Accurate and efficient data-driven psychiatric assessment using machine learning

Accurate and efficient data-driven psychiatric assessment using machine learning | 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 Accurate and efficient data-driven psychiatric assessment using machine learning Kseniia Konishcheva, Bennett Leventhal, Maki Koyama, Sambit Panda, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6270695/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 13 Jan, 2026 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted 10 You are reading this latest preprint version Abstract Background Accurate assessment of mental disorders and learning disabilities is essential for timely intervention. Machine learning and feature selection techniques have potential in improving the accuracy and efficiency of mental health assessments. However, limited research has explored the use of large transdiagnostic datasets, as well as the application of these techniques in developing quick, question-based assessments. The goals of this study are to apply machine learning and feature selection techniques to a large transdiagnostic dataset featuring a high number of assessment items, and to create a tool for the streamlined creation of efficient and effective assessment using existing datasets. Methods This study uses the Healthy Brain Network (HBN) dataset to develop a tool for creation of efficient and effective machine learning-based assessment of mental disorders and learning disabilities. Feature selection algorithms were applied to identify parsimonious item subsets. Modular architecture ensures straightforward application to other datasets. Results Machine learning models trained on the HBN data exhibited improved performance over existing assessments. Using only non-proprietary assessments did not significantly impact model performance. Discussion This study demonstrates the feasibility of using existing large-scale datasets for creating accurate and efficient assessments for mental disorders and learning disabilities. The performance of the machine learning models provide estimates of the performance of the new assessments in a population similar to HBN. The modular architecture of the developed tool ensures seamless application to diverse clinical and research contexts. machine learning mental health assessment psychiatry learning disabilities learning disability assessment feature selection Healthy Brain Network study Figures Figure 1 Key points and relevance - Data-driven methods, such as machine learning and feature selection, have shown promise in improving the accuracy and efficiency of mental health assessment. - Few studies investigated these methods for creation of learning disability assessments, and in application to transdiagnostic datasets consisting of a large number of assessments. - Using the Healthy Brain Network dataset, we have built a tool for the creation of accurate and efficient machine-learning based assessments for common mental disorders and learning disabilities. - The modular design of the tool ensures its easy adaptation to other datasets, accommodating a variety of clinical and research applications. Background Prevalence of Mental Disorders and Learning Disabilities Globally, approximately one in eight people currently have a mental disorder (World Health Organization, n.d.), with roughly 30% having experienced one at some point in their life (Steele, 2004). Learning disabilities (LDs) are estimated to impact between 5 and 15% of the population (Willcutt et al., 2011). Accurate and timely identification of psychiatric and learning disorders leads to more effective intervention (Hetrick et al., 2008; Lange, 2006). Feature Selection for Mental Health and Learning Disability Assessment Increasing availability of large transdiagnostic datasets, coupled with advancement in technology, presents novel approaches for improvement of the assessment of mental disorders and learning disabilities. Previous studies have demonstrated the utility of machine learning in conjunction with feature selection techniques for shortening existing mental health assessments, and creating new ones based on items from existing assessments and manually constructed item banks (Achenie et al., 2019; Bone et al., 2015, 2016; Brodey et al., 2018; Carpenter, Sprechmann, Calderbank, Sapiro, & Egger, 2016; Gibbons, Chattopadhyay, Meltzer, Kane, & Guinart, 2022; Glavin et al., 2023; Liu, Chokka, Cao, & Chokka, 2021; Liu, Hankey, Cao, & Chokka, 2021; Lu & Petkova, 2014; Schultebraucks et al., 2023; Tartarisco et al., 2021; Tutun et al., 2023). To our knowledge, only one study (Wu, Huang, & Meng, 2006) has leveraged feature selection to improve the identification of learning disabilities. Wu used scores from standardized tests as the input, commonly used for diagnosing learning disabilities. However, these tests require to be administered by a trained professional, are time consuming, and often prohibitively expensive (Hayes, Dombrowski, Shefcyk, & Bulat, 2018; Willcutt et al., 2011). To address these challenges, question-based screeners have been developed to identify students at risk of learning disabilities and requiring further evaluation (Willcutt et al., 2011). Feature selection techniques applied to a large item pool assessing learning disability symptoms, and symptoms commonly co-occurring conditions, can facilitate the development of improved question-based screeners, facilitating timely identification of children at risk for learning disorders. Using existing datasets for assessment creation Large transdiagnostic datasets provide an opportunity to use existing data to evaluate extensive item pools for assessment creation, which would be impractical in a traditional instrument development study that relies on data collection. Moreover, such datasets offer an opportunity to test items from instruments aimed at other disorders that assess seemingly unrelated constructs and may not be encompassed within a theory-driven item pool, but could nonetheless contribute to identification of assessed disorders. The Healthy Brain Network Dataset One example of such a dataset is the Healthy Brain Network (HBN) dataset. Comprising a diverse community-based sample of 4136 children and adolescents from the New York City area, it contains a comprehensive item-level psychiatric and learning assessment alongside demographic variables. Each participant undergoes assessment by a team of clinicians and receives one or several diagnoses, based on a diagnostic interview (Kiddie Schedule for Affective Disorders and Schizophrenia; Kaufman et al., 1997) and assessment scores. The dataset includes item-level responses to over 50 assessments (over 1000 items) targeting a broad range of disorders. In the HBN study, a majority of participants diagnosed with one disorder have one or multiple comorbidities, offering a realistic setting to evaluate the improved assessments as opposed to the relatively straightforward task of distinguishing affected individuals from healthy controls. Research aims The present study aims to create a tool for streamlined creation of effective machine learning based assessments using existing datasets, while validating techniques identified in prior studies on a dataset with a high number of input variables, taking into account the high computational complexity of the task. We have developed a tool that leverages existing datasets to identify parsimonious item subsets for efficient and effective mental health and learning disorder assessment, and provides trained machine learning models that can be used for assessment of new patients. We test if the machine learning models trained on the HBN data provide more efficient and effective assessment compared to existing assessment instruments. The instruments used in the HBN datasets include both proprietary and freely available assessments. We suppose that the authors of assessments that made their instruments freely accessible would be more open to allow the use of items from their assessment for creation of new instruments in the future. Due to the large number of instruments used in the HBN study, many constructs are assessed by multiple items from different instruments. We hypothesize that in the presence of a large number of input assessments, limiting the input variables to items from non-proprietary assessments will not significantly reduce the performance of the trained models. Methods Output variables In addition to clinician diagnoses present in the HBN dataset, custom output variables were created for learning difficulties based on performance on standard tests, which are, unlike the clinician diagnoses, independent of the input assessment scores. Following previous literature (Kramer et al., 2020), the criteria for the learning difficulties were defined as the Full Scale IQ over the two standard deviations below the mean, and the corresponding achievement test subscale under one standard deviation below the mean. Wechsler Individual Achievement Test (WIAT; Wechsler, 2009) and Wechsler Intelligence Scale for Children (WISC; Wechsler, 2014) scores were used. Wechsler Individual Achievement Test Spelling scale was used for written expression impairment instead of the commonly used Written Expression scale because the Written Expression scale was not present in the dataset. Besides the Specific Learning Disorders (SLDs), we created test-based variables for Borderline Intellectual Functioning (BIF), Intellectual Disability-Mild (ID), Processing Speed Deficit (PS), and two definitions of Non-Verbal Learning Disability (NVLD and NVLD-no-read). The rules for ID, BIF, and NVLD are based on criteria from Hetland, Braatveit, Hagen, Lundervold, & Erga, 2021, Margolis et al., 2020, and Petterson, Bourke, Leonard, Jacoby, & Bower, 2007 respectively. All participants who were not administered the assessments required to construct the custom output variables were excluded from the analysis. Input variables All items from all self-report and parent-report assessments were used, except items from physical fitness assessments. Additionally, responses to the pre-intake form were included, containing age, sex, and educational and developmental history of the