| Examining the validity of the use of ratio IQs in psychological assessments | IQ tests are amongst the most used psychological assessments, both in research and clinical settings. For participants who cannot complete IQ tests normed for their age, ratio IQ scores (RIQ) are routinely computed and used as a proxy of IQ, especially in large research databases to avoid missing data points. However, because it has never been scientifically validated, this practice is questionable. In the era of big data, it is important to examine the validity of this widely used practice. In this paper, we use the case of autism to examine the differences between standard full-scale IQ (FSIQ) and RIQ. Data was extracted from four databases in which ages, FSIQ scores and subtests raw scores were available for autistic participants between 2 and 17 years old. The IQ tests included were the MSEL (N=12033), DAS-II early years (N=1270), DAS-II school age (N=2848), WISC-IV (N=471) and WISC-V (N=129). RIQs were computed for each participant as well as the discrepancy (DSC) between RIQ and FSIQ. We performed two linear regressions to respectively assess the effect of FSIQ and of age on the DSC for each IQ test, followed by additional analyses comparing age subgroups as well as FSIQ subgroups on DSC. Participants at the extremes of the FSIQ distribution tended to have a greater DSC than participants with average FSIQ. Furthermore, age significantly predicted the DSC, with RIQ superior to FSIQ for younger participants while the opposite was found for older participants. These results question the validity of this widely used alternative scoring method, especially for individuals at the extremes of the normal distribution, with whom RIQs are most often employed. | 38/17423 | Secondary Analysis | Shared |
| Examining Diagnostic Trends and Gender Differences in the ADOS-II | Approximately 3–4 boys for every girl meet the clinical criteria for autism in studies of community diagnostic patterns and studies of autism using samples of convenience. However, girls with autism have been hypothesized to be underdiagnosed, possibly because they may present with differing symptom profiles as compared to boys. This secondary data analysis used the National Database of Autism Research (NDAR) to examine how gender and symptom profiles are associated with one another in a gold standard assessment of autism symptoms, the Autism Diagnostic Observation Schedule II (ADOS-II; Lord, C., Luyster, R., Guthrie, W., & Pickles A. (2012a). Patterns of developmental trajectories in toddlers with autism spectrum disorder. Journal of Consulting and Clinical Psychology, 80(3):477–489. https://doi.org/10.1037/a0027214. Epub 2012 Apr 16. PMID: 22506796, PMCID: PMC3365612). ADOS-II scores from 6183 children ages 6–14 years from 78 different studies in the NDAR indicated that gender was a significant predictor of total algorithm, restrictive and repetitive behavioral, and social communicative difficulties composite severity scores. These findings suggest that gender differences in ADOS scores are common in many samples and may reflect on current diagnostic practices. | 25/5615 | Secondary Analysis | Shared |
| Gender Differences: Confirmatory Factor Analysis of the ADOS-II | Purpose
Recent research has suggested that autism may present differently in girls compared to boys, encouraging the exploration of a sex-differential diagnostic criteria. Gender differences in diagnostic assessments have been shown on the ADOS-II, such that, on average, females score significantly lower than males on all scales and are less likely to show atypicality on most items related to social communicative difficulties. Yet, gender differences in the latent structure of instruments like the ADOS-II have not been examined systematically.
Methods
As such, this secondary data analysis examined 4,100 youth diagnosed with autism (Mage = 9.9, 813 female & 3287 male) examined item response trends by gender on the ADOS-II Module 3.
Results
Multi-Group Confirmatory Factor Analysis results show that the factor loadings of four ADOS-II items differ across the genders. One SCD item and one RRB item are strongly related to the latent factor in the female group, while two RRB items have larger factor loadings in the male group.
