Exploring Phenotypic and Sociodemographic Influences on Cognitive-Adaptive Functioning Gap in Neurodivergent Children

Exploring Phenotypic and Sociodemographic Influences on Cognitive-Adaptive Functioning Gap in Neurodivergent Children | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Exploring Phenotypic and Sociodemographic Influences on Cognitive-Adaptive Functioning Gap in Neurodivergent Children Eric Wan, Jason P. Lerch, Evdokia Anagnostou, Melanie Penner, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5968118/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: In the general population, adaptive functioning typically aligns with cognitive abilities; however, this relationship appears more complex among neurodivergent individuals. Neurodevelopmental conditions (NDCs), including autism, attention-deficit/hyperactivity disorder (ADHD), and obsessive-compulsive disorder (OCD), are associated with significant differences between cognitive and adaptive functioning, described as the cognitive-adaptive functioning gap. While this gap has been examined primarily in autistic individuals, it has also been observed across other NDCs. In mixed neurotypical and neurodivergent samples, individuals exhibit a range of discrepancies, reflecting the heterogeneity of cognitive-adaptive functioning gap profiles. Although the gap tends to be larger in NDCs compared to neurotypical populations, there is limited understanding of the phenotypic and sociodemographic factors linked to these discrepancies in neurodivergent children. The present study explores the features associated with cognitive-adaptive functioning gap in a sample of children and adolescents, including both neurodivergent and neurotypical individuals. Methods: The study used data from the Province of Ontario Neurodevelopmental Disorders Network (POND), comprising 902 participants (autism = 409, ADHD = 210, OCD = 36, neurotypical = 214, other = 33) aged 6-21 years. Cognitive functioning was measured with full-scale IQ (FSIQ) from the Wechsler family of tests, and adaptive functioning was measured with the Adaptive Behavior Assessment System-II (ABAS-II), specifically the General Adaptive Composite (GAC) score. The cognitive-adaptive functioning gap was calculated as the difference between FSIQ and ABAS-II GAC scores. Phenotypic measures included social communication (Social Communication Questionnaire, or SCQ), ADHD symptoms (Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale, or SWAN), OCD symptoms (Toronto Obsessive-Compulsive Scale, or TOCS), and mental health symptoms (Child Behavior Checklist, or CBCL). Sociodemographic data encompassed sex, age, race, household income, and caregiver education. Nine computational models were used to estimate the cognitive-adaptive functioning gap. SHapley Additive exPlanations (SHAP) analysis was used to interpret the contributions of individual features to the cognitive-adaptive functioning gap model. Results : The random forest model demonstrated the highest predictive accuracy (R² = 0.88, p < 0.001), performing better than other models with a mean absolute error of 4.14. SHAP analysis indicated that FSIQ, SCQ, and inattentive traits were the most influential features in estimating cognitive-adaptive functioning gap. Higher FSIQ (FSIQ ≥ 97.0) was linked to larger cognitive-adaptive functioning gaps across the combined sample. Similarly, social communication differences (SCQ ≥ 10.7) and inattentive traits (SWAN ≥ 3.4) were associated with larger gaps. Sociodemographic factors showed smaller but statistically significant associations, with sex (p < 0.001) and age (p < 0.001) showing relations to the cognitive-adaptive functioning gap (male sex and younger age were linked to smaller gaps). Limitations: Key limitations include the use of broad sociodemographic categories, the cross-sectional design limiting insights into developmental changes in the cognitive-adaptive functioning gap, and the absence of an independent test set, which may affect generalizability. Future longitudinal studies and larger sample size are needed to address these limitations. Conclusions : This study identifies the cognitive-adaptive functioning gap as associated with higher FSIQ, social-communication challenges, and inattentive traits, with phenotypic features showing stronger connections than sociodemographic factors. These findings suggest that focusing support on these key factors may help reduce the gap. Further longitudinal research is needed to explore how these gaps evolve and assess potential intervention strategies. Neurodevelopmental conditions cognitive functioning social communication inattention sociodemographic machine learning feature importance. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Neurodevelopmental conditions (NDCs), including autism, attention-deficit/hyperactivity disorder (ADHD), and obsessive-compulsive disorder (OCD) are characterized by differences in brain development that can impact cognitive and adaptive functioning. Cognitive functioning, often measured via the intelligence quotient (IQ), entails mental processes such as perception, attention, memory, language, problem-solving, and reasoning (American Psychiatric Association, 2022). Adaptive functioning quantifies an individual's capacity to perform daily tasks, adapt to environmental changes, and meet the demands of everyday life (American Psychiatric Association, 2022). Cognitive functioning varies widely in NDCs. For example, among autistic children, an estimated 33% have an intellectual disability (ID) (Knopf, 2020 ; Shenouda et al., 2023 ; Zeidan et al., 2022 ), while approximately 13.7% score in the superior intelligence range (Billeiter & Froiland, 2023 ). ADHD, OCD, and other NDCs can also co-occur with ID (Ahuja et al., 2013 ; Mouga et al., 2015 ; Root et al., 2019 ). Adaptive functioning is also highly varied in NDCs. Autism has been associated with decreased adaptive functioning relative to other NDCs, although large variation in these skills exist (Ameis et al., 2016 ; Jacobs et al., 2021 ; Kenworthy et al., 2010 ; Mahendiran et al., 2019 ; Mouga et al., 2015 ; Perry et al., 2009 ). Specific domains of adaptive functioning, such as socialization, communication, and daily living skills, tend to be more impaired in children with autism, particularly those with co-occurring autism and ADHD, compared to children with ADHD, who exhibit more pronounced difficulties in motor adaptive functioning (Scandurra et al., 2019 ). For the general population, adaptive behavior and cognitive ability are assumed to be positively associated, however, in NDCs, a significant gap between cognitive and adaptive functioning may exist (Charman et al., 2011 ; Duncan & Bishop, 2015 ; Kanne et al., 2011 ; Klin et al., 2007 ; Mouga et al., 2015 ; Tillmann et al., 2019 ). This gap often manifests as lower adaptive functioning relative to cognitive abilities and tends to persist from toddlerhood through adolescence and young adulthood, though its trajectory beyond this stage remains unclear (Bradshaw et al., 2019 ; Kanne et al., 2011 ; Kraper et al., 2017 ). Despite the established presence of the cognitive-adaptive functioning gap, the underlying phenotypic and sociodemographic factors to this gap are still not fully clarified. The associations between phenotypic and sociodemographic features and the cognitive-adaptive functioning gap in neurodevelopmental conditions remain complex and inconclusive. While social-communicative symptoms have been consistently associated with larger gaps (Kanne et al., 2011 ; Tillmann et al., 2019 ), findings regarding restricted and repetitive behaviors (RRBs) and mental health challenges like anxiety and depression are mixed. One study reported no significant associations between the gap and anxiety or RRBs (Tillmann et al., 2019 ) whereas another study identified some correlations between the gap and anxiety, and between the gap and RRBs (Kraper et al., 2017 ). Sociodemographic factors further contribute to this complexity, as evidence suggests the gap widens with age (Bradshaw et al., 2019 ), with sex-specific patterns showing larger gap in females than gap in males over time (McQuaid et al., 2021 ). Also, higher maternal education is associated with better adaptive functioning, whereas household income does not show an association with it (Wang et al., 2023 ). These mixed findings presented a research gap in understanding the phenotypic and sociodemographic correlates of cognitive-adaptive functioning gap. The present study aims to fill this research gap by systematically evaluating the phenotypic and sociodemographic correlates of the cognitive-adaptive functioning gap in neurodivergent children and youth. By leveraging computational models, our primary objective is to identify and evaluate the features most associated with the cognitive-adaptive functioning gap and to demonstrate how machine learning approaches can enhance our understanding of these complex relations. To achieve this, we assessed the performance of nine computational models in estimating the cognitive-adaptive functioning gap and analyzed the key associations of phenotypic and demographic features to the model outputs. Methods Participants For this study, we used a subset of the data from the Province of Ontario Neurodevelopmental Network (POND), exported on March 26, 2024. The dataset included 3,380 participants, between ages of 0 and 21, who were neurotypical or neurodivergent (neurotypical = 335, autism = 1,307, ADHD = 1,083, OCD = 291, other = 353, 4 siblings without a diagnosis, 16 missing diagnoses). The “other” category included anxiety, sub-threshold ADHD, sub-threshold OCD, Down syndrome, Fragile X, ID only, Rett syndrome, social communication disorders, and Tourette syndrome. Validated assessments were used to support the clinical diagnoses. Specifically, the Autism Diagnostic Observation Schedule-2 (ADOS-2) and Autism Diagnostic Interview-Revised (ADI-R) were used for autism, Parent Interview for Child Symptoms (PICS) was used for ADHD, and the Kiddie-Schedule for Affective Disorders and Schizophrenia (K-SADS) and the Children’s Yale-Brown Obsessive Compulsive Scale (CYBOCS) were used for OCD. A subset of the above dataset was select for the present analysis based on the following inclusion criteria: (1) age between 6 and 21 ( n = 2,281), (2) neurotypical or a diagnosis of autism, ADHD, OCD, or other ( n = 2,277), (3) complete data for the phenotypic measures used in this study (full scale IQ, adaptive functioning, mental health symptoms, and autism, ADHD, and OCD features) ( n = 1,764), and (4) complete data for race, household income, and primary caregiver’s education ( n = 902). The final sample consisted of 902 participants. Measures Full scale IQ was assessed using the Wechsler family of measures, including the Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II) (Wechsler et al., 2011), the Wechsler Intelligence Scale for Children, Fourth (WISC-IV) (Wechsler, 2003 ) and Fifth Editions (WISC-V) (Raiford, 2018 ) the Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition (WPPSI-IV) (Wechsler & Psychological Corporation, 2012), or the Stanford-Binet Scale (Roid & Pomplun, 2012 ), as appropriate for developmental stage. Adaptive functioning was quantified using the Adaptive Behavior Assessment System (ABAS)-II (Harrison & Oakland, 2003), a parent-report measure with scores along domains of Communication, Community Use, Functional Academics, Home Living, Health/Safety, Leisure, Self-Direction, Social, and Work, through norm-referenced scaled scores. Our analysis used the General Adaptive Composite (GAC) score, representing overall adaptive functioning, encompassing all subscales. The cognitive-adaptive functioning gap was defined as full-scale IQ minus ABAS GAC score. Autism, ADHD, and OCD features were quantified using the Social Communication Questionnaire (SCQ) (Berument et al., 1999 ), the Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale (SWAN) (Swanson et al., 2015 ), and the Toronto Obsessive-Compulsive Scale (TOCS) (Park et al., 2016 ), respectively. Mental health symptoms were characterized using the Child Behavior Checklist (CBCL), internalizing and externalizing subscales (Achenbach & Edelbrock, 1983 ). Sociodemographic characteristics included in the study were sex (biological), age, household income, race, and the education of the primary caregiver. The racial categories were Black, East Asian, Indigenous, Latino, Middle Eastern, other, South Asian, Southeast Asian, and White. Household income was categorized as follows: lower than $ 10,000; $ 10,000 to $ 24,999; $ 25,000 to $ 49,999; $ 50,000 to $ 74,999; $ 75,000 to $ 99,999; $ 100,000 to $ 149,999; $ 150,000 to $ 199,999; and $ 200,000 or more. Primary caregiver’s education was categorized as lower than Bachelor’s degree and Bachelor’s degree and higher. In our study, we categorized the household income into three levels (lower than $ 50,000 as low, between $ 50,000 and $ 100,000 as middle, and $ 100,000 or more as high), race as white or minoritized, and education as lower than Bachelor's degree, Bachelor’s degree or higher. Statistical Analyses Our analyses were conducted using R version 4.3.2 with caret version 6.0–94 (Kuhn & Max, 2008). Group Differences We compared differences in sociodemographic and behavioral measures across diagnostic groups using Kruskal-Wallis tests with Bonferroni correction for continuous variables and Chi-squared tests for categorical variables. The p -values for multiple comparisons were adjusted using the Bonferroni correction. Post-hoc analysis was conducted using Dunn’s procedure. We assessed sociodemographic and IQ identity group differences in cognitive-adaptive functioning gap using the Kruskal-Wallis test for each feature group (e.g., sociodemographic and phenotypic measures), with Bonferroni-adjusted p -values used to account for multiple comparisons. For feature groups showing significant differences (p < 0.05), post-hoc pairwise comparisons were performed using the Wilcoxon rank-sum test. Effect sizes were calculated for both tests to quantify the magnitude of group differences. Modeling Cognitive-Adaptive Functioning Gaps The gap between cognitive and adaptive functioning (dependent variable) was examined in relation to phenotypic characteristics (IQ, SCQ, SWAN, TOCS, CBCL internalizing and externalizing symptoms) and sociodemographic factors (age, sex, race, first caregiver’s education, household income). Nine machine learning regression approaches were used: ordinary least squares (OLS) linear regression, ridge regression (Arashi et al., 2021 ), ElasticNet (Zou & Hastie, 2005 ), LASSO (Tibshirani, 1996 ), decision tree (Song & Lu, 2015 ), gradient boosting (Friedman, 2002 ), support vector regression (Smola & Schölkopf, 2004 ), random forest (Breiman, 2001 ), and neural networks (Abiodun et al., 2018 ). OLS regression was used due to its simplicity and interpretability. Ridge, LASSO, and ElasticNet regression techniques were also considered as they handle the multicollinearity present in the phenotypic and the sociodemographic features better than OLS regression. To address the non-linear relations and interactions among features, which may not be effectively captured by linear models, we incorporated tree-based models. Random forests (Breiman, 2001 ) improve upon single decision trees by aggregating the estimations with multiple trees to reduce variance and avoid overfitting, providing robustness especially in the presence of noise. Another tree-based method, Gradient boosting (Friedman, 2002 ), uses sequential model building to enhance accuracy. Furthermore, we used kernel support vector regression (Smola & Schölkopf, 2004 ) for its ability to model non-linear relations (Abiodun et al., 2018 ), with advantages of flexibility, scalability, and capacity to handle noise. The dataset was evaluated using a 5-fold cross-validation strategy to assess model performance and optimize hyperparameters. In this approach, the dataset was split into five folds, with each fold serving as the test set for evaluation in one iteration, while the remaining 80% (four folds) was used for training. Within each fold, the data was split in a stratified manner to ensure proportional representation of key variables, including diagnosis, ABAS GAC (10-point bins), and sex, across the training and test subsets. Model performance was evaluated using the median absolute error (MAE), with final metrics computed with MAE values across the five folds. This cross-validation strategy ensures that every data point is used for both training and testing, providing a robust estimate of the model’s generalization performance. An additional independent test set was not held out due to the limited size of the dataset, as reserving such a set would have significantly reduced the amount of data available for training and validation. For statistical comparison of model performance, we conducted a repeated measures ANalysis Of VAriance (ANOVA) to determine if there were significant differences in performance across the models. Given the repeated nature of the measurements, this approach accounts for within-subject variability. When the ANOVA indicated significant differences, we performed post-hoc comparisons using Tukey's HSD test and conducted pairwise comparisons with Bonferroni correction to identify specific models that differed from each other while controlling for family-wise error. To assess the effect of individual features and their interactions on the IQ-ABAS gap, SHAP (SHapley Additive exPlanations) values were employed. SHAP, derived from cooperative game theory, provides an interpretable framework for understanding how each feature contributes to the predictions of machine learning models by assigning importance scores (Lundberg & Lee, 2017 ; Molnar, 2020 ). Positive SHAP values indicate that a feature increases the model’s estimation, whereas negative values indicate a decrease in the estimation. SHAP dependence plots further illustrate the relations between individual feature values and the predicted IQ-ABAS gap, allowing us to capture specific contributions across the feature space and explore potential interactions among features. This method has been widely adopted in healthcare research to explain model estimations and understand feature importance (Loh et al., 2022 ; Lundberg et al., 2020 ). To evaluate the associations between different features and the individual feature’s SHAP values, we applied appropriate statistical methods based on the nature of the data. For continuous features (e.g., FSIQ, CBCL scores), we used Spearman’s rank correlation to measure the strength and direction of associations, where correlations greater than +/-0.60 were considered strong. For ordinal features (e.g., household income, caregiver education), we employed Kendall’s rank correlation, applying the same threshold of τ > |0.60| for strong associations. Finally, for nominal features (e.g., sex, race), we conducted ANOVA to compare group differences in the IQ-ABAS gap. These methods allowed us to explore how each feature type was associated with the observed discrepancies in cognitive and adaptive functioning. Results Sample Characteristics The final subset of the POND sample used for the analysis included data from 902 participants (neurotypical = 214, autism = 409, ADHD = 210, OCD = 36, Other = 33). Detailed sample characteristics are presented in Table 1 . Table 1 Sample characteristics with median(IQR) for various features, presented as the entire sample and by diagnosis subgroups. All Autism ADHD OCD NT Other n 902 409 210 36 214 33 Sex b female 265 85 63 15 90 12 male 637 324 147 21 124 21 Age b (years) 11 (6.0) 11 (5.0) 10 (5.0) 12 (4.0) 11 (5.0) 10 (6.0) FSIQ a 100 (27.8) 93 (35.0) 98 (19.8) 113 (15.5) 108 (17.0) 79 (45.0) SCQ a 10 (17) 20 (12) 6 (9.0) 5 (5.3) 2 (2.0) 9 (13) Inattention a 3 (7) 5 (5) 6 (5) 0 (5) 0 (0) 5 (6) Hyperactivity a 1.0 (5) 3.0 (5) 3.5 (6) 0.0 (1) 0.0 (0) 3.5 (7) TOCS a -21.0 (50.0) -8.00 (37.0) -30.5 (49.8) 19.5 (19.5) -50.0 (43.5) -13.0 (48.0) CBCL Internal a 60.0 (18.0) 65.0 (13.0) 63.0 (16.0) 64.5 (10.3) 48.0 (12.0) 63.0 (18.0) CBCL External a 54.0 (18.0) 58.0 (14.0) 61.0 (17.0) 53.0 (13.5) 41.0 (15.0) 56.0 (13.0) ABAS GAC a 77.0 (33.8) 63.0 (21.0) 78.0 (24.0) 92.0 (21.5) 103 (22.0) 72.0 (33.0) IQ-ABAS gap a 17.5 (29.0) 26.0 (33.0) 18.0 (23.8) 20.0 (20.8) 6.50 (22.8) 4.00 (24.0) Income b = $ 50k and = $ 100k 451 174 92 21 149 15 Race white 652 283 151 31 166 21 minoritized 250 126 59 5 48 12 Education b lower than bachelors 435 208 130 20 64 13 Bachelors or higher 467 201 80 16 150 20 Note : a p < .001 (group effect; with Bonferroni correction) , b p<..05 (group effect; with Bonferroni correction); SCQ: Social Communication Questionnaire (raw scores), TOCS: Toronto Obsessive-Compulsive Scale (raw scores), CBCL: Child Behavior Checklist (T-scores), ABAS2 GAC: Adaptive Behavior Assessment System Second Edition General Adaptive Composite (standard scores). Group Comparisons The distribution of the IQ-ABAS gap is illustrated in Fig. 1, disaggregated by diagnosis and sociodemographic identities. Wilcoxon Signed-Rank tests with the Bonferroni correction were used to test differences of median IQ-ABAS gap from zero. The gap was significantly greater than zero for the pooled sample (median = 17.50, W = 346163.5, p < .001) as well as each diagnosis group (autism: median = 26, W = 74899.50, p < .001; ADHD: median = 18, W = 19599.50, p < .001; OCD: median = 20, W = 631.50, p < .001; NT: median = 6.50, W = 15446.00, p < .001), except the Other NDCs group (median = 4, W = 369.50, p = .25). A Kruskal-Wallis test indicated that there was a significant difference in IQ-ABAS gap across the diagnostic groups (Autism: median = 26, IQR = 33; ADHD: median = 18, IQR = 23.75; OCD: median = 20, IQR = 20.75; NT: median T =6.5, IQR = 22.75; Other: median r =4, IQR = 24; χ2 (4) = 118, p < .001 ) . Post-hoc comparisons using Dunn’s Test with Bonferroni correction revealed that the autism group had the highest IQ-ABAS gap, significantly higher than ADHD ( Z = 3.52, p = .002), NT ( Z = 10.40, p < .001), and other NDCs ( Z = 4.68, p < .001) groups. The OCD group had the second-highest gap, significantly higher than both NT ( Z = 3.80, p < .001) and other NDCs ( Z = 2.72, p = .033). ADHD showed a significantly higher gap than NT ( Z = 5.96, p < .001) and other NDCs ( Z = 2.93, p = .017). However, there were no significant differences between the discrepancies of the ADHD and OCD groups ( Z =-0.59, p = 1), autism and OCD groups ( Z = 1.11, p = 1), or between NT and other NDCs group ( Z =-0.165, p = 1). A Kruskal-Wallis test was performed to evaluate group differences in cognitive-adaptive functioning gap across sociodemographic and IQ identity subgroups. Here, we dichotomized the data with an FSIQ value of 70. The test indicated significant differences for IQ ( H (1) = 93.10, adj-p < .001, η² =.10; =70: median = 20), with the lower IQ group having less of a gap than the higher IQ group, and sex ( H (1) = 12.70, adj-p = .0017, η² =.01, female: median = 13, male: median = 19), with females having less of a gap than the males, but not for education ( H (1) = .84, adj-p = 1, η² =0), household income ( H (2) = 7.50, adj-p = .12, η² =0), and race ( H (1) = 1.10, adj-p = 1, η² =0). Model Performance The accuracies of nine regression models in estimating the IQ-ABAS gap are presented graphically in Fig. 2 (A and B) and detailed numerically in Supplementary Tables 8 and 9. The results were obtained using the test folds from the 5-fold cross-validation process described in the Methods section. Performance metrics (e.g., MAE) were averaged across the five folds. The results are disaggregated by sex, race, socioeconomic status, and IQ subgroups. The repeated measures ANOVA showed a statistically significant difference in performance across the nine models ( F (8, 175) = 20.00, p < .001, η² =.82). Post-hoc Tukey's HSD analysis demonstrated that random forest significantly outperformed all other models on MAEs (Mean = 4.14, SD = .25), with differences ranging from gradient boosting ( p < .001, 95% C.I.=[1.95, 5.25]) to ridge regression ( p < .001, 95% C.I.=[3.38, 6.69]). Detailed pairwise comparisons are provided in Supplementary Table 7. The random forest model demonstrated strong predictive power [ R 2 = 0.88, F (1, 175) = 1236, p < .001], indicating that it explained 87.60% of the variance in actual cognitive-adaptive functioning gap values (Fig. 2 (C)). It suggests a high correlation between predicted and actual values, so the model can reliably predict the cognitive-adaptive functioning gap. The significant intercept [ B = 3.25, SE = 0.65, t (175) = 4.97, p < .001] (Fig. 2 (C)). suggests a consistent bias in estimates across all values. This model has been used for SHAP analysis in the subsequent results. Predictive Features The SHAP importance plot (Fig. 2 (D) ) ranks the features by their mean absolute SHAP values, indicating the relative contribution of each feature in predicting IQ-ABAS gap. The features that contributed most to the model’s predictions were FSIQ, SCQ, and inattention. Overall, phenotypic features contributed more to the model than sociodemographic features. The SHAP dependence plots (Fig. 3 ) further explained the association between the features and the model's predictions for IQ-ABAS gap. These plots show the directionality and the strength of a feature’s contribution to IQ-ABAS gap. In particular, the zero-crossing points indicate the threshold or value of the feature at which its contribution shifts from increasing the model’s prediction to decreasing it (or vice versa). Phenotypic features with strong positive relations with their SHAP values include FSIQ, with a zero-crossing point of 97.0 ( R 2 = 0.93, p < .001); SCQ, with a zero-crossing point of 10.7 ( R 2 = 0.87, p < .001); inattention, with a zero-crossing point of 3.4 ( R 2 = 0.87, p < .001); CBCL internalizing symptoms, with a zero-crossing point of 58.1 ( R 2 = 0.76, p < .001); CBCL externalizing symptoms, with a zero-crossing point of 54.2 ( R 2 = 0.80, p < .001); and hyperactivity, with a zero-crossing point of 2.4 ( R 2 = 0.57, p < .001). The relation between TOCS and SHAP values was significant but weak and negative, with a zero-crossing point of -22.8 ( R 2 = 0.34, p < .001). The sociodemographic features, except for age, showed lower mean absolute SHAP values compared to the phenotypic features. All sociodemographic features, except for race, were significantly associated with the gap. Age showed a strong negative correlation with SHAP values ( r =-0.75, p < .001), while both household income ( τ =-0.35, p < .001) and first caregiver's education level ( τ =-0.30, p < .001) exhibited weaker negative correlations. ANOVA showed a significant effect of sex on SHAP values ( F (1, 173) = 299, p < .001), and a marginal effect of race ( F (1, 173) = 2.96, p = .0872). The SHAP interaction results suggested that most of the feature pairs had little to no interactions contributing to cognitive-adaptive functioning gap (Fig. 4 (A)), with the highest SHAP interactions observed in the FSIQ-SCQ feature pair (3.44), the FSIQ-inattention feature pair (1.7), and the SCQ-inattention feature pair (1.61). Figure 4 (B and C) visualize these interactions. Discussion This study examined the correlates of the gap between cognitive and adaptive functioning in a neurodiverse sample of children. This was done using univariate testing and group comparisons, as well as with a multivariate approach that employed various computational models to predict the cognitive-adaptive functioning gap. Our findings demonstrate that the cognitive-adaptive functioning gap is observed across NDCs and typical development, although the widest gaps were found in autism, ADHD, and OCD. This is consistent with previous findings in autism (Bradshaw et al., 2019 ; Zukerman et al., 2021 ), extending the understanding to other neurodevelopmental conditions. While the presence of this gap in the neurotypical group may reflect natural variability in the relation between cognitive ability and adaptive functioning rather than a pathological process. Importantly, this smaller gap in the neurotypical group provides a useful comparison point, with the much larger and more clinically significant gaps observed in NDCs. These findings motivate the investigation of shared, transdiagnostic factors that may impact the IQ-adaptive functioning gap. Our univariate and multivariate analyses indicate that IQ significantly contributes to the cognitive-adaptive gap, with individuals possessing higher IQ scores more likely to experience a higher gap between their cognitive abilities and adaptive functioning. This finding is consistent with previous studies reporting that the IQ-adaptive functioning gap is more pronounced in individuals with higher IQ levels (McQuaid et al., 2021 ; Wang et al., 2023 ; Zukerman et al., 2021 ). From a measurement perspective, this may be related to the fact that at higher IQ levels, a wider range of discrepancies is numerically possible, whereas both our IQ and ABAS measures had minimum values (floor) at 40. Consistent with previous literature, we observed that IQ alone does not fully account for adaptive functioning, particularly in neurodivergent individuals (Wang et al., 2023 ). Specifically, our findings suggest that in addition to IQ, phenotypic features including social communication difficulties, inattentive features and hyperactivity, and internalizing and externalizing symptoms are associated with increases in cognitive-adaptive functioning gap. This is not surprising given the disabling impact of these features on adaptive abilities (Kraper et al., 2017 ; Tillmann et al., 2019 ; Zukerman et al., 2021 ). Interestingly, participants with subthreshold scores (e.g., SCQ scores ≥ 10 or SWAN-inattention scores ≥ 3) exhibited increases in the IQ-ABAS gap. This suggests that even subthreshold traits in these domains are linked to increased gaps between IQ and adaptive functioning Sociodemographic Differences While phenotypic features emerged as the strongest correlates of cognitive-adaptive functioning gap, sociodemographic factors including age, sex, household income, and maternal education also showed notable but smaller associations. Our findings indicated that the cognitive-adaptive functioning gap was larger among older individuals across multiple NDCs. Although our study is cross-sectional, this finding aligns with previous longitudinal research reporting a similar pattern of increasing gaps with age (Bradshaw et al., 2019 ; Wang et al., 2023 ). This trend may be explained by the difference resulting from the stable cognitive abilities of individuals with autism and the decrease in adaptive functioning with age. Specifically, older individuals face increasing expectations for independence in socialization, communication, and daily living skills, which may expose challenges in these areas that are not compensated for by cognitive strengths, thereby widening the gap (Bradshaw et al., 2019 ). Our results also revealed that female participants exhibited a decreased cognitive-adaptive functioning gap compared to males, consistent with previous findings indicating sex-specific differences in the cognitive-adaptive functioning gap (McQuaid et al., 2021 ). Although the contribution of sex to the cognitive-adaptive functioning gap predictions was lower than those for phenotypic features, this modest effect may reflect behavioral masking in females, thereby contributing slightly and negatively to the cognitive-adaptive functioning gap. Masking refers to compensatory behaviors, such as imitating social norms or rehearsing interactions, which may elevate adaptive functioning scores despite underlying broader challenges. This observation is more commonly reported in females and could lead to an attenuated gap by boosting measured adaptive functioning relative to cognitive performance (Dean et al., 2017 ; Lai et al., 2011 ; McQuaid et al., 2021 ). Income, education, and race also showed small contributions to the cognitive-adaptive functioning gap. Lower-income groups were associated with a slightly higher cognitive-adaptive functioning gap, and higher levels of caregiver education were linked to a lower cognitive-adaptive functioning gap. These findings align with prior research suggesting that socioeconomic resources may be linked to differences in adaptive functioning outcomes. For example, previous studies have reported associations between parental education and income and the accessibility of supportive interventions, which could be related to developmental outcomes in children with autism (Bradshaw et al., 2019 ; Wang et al., 2023 ). It is important to note that most sociodemographic variables in our study were dichotomized, which may have reduced their variability and weakened their predictive power. However, this does not fully account for their relatively low importance in our model. For example, age, which was analyzed as a continuous variable, also demonstrated low importance, suggesting that sociodemographic variables as a group may have limited predictive power for the cognitive-adaptive functioning gap in this context. Future research could examine whether alternative variable representations, such as using continuous measures instead of dichotomized ones, yield different results. Although these features were not primary drivers of the cognitive-adaptive functioning gap in our model, these results motivate future avenues for further investigation into how these differences interact with the cognitive-adaptive functioning gap. Interaction Effects Between Phenotypic Features In our study, sociodemographic features (age, sex, household income, race, and first caregiver’s education) showed relatively low SHAP value rankings in predicting cognitive-adaptive functioning gap compared to phenotypic features. Despite their lower overall importance, these features reveal subtle yet significant biases that can influence assessments of the outcomes. This aligns with prior research (Bradshaw et al., 2019 ; McQuaid et al., 2021 ; Tillmann et al., 2019 ), which suggests that while sociodemographic features are not consistent predictors, they can contextually impact outcomes. The strongest interaction effect was observed between cognitive abilities and social communication. Individuals with more social communication challenges exhibited more variability in their interaction with cognitive functioning on the cognitive-adaptive functioning gap, across the scale of FSIQ. Previous studies indicate that cognitive abilities often predict adaptive skills in neurotypical populations, but this association can weaken in the presence of strong social communication differences (i.e. autism) (Kenworthy et al., 2010 ; Klin et al., 2007 ). This suggested that cognitive abilities and greater social communication issues jointly influence the cognitive-adaptive functioning gap inconsistently, and that these features may need to be investigated alongside other traits in the presence of social communication symptoms. For individuals with lower cognitive ability scores, high social communication symptoms contributed to an increase in the cognitive-adaptive functioning gap, indicating that adaptive functioning outcomes are particularly compromised when both cognitive abilities are low and social communication differences are high. Conversely, with higher cognitive abilities, social communication symptoms become less relevant in the contribution of their interaction to the cognitive-adaptive functioning gap; regardless