participant. Due to the nature of the HBN study protocol, most of the assessments were administered only to a subset of participants (i.e. not a single participant was administered all assessments). To minimize the amount of missing data, the most administered assessments were used, and the participants were limited to those who were administered all of these assessments (n=2335, 584 items). The only assessment in the dataset evaluating symptoms of learning disorders (Conners ADHD Rating Scales – Self Report Short Form; Conners, 2008) is only administered to older participants, and was not included in the final list of assessments. An additional dataset was constructed to include a larger set of assessments. Similarly to the original dataset, participants who were not administered all of the assessments were excluded, resulting in a smaller number of participants (n=1406, 1146 items). Assessments included in the two datasets are presented in Table 1. Assessments included in the dataset with only non-proprietary assessments were: Basic_Demos, PreInt_EduHx, PreInt_DevHx, SympChck, ASSQ, ARI_P, SDQ, SWAN, ARI_S, SCARED_P, ICU_P, APQ_P, DTS, ESWAN, MFQ_P, APQ_SR . Table 1: Input assessments and abbreviations used in the paper. Assessment Name Abbreviation Demographics: age, sex Basic_Demos Intake Interview – Education History PreInt_EduHx Intake Interview – Developmental History PreInt_DevHx Child Mind Institute Symptom Checker (“Symptom Checker,” n.d.) SympChck Social Communication Questionnaire (Rutter, Bailey, & Lord, 2003) SCQ Barratt Simplified Measure of Social Status (Barratt, 2012) Barratt Autism Spectrum Screening Questionnaire (Ehlers & Gillberg, 1993) ASSQ Affective Reactivity Index – Parent Report (Stringaris et al., 2012) ARI_P Strengths and Difficulties Questionnaire (Goodman, 1997) SDQ Strengths and Weaknesses Assessment of ADHD and Normal Behavior (Swanson et al., 2001) SWAN Affective Reactivity Index – Self Report (Stringaris et al., 2012) ARI_S Social Responsiveness Scale-2 (Constantino et al., 2003) SRS Child Behavior Checklist (Achenbach, 1991) CBCL Screen for Child Anxiety Related Disorders – Parent report (Birmaher et al., 1997) SCARED_P Inventory of Callous-Unemotional Traits – Parent Report (Essau et al., 2006a) ICU_P Alabama Parenting Questionnaire – Parent Report Report (Essau et al., 2006b) APQ_P Parent-Child Internet Addiction Test (Young, 1998) PCIAT Distress Tolerance Scale (Simons & Gaher, 2005) DTS Extended Strengths and Weaknesses Assessment of Normal Behavior-Parent Report (Alexander, Salum, Milham, & Swanson, n.d.) ESWAN Mood and Feelings Questionnaire – Parent Report (Costello & Angold, 1988) MFQ_P Alabama Parenting Questionnaire (Essau et al., 2006b) APQ_SR Assessments added in the expanded set: WHO Disability Assessment Schedule – Parent Report (Gold, 2014) WHODAS_P The Columbia Impairment Scale – Parent Report (Bird, Shaffer, Fisher, Gould, & et al, 1993) CIS_P Parenting Stress Index Fourth Edition (Abidin, 2012) PSI Social Aptitudes Scale (Liddle, Batty, & Goodman, 2009) SAS Repetitive Behavior Scale (Lam & Aman, 2007) RBS PhenX Neighborhood Safety (Mujahid, Diez Roux, Morenoff, & Raghunathan, 2007) PhenX_Neighborhood WHO Disability Assessment Schedule – Self Report (Gold, 2014) WHODAS_SR The Columbia Impairment Scale – Self Report (Bird et al., 1993) CIS_SR Screen for Child Anxiety Related Disorders – Self report (Birmaher et al., 1997) SCARED_SR Conners ADHD Rating Scales – Self Report Short Form (Conners, 2008) C3SR Mood and Feelings Questionnaire – Self Report (Costello & Angold, 1988) MFQ_SR Modular architecture The developed tool is organized in a modular fashion, to ensure minimal required changes when applying the tool to another dataset. Project structure proposed by the Cookiecutter Data Science package (drivendata, n.d.) was used, which separates data preparation and model training. The results of executing the two modules are machine learning models trained on examined item subsets, recommended number of items, recommended probability threshold (the equivalent of the cut-off score in sum-score based assessments), and the performance of the models. Analysis Ten-fold cross-validation was used to obtain reliable performance values of the models. The models were trained only for diagnoses with more than twenty positive examples in the validation set. Elastic Net model was used, as it achieved good performance in previous research in the mental health domain (Haynos et al., 2021; Liu, Cao, & Chokka, 2023; Liu, Chokka, et al., 2021; Liu, Hankey, et al., 2021) and in preliminary testing, and is highly explainable. To address class imbalance (low number of positive cases compared to negative cases), positive cases were weighted higher than negative cases, the weight being automatically adjusted based on the prevalence. AUROC, specificity, and sensitivity were used as performance measures. A probability threshold was estimated in each cross-validation fold to produce well-balanced sensitivity and specificity values. Final sensitivity and specificity were calculated on a separate test-set, using the median of the identified probability thresholds (the recommended threshold). Recursive Feature Elimination (RFE, sklearn python library; Pedregosa et al., 2011) and Sequential Forward Floating Selection (Raschka, 2018) were used as feature selection algorithms. Due to the high computational complexity of the SFFS algorithm and the high number of input variables, RFE was used first to identify the top twenty-seven items, then SSFS was applied to the selected items to identify smaller item subsets. Twenty-seven was chosen as the maximum acceptable length of an assessment for a single disorder, based on the number of items in the longest input assessment for a single disorder in the HBN dataset (ASSQ). The recommended number of items was calculated by examining the performance of the model with each number of items, and choosing the number of items that reaches 95% of the maximum performance among the evaluated subsets. Performance evaluation To check if the trained machine learning models outperform traditional assessment instruments, AUROC of each total and subscale score of assessments from the HBN dataset for each predicted diagnosis was calculated. Mann-Whitney test was used to compare the performance of the best existing scale for each diagnosis to the performance of trained models using the same number of features. The analysis was repeated using only non-proprietary assessments. Mann-Whitney test was used to compare the performance of the trained models using all assessments to the performance of the models using only non-proprietary assessments. Additional analysis was performed, to check if identified item subsets can be scored using sum-scores, without using the trained machine learning models. Performance of the subset sum-scores was compared to the performance of existing total and subscale scores. Results Table 2 presents the final list of predicted diagnoses. Table 2 List of predicted diagnoses and abbreviations used throughout the paper. Diagnosis Abbreviation Autism Spectrum Disorder ASD ADHD-Combined Type ADHD-C Social Anxiety (Social Phobia) SAD Generalized Anxiety Disorder GAD Oppositional Defiant Disorder ODD Any Diagnosis Any No Diagnosis Given None ADHD-Inattentive Type ADHD-I Specific Learning Disorder with Impairment in Mathematics SLD-Math Language Disorder Language Specific Phobia Phobia Specific Learning Disorder with Impairment in Reading SLD-Reading Specific Learning Disorder with Impairment in Written Expression SLD-Writing Processing Speed Deficit (test-based) PS (test) Borderline Intellectual Functioning (test-based) BIF (test) NVLD (test-based) NVLD (test) NVLD without reading condition (test-based) NVLD no read (test) Specific Learning Disorder with Impairment in Mathematics (test-based) SLD-Math (test) Specific Learning Disorder with Impairment in Written Expression (test-based) SLD-Writing (test) Model performances The comparison of performance of the trained machine learning models across ten cross-validation folds, and the performance of the best existing subscale for each diagnosis is presented in Fig. 1. The Mann-Whitney test showed a significant difference between the performance of the best existing subscale and the performance of the trained machine learning models (p-value = 0.004). There was no significant difference in the performance of models trained on all assessments and the models trained on only non-proprietary assessments. Increase in performance between scores of existing scales and sum-scores of identified subsets was not statistically significant. Table 3 presents the recommended number of items, and test-set performance of the models using the recommended threshold. Table 4 presents the test-set performance of the SLD models trained on the second, extended dataset that includes LD items. Table 3 AUC, sensitivity and specificity for the recommended threshold and recommended number of items (test-set). Diagnosis N features AUC Sensitivity Specificity Autism Spectrum Disorder 4 0.865 0.865 0.738 Generalized Anxiety Disorder 5 0.840 0.811 0.672 Oppositional Defiant Disorder 3 0.810 0.851 0.667 Borderline Intellectual Functioning (test) 8 0.793 0.818 0.652 No Diagnosis Given 4 0.786 0.833 0.557 NVLD without reading condition (test) 7 0.781 0.921 0.495 ADHD-Combined Type 3 0.772 0.756 0.662 SLD with Impairment in Reading (test) 7 0.759 0.710 0.707 Specific Phobia 5 0.758 0.970 0.178 Language Disorder 7 0.752 0.758 0.539 SLD with Impairment in Written Expression (test) 7 0.749 0.886 0.461 Any Diagnosis 5 0.746 0.860 0.444 ADHD-Inattentive Type 7 0.715 0.796 0.535 Processing Speed Deficit (test) 8 0.712 0.817 0.475 SLD with Impairment in Mathematics (test) 6 0.711 0.808 0.509 SLD with Impairment in Reading 7 0.708 0.750 0.541 Social Anxiety (Social Phobia) 6 0.686 0.622 0.708 SLD with Impairment in Mathematics 6 0.668 0.875 0.363 NVLD (test) 6 0.597 0.517 0.509 Table 4 AUC, sensitivity and specificity for the recommended threshold and recommended number of items for SLD diagnoses using the expanded dataset (test-set). Diagnosis N features AUC Sensitivity Specificity SLD with Impairment in Reading (test) 5 0.855 0.838 0.686 