of the specific deviant behaviours required for the diagnosis of autism’ (World Health Organization, 1992, p.255). In short, these diagnoses have not proved to have clinical consistency or utility.

9.14 There have been many other attempts to propose meaningful subgroups within the autism spectrum, some of which have generated interest at times outwith mainstream research and practice. For example, currently the concept of ‘pathological demand avoidance syndrome’ ( PDA) is frequently encountered (Newson, Le Maréchal, & David, 2003) but, in common with other proposals for new diagnoses based on particular features often encountered in autism, it has not met criteria for clinical validity for acceptance in either DSM 5 or the forthcoming ICD 11.

9.15 In relation to identifying different ASD profiles, with or without these mapping on to specific diagnostic subgroups, research in this area has made a significant but limited contribution to segmentation. Three areas of differing profile have proved to be robust in their importance as predictors of later outcome. These have been covered in paras. 3.21-3.28 under the headings of intellectual ability, language and symptom severity. However, all show limitations in their capacity to offer a consistent basis for segmentation.

9.16 With regard to intellectual ability, its contribution to segmenting the ASD population is discussed in paras. 3.21- 3.23 and 5.1 to 5.7 in terms of the principal determinant of differential outcomes, namely, the presence or absence of intellectual disability. Those who match the cognitive profile of moderate and severe intellectual disability, that is, those in the IQ ranges below 50, have the poorest outcomes and the highest needs for service provision. Those who match the profile of mild intellectual disability, that is, the IQ range 50-70, have better outcomes and a lower tariff of needs, but these needs are markedly greater than those without intellectual disability, that is, the IQ range 70+, the latter group including those with the highest levels of independent living, employment and long-term relationships and the lowest level of service needs.

9.17 With regard to language the position is less straightforward. This is because of the extent to which language is a proxy for intellectual status and in turn for assigning diagnostic subgroup, as covered in detail in paras. 3.24- 3.27. As to its relation to intellectual status, linguistic function has always been a core part of intellectual assessment. The most established approaches to assessing intellectual level have language as one of their major domains. The Wechsler-type tests traditionally generated a ‘verbal’ and a ‘performance’ IQ, and although that foundation has now been broadened to include working memory and processing speed domains, the verbal comprehension domain remains central to the definition of IQ. Similarly the Raven-type tests comprise matrices, which relate mainly to largely non-verbal concepts, and vocabulary-based tests designed to assess the ability to recall and use a culture’s store of explicit verbalised concepts. Status in terms of language development therefore cannot be seen as a factor independent of intellectual status, although the overlap is not complete, as shown in the study by Howlin et al. (2004) in which language differentially predicted outcome in children who all had IQ70+ ( para. 3.27).

9.18 As to the relation of language to diagnostic subgroup, there are specific language and communication criteria for childhood autism but not for Asperger’s Syndrome. The first of these is that there may be a delay in or total lack of development of spoken language that is not accompanied by an attempt to compensate through the use of gesture or mime as alternative modes of communication, often preceded by a lack of communicative babbling. The only linguistic criterion relevant to Asperger’s Syndrome, other than in general a weak integration of social, emotional and communicative behaviours, is a criterion of exclusion; that is, there must be no clinically significant general delay in spoken or receptive language. Thus, language cannot be seen as a factor independent of diagnostic subgroup.

9.19 With regard to symptom severity, we have discussed in paras. 3.26 and 3.28 its relation to intellectual and language function, and have anticipated in para. 3.26 our view that the Asperger diagnosis, in its distinction from the autism diagnosis, is comprehended, in terms of diagnostic criteria and practice, within the interplay of these three factors relating to IQ, language and severity of symptoms.

9.20 In relation to diagnostic criteria, it is the first two of these factors, intellectual status and linguistic functioning, that are specified most precisely as diagnostic requirements for Asperger’s Syndrome. In diagnostic practice, it can also be demonstrated that lower symptom severity is a significant factor in assigning Asperger’s Syndrome as opposed to childhood autism.

9.21 The diagnostic requirements are stated both in ICD-10 and DSM-IV. Both state that there is no clinically significant general delay in spoken or receptive language or cognitive function. Diagnosis requires that single words should have developed by two years of age or earlier and that communicative phrases are used by three years of age or earlier. In addition, the criteria require that symptoms which may be seen as reflecting normal intellectual development are present. These are self-help skills, adaptive behaviour and curiosity about the environment during the first three years at a level consistent with normal cognitive function. Thus, there is a degree to which symptom severity, in addition to language function itself, serves as a proxy for intellectual status (see paras. 3.26 and 3.28 ).

