Are Diagnostic Instruments Equally Accurate at Classifying Autism Spectrum Disorder in Males and Females?

Background: In cognitively-able samples, the male to female ratio in Autism Spectrum Disorder (ASD) is up to 10:1 (Fombonne, 2009). One hypothesis for the sex discrepancy is that diagnostic criteria is orientated towards a presentation typically seen in males (Kreiser & White, 2014). It is argued as a result, females are systematically under-diagnosed, a phenomena termed the ‘male bias’ (Kirkovski et al., 2013). With increasing evidence that females presenting with clinically significant traits are less likely to receive a diagnosis, many are highlighting the need for gender-specific thresholds in ASD assessments (Constantino and Chairman, 2012; Dworzynski et al., 2012). However, examination of diagnostic instruments is required to ascertain whether tools are indeed less accurate at diagnosing ASD in females, and to identify the areas of inaccuracy, before such gendered tools may be developed.

Objectives: Three aims were identified; 1) to examine whether diagnostic tools’ classifications of males and females differed in diagnostic accuracy. 2) To compare the accuracy of diagnostic tools by sex i.e. to investigate whether a tool may be more accurate at diagnosing females. 3) To investigate whether behavioural domain classifications (i.e. areas of deficits that relate to diagnostic algorithms) of males and females differed in diagnostic accuracy, in order to identify specific areas of inaccuracy for the development of gender-specific thresholds.

Methods: Autism Diagnostic Observational Schedule (ADOS) (Lord et al., 2000) and Developmental, Dimensional and Diagnostic Interview (3Di) (Skuse et al., 2004) classifications of 256 males and 60 females were compared against the ‘gold standard’ multidisciplinary team assessment diagnosis. Receiver Operating Characteristics curves of ADOS and 3Di classifications of males and females were performed to describe diagnostic accuracy, measured by sensitivity and specificity. Participants data was selected from a database of consecutive referrals for the assessment of high-functioning ASD. ADOS and 3Di assessments took place concurrently as part of routine clinical practice.

Results: Chi-square and Fisher’s exact test found no significant differences in the sensitivity or specificity of overall or behavioural domain classifications of males and females. McNemar’s test found no significant differences in the sensitivity of ADOS and 3Di classifications of males or females but found the 3Di showed significantly lower specificity than the ADOS of males and females.

Conclusions: The 3Di and ADOS are found to be equally accurate at classifying ASD in males and females. The 3Di and ADOS show comparable accuracy classifying both males and females but the 3Di showed comparably lower accuracy classifying non-spectrum individuals. Thus, when ruling out ASD is of greater importance, the ADOS may be a more accurate diagnostic instrutment for both males and females. No behavioural domain was found to be less accurate classifying females, thus the results do not identify areas of inaccuracy to recommend the development of gender-specific thresholds in these tools. In conclusion, the 3Di and ADOS show equal accuracy classifying high-functioning males and females referred for clinical assessment.