Conclusion
The assumption of an identical latent factor structure for the ADOS-II Module 3 for males and females might not be justifiable. Possible diagnostic implications are discussed. | 25/5615 | Secondary Analysis | Shared |
| Prognostic early snapshot stratification of autism based on adaptive functioning | A major goal of precision medicine is to predict prognosis based on individualized information at the earliest possible points in development. Using early snapshots of adaptive functioning and unsupervised data- driven discovery methods, we uncover highly stable early autism subtypes that yield information relevant to later prognosis. Data from the National Institute of Mental Health Data Archive (NDA) (n = 1,098) was used to uncover three early subtypes (<72 months) that generalize with 96% accuracy. Outcome data from NDA (n = 2,561; mean age, 13 years) also reproducibly clusters into three subtypes with 99% generalization accuracy. Early snapshot subtypes predict developmental trajectories in non-verbal cognitive, language and motor domains and are predictive of membership in different adaptive functioning outcome subtypes. Robust and prognosis- relevant subtyping of autism based on early snapshots of adaptive functioning may aid future research work via prediction of these subtypes with our reproducible stratification model. | 5/3517 | Secondary Analysis | Shared |
| Investigating autism etiology and heterogeneity by decision tree algorithm | Autism spectrum disorder (ASD) is a neurodevelopmental disorder that causes deficits in cognition, communication and social skills. ASD, however, is a highly heterogeneous disorder. This heterogeneity has made identifying the etiology of ASD a particularly difficult challenge, as patients exhibit a wide spectrum of symptoms without any unifying genetic or environmental factors to account for the disorder. For better understanding of ASD, it is paramount to identify potential genetic and environmental risk factors that are comorbid with it. Identifying such factors is of great importance to determine potential causes for the disorder, and understand its heterogeneity. Existing large-scale datasets offer an opportunity for computer scientists to undertake this task by utilizing machine learning to reliably and efficiently obtain insight about potential ASD risk factors, which would in turn assist in guiding research in the field. In this study, decision tree algorithms were utilized to analyze related factors in datasets obtained from the National Database for Autism Research (NDAR) consisting of nearly 3000 individuals. We were able to identify 15 medical conditions that were highly associated with ASD diagnoses in patients; furthermore, we extended our analysis to the family medical history of patients and we report six potentially hereditary medical conditions associated with ASD. Associations reported had a 90% accuracy. Meanwhile, gender comparisons highlighted conditions that were unique to each gender and others that overlapped. Those findings were validated by the academic literature, thus opening the way for new directions for the use of decision tree algorithms to further understand the etiology of autism.
| 1/3382 | Secondary Analysis | Shared |
| Imbalanced social-communicative and restricted repetitive behavior subtypes in autism spectrum disorder exhibit different neural circuitry | Social-communication (SC) and restricted repetitive behaviors (RRB) are autism diagnostic symptom domains. SC and RRB severity can markedly differ within and between individuals and may be underpinned by different neural circuitry and genetic mechanisms. Modeling SC-RRB balance could help identify how neural circuitry and genetic mechanisms map onto such phenotypic heterogeneity. Here, we developed a phenotypic stratification model that makes highly accurate (97–99%) out-of-sample SC = RRB, SC > RRB, and RRB > SC subtype predictions. Applying this model to resting state fMRI data from the EU-AIMS LEAP dataset (n = 509), we find that while the phenotypic subtypes share many commonalities in terms of intrinsic functional connectivity, they also show replicable differences within some networks compared to a typically-developing group (TD). Specifically, the somatomotor network is hypoconnected with perisylvian circuitry in SC > RRB and visual association circuitry in SC = RRB. The SC = RRB subtype show hyperconnectivity between medial motor and anterior salience circuitry. Genes that are highly expressed within these networks show a differential enrichment pattern with known autism-associated genes, indicating that such circuits are affected by differing autism-associated genomic mechanisms. These results suggest that SC-RRB imbalance subtypes share many commonalities, but also express subtle differences in functional neural circuitry and the genomic underpinnings behind such circuitry. | 3/1708 | Secondary Analysis | Shared |
| Word Learning and Word Features | Vocabulary composition and word-learning biases are closely interrelated in typical development. Learning new words involves attending to certain properties to facilitate word learning. Such word-learning biases are influenced by perceptually and conceptually salient word features, including high imageability, concreteness, and iconicity. This study examined the association of vocabulary knowledge and word features in young children with ASD (n = 280) and typically developing (TD) toddlers (n = 1,054). Secondary analyses were conducted using data from the National Database for Autism Research and the Wordbank database. Expressive vocabulary was measured using the MacArthur-Bates Communicative Development Inventory. Although the trajectories for concreteness, iconicity, and imageability are similar between children with ASD and TD toddlers, divergences were observed. Differences in imageability are seen early but resolve to a common trajectory; differences in iconicity are small but consistent; and differences in concreteness only emerge after both groups reach a simultaneous peak, before converging again. This study reports unique information about the nonlinear growth patterns associated with each word feature for and distinctions in these growth patterns between the groups. | 1/280 | Primary Analysis | Shared |