9.22 Symptom severity also makes an independent contribution to outcome variance ( para. 3.28), and in doing so it is an important factor in diagnostic practice in determining whether it is the Asperger diagnosis rather than the autism diagnosis that is assigned. Owing to the lack of a clinically valid basis for differentiating Asperger’s Syndrome and high functioning autism in terms of clinical trials, for the purposes of the Scottish Autism Survey dataset these two categories were grouped, since research has shown that their similarities are greater than their differences (Macintosh & Dissanayake, 2004). Thus, the best fit for overall analysis arose from combining these categories.

9.23 However, in practical terms, the literature indicates that for those for whom clinicians have specifically assigned an Asperger diagnosis in preference to an autism diagnosis, even where there is no intellectual disability, this is done on the basis of increased symptom severity. This may be demonstrated by considering studies which have examined ASD groups matched for intellectual ability but differing in the diagnosis clinicians had assigned to them. Szatmari, Bartolucci and Bremner (1989) compared early history and outcome of 28 individuals with Asperger’s Syndrome and 25 with high functioning autism, matched by full-scale IQ. On the basis of parent information about impairments in socialisation, communication and imagination, high functioning autism was distinguished from Asperger’s Syndrome in terms of symptom severity. Prior et al. (1998) used a sample of 135 participants diagnosed with high-functioning autism, Asperger’s disorder, or PDD-not otherwise specified (without intellectual disability). Again, group differences were attributable to variations in severity of symptoms, with Asperger’s disorder less severe. Ozonoff, Rogers and Pennington (1991) noted other indicators of less severe symptomatology in Asperger’s Syndrome, specifically better verbal memory and theory of mind. In addition, a principal reason for the later average age of diagnosis in Asperger’s Syndrome is that symptoms are generally less severe and more subtle than in autism (Howlin & Asgharian, 1999).

9.24 In summary, it may be asserted in terms of the ASD diagnostic categories that, as noted by Macintosh and Dissanayake (2004), ‘a relatively consistent finding has been that differences between groups are largely interpretable as a function of symptom severity, intellectual ability and level of adaptive functioning’ (p.422). This is of importance in relation to microsegmentation of the autism spectrum in terms of interpreting the outcomes literature.

9.25 In relation to co-occurring conditions, it is recognised that the presence of these is important in terms of what it may imply for service provision and economic impact. However, both the literature and the data generated for this research show that, unlike dimensions of intellectual status and symptom severity, their impact cannot be graded in any way that has stability and utility. We can predict service needs within a broad gradation that relates to intellectual status (from high functioning to moderate and severe intellectual disability) and to symptom severity (independently of its status as a proxy for intellectual ability, but in terms of the severity of autistic symptomatology in areas such as the early impact of the autism triad, or ongoing impairments in socialisation). We cannot use co-occurring conditions or associated features as a stable indicator of service needs or economic impact.

9.26 The dataset generated for this research, both in terms of what may be discerned from the descriptive statistics themselves, from the regression analyses carried out on the data and from all of the evidence presented in relation to economic impact (Chapters 7 and 8) supports these assertions. Intellectual disability, as already established in the world literature, confirmed its importance as a stable predictor of cost. Additional co-occurring conditions led also to increased cost. For example, as an overall group those with ADHD were more likely to make use of social care services, and those with the autism diagnosis and ADHD received additional services funded by education. Those with mood disorders made use of additional health services. Similarly, those with OCD/Tourette’s incurred additional health service costs. However, the impact of these conditions occurred in a variable way.

9.27 It is almost axiomatic to say that autism plus additional conditions will have additional economic consequences, as there are known and unknown costs and service needs for the general population associated with the wide variety of relevant conditions and associated features. When the costs of autism have these other costs added to them, it is clear that there will be additional service needs and economic consequences.

9.28 It is the lack of clinical validity of the existing formulations of autism that has resulted in the whole concept of autism spectrum disorder being reformulated in DSM 5 (and likewise as proposed for ICD-11). All of the existing diagnostic categories defining autism, Asperger’s Syndrome, atypical autism or PDD-NOS have been replaced by a single dimension of ‘autism spectrum disorder’. In DSM 5 this is then specified in terms of intellectual impairment, language function, symptom severity and whether or not there are additional disorders. Thus the new formulation reflects the key findings of the research literature and offers a model which is fully consistent with the dataset for this research.