of social communication symptoms, the cognitive-adaptive functioning gap continues to increase with higher cognitive abilities. This pattern is consistent with findings showing that higher cognitive abilities do not consistently result in better adaptive functioning in autistic individuals (Alvares et al., 2020 ). For the interactions between cognitive functioning and inattention, low inattention levels were predominantly observed in individuals with high cognitive abilities, indicating that greater cognitive abilities are associated with fewer inattentive features. Among individuals with higher inattention levels, a quadratic relation emerged with the cognitive-adaptive functioning gap across the range of cognitive abilities. Specifically, those with higher inattention levels and average cognitive functioning contributed to an increase in the cognitive-adaptive functioning gap. In contrast, individuals with higher inattention levels but either low or high cognitive functioning contributed to a decrease in the gap. For individuals with lower inattention levels, a linear relation was observed, with contributions to the cognitive-adaptive functioning gap increasing steadily with cognitive abilities. This pattern indicates a consistent amplifying effect of cognitive functioning on the gap in this subgroup. The interaction between cognitive functioning and inattention continued to drive an increased cognitive-adaptive functioning gap in the low inattention group, supporting that high cognitive abilities alone are insufficient to guarantee positive adaptive functioning outcomes (Alvares et al., 2020 ). Limitations Three key limitations should be considered in interpreting the results of this study. First, the use of broad categories of sociodemographic and race may have impacted our ability to detect the impacts of these variables on the cognitive-adaptive functioning gap. Second, the study’s cross-sectional design limited our capacity to examine developmental changes in the cognitive-adaptive functioning gap. Third, while the 5-fold cross-validation strategy maximized the use of available data and ensured robust performance estimates, the absence of an independent test set may limit the assessment of generalization to entirely unseen datasets. Future longitudinal studies with larger sample sizes could address these limitations by incorporating external validation or independent test sets to further validate model performance and clarify developmental trajectories. Conclusion This study identifies that cognitive-adaptive functioning gap was associated with higher FSIQ, social-communication challenges, and inattentive traits across diagnostic groups, with phenotypic features showing stronger connections than sociodemographic features. Sociodemographic factors, including sex and age, were also related to the gap, reflecting the diverse influences on cognitive and adaptive functioning. These findings improve our understanding of the cognitive-adaptive gap in neurodivergent populations and suggest that focusing support on key associated factors, such as cognitive abilities, social-communication challenges, and inattentive traits, may help to reduce this gap. Given the cross-sectional nature of the study, further longitudinal research is needed to examine how these gaps and their associated factors change over time and to explore the potential for interventions targeting these factors. Abbreviations NDCs: Neurodevelopmental Conditions; ADHD: Attention-Deficit/Hyperactivity Disorder; OCD: Obsessive-Compulsive Disorder; POND: Province of Ontario Neurodevelopmental Network; FSIQ: Full-Scale Intelligence Quotient; ABAS-II: Adaptive Behavior Assessment System Second Edition; GAC: General Adaptive Composite; SCQ: Social Communication Questionnaire; SWAN: Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale; TOCS: Toronto Obsessive-Compulsive Scale; CBCL: Child Behavior Checklist; SHAP: SHapley Additive exPlanations; OLS: Ordinary Least Squares; ANOVA: Analysis of Variance; MAE: Mean Absolute Error; DSM-5-TR: Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision; WASI-II: Wechsler Abbreviated Scale of Intelligence, Second Edition; WISC-IV/V: Wechsler Intelligence Scale for Children, Fourth/Fifth Editions; WPPSI-IV: Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition; ADOS-2: Autism Diagnostic Observation Schedule-2; ADI-R: Autism Diagnostic Interview-Revised; PICS: Parent Interview for Child Symptoms; K-SADS: Kiddie-Schedule for Affective Disorders and Schizophrenia; CYBOCS: Children’s Yale-Brown Obsessive Compulsive Scale; ID: Intellectual Disability; Tukey’s HSD: Tukey’s Honest Significant Difference Declarations Acknowledgements We gratefully acknowledge the financial support of the Canadian Institutes of Health Research (CIHR), which made this research possible. Authors’ contributions EW designed the study, conducted data analysis, interpreted the results, and drafted the manuscript. AK supervised EW, provided input on and co-interpreted the results. JPL, EA, ML, EK, RS, and RN reviewed and approved the final manuscript. Funding This research was supported by the Canadian Institutes of Health Research (CIHR), Availability of data and materials Data from POND is available through an Ontario Brain Institute’s controlled data release through Brain-CODE by request. Ethics approval and consent to participate The Research Ethics Board at each participating site approved the POND study protocol. Participants provided informed consent when able, while those who could not provided assent, and their caregiver gave consent. Competing interests A. Kushki and E. Anagnostou have a patent for hollyTM (formerly Anxiety Meter) with royalties paid from Awake Labs. A. Kushki has received consulting fees from DNAStack and Shaftesbury. E. Anagnostou has received grants from Roche and Anavex, served as a consultant to Roche, Quadrant Therapeutics, Ono, and Impel Pharmaceuticals, has received in-kind support from AMO Pharma and CRA-Simons Foundation, received royalties from APPI and Springer, received an editorial honorarium from Wiley, and has a patent for hollyTM (formerly Anxiety Meter). R. Nicolson received grant funding from Hoffman - La Roche Limited and MapLight Therapeutics. The remaining authors have reported no financial interests or potential conflicts of interest. Author details 1 Institute of Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada. 2 Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom. 3 Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada. 4 Autism Research Centre, Bloorview Research Institute, Holland Bloorview Kids Rehabilitation Hospital, Toronto, Ontario, Canada. 5 Departments of Psychology and Psychiatry, Queen's University, Kingston, Ontario, Canada. 6 Department of Psychiatry, Neurosciences and Mental Health, The Hospital for Sick Children, Toronto, Ontario, Canada. 7 Department of Psychiatry, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada. 8 Department of Psychiatry, University of Western Ontario, London, Ontario, Canada References Abiodun, O. I., Jantan, A., Omolara, A. E., Dada, K. V., Mohamed, N. A., & Arshad, H. (2018). State-of-the-art in artificial neural network applications: A survey. Heliyon , 4 (11), e00938. https://doi.org/10.1016/j.heliyon.2018.e00938 Achenbach, T. M., & Edelbrock, C. S. (1983). Manual for the child behavior checklist and revised child behavior profile. (No Title) . Ahuja, A., Martin, J., Langley, K., & Thapar, A. (2013). Intellectual Disability in Children with Attention Deficit Hyperactivity Disorder. The Journal of Pediatrics , 163 (3), 890-895.e1. https://doi.org/10.1016/j.jpeds.2013.02.043 Alvares, G. A., Bebbington, K., Cleary, D., Evans, K., Glasson, E. J., Maybery, M. T., Pillar, S., Uljarević, M., Varcin, K., Wray, J., & Whitehouse, A. J. (2020). The misnomer of ‘high functioning autism’: Intelligence is an imprecise predictor of functional abilities at diagnosis. Autism , 24 (1), 221–232. https://doi.org/10.1177/1362361319852831 Ameis, S. H., Lerch, J. P., Taylor, M. J., Lee, W., Viviano, J. D., Pipitone, J., Nazeri, A., Croarkin, P. E., Voineskos, A. N., Lai, M.-C., Crosbie, J., Brian, J., Soreni, N., Schachar, R., Szatmari, P., Arnold, P. D., & Anagnostou, E. (2016). A Diffusion Tensor Imaging Study in Children With ADHD, Autism Spectrum Disorder, OCD, and Matched Controls: Distinct and Non-Distinct White Matter Disruption and Dimensional Brain-Behavior Relationships. American Journal of Psychiatry , 173 (12), 1213–1222. https://doi.org/10.1176/appi.ajp.2016.15111435 American Psychiatric Association (Ed.). (2022). Diagnostic and statistical manual of mental disorders, Fifth Edition, Text Revision (DSM-5-TR) (5th ed). American Psychiatric Association. https://psychiatryonline.org/doi/book/10.1176/appi.books.9780890425787 Arashi, M., Roozbeh, M., Hamzah, N. A., & Gasparini, M. (2021). Ridge regression and its applications in genetic studies. PLOS ONE , 16 (4), e0245376. https://doi.org/10.1371/journal.pone.0245376 Berument, S. K., Rutter, M., Lord, C., Pickles, A., & Bailey, A. (1999). Autism screening questionnaire: Diagnostic validity. British Journal of Psychiatry , 175 (5), 444–451. https://doi.org/10.1192/bjp.175.5.444 Billeiter, K. B., & Froiland, J. M. (2023). Diversity of Intelligence is the Norm Within the Autism Spectrum: Full Scale Intelligence Scores Among Children with ASD. Child Psychiatry & Human Development , 54 (4), 1094–1101. https://doi.org/10.1007/s10578-021-01300-9 Bradshaw, J., Gillespie, S., Klaiman, C., Klin, A., & Saulnier, C. (2019). Early emergence of discrepancy in adaptive behavior and cognitive skills in toddlers with autism spectrum disorder. Autism , 23 (6), 1485–1496. https://doi.org/10.1177/1362361318815662 Breiman, L. (2001). Random Forests. Machine Learning , 45 (1), 5–32. https://doi.org/10.1023/A:1010933404324 Charman, T., Pickles, A., Simonoff, E., Chandler, S., Loucas, T., & Baird, G. (2011). IQ in children with autism spectrum disorders: Data from the Special Needs and Autism Project (SNAP). Psychological Medicine , 41 (3), 619–627. https://doi.org/10.1017/S0033291710000991 Dean, M., Harwood, R., & Kasari, C. (2017). The art of camouflage: Gender differences in the social behaviors of girls and boys with autism spectrum disorder. Autism , 21 (6), 678–689. https://doi.org/10.1177/1362361316671845 Duncan, A. W., & Bishop, S. L. (2015). Understanding the gap between cognitive abilities and daily living skills in adolescents with autism spectrum disorders with average intelligence. Autism , 19 (1), 64–72. https://doi.org/10.1177/1362361313510068 Friedman, J. H. (2002). Stochastic gradient boosting. Computational Statistics & Data Analysis , 38 (4), 367–378. https://doi.org/10.1016/S0167-9473(01)00065-2 Jacobs, G. R., Voineskos, A. N., Hawco, C., Stefanik, L., Forde, N. J., Dickie, E. W., Lai, M.-C., Szatmari, P., Schachar, R., Crosbie, J., Arnold, P. D., Goldenberg, A., Erdman, L., & Ameis, S. H. (2021). Integration of brain and behavior measures for identification of data-driven groups cutting across children with ASD, ADHD, or OCD. Neuropsychopharmacology , 46 (3), 643–653. https://doi.org/10.1038/s41386-020-00902-6 Kanne, S., Gerber, A. J., Quirmbach, L. M., Sparrow, S. S., Cicchetti, D. V., & Saulnier, C. (2011). The Role of Adaptive Behavior in Autism Spectrum Disorders: Implications for Functional Outcome . Kenworthy, L., Case, L., Harms, M. B., Martin, A., & Wallace, G. L. (2010). Adaptive Behavior Ratings Correlate With Symptomatology and IQ Among Individuals With High-Functioning Autism Spectrum Disorders. Journal of Autism and Developmental Disorders , 40 (4), 416–423. https://doi.org/10.1007/s10803-009-0911-4 Klin, A., Saulnier, C. A., Sparrow, S. S., Cicchetti, D. V., Volkmar, F. R., & Lord, C. (2007). Social and Communication Abilities and Disabilities in Higher Functioning Individuals with Autism Spectrum Disorders: The Vineland and the ADOS. Journal of Autism and Developmental Disorders , 37 (4), 748–759. https://doi.org/10.1007/s10803-006-0229-4 Knopf, A. (2020). Autism prevalence increases from 1 in 60 to 1 in 54: CDC. The Brown University Child and Adolescent Behavior Letter , 36 (6), 4–4. https://doi.org/10.1002/cbl.30470 Kraper, C. K., Kenworthy, L., Popal, H., Martin, A., & Wallace, G. L. (2017). The Gap Between Adaptive Behavior and Intelligence in Autism Persists into Young Adulthood and is Linked to Psychiatric Co-morbidities. Journal of Autism and Developmental Disorders , 47 (10), 3007–3017. https://doi.org/10.1007/s10803-017-3213-2 Kuhn & Max. (2008). Building Predictive Models in R Using the caret Package. Journal of Statistical Software , 28 (5), 1–26. https://doi.org/10.18637/jss.v028.i05 Lai, M.-C., Lombardo, M. V., Pasco, G., Ruigrok, A. N. V., Wheelwright, S. J., Sadek, S. A., Chakrabarti, B., MRC AIMS Consortium, & Baron-Cohen, S. (2011). A Behavioral Comparison of Male and Female Adults with High Functioning Autism Spectrum Conditions. PLoS ONE , 6 (6), e20835. https://doi.org/10.1371/journal.pone.0020835 Loh, H. W., Ooi, C. P., Seoni, S., Barua, P. D., Molinari, F., & Acharya, U. R. (2022). Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011–2022). Computer Methods and Programs in Biomedicine , 226 , 107161. https://doi.org/10.1016/j.cmpb.2022.107161 Lundberg, S. M., Erion, G., Chen, H., DeGrave, A., Prutkin, J. M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., & Lee, S.-I. (2020). From local explanations to global understanding with explainable AI for trees. Nature Machine Intelligence , 2 (1), 56–67. https://doi.org/10.1038/s42256-019-0138-9 Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems , 30 . Mahendiran, T., Dupuis, A., Crosbie, J., Georgiades, S., Kelley, E., Liu, X., Nicolson, R., Schachar, R., Anagnostou, E., & Brian, J. (2019). Sex Differences in Social Adaptive Function in Autism Spectrum Disorder and Attention-Deficit Hyperactivity Disorder. Frontiers in Psychiatry , 10 , 607. https://doi.org/10.3389/fpsyt.2019.00607 McQuaid, G. A., Pelphrey, K. A., Bookheimer, S. Y., Dapretto, M., Webb, S. J., Bernier, R. A., McPartland, J. C., Van Horn, J. D., & Wallace, G. L. (2021). The gap between IQ and adaptive functioning in autism spectrum disorder: Disentangling diagnostic and sex differences. Autism , 25 (6), 1565–1579. https://doi.org/10.1177/1362361321995620 Molnar, C. (2020). Interpretable Machine Learning . Mouga, S., Almeida, J., Café, C., Duque, F., & Oliveira, G. (2015). Adaptive Profiles in Autism and Other Neurodevelopmental Disorders. Journal of Autism and Developmental Disorders , 45 (4), 1001–1012. https://doi.org/10.1007/s10803-014-2256-x Park, L. S., Burton, C. L., Dupuis, A., Shan, J., Storch, E. A., Crosbie, J., Schachar, R. J., & Arnold, P. D. (2016). The Toronto Obsessive-Compulsive Scale: Psychometrics of a Dimensional Measure of Obsessive-Compulsive Traits. Journal of the American Academy of Child & Adolescent Psychiatry , 55 (4), 310-318.e4. https://doi.org/10.1016/j.jaac.2016.01.008 Perry, A., Flanagan, H. E., Dunn Geier, J., & Freeman, N. L. (2009). Brief Report: The Vineland Adaptive Behavior Scales in Young Children with Autism Spectrum Disorders at Different Cognitive Levels. Journal of Autism and Developmental Disorders , 39 (7), 1066–1078. https://doi.org/10.1007/s10803-009-0704-9 Raiford, S. E. (2018). The Wechsler Intelligence Scale for Children—Fifth Edition Integrated. In Contemporary intellectual assessment: Theories, tests, and issues, 4th ed. (pp. 303–332). The Guilford Press. Roid, G. H., & Pomplun, M. (2012). The Stanford-Binet Intelligence Scales, Fifth Edition. In Contemporary intellectual assessment: Theories, tests, and issues, 3rd ed. (pp. 249–268). The Guilford Press. Root, A., Brown, J. P., Forbes, H. J., Bhaskaran, K., Hayes, J., Smeeth, L., & Douglas, I. J. (2019). Association of Relative Age in the School Year With Diagnosis of Intellectual Disability, Attention-Deficit/Hyperactivity Disorder, and Depression. JAMA Pediatrics , 173 (11), 1068. https://doi.org/10.1001/jamapediatrics.2019.3194 Scandurra, V., Emberti Gialloreti, L., Barbanera, F., Scordo, M. R., Pierini, A., & Canitano, R. (2019). Neurodevelopmental Disorders and Adaptive Functions: A Study of Children With Autism Spectrum Disorders (ASD) and/or Attention Deficit and Hyperactivity Disorder (ADHD). Frontiers in Psychiatry , 10 , 673. https://doi.org/10.3389/fpsyt.2019.00673 Shenouda, J., Barrett, E., Davidow, A. L., Sidwell, K., Lescott, C., Halperin, W., Silenzio, V. M. B., & Zahorodny, W. (2023). Prevalence and Disparities in the Detection of Autism Without Intellectual Disability. Pediatrics , 151 (2), e2022056594. https://doi.org/10.1542/peds.2022-056594 Smola, A. J., & Schölkopf, B. (2004). A tutorial on support vector regression. Statistics and Computing , 14 (3), 199–222. https://doi.org/10.1023/B:STCO.0000035301.49549.88 Song, Y., & Lu, Y. (2015). Decision tree methods: Applications for classification and prediction . 27 (2). Swanson, J. M., Schuck, S., Porter, M. M., Carlson, C., Hartman, C. A., Sergeant, J. A., Clevenger, W., Wasdell, M., McCleary, R., Lakes, K., & Wigal, T. (2015). Categorical and Dimensional Definitions and Evaluations of Symptoms of ADHD: History of the SNAP and the SWAN Rating Scales . Tibshirani, R. (1996). Regression Shrinkage and Selection Via the Lasso. Journal of the Royal Statistical Society: Series B (Methodological) , 58 (1), 267–288. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x Tillmann, J., San José Cáceres, A., Chatham, C. H., Crawley, D., Holt, R., Oakley, B., Banaschewski, T., Baron‐Cohen, S., Bölte, S., Buitelaar, J. K., Durston, S., Ham, L., Loth, E., Simonoff, E., Spooren, W., Murphy, D. G., Charman, T., the EU‐AIMS LEAP group, Ahmad, J., … Zwiers, M. P. (2019). Investigating the factors underlying adaptive functioning in autism in the EU‐AIMS Longitudinal European Autism Project. Autism Research , 12 (4), 645–657. https://doi.org/10.1002/aur.2081 Wang, S., Zhou, Q., Chen, K., Ceng, C., Zhan, G., You, C., Xing, Y., Zou, Y., & Deng, H. (2023). Cognitive-adaptive Functioning Gap and Mediating Factors that Impact Adaptive Functioning in Chinese Preschool-aged Children with Autism Spectrum Disorder. Journal of Autism and Developmental Disorders . https://doi.org/10.1007/s10803-023-06029-z Wechsler, D. (2003). WISC-IV: Wechsler Intelligence Scale for Children–fourth edition. In Technical and interpretive manual . TX Psychol Corp San Antonio. Wechsler, D. & Psychological Corporation. (2012). WPPSI-IV : Wechsler preschool and primary scale of intelligence—Fourth edition . 1 technical and interpretive manual (xi, 256 pages ; 28 cm), 1 administration and scoring manual (xi, 331 pages 24 cm), 3 stimulus books, 1 box of 14 red and white blocks, 1 box of 13 cardboard object assembly puzzles, 1 box zoo locations card set with ink dauber, 1 extra ink dauber, 1 folder of 3 zoo locations layouts, 3 response booklets (Bug search, Cancellation, Animal coding), 2 record forms (ages 2:6-3:11, ages 4:0-7:7), 3 bug search scoring key cards, 1 animal coding scoring key, 1 cancellation scoring template ; in suitcase 50 x 37 x 26 cm. WorldCat. Wechsler, D., Zhou, X., Psychological Corporation., & Assessment Library Materials (University of Lethbridge. Faculty of Education. Curriculum Laboratory). (2011). WASI-II : Wechsler abbreviated scale of intelligence. Wechsler Abbreviated Scale of Intelligence . Zeidan, J., Fombonne, E., Scorah, J., Ibrahim, A., Durkin, M. S., Saxena, S., Yusuf, A., Shih, A., & Elsabbagh, M. (2022). Global prevalence of autism: A systematic review update. Autism Research , 15 (5), 778–790. https://doi.org/10.1002/aur.2696 Zou, H., & Hastie, T. (2005). Regularization and Variable Selection Via the Elastic Net. Journal of the Royal Statistical Society Series B: Statistical Methodology , 67 (2), 301–320. https://doi.org/10.1111/j.1467-9868.2005.00503.x Zukerman, G., Yahav, G., & Ben-Itzchak, E. (2021). The Gap Between Cognition and Adaptive Behavior in Students with Autism Spectrum Disorder: Implications for Social Anxiety and the Moderating Effect of Autism Traits. Journal of Autism and Developmental Disorders , 51 (5), 1466–1478. https://doi.org/10.1007/s10803-020-04632-y Additional Declarations Competing interest reported. A. Kushki and E. Anagnostou have a patent for hollyTM (formerly Anxiety Meter) with royalties paid from Awake Labs. A. Kushki has received consulting fees from DNAStack and Shaftesbury. E. Anagnostou has received grants from Roche and Anavex, served as a consultant to Roche, Quadrant Therapeutics, Ono, and Impel Pharmaceuticals, has received in-kind support from AMO Pharma and CRA-Simons Foundation, received royalties from APPI and Springer, received an editorial honorarium from Wiley, and has a patent for hollyTM (formerly Anxiety Meter). R. Nicolson received grant funding from Hoffman - La Roche Limited and MapLight Therapeutics. The remaining authors have reported no financial interests or potential conflicts of interest. Supplementary Files SupplementaryMaterials.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5968118","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":412550420,"identity":"42489809-8f8e-4809-ad1f-b9f03766bcc3","order_by":0,"name":"Eric Wan","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Eric","middleName":"","lastName":"Wan","suffix":""},{"id":412550421,"identity":"a25e8850-9069-4c70-92c5-2cd14b5b2774","order_by":1,"name":"Jason P. Lerch","email":"","orcid":"","institution":"University of Oxford","correspondingAuthor":false,"prefix":"","firstName":"Jason","middleName":"P.","lastName":"Lerch","suffix":""},{"id":412550422,"identity":"9c2e1173-9b98-4479-a6fa-10d06ce5fc03","order_by":2,"name":"Evdokia Anagnostou","email":"","orcid":"","institution":"Holland Bloorview Kids Rehabilitation Hospital","correspondingAuthor":false,"prefix":"","firstName":"Evdokia","middleName":"","lastName":"Anagnostou","suffix":""},{"id":412550423,"identity":"e762759b-b0a5-4075-868d-896e033eb466","order_by":3,"name":"Melanie Penner","email":"","orcid":"","institution":"Holland Bloorview Kids Rehabilitation Hospital","correspondingAuthor":false,"prefix":"","firstName":"Melanie","middleName":"","lastName":"Penner","suffix":""},{"id":412550424,"identity":"42dbbba1-388a-4527-81e0-cccb749172e0","order_by":4,"name":"Elizabeth Kelley","email":"","orcid":"","institution":"Queen's University","correspondingAuthor":false,"prefix":"","firstName":"Elizabeth","middleName":"","lastName":"Kelley","suffix":""},{"id":412550425,"identity":"cd352937-067f-4556-8368-5749936157a2","order_by":5,"name":"Russell Schachar","email":"","orcid":"","institution":"Hospital for Sick Children","correspondingAuthor":false,"prefix":"","firstName":"Russell","middleName":"","lastName":"Schachar","suffix":""},{"id":412550426,"identity":"b74e3acf-6f34-471c-9bba-5ef225da2119","order_by":6,"name":"Rob Nicolson","email":"","orcid":"","institution":"Western University","correspondingAuthor":false,"prefix":"","firstName":"Rob","middleName":"","lastName":"Nicolson","suffix":""},{"id":412550429,"identity":"ff88a544-9474-4c93-a24a-14c9847e52a0","order_by":7,"name":"Azadeh Kushki","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFElEQVRIie3RwUrDMBjA8YxCc/m010Bxe4WMwBAmPktLYb2soyB4nBmFnup94EsUPKi3hkB3ydi1R30AodCToGg2EREy59FD/qcQ+PHlIwjZbP8xQMhBFelT5CwqfQ6Od9f0MGEU9fiOuH8kKCx/kl/yCsW6VJ3Gd3jBq5f78czFWd1BehZyLB9NhKyLkb9sSPJQCC6uVXzhQh3dAJ2EHCbm521g5EBLkrIJeXWUyzAnU9ZbUplwYt5osAHWaRJTTcTbN3nXBLcmQtcF9aEhwZbIrymopZUmYJwyVPWlD4oMSyW4PMk/d9EkusphmppIX0W3HdTzAV1l4uk5H888nEkUvJ4zD69K4/r7O/hBNpvNZtvbB7rmZoZmwyeeAAAAAElFTkSuQmCC","orcid":"","institution":"Holland Bloorview Kids Rehabilitation Hospital","correspondingAuthor":true,"prefix":"","firstName":"Azadeh","middleName":"","lastName":"Kushki","suffix":""}],"badges":[],"createdAt":"2025-02-05 18:53:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5968118/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5968118/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":75999464,"identity":"00980e7d-17df-4ca0-b2c9-0800c7f8f2e6","added_by":"auto","created_at":"2025-02-11 10:20:04","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":59596,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eDistribution of the IQ-ABAS gap disaggregated by (A) diagnosis and (B) sociodemographic features and IQ identity. Note: \u003c/em\u003e\u003csup\u003e\u003cem\u003e* \u003c/em\u003e\u003c/sup\u003e\u003cem\u003ep\u0026lt;.05 (subgroup difference)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/0b9342bf471a991585963829.png"},{"id":76001337,"identity":"e5ff6883-acb7-462e-a9d3-77a2f8db8119","added_by":"auto","created_at":"2025-02-11 10:36:04","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":58447,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003ePerformance of the regression models disaggregated by diagnosis (A), and sociodemographic characteristics and intellectual disability (B). The association between predicted and actual values for IQ-ABAS gap in a random forest model (r = 0.94, p \u0026lt; .001), with the dashed line representing the regression and the solid line representing the perfect prediction (C). SHAP importance plot for the random forest model, showing the average absolute SHAP values for each phenotypic and sociodemographic feature. The features are ranked by their importance (D).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/7cc290292a59ac1792833136.png"},{"id":75999467,"identity":"1e33655b-2c9a-4e0e-88bf-5cd09a89a4ab","added_by":"auto","created_at":"2025-02-11 10:20:04","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":95857,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eSHAP dependence plots showing the impact of phenotypic features on a random forest regression model's predictions for IQ-ABAS gap (A - G).\u003c/em\u003e\u003cem\u003e\u003cstrong\u003e \u003c/strong\u003e\u003c/em\u003e\u003cem\u003eSHAP dependence plots showing the impact of sociodemographic features on a random forest regression model's predictions for IQ-ABAS gap (H - L)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/fdd69a351e0be9027f2fdb01.png"},{"id":76001008,"identity":"84815d12-7941-4e92-ad3e-0a1418c308d2","added_by":"auto","created_at":"2025-02-11 10:28:04","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":80868,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe main and interaction effects based on a random forest regression model, with the mean absolute SHAP interaction values (A). The SHAP interaction dependence plots with interactions between FSIQ and SCQ (B), and between FSIQ and inattention (C). \u003c/em\u003ePlots are visualized with two subgroups: those with scores above (orange) and below (purple) the median.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/5dc856346614d2a074695292.png"},{"id":78716570,"identity":"0d8787d7-993c-46d1-9355-795fbda5c1b2","added_by":"auto","created_at":"2025-03-18 03:16:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1249997,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/006039cc-6429-4911-94ea-66b346f6e6be.pdf"},{"id":75999465,"identity":"1a993b45-1544-4655-9b20-fe8c240b5a69","added_by":"auto","created_at":"2025-02-11 10:20:04","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":39250,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-5968118/v1/ca6d0d2284b6f426c54927dd.docx"}],"financialInterests":"Competing interest reported. A. Kushki and E. Anagnostou have a patent for hollyTM (formerly Anxiety Meter) with royalties paid from Awake Labs. A. Kushki has received consulting fees from DNAStack and Shaftesbury. E. Anagnostou has received grants from Roche and Anavex, served as a consultant to Roche, Quadrant Therapeutics, Ono, and Impel Pharmaceuticals, has received in-kind support from AMO Pharma and CRA-Simons Foundation, received royalties from APPI and Springer, received an editorial honorarium from Wiley, and has a patent for hollyTM (formerly Anxiety Meter). R. Nicolson received grant funding from Hoffman - La Roche Limited and MapLight Therapeutics. The remaining authors have reported no financial interests or potential conflicts of interest.","formattedTitle":"Exploring Phenotypic and Sociodemographic Influences on Cognitive-Adaptive Functioning Gap in Neurodivergent Children","fulltext":[{"header":"Introduction","content":"\u003cp\u003eNeurodevelopmental conditions (NDCs), including autism, attention-deficit/hyperactivity disorder (ADHD), and obsessive-compulsive disorder (OCD) are characterized by differences in brain development that can impact cognitive and adaptive functioning. Cognitive functioning, often measured via the intelligence quotient (IQ), entails mental processes such as perception, attention, memory, language, problem-solving, and reasoning (American Psychiatric Association, 2022). Adaptive functioning quantifies an individual's capacity to perform daily tasks, adapt to environmental changes, and meet the demands of everyday life (American Psychiatric Association, 2022). Cognitive functioning varies widely in NDCs. For example, among autistic children, an estimated 33% have an intellectual disability (ID) (Knopf, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Shenouda et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zeidan et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), while approximately 13.7% score in the superior intelligence range (Billeiter \u0026amp; Froiland, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). ADHD, OCD, and other NDCs can also co-occur with ID (Ahuja et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Mouga et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Root et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Adaptive functioning is also highly varied in NDCs. Autism has been associated with decreased adaptive functioning relative to other NDCs, although large variation in these skills exist (Ameis et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Jacobs et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Kenworthy et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Mahendiran et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Mouga et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Perry et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Specific domains of adaptive functioning, such as socialization, communication, and daily living skills, tend to be more impaired in children with autism, particularly those with co-occurring autism and ADHD, compared to children with ADHD, who exhibit more pronounced difficulties in motor adaptive functioning (Scandurra et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor the general population, adaptive behavior and cognitive ability are assumed to be positively associated, however, in NDCs, a significant gap between cognitive and adaptive functioning may exist (Charman et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Duncan \u0026amp; Bishop, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Kanne et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Klin et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Mouga et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Tillmann et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This gap often manifests as lower adaptive functioning relative to cognitive abilities and tends to persist from toddlerhood through adolescence and young adulthood, though its trajectory beyond this stage remains unclear (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Kanne et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Kraper et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Despite the established presence of the cognitive-adaptive functioning gap, the underlying phenotypic and sociodemographic factors to this gap are still not fully clarified.\u003c/p\u003e \u003cp\u003eThe associations between phenotypic and sociodemographic features and the cognitive-adaptive functioning gap in neurodevelopmental conditions remain complex and inconclusive. While social-communicative symptoms have been consistently associated with larger gaps (Kanne et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Tillmann et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), findings regarding restricted and repetitive behaviors (RRBs) and mental health challenges like anxiety and depression are mixed. One study reported no significant associations between the gap and anxiety or RRBs (Tillmann et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) whereas another study identified some correlations between the gap and anxiety, and between the gap and RRBs (Kraper et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Sociodemographic factors further contribute to this complexity, as evidence suggests the gap widens with age (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), with sex-specific patterns showing larger gap in females than gap in males over time (McQuaid et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Also, higher maternal education is associated with better adaptive functioning, whereas household income does not show an association with it (Wang et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). These mixed findings presented a research gap in understanding the phenotypic and sociodemographic correlates of cognitive-adaptive functioning gap.\u003c/p\u003e \u003cp\u003eThe present study aims to fill this research gap by systematically evaluating the phenotypic and sociodemographic correlates of the cognitive-adaptive functioning gap in neurodivergent children and youth. By leveraging computational models, our primary objective is to identify and evaluate the features most associated with the cognitive-adaptive functioning gap and to demonstrate how machine learning approaches can enhance our understanding of these complex relations. To achieve this, we assessed the performance of nine computational models in estimating the cognitive-adaptive functioning gap and analyzed the key associations of phenotypic and demographic features to the model outputs.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eFor this study, we used a subset of the data from the Province of Ontario Neurodevelopmental Network (POND), exported on March 26, 2024. The dataset included 3,380 participants, between ages of 0 and 21, who were neurotypical or neurodivergent (neurotypical\u0026thinsp;=\u0026thinsp;335, autism\u0026thinsp;=\u0026thinsp;1,307, ADHD\u0026thinsp;=\u0026thinsp;1,083, OCD\u0026thinsp;=\u0026thinsp;291, other\u0026thinsp;=\u0026thinsp;353, 4 siblings without a diagnosis, 16 missing diagnoses). The \u0026ldquo;other\u0026rdquo; category included anxiety, sub-threshold ADHD, sub-threshold OCD, Down syndrome, Fragile X, ID only, Rett syndrome, social communication disorders, and Tourette syndrome. Validated assessments were used to support the clinical diagnoses. Specifically, the Autism Diagnostic Observation Schedule-2 (ADOS-2) and Autism Diagnostic Interview-Revised (ADI-R) were used for autism, Parent Interview for Child Symptoms (PICS) was used for ADHD, and the Kiddie-Schedule for Affective Disorders and Schizophrenia (K-SADS) and the Children\u0026rsquo;s Yale-Brown Obsessive Compulsive Scale (CYBOCS) were used for OCD.\u003c/p\u003e \u003cp\u003eA subset of the above dataset was select for the present analysis based on the following inclusion criteria: (1) age between 6 and 21 (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2,281), (2) neurotypical or a diagnosis of autism, ADHD, OCD, or other (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2,277), (3) complete data for the phenotypic measures used in this study (full scale IQ, adaptive functioning, mental health symptoms, and autism, ADHD, and OCD features) (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1,764), and (4) complete data for race, household income, and primary caregiver\u0026rsquo;s education (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;902). The final sample consisted of 902 participants.