SLD with Impairment in Reading 5 0.791 0.762 0.721 SLD with Impairment in Written Expression (test) 6 0.764 0.766 0.609 SLD with Impairment in Mathematics 3 0.687 0.600 0.672 SLD with Impairment in Mathematics (test) 6 0.670 0.679 0.572 Discussion The goal of this study was to test a standardized process for creating new assessments for mental and learning disorders using existing datasets, and to validate effectiveness of techniques identified in prior work on datasets with a high number of input variables. We used the HBN dataset to train machine learning models on parsimonious item subsets. These models can be used for assessment of common mental health and learning disorders, after acquiring consent of the authors of the original assessments. The performance values of the machine learning models provide preliminary estimates of the performance of the new assessment in a population similar to the population in the HBN study. The approach demonstrated improved performance over existing assessments on the HBN dataset, making it a valuable resource for future researchers to build new assessment instruments with improved efficacy. The code is available on github.com (Konishcheva, 2023). To apply the method to another dataset it is sufficient to update the dataset preparation package to match the standard data format expected by the model-training package. Using only items from non-proprietary assessments did not significantly affect performance, possibly due to the large item pool containing similar items. The trained model for SLD-Reading showed acceptable performance for a screening instrument and could potentially be used for question-based screening in populations similar to HBN, such as schools for children with learning differences. The model needs to be validated on the general population before being used in a more general setting. It is crucial to recalculate the recommended probability thresholds when applying the trained models to new populations with different prevalences. Since the assessment scores used as the input for the trained models are used by the clinicians in the HBN study to evaluate the consensus diagnoses, the performance of the trained models can be overestimated for consensus diagnoses. The performance of these models should only be used for comparison to the performance of other assessments scores present in the same dataset. This limitation does not apply to custom output variables based on performance on achievement tests, which are independent of the input assessment scores. Using standardized tests instead of consensus diagnoses also provides a more objective gold standard for the presence/absence of learning disorders, harmonizing the approach with existing research. Another limitation of the study is that the most informative item for SLD-Reading, and other SLD diagnoses, was from Conners self-report questionnaire, which is typically administered to children over the age of 8. Generally, it is preferable to screen for learning disorders at an earlier age. Nevertheless, having a screener for older children remains valuable, as not all communities have access to early screening, and many individuals are only diagnosed as adults or not at all. Additionally, the use of items from proprietary assessments like the Conners questionnaire raises potential licensing issues. Non-proprietary instruments that assess learning disability symptoms (e.g. Colorado Learning Difficulties Questionnaire; Willcutt et al., 2011 ) could be included in transdiagnostic studies in the future. We expect better performance for learning disorders when using a dataset containing more items assessing LD symptoms. Compared to the Colorado Learning Difficulties Questionnaire, which includes 6 items evaluating different aspects of reading difficulty, only 1 item in the whole body of items from HBN assessments includes an item directly assessing reading difficulty ("I have trouble with reading" from the Conners self-report questionnaire). It is important to acknowledge the limitations of applying machine learning methods in the mental health domain. Existing bias in diagnosis, such as gender and ethnicity bias in attributing certain diagnoses, can be exacerbated when using machine learning algorithms, as these models learn from existing diagnostic data. Furthermore, the absence of an objective gold standard test for mental health diagnoses is a fundamental challenge. Moreover, the creation of assessment tools using machine learning may inadvertently overlook rare symptoms, potentially leading to gaps in the diagnostic process. As machine learning models predominantly learn from prevalent patterns in the data, rare occurrences might not be adequately represented. One of the ways to mitigate these limitations is ongoing validation of the models using large, diverse, and representative datasets. Moving forward, we propose exploring multi-label feature selection to identify item subsets for several disorders simultaneously. Additionally, stratifying the training set by age and building separate models for different age ranges could result in more precise assessment across different age ranges. This analysis would require a larger dataset, and will become possible with new HBN releases. A larger dataset is also likely to result in a lower variance in cross-validation performance estimates. To facilitate the application of the trained models to assess new cases, we recommend the development of a user-friendly interface for assessment administration. Funding declaration Partial support of the work conducted in Paris came from Fondation Bettencourt Schueller, and partial support of the work conducted in New York at the Child Mind Institute came from Oak Foundation. Declarations Author Contribution KK wrote the main manuscript text and prepared the figures and tables. All authors contributed to the editing and review of the manuscript. Data Availability All software is available in a GitHub repository (https://github.com/charlie42/diagnosis-predictor). This software was used to analyze publicly available data from the Healthy Brain Network study (https://healthybrainnetwork.org). Funding declaration Partial support of the work conducted in Paris came from Fondation Bettencourt Schueller, and partial support of the work conducted in New York at the Child Mind Institute came from Oak Foundation. References Abidin RR. (2012). Parenting stress index–fourth edition (PSI-4). Lutz, FL: Psychological Assessment Resources . Achenbach TM. US-1st Printing edition). Integrative Guide for the 1991 CBCL/4–18, Ysr, and Trf Profiles. 1 ed. Univ Vermont/Dept Psychiatry; 1991. Achenie LEK, Scarpa A, Factor RS, Wang T, Robins DL, McCrickard DS. A Machine Learning Strategy for Autism Screening in Toddlers. J Dev Behav pediatrics: JDBP. 2019;40(5):369–76. Alexander L, Salum G, Milham M, Swanson J. (n.d.). Extended Strengths and Weaknesses Assessment of Normal Behavior (E-SWAN). Retrieved November 16, 2023, from http://www.eswan.org/ Barratt W. (2012, June 14). The Barratt Simplified Measure of Social Status (BSMSS). Social class on campus . Retrieved November 16, 2023, from http://socialclassoncampus.blogspot.com/2012/06/barratt-simplified-measure-of-social.html Bird HR, Shaffer D, Fisher P, Gould MS et al. (1993). The Columbia Impairment Scale (CIS): Pilot findings on a measure of global impairment for children and adolescents. International Journal of Methods in Psychiatric Research , 3 (3), 167–176. US: John Wiley & Sons. Birmaher B, Khetarpal S, Brent D, Cully M, Balach L, Kaufman J, Neer SM. The Screen for Child Anxiety Related Emotional Disorders (SCARED): Scale construction and psychometric characteristics. J Am Acad Child Adolesc Psychiatry. 1997;36(4):545–53. Bone D, Bishop SL, Black MP, Goodwin MS, Lord C, Narayanan SS. Use of machine learning to improve autism screening and diagnostic instruments: Effectiveness, efficiency, and multi-instrument fusion. J Child Psychol Psychiatry. 2016;57(8):927–37. Bone D, Goodwin MS, Black MP, Lee C-C, Audhkhasi K, Narayanan S. Applying machine learning to facilitate autism diagnostics: Pitfalls and promises. J Autism Dev Disord. 2015;45(5):1121–36. Brodey BB, Girgis RR, Favorov OV, Addington J, Perkins DO, Bearden CE, Woods SW, et al. The Early Psychosis Screener (EPS): Quantitative validation against the SIPS using machine learning. Schizophr Res. 2018;197:516–21. Carpenter KLH, Sprechmann P, Calderbank R, Sapiro G, Egger HL. (2016). Quantifying Risk for Anxiety Disorders in Preschool Children: A Machine Learning Approach. PLoS ONE, 11(11), e0165524. Conners K. Conners 3rd edition manual. New York: Multi-Health Systems. Inc.[Google Scholar]; 2008. Constantino JN, Davis SA, Todd RD, Schindler MK, Gross MM, Brophy SL, Metzger LM, et al. Validation of a brief quantitative measure of autistic traits: Comparison of the social responsiveness scale with the autism diagnostic interview-revised. J Autism Dev Disord. 2003;33(4):427–33. Costello EJ, Angold A. Scales to assess child and adolescent depression: Checklists, screens, and nets. J Am Acad Child Adolesc Psychiatry. 1988;27(6):726–37. drivendata. (n.d.). drivendata/cookiecutter-data-science: A logical, reasonably standardized, but flexible project structure for doing and sharing data science work. Retrieved November 16. 2023, from https://github.com/drivendata/cookiecutter-data-science Ehlers S, Gillberg C. The epidemiology of Asperger syndrome. A total population study. J Child Psychol Psychiatry Allied Discip. 1993;34(8):1327–50. Essau CA, Sasagawa S, Frick PJ. Callous-unemotional traits in a community sample of adolescents. Assessment. 2006a;13(4):454–69. Essau CA, Sasagawa S, Frick PJ. (2006b). Psychometric Properties of the Alabama Parenting Questionnaire. Journal of Child and Family Studies , 15 (5), 597–616. Germany: Springer. Gibbons RD, Chattopadhyay I, Meltzer H, Kane JM, Guinart D. Development of a Computerized Adaptive Diagnostic Screening Tool for Psychosis. Schizophr Res. 2022;245:116–21. Glavin D, Grua EM, Nakamura CA, Scazufca M, dos Santos ER, Wong GHY, Hollingworth W, et al. Patient Health Questionnaire-9 Item Pairing Predictiveness for Prescreening Depressive Symptomatology: Machine Learning Analysis. JMIR Mental Health. 