9.29 On the basis of all of these considerations, we have now been able to construct a conceptual map – a microsegmentation – of the autism spectrum. All of the evidence pointed to three essential factors: intellectual status, which could be graded in terms of normal intelligence through mild disability to moderate/severe disability; symptom severity as reflected in current diagnostic assignment, with those who fitted the Asperger profile showing a more favourable position to those with autism and other diagnoses after controlling for intellectual status, and co-occurring conditions. The first two – intellectual status and symptom severity may be described as stable moderators, in that they imply a gradation from normal or mild to moderate or severe; the last – co-occurring conditions – may be described as a variable moderator, since while it is evident that the more conditions present the greater the additive risk factors, the impact of the presence of these conditions varies extensively.

9.30 Figure 9.1 shows the resultant ‘microsegmentation matrix’ for the autism spectrum.

Figure 9.1 The autism spectrum: microsegmentation matrix

9.31 This is the model which we recommend as a basis for setting priorities for research, resource and budget planning, designing service provision and tailoring interventions to address needs.

9.32 The microsegmentation matrix may be used to offer an evidence-based template for a structured approach to future research and provision. It may be combined with any other framework to provide a model best suited to addressing the issues which will most affect the quality of life of individuals on the autism spectrum and their parents and carers, leading to positive impacts both for individuals and for the economy as a whole.

9.33 The concept of using a matrix as a template which may be combined with any other framework as a structure for planning future research or assessing the quality of service provision may be illustrated by reference to the Scottish Government’s review of educational psychology services in Scotland. MacKay (1989) proposed five core functions for the profession, consultation, assessment, intervention, training and research, and later in establishing performance indicators for the profession proposed that each of these should operate at three levels, the level of the individual child or family, the level of the school or establishment, and the strategic level of the local authority or nationally (MacKay, 1999). This produced in the first instance a 15 cell matrix of five functions at three levels, and was endorsed by the Scottish Ministers as the basis for the operation of psychological services (Scottish Executive, 2002). The matrix was then able to be combined with other frameworks such as key questions for quality assessment in the European Foundation for Quality Management - What key outcomes has the service achieved? How well does it meet the needs of its stakeholders? How good is the leadership of the service? What is its capacity for improvement? Assessing each cell in the matrix against these four questions thus allowed a detailed and comprehensive microsegmentation of this area to support quality assessment and future planning that would address every relevant area of practice.

9.34 In terms of priorities for research, Recommendation 12 of the Scottish Strategy for Autism was: ‘that an evaluation of existing research is commissioned by the ASD Reference Group as well as consideration given to what further research is necessary with a view to disseminating what is available and to the commissioning some pieces that would be of particular practical value to people with ASD and their carers’.

9.35 The microsegmentation matrix may be applied and developed in a similar way to that described above by using the segments as a template to be combined with any research agenda or set of requirements. For example, Pellicano, Dinsmore and Charman (2013), in setting out an agenda for shaping autism research in the UK, considered current research priorities as reflected by funding assigned to six categories: diagnosis; biology, brain and cognition; causes; treatment and interventions; services; and societal issues. They concluded that UK autism research is mostly focussed on children, that it is dominated by funding for the category of biology, brain and cognition with much lower funding for the other five categories and that its priorities are to a large extent divorced from the real needs and aspirations of those on the autism spectrum.

9.36 Using this example in relation to the microsegmentation matrix, its application to these six categories would produce an 8 x 6 matrix of 48 cells. This could then be used as a template for taking forward a research agenda for the Scottish Strategy for Autism, by identifying the spread of existing research and funding across the matrix, ascertaining gaps, agreeing on priorities and planning the projects that would address these priorities. A matrix of this kind can be used flexibly according to differing needs, with cells being combined or subdivided for particular purposes as they arise.

9.37 Similarly, the matrix may be combined with any existing approach or structure to provide a framework for developing ASD provision and support or for planning interventions. For example, using a simple approach based on age and using the broad categories which education authorities, health services and other agencies find to have most utility, namely, preschool, primary school, secondary school and post-school/adult, these four categories combined with the microsegmentation matrix would generate a more detailed matrix of 32 cells. Again, flexible use of the matrix would allow particular cells to be combined or further subdivided to suit the specific purpose for which it was being used.

9.38 However, populating the cells of a matrix of the types exemplified above requires more than providing an evidence-based framework. It requires a rationale to inform content as well as structure. Without such a rationale there is an insufficient basis to guide the question of what the research priorities should be, or what should be the focus of interventions. This issue is addressed in Chapter 10: Microsegmentation and future research and provision for ASD in Scotland.