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMeasures\u003c/h3\u003e\n\u003cp\u003eFull scale IQ was assessed using the Wechsler family of measures, including the Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II) (Wechsler et al., 2011), the Wechsler Intelligence Scale for Children, Fourth (WISC-IV) (Wechsler, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) and Fifth Editions (WISC-V) (Raiford, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) the Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition (WPPSI-IV) (Wechsler \u0026amp; Psychological Corporation, 2012), or the Stanford-Binet Scale (Roid \u0026amp; Pomplun, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), as appropriate for developmental stage.\u003c/p\u003e \u003cp\u003eAdaptive functioning was quantified using the Adaptive Behavior Assessment System (ABAS)-II (Harrison \u0026amp; Oakland, 2003), a parent-report measure with scores along domains of Communication, Community Use, Functional Academics, Home Living, Health/Safety, Leisure, Self-Direction, Social, and Work, through norm-referenced scaled scores. Our analysis used the General Adaptive Composite (GAC) score, representing overall adaptive functioning, encompassing all subscales. The cognitive-adaptive functioning gap was defined as full-scale IQ minus ABAS GAC score.\u003c/p\u003e \u003cp\u003eAutism, ADHD, and OCD features were quantified using the Social Communication Questionnaire (SCQ) (Berument et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1999\u003c/span\u003e), the Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale (SWAN) (Swanson et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), and the Toronto Obsessive-Compulsive Scale (TOCS) (Park et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), respectively. Mental health symptoms were characterized using the Child Behavior Checklist (CBCL), internalizing and externalizing subscales (Achenbach \u0026amp; Edelbrock, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1983\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSociodemographic characteristics included in the study were sex (biological), age, household income, race, and the education of the primary caregiver. The racial categories were Black, East Asian, Indigenous, Latino, Middle Eastern, other, South Asian, Southeast Asian, and White. Household income was categorized as follows: lower than \u003cspan\u003e$\u003c/span\u003e10,000; \u003cspan\u003e$\u003c/span\u003e10,000 to \u003cspan\u003e$\u003c/span\u003e24,999; \u003cspan\u003e$\u003c/span\u003e25,000 to \u003cspan\u003e$\u003c/span\u003e49,999; \u003cspan\u003e$\u003c/span\u003e50,000 to \u003cspan\u003e$\u003c/span\u003e74,999; \u003cspan\u003e$\u003c/span\u003e75,000 to \u003cspan\u003e$\u003c/span\u003e99,999; \u003cspan\u003e$\u003c/span\u003e100,000 to \u003cspan\u003e$\u003c/span\u003e149,999; \u003cspan\u003e$\u003c/span\u003e150,000 to \u003cspan\u003e$\u003c/span\u003e199,999; and \u003cspan\u003e$\u003c/span\u003e200,000 or more. Primary caregiver\u0026rsquo;s education was categorized as lower than Bachelor\u0026rsquo;s degree and Bachelor\u0026rsquo;s degree and higher. In our study, we categorized the household income into three levels (lower than \u003cspan\u003e$\u003c/span\u003e50,000 as low, between \u003cspan\u003e$\u003c/span\u003e50,000 and \u003cspan\u003e$\u003c/span\u003e100,000 as middle, and \u003cspan\u003e$\u003c/span\u003e100,000 or more as high), race as white or minoritized, and education as lower than Bachelor's degree, Bachelor\u0026rsquo;s degree or higher.\u003c/p\u003e\n\u003ch3\u003eStatistical Analyses\u003c/h3\u003e\n\u003cp\u003eOur analyses were conducted using R version 4.3.2 with caret version 6.0\u0026ndash;94 (Kuhn \u0026amp; Max, 2008).\u003c/p\u003e\n\u003ch3\u003eGroup Differences\u003c/h3\u003e\n\u003cp\u003eWe compared differences in sociodemographic and behavioral measures across diagnostic groups using Kruskal-Wallis tests with Bonferroni correction for continuous variables and Chi-squared tests for categorical variables. The \u003cem\u003ep\u003c/em\u003e-values for multiple comparisons were adjusted using the Bonferroni correction. Post-hoc analysis was conducted using Dunn\u0026rsquo;s procedure.\u003c/p\u003e \u003cp\u003eWe assessed sociodemographic and IQ identity group differences in cognitive-adaptive functioning gap using the Kruskal-Wallis test for each feature group (e.g., sociodemographic and phenotypic measures), with Bonferroni-adjusted \u003cem\u003ep\u003c/em\u003e-values used to account for multiple comparisons. For feature groups showing significant differences \u003cem\u003e(p\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), post-hoc pairwise comparisons were performed using the Wilcoxon rank-sum test. Effect sizes were calculated for both tests to quantify the magnitude of group differences.\u003c/p\u003e\n\u003ch3\u003eModeling Cognitive-Adaptive Functioning Gaps\u003c/h3\u003e\n\u003cp\u003eThe gap between cognitive and adaptive functioning (dependent variable) was examined in relation to phenotypic characteristics (IQ, SCQ, SWAN, TOCS, CBCL internalizing and externalizing symptoms) and sociodemographic factors (age, sex, race, first caregiver\u0026rsquo;s education, household income). Nine machine learning regression approaches were used: ordinary least squares (OLS) linear regression, ridge regression (Arashi et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), ElasticNet (Zou \u0026amp; Hastie, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), LASSO (Tibshirani, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e1996\u003c/span\u003e), decision tree (Song \u0026amp; Lu, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), gradient boosting (Friedman, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2002\u003c/span\u003e), support vector regression (Smola \u0026amp; Sch\u0026ouml;lkopf, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), random forest (Breiman, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2001\u003c/span\u003e), and neural networks (Abiodun et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOLS regression was used due to its simplicity and interpretability. Ridge, LASSO, and ElasticNet regression techniques were also considered as they handle the multicollinearity present in the phenotypic and the sociodemographic features better than OLS regression. To address the non-linear relations and interactions among features, which may not be effectively captured by linear models, we incorporated tree-based models. Random forests (Breiman, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) improve upon single decision trees by aggregating the estimations with multiple trees to reduce variance and avoid overfitting, providing robustness especially in the presence of noise. Another tree-based method, Gradient boosting (Friedman, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2002\u003c/span\u003e), uses sequential model building to enhance accuracy. Furthermore, we used kernel support vector regression (Smola \u0026amp; Sch\u0026ouml;lkopf, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) for its ability to model non-linear relations (Abiodun et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), with advantages of flexibility, scalability, and capacity to handle noise.\u003c/p\u003e \u003cp\u003eThe dataset was evaluated using a 5-fold cross-validation strategy to assess model performance and optimize hyperparameters. In this approach, the dataset was split into five folds, with each fold serving as the test set for evaluation in one iteration, while the remaining 80% (four folds) was used for training. Within each fold, the data was split in a stratified manner to ensure proportional representation of key variables, including diagnosis, ABAS GAC (10-point bins), and sex, across the training and test subsets.\u003c/p\u003e \u003cp\u003eModel performance was evaluated using the median absolute error (MAE), with final metrics computed with MAE values across the five folds. This cross-validation strategy ensures that every data point is used for both training and testing, providing a robust estimate of the model\u0026rsquo;s generalization performance. An additional independent test set was not held out due to the limited size of the dataset, as reserving such a set would have significantly reduced the amount of data available for training and validation.\u003c/p\u003e \u003cp\u003eFor statistical comparison of model performance, we conducted a repeated measures ANalysis Of VAriance (ANOVA) to determine if there were significant differences in performance across the models. Given the repeated nature of the measurements, this approach accounts for within-subject variability. When the ANOVA indicated significant differences, we performed post-hoc comparisons using Tukey's HSD test and conducted pairwise comparisons with Bonferroni correction to identify specific models that differed from each other while controlling for family-wise error.\u003c/p\u003e \u003cp\u003eTo assess the effect of individual features and their interactions on the IQ-ABAS gap, SHAP (SHapley Additive exPlanations) values were employed. SHAP, derived from cooperative game theory, provides an interpretable framework for understanding how each feature contributes to the predictions of machine learning models by assigning importance scores (Lundberg \u0026amp; Lee, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Molnar, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Positive SHAP values indicate that a feature increases the model\u0026rsquo;s estimation, whereas negative values indicate a decrease in the estimation. SHAP dependence plots further illustrate the relations between individual feature values and the predicted IQ-ABAS gap, allowing us to capture specific contributions across the feature space and explore potential interactions among features. This method has been widely adopted in healthcare research to explain model estimations and understand feature importance (Loh et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lundberg et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo evaluate the associations between different features and the individual feature\u0026rsquo;s SHAP values, we applied appropriate statistical methods based on the nature of the data. For continuous features (e.g., FSIQ, CBCL scores), we used Spearman\u0026rsquo;s rank correlation to measure the strength and direction of associations, where correlations greater than +/-0.60 were considered strong. For ordinal features (e.g., household income, caregiver education), we employed Kendall\u0026rsquo;s rank correlation, applying the same threshold of τ \u0026gt; |0.60| for strong associations. Finally, for nominal features (e.g., sex, race), we conducted ANOVA to compare group differences in the IQ-ABAS gap. These methods allowed us to explore how each feature type was associated with the observed discrepancies in cognitive and adaptive functioning.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003eSample Characteristics\u003c/h2\u003e\n \u003cp\u003eThe final subset of the POND sample used for the analysis included data from 902 participants (neurotypical\u0026thinsp;=\u0026thinsp;214, autism\u0026thinsp;=\u0026thinsp;409, ADHD\u0026thinsp;=\u0026thinsp;210, OCD\u0026thinsp;=\u0026thinsp;36, Other\u0026thinsp;=\u0026thinsp;33). Detailed sample characteristics are presented in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSample characteristics with median(IQR) for various features, presented as the entire sample and by diagnosis subgroups.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"7\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAutism\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eADHD\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eOCD\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNT\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003en\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e902\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e409\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e210\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e214\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003efemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e265\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e637\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e324\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e147\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e124\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003csup\u003eb\u003c/sup\u003e \u003cstrong\u003e(years)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (6.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (5.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (5.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12 (4.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (5.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (6.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eFSIQ\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100 (27.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93 (35.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98 (19.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e113 (15.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e108 (17.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e79 (45.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSCQ\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20 (12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (9.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (5.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (2.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9 (13)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInattention\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0 (0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eHyperactivity\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.0 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.0 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.5 (6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0 (1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0 (0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.5 (7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTOCS\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-21.0 (50.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-8.00 (37.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-30.5 (49.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e19.5 (19.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-50.0 (43.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-13.0 (48.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCBCL Internal\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60.0 (18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e65.0 (13.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63.0 (16.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e64.5 (10.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e48.0 (12.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63.0 (18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCBCL External\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e54.0 (18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e58.0 (14.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e61.0 (17.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e53.0 (13.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e41.0 (15.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56.0 (13.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eABAS GAC\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e77.0 (33.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63.0 (21.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e78.0 (24.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92.0 (21.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e103 (22.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e72.0 (33.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eIQ-ABAS gap\u003c/strong\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17.5 (29.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e26.0 (33.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e18.0 (23.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20.0 (20.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.50 (22.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4.00 (24.