2023;10(1):e48444. Gold LH. DSM-5 and the assessment of functioning: The World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0). J Am Acad Psychiatry Law. 2014;42(2):173–81. Goodman R. The Strengths and Difficulties Questionnaire: A research note. J Child Psychol Psychiatry Allied Discip. 1997;38(5):581–6. Hayes AM, Dombrowski E, Shefcyk A, Bulat J. (2018). Learning Disabilities Screening and Evaluation Guide for Low- and Middle-Income Countries . RTI Press Occasional Papers. Research Triangle Park (NC): RTI Press. Retrieved November 14, 2023, from http://www.ncbi.nlm.nih.gov/books/NBK545498/ Haynos AF, Wang SB, Lipson S, Peterson CB, Mitchell JE, Halmi KA, Agras WS, et al. Machine learning enhances prediction of illness course: A longitudinal study in eating disorders. Psychol Med. 2021;51(8):1392–402. Hetland J, Braatveit KJ, Hagen E, Lundervold AJ, Erga AH. Prevalence and Characteristics of Borderline Intellectual Functioning in a Cohort of Patients With Polysubstance Use Disorder. Front Psychiatry. 2021;12:651028. Hetrick SE, Parker AG, Hickie IB, Purcell R, Yung AR, McGorry PD. Early identification and intervention in depressive disorders: Towards a clinical staging model. Psychother Psychosom. 2008;77(5):263–70. Kaufman J, Birmaher B, Brent D, Rao U, Flynn C, Moreci P, Williamson D, et al. Initial Reliability and Validity Data. J Am Acad Child Adolesc Psychiatry. 1997;36(7):980–8. Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL). Konishcheva K. (2023, October 1). Diagnosis-predictor. Python,. Retrieved November 16, 2023, from https://github.com/charlie42/diagnosis-predictor Kramer E, Koo B, Restrepo A, Koyama M, Neuhaus R, Pugh K, Andreotti C et al. (2020). Diagnostic Associations of Processing Speed in a Transdiagnostic, Pediatric Sample. Scientific Reports , 10 (1), 10114. Nature Publishing Group. Lam KSL, Aman MG. The Repetitive Behavior Scale-Revised: Independent validation in individuals with autism spectrum disorders. J Autism Dev Disord. 2007;37(5):855–66. Lange SM. (2006). Early Identification And Interventions For Children At Risk For Learning Disabilities. Int J Special Educ, 21 . Liddle EB, Batty MJ, Goodman R. The Social Aptitudes Scale: An initial validation. Soc Psychiatry Psychiatr Epidemiol. 2009;44(6):508–13. Liu Y, Cao B, Chokka PR. Screening for Adulthood ADHD and Comorbidities in a Tertiary Mental Health Center Using EarlyDetect: A Machine Learning-Based Pilot Study. J Atten Disord. 2023;27(3):324–31. Liu Y, Chokka S, Cao B, Chokka PR. Screening for bipolar disorder in a tertiary mental health centre using EarlyDetect: A machine learning-based pilot study. J Affect Disorders Rep. 2021;6:100215. Liu Y, Hankey J, Cao B, Chokka P. Screening for major depressive disorder in a tertiary mental health centre using EarlyDetect: A machine learning-based pilot study. J Affect Disorders Rep. 2021;3:100062. Lu F, Petkova E. A comparative study of variable selection methods in the context of developing psychiatric screening instruments. Stat Med. 2014;33(3):401–21. Margolis AE, Broitman J, Davis JM, Alexander L, Hamilton A, Liao Z, Banker S et al. (2020). Estimated Prevalence of Nonverbal Learning Disability Among North American Children and Adolescents. JAMA Netw open, 3(4), e202551. Mujahid MS, Roux D, Morenoff AV, J. D., Raghunathan T. Assessing the measurement properties of neighborhood scales: From psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858–67. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;12(85):2825–30. Petterson B, Bourke J, Leonard H, Jacoby P, Bower C. Co-occurrence of birth defects and intellectual disability. Paediatr Perinat Epidemiol. 2007;21(1):65–75. Raschka S. MLxtend: Providing machine learning and data science utilities and extensions to Python’s scientific computing stack. J Open Source Softw. 2018;3(24):638. Rutter M, Bailey A, Lord C. (2003). (SCQ) Social Communication Questionnaire. Retrieved November 16, 2023, from https://www.wpspublish.com/scq-social-communication-questionnaire.html Schultebraucks K, Stevens JS, Michopoulos V, Maples-Keller J, Lyu J, Smith RN, Rothbaum BO, et al. Development and validation of a brief screener for posttraumatic stress disorder risk in emergency medical settings. Gen Hosp Psychiatry. 2023;81:46–50. Simons JS, Gaher RM. The Distress Tolerance Scale: Development and Validation of a Self-Report Measure. Motivation Emot. 2005;29(2):83–102. Steele MM. Making the Case for Early Identification and Intervention for Young Children at Risk for Learning Disabilities. Early Childhood Educ J. 2004;32(2):75–9. Stringaris A, Goodman R, Ferdinando S, Razdan V, Muhrer E, Leibenluft E, Brotman MA. The Affective Reactivity Index: A concise irritability scale for clinical and research settings. J Child Psychol Psychiatry Allied Discip. 2012;53(11):1109–17. Swanson JM, Kraemer HC, Hinshaw SP, Arnold LE, Conners CK, Abikoff HB, Clevenger W, et al. Clinical relevance of the primary findings of the MTA: Success rates based on severity of ADHD and ODD symptoms at the end of treatment. J Am Acad Child Adolesc Psychiatry. 2001;40(2):168–79. Symptom Checker. (n.d.). Child Mind Institute. Retrieved November 14. 2023, from https://childmind.org/symptomchecker/ Tartarisco G, Cicceri G, Di Pietro D, Leonardi E, Aiello S, Marino F, Chiarotti F, et al. Use of Machine Learning to Investigate the Quantitative Checklist for Autism in Toddlers (Q-CHAT) towards Early Autism Screening. Diagnostics (Basel Switzerland). 2021;11(3):574. Tutun S, Johnson ME, Ahmed A, Albizri A, Irgil S, Yesilkaya I, Ucar EN, et al. An AI-based Decision Support System for Predicting Mental Health Disorders. Inform Syst Front. 2023;25(3):1261–76. Wechsler D. (2009). Wechsler Individual Achievement Test | Third Edition. Retrieved November 16, 2023, from https://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Academic-Learning/Reading/Wechsler-Individual-Achievement-Test-%7C-Third-Edition/p/100000463.html Wechsler D. (2014). Wechsler Intelligence Scale for Children | Fifth Edition. Retrieved November 16, 2023, from https://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Cognition-%26-Neuro/Gifted-%26-Talented/Wechsler-Intelligence-Scale-for-Children-%7C-Fifth-Edition-/p/100000771.html Willcutt EG, Boada R, Riddle MW, Chhabildas N, DeFries JC, Pennington BF. Colorado Learning Difficulties Questionnaire:Validation of a parent-report screening measure. Psychol Assess. 2011;23(3):778–91. World Health Organization. (n.d.). Mental disorders. World Health Organization. Retrieved March 27. 2024, from https://www.who.int/news-room/fact-sheets/detail/mental-disorders Wu T-K, Huang S-C, Meng Y-R. (2006). Effects of Feature Selection on the Identification of Students with Learning Disabilities Using ANN. In L. Jiao, L. Wang, X. Gao, J. Liu, & F. Wu, editors, Advances in Natural Computation , Lecture Notes in Computer Science (Vol. 4221, pp. 565–574). Berlin, Heidelberg: Springer Berlin Heidelberg. Retrieved March 27, 2024, from http://link.springer.com/10.1007/11881070_77 Young KS. (1998). Internet addiction: The emergence of a new clinical disorder. CyberPsychology & Behavior , 1 (3), 237–244. Belgium: Virtual Reality Medical Institute BVBA. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 13 Jan, 2026 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted Editorial decision: Revision requested 12 May, 2025 Reviews received at journal 09 May, 2025 Reviewers agreed at journal 01 May, 2025 Reviews received at journal 07 Apr, 2025 Reviewers agreed at journal 26 Mar, 2025 Reviewers invited by journal 25 Mar, 2025 Editor invited by journal 24 Mar, 2025 Editor assigned by journal 23 Mar, 2025 Submission checks completed at journal 23 Mar, 2025 First submitted to journal 20 Mar, 2025 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-6270695","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":434298060,"identity":"200b65c2-e1a9-4ee8-a70b-2d7f7c5b0ca0","order_by":0,"name":"Kseniia Konishcheva","email":"","orcid":"","institution":"Learning Planet Institute","correspondingAuthor":false,"prefix":"","firstName":"Kseniia","middleName":"","lastName":"Konishcheva","suffix":""},{"id":434298061,"identity":"ec758e05-c924-4301-844f-2f07db0c78f5","order_by":1,"name":"Bennett Leventhal","email":"","orcid":"","institution":"University of Chicago","correspondingAuthor":false,"prefix":"","firstName":"Bennett","middleName":"","lastName":"Leventhal","suffix":""},{"id":434298062,"identity":"76b8c774-29d8-4568-816c-66a6b51b87f7","order_by":2,"name":"Maki Koyama","email":"","orcid":"","institution":"Child Mind Institute","correspondingAuthor":false,"prefix":"","firstName":"Maki","middleName":"","lastName":"Koyama","suffix":""},{"id":434298063,"identity":"ee6aa677-bae9-4141-a45d-2b98613172c4","order_by":3,"name":"Sambit Panda","email":"","orcid":"","institution":"Johns Hopkins Hospital","correspondingAuthor":false,"prefix":"","firstName":"Sambit","middleName":"","lastName":"Panda","suffix":""},{"id":434298064,"identity":"6675081d-bda0-44d7-b435-6cdddc2bd338","order_by":4,"name":"Joshua Vogelstein","email":"","orcid":"","institution":"Johns Hopkins Hospital","correspondingAuthor":false,"prefix":"","firstName":"Joshua","middleName":"","lastName":"Vogelstein","suffix":""},{"id":434298065,"identity":"74ecdf98-d725-4481-81e2-7cff06e8474d","order_by":5,"name":"Michael Milham","email":"","orcid":"","institution":"Child Mind Institute","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Milham","suffix":""},{"id":434298066,"identity":"3c426e18-b3c3-4b6b-ab63-1c58b91fa47c","order_by":6,"name":"Ariel Lindner","email":"","orcid":"","institution":"Université Paris Cité, INSERM U1284","correspondingAuthor":false,"prefix":"","firstName":"Ariel","middleName":"","lastName":"Lindner","suffix":""},{"id":434298067,"identity":"3d25ad1d-6dad-4400-af6b-0fa6459b421f","order_by":7,"name":"Arno Klein","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+UlEQVRIie3PsWrCQBzH8X8I3HS16w+i5hUuHJQMYl/lwoGTFcHFrZnSpQ8g+AydCpkVwels10xil84prg7G4CLtWbsJve/45z7/P0fkcl1ts5gY+TOfROylPhEuIKgIUxXBHwgRF4fHXkq/kNvpar4hg3YjeP7aDodovTzdzAsar7upheBtoAUVkKy5eg0mAjJfNHRMZqRthAy/A5VIMjzkPhdI8kU18TKlbSI8kscM/c/LiKhJAcXQZyekayOR4VIogyhDTwa8/guXsTJK2Ujb8GhTLjthONEfW77rtPJ3ExXlWN3bSN33hdUkSc+anzp/xeVyuf5TexbUTWphNBMSAAAAAElFTkSuQmCC","orcid":"","institution":"Child Mind Institute","correspondingAuthor":true,"prefix":"","firstName":"Arno","middleName":"","lastName":"Klein","suffix":""}],"badges":[],"createdAt":"2025-03-20 14:53:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6270695/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6270695/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12911-025-03329-5","type":"published","date":"2026-01-13T16:29:35+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":79711564,"identity":"4c7e5dd0-d7b7-41ed-bcdd-b49f2c425cd5","added_by":"auto","created_at":"2025-04-01 20:17:25","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":66010,"visible":true,"origin":"","legend":"\u003cp\u003eA. Comparison of the classification performance of machine learning models across ten cross-validation folds to the performance of the best HBN scale for the diagnosis, using the same number of items, sorted by the difference between the mean machine learning performance across folds and the performance of the best HBN scale. B. Difference between the mean machine learning performance across folds and the performance of the best HBN scale across all diagnoses.