| Examining the Shape Bias in Young Autistic Children: A Vocabulary Composition Analysis | Shape is a salient object property and one of the first that children use to categorize objects under one label. Colunga and Sims (2017) suggest that noun vocabulary composition and word learning biases are closely interrelated in typical development. The current study examined the association between noun vocabulary knowledge and perceptual word features, specifically shape and material features. Participants included 249 autistic children and 1,245 non-autistic toddlers who were matched on expressive noun vocabulary size and gender. Nouns were categorized using the Samuelson and Smith (1999) noun feature database. A simple group comparison revealed no group differences in shape bias; both groups evidenced developing noun vocabularies that favored shape+solid and nonsolid+material nouns. However, the trajectory of evidence of shape bias as a function of vocabulary size differed between the groups, with autistic children
demonstrating a reduced shape-bias initially. Future work should examine how children’s learning biases shift over development and whether the shape bias promotes lexical development to the same degree across groups. | 3/249 | Secondary Analysis | Shared |
| Modeling Vocabulary Growth in Autistic and Non-Autistic Children | We assessed the goodness of fit of three models of vocabulary growth, with varying sensitivity to the structure of the environment and the learner’s internal state, to estimated vocabulary growth trajectories in autistic and non-autistic children. We first computed word-level acquisition norms that indicate the vocabulary size at which individual words tend to be learned by each group. We then evaluated how well network growth models based on natural language co-occurrence structure and word associations account for variance in the autistic and non-autistic acquisition norms. In addition to replicating key observations from prior work and observing that the growth models explained similar amounts of variance in each group, we found that autistic vocabulary growth also exhibits growth consistent with “the lure of the associates” model. Thus, both groups leverage semantic structure in the learning environment for vocabulary development, but autistic vocabulary growth is also strongly influenced by existing vocabulary knowledge. | 2/247 | Secondary Analysis | Shared |
| Semantic modeling 2023 | Although it is well documented that children with ASD are slower to develop their lexicons, we still have a limited understanding of the structure of early lexical knowledge in children with ASD. The current study uses network analysis and differential item functioning anlaysis to examine the structure of semantic knowledge, which may provide insight into the learning processes that influence early word learning. | 4/208 | Secondary Analysis | Shared |
| Semantic Network Modeling in Young Autistic Children | Background: Most young autistic children have delayed vocabulary growth relative to their non-autistic peers. Additionally, previous studies have revealed that autistic children are less likely to encode associated features of novel objects, suggesting inefficient encoding or different processes for acquiring semantic information about words. Recent network analyses of vocabulary growth revealed important relationships between early vocabulary acquisition and the structure of the sematic environment.
Methods: We studied the expressive vocabularies of 970 non-autistic toddlers (Mage = 20.82 months) and 194 autistic children (Mage = 54.58 months) in two studies. The groups were vocabulary-matched (words produced: MAutistic = 213.60, MNon-autistic = 213.72). In study 1, we estimated their trajectories of semantic development using network analyses. Network structure was based on child-oriented adult-generated word associations. We compared child semantic networks according to indegree, average shortest path length, and clustering coefficient (features that holistically contribute to well-connected network structure). Then, in study 2, we attempted to relate vocabulary-level effects to word-level learning biases.
Results: Study 1 revealed that autistic and non-autistic children are sensitive to the structure of their semantic environment. Both groups demonstrated nonlinear vocabulary trajectories that differed from random acquisition networks. Despite similarities, group differences were observed for each network metric. Differences were most pronounced for clustering coefficient (how closely connected groups of words are), with earlier peaks for autistic children. Study 2 demonstrated that many words differ in their expected vocabulary size of acquisition.
Conclusions: Group differences at the vocabulary- and word-levels indicate that, although autistic children are learning from their semantic environment, they may be processing their semantic environment differently. These deviations indicate that autistic children have distinctive learning biases that may align with core autism features. | 4/194 | Secondary Analysis | Shared |