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eIncome\u003c/strong\u003e\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003e\u0026lt; \u003cspan\u003e$\u003c/span\u003e50k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e224\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026gt;=\u003cspan\u003e$\u003c/span\u003e50k and \u0026lt; \u003cspan\u003e$\u003c/span\u003e100k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e227\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e120\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026gt;= \u003cspan\u003e$\u003c/span\u003e100k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e451\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e174\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e149\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eRace\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003ewhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e652\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e283\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e151\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e166\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eminoritized\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e250\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e126\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eEducation\u003c/strong\u003e\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003elower than bachelors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e435\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e208\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e130\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBachelors or higher\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e467\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e201\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e150\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cem\u003eNote\u003c/em\u003e: \u003csup\u003ea\u003c/sup\u003e \u003cem\u003ep\u0026thinsp;\u0026lt;\u0026thinsp;.001 (group effect; with Bonferroni correction)\u003c/em\u003e, \u003csup\u003eb\u003c/sup\u003e\u003cem\u003ep\u0026lt;..05 (group effect; with Bonferroni correction); SCQ: Social Communication Questionnaire (raw scores), TOCS: Toronto Obsessive-Compulsive Scale (raw scores), CBCL: Child Behavior Checklist (T-scores), ABAS2 GAC: Adaptive Behavior Assessment System Second Edition General Adaptive Composite (standard scores).\u003c/em\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eGroup Comparisons\u003c/h3\u003e\n\u003cp\u003eThe distribution of the IQ-ABAS gap is illustrated in Fig. 1, disaggregated by diagnosis and sociodemographic identities. Wilcoxon Signed-Rank tests with the Bonferroni correction were used to test differences of median IQ-ABAS gap from zero. The gap was significantly greater than zero for the pooled sample (median\u0026thinsp;=\u0026thinsp;17.50, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;346163.5, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) as well as each diagnosis group (autism: median\u0026thinsp;=\u0026thinsp;26, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;74899.50, p\u0026thinsp;\u0026lt;\u0026thinsp;.001; ADHD: median\u0026thinsp;=\u0026thinsp;18, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;19599.50, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001; OCD: median\u0026thinsp;=\u0026thinsp;20, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;631.50, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001; NT: median\u0026thinsp;=\u0026thinsp;6.50, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;15446.00, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), except the Other NDCs group (median\u0026thinsp;=\u0026thinsp;4, \u003cem\u003eW\u003c/em\u003e\u0026thinsp;=\u0026thinsp;369.50, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.25).\u003c/p\u003e\n\u003cp\u003eA Kruskal-Wallis test indicated that there was a significant difference in IQ-ABAS gap across the diagnostic groups (Autism: median\u0026thinsp;=\u0026thinsp;26, \u003cem\u003eIQR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;33; ADHD: median\u0026thinsp;=\u0026thinsp;18, \u003cem\u003eIQR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;23.75; OCD: median\u0026thinsp;=\u0026thinsp;20, \u003cem\u003eIQR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;20.75; NT: median\u003csub\u003eT\u003c/sub\u003e=6.5, \u003cem\u003eIQR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;22.75; Other: median\u003csub\u003er\u003c/sub\u003e=4, \u003cem\u003eIQR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;24; \u003cem\u003e\u0026chi;2\u003c/em\u003e(4)\u0026thinsp;=\u0026thinsp;118, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001\u003cem\u003e)\u003c/em\u003e. Post-hoc comparisons using Dunn\u0026rsquo;s Test with Bonferroni correction revealed that the autism group had the highest IQ-ABAS gap, significantly higher than ADHD (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.52, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.002), NT (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;10.40, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), and other NDCs (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;4.68, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) groups. The OCD group had the second-highest gap, significantly higher than both NT (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.80, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) and other NDCs (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.72, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.033). ADHD showed a significantly higher gap than NT (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;5.96, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) and other NDCs (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.93, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.017). However, there were no significant differences between the discrepancies of the ADHD and OCD groups (\u003cem\u003eZ\u003c/em\u003e=-0.59, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1), autism and OCD groups (\u003cem\u003eZ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.11, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1), or between NT and other NDCs group (\u003cem\u003eZ\u003c/em\u003e=-0.165, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1).\u003c/p\u003e\n\u003cp\u003eA Kruskal-Wallis test was performed to evaluate group differences in cognitive-adaptive functioning gap across sociodemographic and IQ identity subgroups. Here, we dichotomized the data with an FSIQ value of 70. The test indicated significant differences for IQ (\u003cem\u003eH\u003c/em\u003e(1)\u0026thinsp;=\u0026thinsp;93.10, \u003cem\u003eadj-p\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=.10; \u0026lt;70: median\u0026thinsp;=\u0026thinsp;0; \u0026gt;=70: median\u0026thinsp;=\u0026thinsp;20), with the lower IQ group having less of a gap than the higher IQ group, and sex (\u003cem\u003eH\u003c/em\u003e(1)\u0026thinsp;=\u0026thinsp;12.70, \u003cem\u003eadj-p\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.0017, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=.01, female: median\u0026thinsp;=\u0026thinsp;13, male: median\u0026thinsp;=\u0026thinsp;19), with females having less of a gap than the males, but not for education (\u003cem\u003eH\u003c/em\u003e(1)\u0026thinsp;=\u0026thinsp;.84, \u003cem\u003eadj-p\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=0), household income (\u003cem\u003eH\u003c/em\u003e(2)\u0026thinsp;=\u0026thinsp;7.50, \u003cem\u003eadj-p\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.12, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=0), and race (\u003cem\u003eH\u003c/em\u003e(1)\u0026thinsp;=\u0026thinsp;1.10, \u003cem\u003eadj-p\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=0).\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eModel Performance\u003c/h2\u003e\n \u003cp\u003eThe accuracies of nine regression models in estimating the IQ-ABAS gap are presented graphically in Fig.\u0026nbsp;2 (A and B) and detailed numerically in Supplementary Tables\u0026nbsp;8 and 9. The results were obtained using the test folds from the 5-fold cross-validation process described in the Methods section. Performance metrics (e.g., MAE) were averaged across the five folds. The results are disaggregated by sex, race, socioeconomic status, and IQ subgroups.\u003c/p\u003e\n \u003cp\u003eThe repeated measures ANOVA showed a statistically significant difference in performance across the nine models (\u003cem\u003eF\u003c/em\u003e(8, 175)\u0026thinsp;=\u0026thinsp;20.00, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, \u003cem\u003e\u0026eta;\u0026sup2;\u003c/em\u003e=.82). Post-hoc Tukey\u0026apos;s HSD analysis demonstrated that random forest significantly outperformed all other models on MAEs (Mean\u0026thinsp;=\u0026thinsp;4.14, SD\u0026thinsp;=\u0026thinsp;.25), with differences ranging from gradient boosting (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, 95% C.I.=[1.95, 5.25]) to ridge regression (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, 95% C.I.=[3.38, 6.69]). Detailed pairwise comparisons are provided in Supplementary Table\u0026nbsp;7.\u003c/p\u003e\n \u003cp\u003eThe random forest model demonstrated strong predictive power [\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.88, \u003cem\u003eF\u003c/em\u003e(1, 175)\u0026thinsp;=\u0026thinsp;1236, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001], indicating that it explained 87.60% of the variance in actual cognitive-adaptive functioning gap values (Fig.\u0026nbsp;2 (C)). It suggests a high correlation between predicted and actual values, so the model can reliably predict the cognitive-adaptive functioning gap. The significant intercept [\u003cem\u003eB\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.25, SE\u0026thinsp;=\u0026thinsp;0.65, \u003cem\u003et\u003c/em\u003e(175)\u0026thinsp;=\u0026thinsp;4.97, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001] (Fig.\u0026nbsp;2 (C)). suggests a consistent bias in estimates across all values. This model has been used for SHAP analysis in the subsequent results.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003ePredictive Features\u003c/h2\u003e\n \u003cp\u003eThe SHAP importance plot (Fig.\u0026nbsp;2 (D)\u003cstrong\u003e)\u003c/strong\u003e ranks the features by their mean absolute SHAP values, indicating the relative contribution of each feature in predicting IQ-ABAS gap. The features that contributed most to the model\u0026rsquo;s predictions were FSIQ, SCQ, and inattention. Overall, phenotypic features contributed more to the model than sociodemographic features.\u003c/p\u003e\n \u003cp\u003eThe SHAP dependence plots (Fig.\u0026nbsp;3\u003cstrong\u003e)\u003c/strong\u003e further explained the association between the features and the model\u0026apos;s predictions for IQ-ABAS gap. These plots show the directionality and the strength of a feature\u0026rsquo;s contribution to IQ-ABAS gap. In particular, the zero-crossing points indicate the threshold or value of the feature at which its contribution shifts from increasing the model\u0026rsquo;s prediction to decreasing it (or vice versa). Phenotypic features with strong positive relations with their SHAP values include FSIQ, with a zero-crossing point of 97.0 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.93, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001); SCQ, with a zero-crossing point of 10.7 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.87, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001); inattention, with a zero-crossing point of 3.4 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.87, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001); CBCL internalizing symptoms, with a zero-crossing point of 58.1 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.76, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001); CBCL externalizing symptoms, with a zero-crossing point of 54.2 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.80, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001); and hyperactivity, with a zero-crossing point of 2.4 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.57, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001). The relation between TOCS and SHAP values was significant but weak and negative, with a zero-crossing point of -22.8 (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.34, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001).\u003c/p\u003e\n \u003cp\u003eThe sociodemographic features, except for age, showed lower mean absolute SHAP values compared to the phenotypic features. All sociodemographic features, except for race, were significantly associated with the gap. Age showed a strong negative correlation with SHAP values (\u003cem\u003er\u003c/em\u003e=-0.75, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), while both household income (\u003cem\u003e\u0026tau;\u003c/em\u003e=-0.35, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) and first caregiver\u0026apos;s education level (\u003cem\u003e\u0026tau;\u003c/em\u003e=-0.30, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) exhibited weaker negative correlations. ANOVA showed a significant effect of sex on SHAP values (\u003cem\u003eF\u003c/em\u003e(1, 173)\u0026thinsp;=\u0026thinsp;299, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), and a marginal effect of race (\u003cem\u003eF\u003c/em\u003e(1, 173)\u0026thinsp;=\u0026thinsp;2.96, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.0872).\u003c/p\u003e\n \u003cp\u003eThe SHAP interaction results suggested that most of the feature pairs had little to no interactions contributing to cognitive-adaptive functioning gap (Fig. 4 (A)), with the highest SHAP interactions observed in the FSIQ-SCQ feature pair (3.44), the FSIQ-inattention feature pair (1.7), and the SCQ-inattention feature pair (1.61). Figure 4 (B and C) visualize these interactions.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study examined the correlates of the gap between cognitive and adaptive functioning in a neurodiverse sample of children. This was done using univariate testing and group comparisons, as well as with a multivariate approach that employed various computational models to predict the cognitive-adaptive functioning gap.\u003c/p\u003e \u003cp\u003eOur findings demonstrate that the cognitive-adaptive functioning gap is observed across NDCs and typical development, although the widest gaps were found in autism, ADHD, and OCD. This is consistent with previous findings in autism (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Zukerman et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), extending the understanding to other neurodevelopmental conditions. While the presence of this gap in the neurotypical group may reflect natural variability in the relation between cognitive ability and adaptive functioning rather than a pathological process. Importantly, this smaller gap in the neurotypical group provides a useful comparison point, with the much larger and more clinically significant gaps observed in NDCs. These findings motivate the investigation of shared, transdiagnostic factors that may impact the IQ-adaptive functioning gap.\u003c/p\u003e \u003cp\u003eOur univariate and multivariate analyses indicate that IQ significantly contributes to the cognitive-adaptive gap, with individuals possessing higher IQ scores more likely to experience a higher gap between their cognitive abilities and adaptive functioning. This finding is consistent with previous studies reporting that the IQ-adaptive functioning gap is more pronounced in individuals with higher IQ levels (McQuaid et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zukerman et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). From a measurement perspective, this may be related to the fact that at higher IQ levels, a wider range of discrepancies is numerically possible, whereas both our IQ and ABAS measures had minimum values (floor) at 40.\u003c/p\u003e \u003cp\u003eConsistent with previous literature, we observed that IQ alone does not fully account for adaptive functioning, particularly in neurodivergent individuals (Wang et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Specifically, our findings suggest that in addition to IQ, phenotypic features including social communication difficulties, inattentive features and hyperactivity, and internalizing and externalizing symptoms are associated with increases in cognitive-adaptive functioning gap. This is not surprising given the disabling impact of these features on adaptive abilities (Kraper et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Tillmann et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Zukerman et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Interestingly, participants with subthreshold scores (e.g., SCQ scores\u0026thinsp;\u0026ge;\u0026thinsp;10 or SWAN-inattention scores\u0026thinsp;\u0026ge;\u0026thinsp;3) exhibited increases in the IQ-ABAS gap. This suggests that even subthreshold traits in these domains are linked to increased gaps between IQ and adaptive functioning\u003c/p\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSociodemographic Differences\u003c/h2\u003e \u003cp\u003eWhile phenotypic features emerged as the strongest correlates of cognitive-adaptive functioning gap, sociodemographic factors including age, sex, household income, and maternal education also showed notable but smaller associations.