\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6270695/v1/d3100147f88d8ca63e54d24e.jpg"},{"id":100616094,"identity":"cb6af57f-26d5-44da-a5fb-f3cda1bb9044","added_by":"auto","created_at":"2026-01-19 17:39:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":871328,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6270695/v1/2ffe8abe-8e35-4791-8e56-3226f4ba3ba3.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Accurate and efficient data-driven psychiatric assessment using machine learning","fulltext":[{"header":"Key points and relevance ","content":"\u003cp\u003e- Data-driven methods, such as machine learning and feature selection, have shown promise in improving the accuracy and efficiency of mental health assessment.\u003c/p\u003e\n\u003cp\u003e- Few studies investigated these methods for creation of learning disability assessments, and in application to transdiagnostic datasets consisting of a large number of assessments.\u003c/p\u003e\n\u003cp\u003e- Using the Healthy Brain Network dataset, we have built a tool for the creation of accurate and efficient machine-learning based assessments for common mental disorders and learning disabilities.\u003c/p\u003e\n\u003cp\u003e- The modular design of the tool ensures its easy adaptation to other datasets, accommodating a variety of clinical and research applications.\u003c/p\u003e"},{"header":"Background","content":"\u003ch2\u003ePrevalence of Mental Disorders and Learning Disabilities\u003c/h2\u003e\n\u003cp\u003eGlobally, approximately one in eight people currently have a mental disorder (World Health Organization, n.d.), with roughly 30% having experienced one at some point in their life (Steele, 2004). Learning disabilities (LDs) are estimated to impact between 5 and 15% of the population (Willcutt et al., 2011). Accurate and timely identification of psychiatric and learning disorders leads to more effective intervention (Hetrick et al., 2008; Lange, 2006).\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eFeature Selection for Mental Health and Learning Disability Assessment\u003c/h2\u003e\n\u003cp\u003eIncreasing availability of large transdiagnostic datasets, coupled with advancement in technology, presents novel approaches for improvement of the assessment of mental disorders and learning disabilities. Previous studies have demonstrated the utility of machine learning in conjunction with feature selection techniques for shortening existing mental health assessments, and creating new ones based on items from existing assessments and manually constructed item banks (Achenie et al., 2019; Bone et al., 2015, 2016; Brodey et al., 2018; Carpenter, Sprechmann, Calderbank, Sapiro, \u0026amp; Egger, 2016; Gibbons, Chattopadhyay, Meltzer, Kane, \u0026amp; Guinart, 2022; Glavin et al., 2023; Liu, Chokka, Cao, \u0026amp; Chokka, 2021; Liu, Hankey, Cao, \u0026amp; Chokka, 2021; Lu \u0026amp; Petkova, 2014; Schultebraucks et al., 2023; Tartarisco et al., 2021; Tutun et al., 2023).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo our knowledge, only one study (Wu, Huang, \u0026amp; Meng, 2006) has leveraged feature selection to improve the identification of learning disabilities. Wu used scores from standardized tests as the input, commonly used for diagnosing learning disabilities. However, these tests require to be administered by a trained professional, are time consuming, and often prohibitively expensive (Hayes, Dombrowski, Shefcyk, \u0026amp; Bulat, 2018; Willcutt et al., 2011). To address these challenges, question-based screeners have been developed to identify students at risk of learning disabilities and requiring further evaluation (Willcutt et al., 2011). Feature selection techniques applied to a large item pool assessing learning disability symptoms, and symptoms commonly co-occurring conditions, can facilitate the development of improved question-based screeners, facilitating timely identification of children at risk for learning disorders.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eUsing existing datasets for assessment creation\u003c/h2\u003e\n\u003cp\u003eLarge transdiagnostic datasets provide an opportunity to use existing data to evaluate extensive item pools for assessment creation, which would be impractical in a traditional instrument development study that relies on data collection. Moreover, such datasets offer an opportunity to test items from instruments aimed at other disorders that assess seemingly unrelated constructs and may not be encompassed within a theory-driven item pool, but could nonetheless contribute to identification of assessed disorders. \u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eThe Healthy Brain Network Dataset\u003c/h2\u003e\n\u003cp\u003eOne example of such a dataset is the Healthy Brain Network (HBN) dataset. Comprising a diverse community-based sample of 4136 children and adolescents from the New York City area, it contains a comprehensive item-level psychiatric and learning assessment alongside demographic variables. Each participant undergoes assessment by a team of clinicians and receives one or several diagnoses, based on a diagnostic interview (Kiddie Schedule for Affective Disorders and Schizophrenia; Kaufman et al., 1997) and assessment scores. The dataset includes item-level responses to over 50 assessments (over 1000 items) targeting a broad range of disorders.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the HBN study, a majority of participants diagnosed with one disorder have one or multiple comorbidities, offering a realistic setting to evaluate the improved assessments as opposed to the relatively straightforward task of distinguishing affected individuals from healthy controls.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eResearch aims\u003c/h2\u003e\n\u003cp\u003eThe present study aims to create a tool for streamlined creation of effective machine learning based assessments using existing datasets, while validating techniques identified in prior studies on a dataset with a high number of input variables, taking into account the high computational complexity of the task.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe have developed a tool that leverages existing datasets to identify parsimonious item subsets for efficient and effective mental health and learning disorder assessment, and provides trained machine learning models that can be used for assessment of new patients. We test if the machine learning models trained on the HBN data provide more efficient and effective assessment compared to existing assessment instruments.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe instruments used in the HBN datasets include both proprietary and freely available assessments. We suppose that the authors of assessments that made their instruments freely accessible would be more open to allow the use of items from their assessment for creation of new instruments in the future. Due to the large number of instruments used in the HBN study, many constructs are assessed by multiple items from different instruments. We hypothesize that in the presence of a large number of input assessments, limiting the input variables to items from non-proprietary assessments will not significantly reduce the performance of the trained models. \u0026nbsp;\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eOutput variables\u003c/h2\u003e\n\u003cp\u003eIn addition to clinician diagnoses present in the HBN dataset, custom output variables were created for learning difficulties based on performance on standard tests, which are, unlike the clinician diagnoses, independent of the input assessment scores.\u003c/p\u003e\n\u003cp\u003eFollowing previous literature (Kramer et al., 2020), the criteria for the learning difficulties were defined as the Full Scale IQ over the two standard deviations below the mean, and the corresponding achievement test subscale under one standard deviation below the mean. Wechsler Individual Achievement Test (WIAT; Wechsler, 2009) and Wechsler Intelligence Scale for Children (WISC; Wechsler, 2014) scores were used. Wechsler Individual Achievement Test Spelling scale was used for written expression impairment instead of the commonly used Written Expression scale because the Written Expression scale was not present in the dataset.\u003c/p\u003e\n\u003cp\u003eBesides the Specific Learning Disorders (SLDs), we created test-based variables for Borderline Intellectual Functioning (BIF), Intellectual Disability-Mild (ID), Processing Speed Deficit (PS), and two definitions of Non-Verbal Learning Disability (NVLD and NVLD-no-read). The rules for ID, BIF, and NVLD are based on criteria from Hetland, Braatveit, Hagen, Lundervold, \u0026amp; Erga, 2021, Margolis et al., 2020, and Petterson, Bourke, Leonard, Jacoby, \u0026amp; Bower, 2007 respectively.\u003c/p\u003e\n\u003cp\u003eAll participants who were not administered the assessments required to construct the custom output variables were excluded from the analysis.