\u003c/p\u003e \u003cp\u003eOur findings indicated that the cognitive-adaptive functioning gap was larger among older individuals across multiple NDCs. Although our study is cross-sectional, this finding aligns with previous longitudinal research reporting a similar pattern of increasing gaps with age (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This trend may be explained by the difference resulting from the stable cognitive abilities of individuals with autism and the decrease in adaptive functioning with age. Specifically, older individuals face increasing expectations for independence in socialization, communication, and daily living skills, which may expose challenges in these areas that are not compensated for by cognitive strengths, thereby widening the gap (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur results also revealed that female participants exhibited a decreased cognitive-adaptive functioning gap compared to males, consistent with previous findings indicating sex-specific differences in the cognitive-adaptive functioning gap (McQuaid et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Although the contribution of sex to the cognitive-adaptive functioning gap predictions was lower than those for phenotypic features, this modest effect may reflect behavioral masking in females, thereby contributing slightly and negatively to the cognitive-adaptive functioning gap. Masking refers to compensatory behaviors, such as imitating social norms or rehearsing interactions, which may elevate adaptive functioning scores despite underlying broader challenges. This observation is more commonly reported in females and could lead to an attenuated gap by boosting measured adaptive functioning relative to cognitive performance (Dean et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lai et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; McQuaid et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIncome, education, and race also showed small contributions to the cognitive-adaptive functioning gap. Lower-income groups were associated with a slightly higher cognitive-adaptive functioning gap, and higher levels of caregiver education were linked to a lower cognitive-adaptive functioning gap. These findings align with prior research suggesting that socioeconomic resources may be linked to differences in adaptive functioning outcomes. For example, previous studies have reported associations between parental education and income and the accessibility of supportive interventions, which could be related to developmental outcomes in children with autism (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIt is important to note that most sociodemographic variables in our study were dichotomized, which may have reduced their variability and weakened their predictive power. However, this does not fully account for their relatively low importance in our model. For example, age, which was analyzed as a continuous variable, also demonstrated low importance, suggesting that sociodemographic variables as a group may have limited predictive power for the cognitive-adaptive functioning gap in this context. Future research could examine whether alternative variable representations, such as using continuous measures instead of dichotomized ones, yield different results. Although these features were not primary drivers of the cognitive-adaptive functioning gap in our model, these results motivate future avenues for further investigation into how these differences interact with the cognitive-adaptive functioning gap.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eInteraction Effects Between Phenotypic Features\u003c/h2\u003e \u003cp\u003e In our study, sociodemographic features (age, sex, household income, race, and first caregiver\u0026rsquo;s education) showed relatively low SHAP value rankings in predicting cognitive-adaptive functioning gap compared to phenotypic features. Despite their lower overall importance, these features reveal subtle yet significant biases that can influence assessments of the outcomes. This aligns with prior research (Bradshaw et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; McQuaid et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Tillmann et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), which suggests that while sociodemographic features are not consistent predictors, they can contextually impact outcomes.\u003c/p\u003e \u003cp\u003eThe strongest interaction effect was observed between cognitive abilities and social communication. Individuals with more social communication challenges exhibited more variability in their interaction with cognitive functioning on the cognitive-adaptive functioning gap, across the scale of FSIQ. Previous studies indicate that cognitive abilities often predict adaptive skills in neurotypical populations, but this association can weaken in the presence of strong social communication differences (i.e. autism) (Kenworthy et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Klin et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). This suggested that cognitive abilities and greater social communication issues jointly influence the cognitive-adaptive functioning gap inconsistently, and that these features may need to be investigated alongside other traits in the presence of social communication symptoms. For individuals with lower cognitive ability scores, high social communication symptoms contributed to an increase in the cognitive-adaptive functioning gap, indicating that adaptive functioning outcomes are particularly compromised when both cognitive abilities are low and social communication differences are high. Conversely, with higher cognitive abilities, social communication symptoms become less relevant in the contribution of their interaction to the cognitive-adaptive functioning gap; regardless of social communication symptoms, the cognitive-adaptive functioning gap continues to increase with higher cognitive abilities. This pattern is consistent with findings showing that higher cognitive abilities do not consistently result in better adaptive functioning in autistic individuals (Alvares et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor the interactions between cognitive functioning and inattention, low inattention levels were predominantly observed in individuals with high cognitive abilities, indicating that greater cognitive abilities are associated with fewer inattentive features. Among individuals with higher inattention levels, a quadratic relation emerged with the cognitive-adaptive functioning gap across the range of cognitive abilities. Specifically, those with higher inattention levels and average cognitive functioning contributed to an increase in the cognitive-adaptive functioning gap. In contrast, individuals with higher inattention levels but either low or high cognitive functioning contributed to a decrease in the gap. For individuals with lower inattention levels, a linear relation was observed, with contributions to the cognitive-adaptive functioning gap increasing steadily with cognitive abilities. This pattern indicates a consistent amplifying effect of cognitive functioning on the gap in this subgroup. The interaction between cognitive functioning and inattention continued to drive an increased cognitive-adaptive functioning gap in the low inattention group, supporting that high cognitive abilities alone are insufficient to guarantee positive adaptive functioning outcomes (Alvares et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThree key limitations should be considered in interpreting the results of this study. First, the use of broad categories of sociodemographic and race may have impacted our ability to detect the impacts of these variables on the cognitive-adaptive functioning gap. Second, the study\u0026rsquo;s cross-sectional design limited our capacity to examine developmental changes in the cognitive-adaptive functioning gap. Third, while the 5-fold cross-validation strategy maximized the use of available data and ensured robust performance estimates, the absence of an independent test set may limit the assessment of generalization to entirely unseen datasets. Future longitudinal studies with larger sample sizes could address these limitations by incorporating external validation or independent test sets to further validate model performance and clarify developmental trajectories.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study identifies that cognitive-adaptive functioning gap was associated with higher FSIQ, social-communication challenges, and inattentive traits across diagnostic groups, with phenotypic features showing stronger connections than sociodemographic features. Sociodemographic factors, including sex and age, were also related to the gap, reflecting the diverse influences on cognitive and adaptive functioning.\u003c/p\u003e \u003cp\u003eThese findings improve our understanding of the cognitive-adaptive gap in neurodivergent populations and suggest that focusing support on key associated factors, such as cognitive abilities, social-communication challenges, and inattentive traits, may help to reduce this gap. Given the cross-sectional nature of the study, further longitudinal research is needed to examine how these gaps and their associated factors change over time and to explore the potential for interventions targeting these factors.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eNDCs: Neurodevelopmental Conditions; ADHD: Attention-Deficit/Hyperactivity Disorder; OCD: Obsessive-Compulsive Disorder; POND: Province of Ontario Neurodevelopmental Network; FSIQ: Full-Scale Intelligence Quotient; ABAS-II: Adaptive Behavior Assessment System Second Edition; GAC: General Adaptive Composite; SCQ: Social Communication Questionnaire; SWAN: Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale; TOCS: Toronto Obsessive-Compulsive Scale; CBCL: Child Behavior Checklist; SHAP: SHapley Additive exPlanations; OLS: Ordinary Least Squares; ANOVA: Analysis of Variance; MAE: Mean Absolute Error; DSM-5-TR: Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision; WASI-II: Wechsler Abbreviated Scale of Intelligence, Second Edition; WISC-IV/V: Wechsler Intelligence Scale for Children, Fourth/Fifth Editions; WPPSI-IV: Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition; ADOS-2: Autism Diagnostic Observation Schedule-2; ADI-R: Autism Diagnostic Interview-Revised; PICS: Parent Interview for Child Symptoms; K-SADS: Kiddie-Schedule for Affective Disorders and Schizophrenia; CYBOCS: Children\u0026rsquo;s Yale-Brown Obsessive Compulsive Scale; ID: Intellectual Disability; Tukey\u0026rsquo;s HSD: Tukey\u0026rsquo;s Honest Significant Difference\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch1\u003eAcknowledgements\u003c/h1\u003e\n\u003cp\u003eWe gratefully acknowledge the financial support of the Canadian Institutes of Health Research (CIHR), which made this research possible.\u003c/p\u003e\n\u003ch1\u003eAuthors\u0026rsquo; contributions\u003c/h1\u003e\n\u003cp\u003eEW designed the study, conducted data analysis, interpreted the results, and drafted the manuscript. AK supervised EW, provided input on and co-interpreted the results. JPL, EA, ML, EK, RS, and RN reviewed and approved the final manuscript.\u003c/p\u003e\n\u003ch1\u003eFunding\u003c/h1\u003e\n\u003cp\u003eThis research was supported by the Canadian Institutes of Health Research (CIHR),\u003c/p\u003e\n\u003ch1\u003eAvailability of data and materials\u003c/h1\u003e\n\u003cp\u003eData from POND is available through an Ontario Brain Institute\u0026rsquo;s controlled data release through Brain-CODE by request.\u003c/p\u003e\n\u003ch1\u003eEthics approval and consent to participate\u003c/h1\u003e\n\u003cp\u003eThe Research Ethics Board at each participating site approved the POND study protocol. Participants provided informed consent when able, while those who could not provided assent, and their caregiver gave consent.\u003c/p\u003e\n\u003ch1\u003eCompeting interests\u003c/h1\u003e\n\u003cp\u003eA. Kushki and E. Anagnostou have a patent for hollyTM (formerly Anxiety Meter) with royalties paid from Awake Labs. A. Kushki has received consulting fees from DNAStack and Shaftesbury. E. Anagnostou has received grants from Roche and Anavex, served as a consultant to Roche, Quadrant Therapeutics, Ono, and Impel Pharmaceuticals, has received in-kind support from AMO Pharma and CRA-Simons Foundation, received royalties from APPI and Springer, received an editorial honorarium from Wiley, and has a patent for hollyTM (formerly Anxiety Meter). R. Nicolson received grant funding from Hoffman - La Roche Limited and MapLight Therapeutics. The remaining authors have reported no financial interests or potential conflicts of interest.\u003c/p\u003e\n\u003ch1\u003eAuthor details\u003c/h1\u003e\n\u003cp\u003e\u003csup\u003e1\u003c/sup\u003eInstitute of Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada. \u003csup\u003e2\u003c/sup\u003eWellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom. \u003csup\u003e3\u003c/sup\u003eDepartment of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada. \u003csup\u003e4\u003c/sup\u003eAutism Research Centre, Bloorview Research Institute, Holland Bloorview Kids Rehabilitation Hospital, Toronto, Ontario, Canada. \u003csup\u003e5\u003c/sup\u003eDepartments of Psychology and Psychiatry, Queen\u0026apos;s University, Kingston, Ontario, Canada. \u003csup\u003e6\u003c/sup\u003eDepartment of Psychiatry, Neurosciences and Mental Health, The Hospital for Sick Children, Toronto, Ontario, Canada. \u003csup\u003e7\u003c/sup\u003eDepartment of Psychiatry, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada. \u003csup\u003e8\u003c/sup\u003eDepartment of Psychiatry, University of Western Ontario, London, Ontario, Canada\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbiodun, O. I., Jantan, A., Omolara, A. E., Dada, K. V., Mohamed, N. A., \u0026amp; Arshad, H. (2018). State-of-the-art in artificial neural network applications: A survey. \u003cem\u003eHeliyon\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(11), e00938. https://doi.org/10.1016/j.heliyon.2018.e00938\u003c/li\u003e\n\u003cli\u003eAchenbach, T. M., \u0026amp; Edelbrock, C. S. (1983). Manual for the child behavior checklist and revised child behavior profile. \u003cem\u003e(No Title)\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eAhuja, A., Martin, J., Langley, K., \u0026amp; Thapar, A. (2013). Intellectual Disability in Children with Attention Deficit Hyperactivity Disorder. \u003cem\u003eThe Journal of Pediatrics\u003c/em\u003e, \u003cem\u003e163\u003c/em\u003e(3), 890-895.e1. https://doi.org/10.1016/j.jpeds.2013.02.043\u003c/li\u003e\n\u003cli\u003eAlvares, G. A., Bebbington, K., Cleary, D., Evans, K., Glasson, E. J., Maybery, M. T., Pillar, S., Uljarević, M., Varcin, K., Wray, J., \u0026amp; Whitehouse, A. J. (2020). The misnomer of \u0026lsquo;high functioning autism\u0026rsquo;: Intelligence is an imprecise predictor of functional abilities at diagnosis. \u003cem\u003eAutism\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(1), 221\u0026ndash;232. https://doi.org/10.1177/1362361319852831\u003c/li\u003e\n\u003cli\u003eAmeis, S. H., Lerch, J. P., Taylor, M. J., Lee, W., Viviano, J. D., Pipitone, J., Nazeri, A., Croarkin, P. E., Voineskos, A. N., Lai, M.-C., Crosbie, J., Brian, J., Soreni, N., Schachar, R., Szatmari, P., Arnold, P. D., \u0026amp; Anagnostou, E. (2016). A Diffusion Tensor Imaging Study in Children With ADHD, Autism Spectrum Disorder, OCD, and Matched Controls: Distinct and Non-Distinct White Matter Disruption and Dimensional Brain-Behavior Relationships. \u003cem\u003eAmerican Journal of Psychiatry\u003c/em\u003e, \u003cem\u003e173\u003c/em\u003e(12), 1213\u0026ndash;1222. https://doi.org/10.1176/appi.ajp.2016.15111435\u003c/li\u003e\n\u003cli\u003eAmerican Psychiatric Association (Ed.). (2022). \u003cem\u003eDiagnostic and statistical manual of mental disorders, Fifth Edition, Text Revision (DSM-5-TR)\u003c/em\u003e (5th ed). American Psychiatric Association. https://psychiatryonline.org/doi/book/10.1176/appi.books.9780890425787\u003c/li\u003e\n\u003cli\u003eArashi, M., Roozbeh, M., Hamzah, N. A., \u0026amp; Gasparini, M. (2021). Ridge regression and its applications in genetic studies. \u003cem\u003ePLOS ONE\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(4), e0245376. https://doi.org/10.1371/journal.pone.0245376\u003c/li\u003e\n\u003cli\u003eBerument, S. K., Rutter, M., Lord, C., Pickles, A., \u0026amp; Bailey, A. (1999). Autism screening questionnaire: Diagnostic validity. \u003cem\u003eBritish Journal of Psychiatry\u003c/em\u003e, \u003cem\u003e175\u003c/em\u003e(5), 444\u0026ndash;451. https://doi.org/10.1192/bjp.175.5.444\u003c/li\u003e\n\u003cli\u003eBilleiter, K. B., \u0026amp; Froiland, J. M. (2023). Diversity of Intelligence is the Norm Within the Autism Spectrum: Full Scale Intelligence Scores Among Children with ASD. \u003cem\u003eChild Psychiatry \u0026amp; Human Development\u003c/em\u003e, \u003cem\u003e54\u003c/em\u003e(4), 1094\u0026ndash;1101. https://doi.org/10.1007/s10578-021-01300-9\u003c/li\u003e\n\u003cli\u003eBradshaw, J., Gillespie, S., Klaiman, C., Klin, A., \u0026amp; Saulnier, C. (2019). Early emergence of discrepancy in adaptive behavior and cognitive skills in toddlers with autism spectrum disorder. \u003cem\u003eAutism\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(6), 1485\u0026ndash;1496. https://doi.org/10.1177/1362361318815662\u003c/li\u003e\n\u003cli\u003eBreiman, L. (2001). Random Forests. \u003cem\u003eMachine Learning\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(1), 5\u0026ndash;32. https://doi.org/10.1023/A:1010933404324\u003c/li\u003e\n\u003cli\u003eCharman, T., Pickles, A., Simonoff, E., Chandler, S., Loucas, T., \u0026amp; Baird, G. (2011). IQ in children with autism spectrum disorders: Data from the Special Needs and Autism Project (SNAP). \u003cem\u003ePsychological Medicine\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(3), 619\u0026ndash;627. https://doi.org/10.1017/S0033291710000991\u003c/li\u003e\n\u003cli\u003eDean, M., Harwood, R., \u0026amp; Kasari, C. (2017). The art of camouflage: Gender differences in the social behaviors of girls and boys with autism spectrum disorder. \u003cem\u003eAutism\u003c/em\u003e, \u003cem\u003e21\u003c/em\u003e(6), 678\u0026ndash;689. https://doi.org/10.1177/1362361316671845\u003c/li\u003e\n\u003cli\u003eDuncan, A. W., \u0026amp; Bishop, S. L. (2015). Understanding the gap between cognitive abilities and daily living skills in adolescents with autism spectrum disorders with average intelligence. \u003cem\u003eAutism\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(1), 64\u0026ndash;72. https://doi.org/10.1177/1362361313510068\u003c/li\u003e\n\u003cli\u003eFriedman, J. H. (2002). Stochastic gradient boosting. \u003cem\u003eComputational Statistics \u0026amp; Data Analysis\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e(4), 367\u0026ndash;378. https://doi.org/10.1016/S0167-9473(01)00065-2\u003c/li\u003e\n\u003cli\u003eJacobs, G. R., Voineskos, A. N., Hawco, C., Stefanik, L., Forde, N. J., Dickie, E. W., Lai, M.