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eInput variables\u003c/h2\u003e\n\u003cp\u003eAll items from all self-report and parent-report assessments were used, except items from physical fitness assessments. Additionally, responses to the pre-intake form were included, containing age, sex, and educational and developmental history of the participant.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDue to the nature of the HBN study protocol, most of the assessments were administered only to a subset of participants (i.e. not a single participant was administered all assessments). To minimize the amount of missing data, the most administered assessments were used, and the participants were limited to those who were administered all of these assessments (n=2335, 584 items).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe only assessment in the dataset evaluating symptoms of learning disorders (Conners ADHD Rating Scales \u0026ndash; Self Report Short Form; Conners, 2008) is only administered to older participants, and was not included in the final list of assessments. An additional dataset was constructed to include a larger set of assessments. Similarly to the original dataset, participants who were not administered all of the assessments were excluded, resulting in a smaller number of participants (n=1406, 1146 items).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAssessments included in the two datasets are presented in Table 1. Assessments included in the dataset with only non-proprietary assessments were: \u003cem\u003eBasic_Demos, PreInt_EduHx, PreInt_DevHx, SympChck, ASSQ, ARI_P, SDQ, SWAN, ARI_S, SCARED_P, ICU_P, APQ_P, DTS, ESWAN, MFQ_P, APQ_SR\u003c/em\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTable 1: Input assessments and abbreviations used in the paper.\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"602\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAssessment Name\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAbbreviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eDemographics: age, sex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eBasic_Demos\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eIntake Interview \u0026ndash; Education History\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003ePreInt_EduHx\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eIntake Interview \u0026ndash; Developmental History\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003ePreInt_DevHx\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eChild Mind Institute Symptom Checker (\u0026ldquo;Symptom Checker,\u0026rdquo; n.d.)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSympChck\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eSocial Communication Questionnaire (Rutter, Bailey, \u0026amp; Lord, 2003)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSCQ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eBarratt Simplified Measure of Social Status (Barratt, 2012)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eBarratt\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eAutism Spectrum Screening Questionnaire (Ehlers \u0026amp; Gillberg, 1993)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eASSQ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eAffective Reactivity Index \u0026ndash; Parent Report (Stringaris et al., 2012)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eARI_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eStrengths and Difficulties Questionnaire (Goodman, 1997)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSDQ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eStrengths and Weaknesses Assessment of ADHD and Normal Behavior (Swanson et al., 2001)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSWAN\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eAffective Reactivity Index \u0026ndash; Self Report (Stringaris et al., 2012)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eARI_S\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eSocial Responsiveness Scale-2 (Constantino et al., 2003)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSRS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eChild Behavior Checklist (Achenbach, 1991)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eCBCL\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eScreen for Child Anxiety Related Disorders \u0026ndash; Parent report (Birmaher et al., 1997)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSCARED_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eInventory of Callous-Unemotional Traits \u0026ndash; Parent Report \u003cu\u003e(Essau et al., 2006a)\u003c/u\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eICU_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eAlabama Parenting Questionnaire \u0026ndash; Parent Report Report \u003cu\u003e(Essau et al., 2006b)\u003c/u\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eAPQ_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eParent-Child Internet Addiction Test (Young, 1998)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003ePCIAT\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eDistress Tolerance Scale (Simons \u0026amp; Gaher, 2005)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eDTS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eExtended Strengths and Weaknesses Assessment of Normal Behavior-Parent Report (Alexander, Salum, Milham, \u0026amp; Swanson, n.d.)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eESWAN\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eMood and Feelings Questionnaire \u0026ndash; Parent Report (Costello \u0026amp; Angold, 1988)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eMFQ_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eAlabama Parenting Questionnaire \u003cu\u003e(Essau et al., 2006b)\u003c/u\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eAPQ_SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 602px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAssessments added in the expanded set:\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eWHO Disability Assessment Schedule \u0026ndash; Parent Report (Gold, 2014)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eWHODAS_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eThe Columbia Impairment Scale \u0026ndash; Parent Report (Bird, Shaffer, Fisher, Gould, \u0026amp; et al, 1993)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eCIS_P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eParenting Stress Index Fourth Edition (Abidin, 2012)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003ePSI\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eSocial Aptitudes Scale (Liddle, Batty, \u0026amp; Goodman, 2009)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSAS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eRepetitive Behavior Scale (Lam \u0026amp; Aman, 2007)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eRBS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003ePhenX Neighborhood Safety (Mujahid, Diez Roux, Morenoff, \u0026amp; Raghunathan, 2007)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003ePhenX_Neighborhood\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eWHO Disability Assessment Schedule \u0026ndash; Self Report (Gold, 2014)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eWHODAS_SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eThe Columbia Impairment Scale \u0026ndash; Self Report (Bird et al., 1993)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eCIS_SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eScreen for Child Anxiety Related Disorders \u0026ndash; Self report (Birmaher et al., 1997)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eSCARED_SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eConners ADHD Rating Scales \u0026ndash; Self Report Short Form (Conners, 2008)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eC3SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 488px;\"\u003e\n \u003cp\u003eMood and Feelings Questionnaire \u0026ndash; Self Report (Costello \u0026amp; Angold, 1988)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 114px;\"\u003e\n \u003cp\u003eMFQ_SR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003eModular architecture\u003c/h2\u003e\n\u003cp\u003eThe developed tool is organized in a modular fashion, to ensure minimal required changes when applying the tool to another dataset. Project structure proposed by the Cookiecutter Data Science package (drivendata, n.d.) was used, which separates data preparation and model training.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe results of executing the two modules are machine learning models trained on examined item subsets, recommended number of items, recommended probability threshold (the equivalent of the cut-off score in sum-score based assessments), and the performance of the models.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eAnalysis\u003c/h2\u003e\n\u003cp\u003eTen-fold cross-validation was used to obtain reliable performance values of the models. The models were trained only for diagnoses with more than twenty positive examples in the validation set. Elastic Net model was used, as it achieved good performance in previous research in the mental health domain (Haynos et al., 2021; Liu, Cao, \u0026amp; Chokka, 2023; Liu, Chokka, et al., 2021; Liu, Hankey, et al., 2021) and in preliminary testing, and is highly explainable. To address class imbalance (low number of positive cases compared to negative cases), positive cases were weighted higher than negative cases, the weight being automatically adjusted based on the prevalence. AUROC, specificity, and sensitivity were used as performance measures. A probability threshold was estimated in each cross-validation fold to produce well-balanced sensitivity and specificity values. Final sensitivity and specificity were calculated on a separate test-set, using the median of the identified probability thresholds (the recommended threshold).