-C., Szatmari, P., Schachar, R., Crosbie, J., Arnold, P. D., Goldenberg, A., Erdman, L., \u0026amp; Ameis, S. H. (2021). Integration of brain and behavior measures for identification of data-driven groups cutting across children with ASD, ADHD, or OCD. \u003cem\u003eNeuropsychopharmacology\u003c/em\u003e, \u003cem\u003e46\u003c/em\u003e(3), 643\u0026ndash;653. https://doi.org/10.1038/s41386-020-00902-6\u003c/li\u003e\n\u003cli\u003eKanne, S., Gerber, A. J., Quirmbach, L. M., Sparrow, S. S., Cicchetti, D. V., \u0026amp; Saulnier, C. (2011). \u003cem\u003eThe Role of Adaptive Behavior in Autism Spectrum Disorders: Implications for Functional Outcome\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eKenworthy, L., Case, L., Harms, M. B., Martin, A., \u0026amp; Wallace, G. L. (2010). Adaptive Behavior Ratings Correlate With Symptomatology and IQ Among Individuals With High-Functioning Autism Spectrum Disorders. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(4), 416\u0026ndash;423. https://doi.org/10.1007/s10803-009-0911-4\u003c/li\u003e\n\u003cli\u003eKlin, A., Saulnier, C. A., Sparrow, S. S., Cicchetti, D. V., Volkmar, F. R., \u0026amp; Lord, C. (2007). Social and Communication Abilities and Disabilities in Higher Functioning Individuals with Autism Spectrum Disorders: The Vineland and the ADOS. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(4), 748\u0026ndash;759. https://doi.org/10.1007/s10803-006-0229-4\u003c/li\u003e\n\u003cli\u003eKnopf, A. (2020). Autism prevalence increases from 1 in 60 to 1 in 54: CDC. \u003cem\u003eThe Brown University Child and Adolescent Behavior Letter\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(6), 4\u0026ndash;4. https://doi.org/10.1002/cbl.30470\u003c/li\u003e\n\u003cli\u003eKraper, C. K., Kenworthy, L., Popal, H., Martin, A., \u0026amp; Wallace, G. L. (2017). The Gap Between Adaptive Behavior and Intelligence in Autism Persists into Young Adulthood and is Linked to Psychiatric Co-morbidities. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e47\u003c/em\u003e(10), 3007\u0026ndash;3017. https://doi.org/10.1007/s10803-017-3213-2\u003c/li\u003e\n\u003cli\u003eKuhn \u0026amp; Max. (2008). Building Predictive Models in R Using the caret Package. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(5), 1\u0026ndash;26. https://doi.org/10.18637/jss.v028.i05\u003c/li\u003e\n\u003cli\u003eLai, M.-C., Lombardo, M. V., Pasco, G., Ruigrok, A. N. V., Wheelwright, S. J., Sadek, S. A., Chakrabarti, B., MRC AIMS Consortium, \u0026amp; Baron-Cohen, S. (2011). A Behavioral Comparison of Male and Female Adults with High Functioning Autism Spectrum Conditions. \u003cem\u003ePLoS ONE\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(6), e20835. https://doi.org/10.1371/journal.pone.0020835\u003c/li\u003e\n\u003cli\u003eLoh, H. W., Ooi, C. P., Seoni, S., Barua, P. D., Molinari, F., \u0026amp; Acharya, U. R. (2022). Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011\u0026ndash;2022). \u003cem\u003eComputer Methods and Programs in Biomedicine\u003c/em\u003e, \u003cem\u003e226\u003c/em\u003e, 107161. https://doi.org/10.1016/j.cmpb.2022.107161\u003c/li\u003e\n\u003cli\u003eLundberg, S. M., Erion, G., Chen, H., DeGrave, A., Prutkin, J. M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., \u0026amp; Lee, S.-I. (2020). From local explanations to global understanding with explainable AI for trees. \u003cem\u003eNature Machine Intelligence\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 56\u0026ndash;67. https://doi.org/10.1038/s42256-019-0138-9\u003c/li\u003e\n\u003cli\u003eLundberg, S. M., \u0026amp; Lee, S.-I. (2017). A unified approach to interpreting model predictions. \u003cem\u003eAdvances in Neural Information Processing Systems\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eMahendiran, T., Dupuis, A., Crosbie, J., Georgiades, S., Kelley, E., Liu, X., Nicolson, R., Schachar, R., Anagnostou, E., \u0026amp; Brian, J. (2019). Sex Differences in Social Adaptive Function in Autism Spectrum Disorder and Attention-Deficit Hyperactivity Disorder. \u003cem\u003eFrontiers in Psychiatry\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 607. https://doi.org/10.3389/fpsyt.2019.00607\u003c/li\u003e\n\u003cli\u003eMcQuaid, G. A., Pelphrey, K. A., Bookheimer, S. Y., Dapretto, M., Webb, S. J., Bernier, R. A., McPartland, J. C., Van Horn, J. D., \u0026amp; Wallace, G. L. (2021). The gap between IQ and adaptive functioning in autism spectrum disorder: Disentangling diagnostic and sex differences. \u003cem\u003eAutism\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(6), 1565\u0026ndash;1579. https://doi.org/10.1177/1362361321995620\u003c/li\u003e\n\u003cli\u003eMolnar, C. (2020). \u003cem\u003eInterpretable Machine Learning\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eMouga, S., Almeida, J., Caf\u0026eacute;, C., Duque, F., \u0026amp; Oliveira, G. (2015). Adaptive Profiles in Autism and Other Neurodevelopmental Disorders. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(4), 1001\u0026ndash;1012. https://doi.org/10.1007/s10803-014-2256-x\u003c/li\u003e\n\u003cli\u003ePark, L. S., Burton, C. L., Dupuis, A., Shan, J., Storch, E. A., Crosbie, J., Schachar, R. J., \u0026amp; Arnold, P. D. (2016). The Toronto Obsessive-Compulsive Scale: Psychometrics of a Dimensional Measure of Obsessive-Compulsive Traits. \u003cem\u003eJournal of the American Academy of Child \u0026amp; Adolescent Psychiatry\u003c/em\u003e, \u003cem\u003e55\u003c/em\u003e(4), 310-318.e4. https://doi.org/10.1016/j.jaac.2016.01.008\u003c/li\u003e\n\u003cli\u003ePerry, A., Flanagan, H. E., Dunn Geier, J., \u0026amp; Freeman, N. L. (2009). Brief Report: The Vineland Adaptive Behavior Scales in Young Children with Autism Spectrum Disorders at Different Cognitive Levels. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e39\u003c/em\u003e(7), 1066\u0026ndash;1078. https://doi.org/10.1007/s10803-009-0704-9\u003c/li\u003e\n\u003cli\u003eRaiford, S. E. (2018). The Wechsler Intelligence Scale for Children\u0026mdash;Fifth Edition Integrated. In \u003cem\u003eContemporary intellectual assessment: Theories, tests, and issues, 4th ed.\u003c/em\u003e (pp. 303\u0026ndash;332). The Guilford Press.\u003c/li\u003e\n\u003cli\u003eRoid, G. H., \u0026amp; Pomplun, M. (2012). The Stanford-Binet Intelligence Scales, Fifth Edition. In \u003cem\u003eContemporary intellectual assessment: Theories, tests, and issues, 3rd ed.\u003c/em\u003e (pp. 249\u0026ndash;268). The Guilford Press.\u003c/li\u003e\n\u003cli\u003eRoot, A., Brown, J. P., Forbes, H. J., Bhaskaran, K., Hayes, J., Smeeth, L., \u0026amp; Douglas, I. J. (2019). Association of Relative Age in the School Year With Diagnosis of Intellectual Disability, Attention-Deficit/Hyperactivity Disorder, and Depression. \u003cem\u003eJAMA Pediatrics\u003c/em\u003e, \u003cem\u003e173\u003c/em\u003e(11), 1068. https://doi.org/10.1001/jamapediatrics.2019.3194\u003c/li\u003e\n\u003cli\u003eScandurra, V., Emberti Gialloreti, L., Barbanera, F., Scordo, M. R., Pierini, A., \u0026amp; Canitano, R. (2019). Neurodevelopmental Disorders and Adaptive Functions: A Study of Children With Autism Spectrum Disorders (ASD) and/or Attention Deficit and Hyperactivity Disorder (ADHD). \u003cem\u003eFrontiers in Psychiatry\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 673. https://doi.org/10.3389/fpsyt.2019.00673\u003c/li\u003e\n\u003cli\u003eShenouda, J., Barrett, E., Davidow, A. L., Sidwell, K., Lescott, C., Halperin, W., Silenzio, V. M. B., \u0026amp; Zahorodny, W. (2023). Prevalence and Disparities in the Detection of Autism Without Intellectual Disability. \u003cem\u003ePediatrics\u003c/em\u003e, \u003cem\u003e151\u003c/em\u003e(2), e2022056594. https://doi.org/10.1542/peds.2022-056594\u003c/li\u003e\n\u003cli\u003eSmola, A. J., \u0026amp; Sch\u0026ouml;lkopf, B. (2004). A tutorial on support vector regression. \u003cem\u003eStatistics and Computing\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 199\u0026ndash;222. https://doi.org/10.1023/B:STCO.0000035301.49549.88\u003c/li\u003e\n\u003cli\u003eSong, Y., \u0026amp; Lu, Y. (2015). \u003cem\u003eDecision tree methods: Applications for classification and prediction\u003c/em\u003e. \u003cem\u003e27\u003c/em\u003e(2).\u003c/li\u003e\n\u003cli\u003eSwanson, J. M., Schuck, S., Porter, M. M., Carlson, C., Hartman, C. A., Sergeant, J. A., Clevenger, W., Wasdell, M., McCleary, R., Lakes, K., \u0026amp; Wigal, T. (2015). \u003cem\u003eCategorical and Dimensional Definitions and Evaluations of Symptoms of ADHD: History of the SNAP and the SWAN Rating Scales\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eTibshirani, R. (1996). Regression Shrinkage and Selection Via the Lasso. \u003cem\u003eJournal of the Royal Statistical Society: Series B (Methodological)\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e(1), 267\u0026ndash;288. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x\u003c/li\u003e\n\u003cli\u003eTillmann, J., San Jos\u0026eacute; C\u0026aacute;ceres, A., Chatham, C. H., Crawley, D., Holt, R., Oakley, B., Banaschewski, T., Baron‐Cohen, S., B\u0026ouml;lte, S., Buitelaar, J. K., Durston, S., Ham, L., Loth, E., Simonoff, E., Spooren, W., Murphy, D. G., Charman, T., the EU‐AIMS LEAP group, Ahmad, J., \u0026hellip; Zwiers, M. P. (2019). Investigating the factors underlying adaptive functioning in autism in the EU‐AIMS Longitudinal European Autism Project. \u003cem\u003eAutism Research\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(4), 645\u0026ndash;657. https://doi.org/10.1002/aur.2081\u003c/li\u003e\n\u003cli\u003eWang, S., Zhou, Q., Chen, K., Ceng, C., Zhan, G., You, C., Xing, Y., Zou, Y., \u0026amp; Deng, H. (2023). Cognitive-adaptive Functioning Gap and Mediating Factors that Impact Adaptive Functioning in Chinese Preschool-aged Children with Autism Spectrum Disorder. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e. https://doi.org/10.1007/s10803-023-06029-z\u003c/li\u003e\n\u003cli\u003eWechsler, D. (2003). WISC-IV: Wechsler Intelligence Scale for Children\u0026ndash;fourth edition. In \u003cem\u003eTechnical and interpretive manual\u003c/em\u003e. TX Psychol Corp San Antonio.\u003c/li\u003e\n\u003cli\u003eWechsler, D. \u0026amp; Psychological Corporation. (2012). \u003cem\u003eWPPSI-IV : Wechsler preschool and primary scale of intelligence\u0026mdash;Fourth edition\u003c/em\u003e. 1 technical and interpretive manual (xi, 256 pages ; 28 cm), 1 administration and scoring manual (xi, 331 pages 24 cm), 3 stimulus books, 1 box of 14 red and white blocks, 1 box of 13 cardboard object assembly puzzles, 1 box zoo locations card set with ink dauber, 1 extra ink dauber, 1 folder of 3 zoo locations layouts, 3 response booklets (Bug search, Cancellation, Animal coding), 2 record forms (ages 2:6-3:11, ages 4:0-7:7), 3 bug search scoring key cards, 1 animal coding scoring key, 1 cancellation scoring template ; in suitcase 50 x 37 x 26 cm. WorldCat.\u003c/li\u003e\n\u003cli\u003eWechsler, D., Zhou, X., Psychological Corporation., \u0026amp; Assessment Library Materials (University of Lethbridge. Faculty of Education. Curriculum Laboratory). (2011). WASI-II : Wechsler abbreviated scale of intelligence. \u003cem\u003eWechsler Abbreviated Scale of Intelligence\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eZeidan, J., Fombonne, E., Scorah, J., Ibrahim, A., Durkin, M. S., Saxena, S., Yusuf, A., Shih, A., \u0026amp; Elsabbagh, M. (2022). Global prevalence of autism: A systematic review update. \u003cem\u003eAutism Research\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(5), 778\u0026ndash;790. https://doi.org/10.1002/aur.2696\u003c/li\u003e\n\u003cli\u003eZou, H., \u0026amp; Hastie, T. (2005). Regularization and Variable Selection Via the Elastic Net. \u003cem\u003eJournal of the Royal Statistical Society Series B: Statistical Methodology\u003c/em\u003e, \u003cem\u003e67\u003c/em\u003e(2), 301\u0026ndash;320. https://doi.org/10.1111/j.1467-9868.2005.00503.x\u003c/li\u003e\n\u003cli\u003eZukerman, G., Yahav, G., \u0026amp; Ben-Itzchak, E. (2021). The Gap Between Cognition and Adaptive Behavior in Students with Autism Spectrum Disorder: Implications for Social Anxiety and the Moderating Effect of Autism Traits. \u003cem\u003eJournal of Autism and Developmental Disorders\u003c/em\u003e, \u003cem\u003e51\u003c/em\u003e(5), 1466\u0026ndash;1478. https://doi.org/10.1007/s10803-020-04632-y\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Neurodevelopmental conditions, cognitive functioning, social communication, inattention, sociodemographic, machine learning, feature importance.","lastPublishedDoi":"10.21203/rs.3.rs-5968118/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5968118/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground: \u003c/strong\u003eIn the general population, adaptive functioning typically aligns with cognitive abilities; however, this relationship appears more complex among neurodivergent individuals. Neurodevelopmental conditions (NDCs), including autism, attention-deficit/hyperactivity disorder (ADHD), and obsessive-compulsive disorder (OCD), are associated with significant differences between cognitive and adaptive functioning, described as the cognitive-adaptive functioning gap. While this gap has been examined primarily in autistic individuals, it has also been observed across other NDCs. In mixed neurotypical and neurodivergent samples, individuals exhibit a range of discrepancies, reflecting the heterogeneity of cognitive-adaptive functioning gap profiles. Although the gap tends to be larger in NDCs compared to neurotypical populations, there is limited understanding of the phenotypic and sociodemographic factors linked to these discrepancies in neurodivergent children. The present study explores the features associated with cognitive-adaptive functioning gap in a sample of children and adolescents, including both neurodivergent and neurotypical individuals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eThe study used data from the Province of Ontario Neurodevelopmental Disorders Network (POND), comprising 902 participants (autism = 409, ADHD = 210, OCD = 36, neurotypical = 214, other = 33) aged 6-21 years. Cognitive functioning was measured with full-scale IQ (FSIQ) from the Wechsler family of tests, and adaptive functioning was measured with the Adaptive Behavior Assessment System-II (ABAS-II), specifically the General Adaptive Composite (GAC) score. The cognitive-adaptive functioning gap was calculated as the difference between FSIQ and ABAS-II GAC scores. Phenotypic measures included social communication (Social Communication Questionnaire, or SCQ), ADHD symptoms (Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale, or SWAN), OCD symptoms (Toronto Obsessive-Compulsive Scale, or TOCS), and mental health symptoms (Child Behavior Checklist, or CBCL). Sociodemographic data encompassed sex, age, race, household income, and caregiver education. Nine computational models were used to estimate the cognitive-adaptive functioning gap. SHapley Additive exPlanations (SHAP) analysis was used to interpret the contributions of individual features to the cognitive-adaptive functioning gap model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: The random forest model demonstrated the highest predictive accuracy (R² = 0.88, p \u0026lt; 0.001), performing better than other models with a mean absolute error of 4.14. SHAP analysis indicated that FSIQ, SCQ, and inattentive traits were the most influential features in estimating cognitive-adaptive functioning gap. Higher FSIQ (FSIQ ≥ 97.0) was linked to larger cognitive-adaptive functioning gaps across the combined sample. Similarly, social communication differences (SCQ ≥ 10.7) and inattentive traits (SWAN ≥ 3.4) were associated with larger gaps. Sociodemographic factors showed smaller but statistically significant associations, with sex (p \u0026lt; 0.001) and age (p \u0026lt; 0.001) showing relations to the cognitive-adaptive functioning gap (male sex and younger age were linked to smaller gaps).\u003c/p\u003e\n\u003cp\u003eLimitations: Key limitations include the use of broad sociodemographic categories, the cross-sectional design limiting insights into developmental changes in the cognitive-adaptive functioning gap, and the absence of an independent test set, which may affect generalizability. Future longitudinal studies and larger sample size are needed to address these limitations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e: This study identifies the cognitive-adaptive functioning gap as associated with higher FSIQ, social-communication challenges, and inattentive traits, with phenotypic features showing stronger connections than sociodemographic factors. These findings suggest that focusing support on these key factors may help reduce the gap. Further longitudinal research is needed to explore how these gaps evolve and assess potential intervention strategies.\u003c/p\u003e","manuscriptTitle":"Exploring Phenotypic and Sociodemographic Influences on Cognitive-Adaptive Functioning Gap in Neurodivergent Children","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-11 10:19:59","doi":"10.21203/rs.3.rs-5968118/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2d80b81d-b1e7-4f19-a69d-e6cba3cc35ef","owner":[],"postedDate":"February 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-03-18T03:08:50+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-11 10:19:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5968118","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5968118","identity":"rs-5968118","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}