\u003c/p\u003e\n\u003cp\u003eRecursive Feature Elimination (RFE, \u003cem\u003esklearn\u003c/em\u003e python library; Pedregosa et al., 2011) and Sequential Forward Floating Selection (Raschka, 2018) were used as feature selection algorithms. Due to the high computational complexity of the SFFS algorithm and the high number of input variables, RFE was used first to identify the top twenty-seven items, then SSFS was applied to the selected items to identify smaller item subsets. Twenty-seven was chosen as the maximum acceptable length of an assessment for a single disorder, based on the number of items in the longest input assessment for a single disorder in the HBN dataset (ASSQ).\u003c/p\u003e\n\u003cp\u003eThe recommended number of items was calculated by examining the performance of the model with each number of items, and choosing the number of items that reaches 95% of the maximum performance among the evaluated subsets.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003ePerformance evaluation\u003c/h2\u003e\n\u003cp\u003eTo check if the trained machine learning models outperform traditional assessment instruments, AUROC of each total and subscale score of assessments from the HBN dataset for each predicted diagnosis was calculated. Mann-Whitney test was used to compare the performance of the best existing scale for each diagnosis to the performance of trained models using the same number of features.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe analysis was repeated using only non-proprietary assessments. Mann-Whitney test was used to compare the performance of the trained models using all assessments to the performance of the models using only non-proprietary assessments.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditional analysis was performed, to check if identified item subsets can be scored using sum-scores, without using the trained machine learning models. Performance of the subset sum-scores was compared to the performance of existing total and subscale scores.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eTable\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e presents the final list of predicted diagnoses.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eList of predicted diagnoses and abbreviations used throughout the paper.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"2\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDiagnosis\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAbbreviation\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\u003eAutism Spectrum Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eASD\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-Combined Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-C\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSocial Anxiety (Social Phobia)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSAD\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGeneralized Anxiety Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGAD\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOppositional Defiant Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eODD\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAny Diagnosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAny\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo Diagnosis Given\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-Inattentive Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-I\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Learning Disorder with Impairment in Mathematics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD-Math\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLanguage Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLanguage\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Phobia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePhobia\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Learning Disorder with Impairment in Reading\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD-Reading\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Learning Disorder with Impairment in Written Expression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD-Writing\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProcessing Speed Deficit (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePS (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBorderline Intellectual Functioning (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBIF (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD without reading condition (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD no read (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Learning Disorder with Impairment in Mathematics (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD-Math (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Learning Disorder with Impairment in Written Expression (test-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD-Writing (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eModel performances\u003c/h2\u003e\n \u003cp\u003eThe comparison of performance of the trained machine learning models across ten cross-validation folds, and the performance of the best existing subscale for each diagnosis is presented in Fig. 1.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003cp\u003eThe Mann-Whitney test showed a significant difference between the performance of the best existing subscale and the performance of the trained machine learning models (p-value\u0026thinsp;=\u0026thinsp;0.004). There was no significant difference in the performance of models trained on all assessments and the models trained on only non-proprietary assessments. Increase in performance between scores of existing scales and sum-scores of identified subsets was not statistically significant.\u003c/p\u003e\n \u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e presents the recommended number of items, and test-set performance of the models using the recommended threshold. Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e presents the test-set performance of the SLD models trained on the second, extended dataset that includes LD items.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eAUC, sensitivity and specificity for the recommended threshold and recommended number of items (test-set).\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDiagnosis\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eN features\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\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\u003eAutism Spectrum Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.865\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.865\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.738\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGeneralized Anxiety Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.840\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.672\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOppositional Defiant Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.810\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.851\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.667\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBorderline Intellectual Functioning (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.793\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.652\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo Diagnosis Given\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.786\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.833\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.557\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD without reading condition (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.781\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.921\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.495\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-Combined Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.772\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.756\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.662\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Reading (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.759\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.710\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.707\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecific Phobia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.758\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.970\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.178\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLanguage Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.752\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.758\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.539\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Written Expression (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.749\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.886\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.461\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAny Diagnosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.746\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.860\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.444\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eADHD-Inattentive Type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.715\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.796\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.535\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProcessing Speed Deficit (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.712\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.817\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.475\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Mathematics (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.711\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.808\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.509\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Reading\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.708\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.750\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.541\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSocial Anxiety (Social Phobia)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.686\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.708\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Mathematics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.668\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.875\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.363\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNVLD (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.597\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.517\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.509\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eAUC, sensitivity and specificity for the recommended threshold and recommended number of items for SLD diagnoses using the expanded dataset (test-set).\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDiagnosis\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eN features\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\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\u003eSLD with Impairment in Reading (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.855\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.838\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.686\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Reading\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.791\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.762\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.721\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Written Expression (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.766\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.609\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Mathematics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.687\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.672\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSLD with Impairment in Mathematics (test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.670\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.679\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.572\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe goal of this study was to test a standardized process for creating new assessments for mental and learning disorders using existing datasets, and to validate effectiveness of techniques identified in prior work on datasets with a high number of input variables. We used the HBN dataset to train machine learning models on parsimonious item subsets. These models can be used for assessment of common mental health and learning disorders, after acquiring consent of the authors of the original assessments. The performance values of the machine learning models provide preliminary estimates of the performance of the new assessment in a population similar to the population in the HBN study.\u003c/p\u003e \u003cp\u003eThe approach demonstrated improved performance over existing assessments on the HBN dataset, making it a valuable resource for future researchers to build new assessment instruments with improved efficacy. The code is available on github.com (Konishcheva, 2023). To apply the method to another dataset it is sufficient to update the dataset preparation package to match the standard data format expected by the model-training package.\u003c/p\u003e \u003cp\u003eUsing only items from non-proprietary assessments did not significantly affect performance, possibly due to the large item pool containing similar items.\u003c/p\u003e \u003cp\u003eThe trained model for SLD-Reading showed acceptable performance for a screening instrument and could potentially be used for question-based screening in populations similar to HBN, such as schools for children with learning differences. The model needs to be validated on the general population before being used in a more general setting. It is crucial to recalculate the recommended probability thresholds when applying the trained models to new populations with different prevalences.\u003c/p\u003e \u003cp\u003eSince the assessment scores used as the input for the trained models are used by the clinicians in the HBN study to evaluate the consensus diagnoses, the performance of the trained models can be overestimated for consensus diagnoses. The performance of these models should only be used for comparison to the performance of other assessments scores present in the same dataset. This limitation does not apply to custom output variables based on performance on achievement tests, which are independent of the input assessment scores. Using standardized tests instead of consensus diagnoses also provides a more objective gold standard for the presence/absence of learning disorders, harmonizing the approach with existing research.\u003c/p\u003e \u003cp\u003eAnother limitation of the study is that the most informative item for SLD-Reading, and other SLD diagnoses, was from Conners self-report questionnaire, which is typically administered to children over the age of 8. Generally, it is preferable to screen for learning disorders at an earlier age. Nevertheless, having a screener for older children remains valuable, as not all communities have access to early screening, and many individuals are only diagnosed as adults or not at all. Additionally, the use of items from proprietary assessments like the Conners questionnaire raises potential licensing issues. Non-proprietary instruments that assess learning disability symptoms (e.g. Colorado Learning Difficulties Questionnaire; Willcutt et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) could be included in transdiagnostic studies in the future. We expect better performance for learning disorders when using a dataset containing more items assessing LD symptoms. Compared to the Colorado Learning Difficulties Questionnaire, which includes 6 items evaluating different aspects of reading difficulty, only 1 item in the whole body of items from HBN assessments includes an item directly assessing reading difficulty (\"I have trouble with reading\" from the Conners self-report questionnaire).\u003c/p\u003e \u003cp\u003eIt is important to acknowledge the limitations of applying machine learning methods in the mental health domain. Existing bias in diagnosis, such as gender and ethnicity bias in attributing certain diagnoses, can be exacerbated when using machine learning algorithms, as these models learn from existing diagnostic data. Furthermore, the absence of an objective gold standard test for mental health diagnoses is a fundamental challenge. Moreover, the creation of assessment tools using machine learning may inadvertently overlook rare symptoms, potentially leading to gaps in the diagnostic process. As machine learning models predominantly learn from prevalent patterns in the data, rare occurrences might not be adequately represented. One of the ways to mitigate these limitations is ongoing validation of the models using large, diverse, and representative datasets.\u003c/p\u003e \u003cp\u003eMoving forward, we propose exploring multi-label feature selection to identify item subsets for several disorders simultaneously. Additionally, stratifying the training set by age and building separate models for different age ranges could result in more precise assessment across different age ranges. This analysis would require a larger dataset, and will become possible with new HBN releases. A larger dataset is also likely to result in a lower variance in cross-validation performance estimates.\u003c/p\u003e \u003cp\u003eTo facilitate the application of the trained models to assess new cases, we recommend the development of a user-friendly interface for assessment administration.\u003c/p\u003e \u003cp\u003eFunding declaration\u003c/p\u003e \u003cp\u003ePartial support of the work conducted in Paris came from Fondation Bettencourt Schueller, and partial support of the work conducted in New York at the Child Mind Institute came from Oak Foundation.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eKK wrote the main manuscript text and prepared the figures and tables. All authors contributed to the editing and review of the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll software is available in a GitHub repository (https://github.com/charlie42/diagnosis-predictor). This software was used to analyze publicly available data from the Healthy Brain Network study (https://healthybrainnetwork.org).\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eFunding declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePartial support of the work conducted in Paris came from Fondation Bettencourt Schueller, and partial support of the work conducted in New York at the Child Mind Institute came from Oak Foundation.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbidin RR. (2012). Parenting stress index\u0026ndash;fourth edition (PSI-4). \u003cem\u003eLutz, FL: Psychological Assessment Resources\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAchenbach TM. US-1st Printing edition). Integrative Guide for the 1991 CBCL/4\u0026ndash;18, Ysr, and Trf Profiles. 1 ed. Univ Vermont/Dept Psychiatry; 1991.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAchenie LEK, Scarpa A, Factor RS, Wang T, Robins DL, McCrickard DS. A Machine Learning Strategy for Autism Screening in Toddlers. J Dev Behav pediatrics: JDBP. 2019;40(5):369\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlexander L, Salum G, Milham M, Swanson J. (n.d.). Extended Strengths and Weaknesses Assessment of Normal Behavior (E-SWAN). Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.eswan.org/\u003c/span\u003e\u003cspan address=\"http://www.eswan.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarratt W. (2012, June 14). The Barratt Simplified Measure of Social Status (BSMSS). \u003cem\u003eSocial class on campus\u003c/em\u003e. Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://socialclassoncampus.blogspot.com/2012/06/barratt-simplified-measure-of-social.html\u003c/span\u003e\u003cspan address=\"http://socialclassoncampus.blogspot.com/2012/06/barratt-simplified-measure-of-social.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBird HR, Shaffer D, Fisher P, Gould MS et al. (1993). The Columbia Impairment Scale (CIS): Pilot findings on a measure of global impairment for children and adolescents. \u003cem\u003eInternational Journal of Methods in Psychiatric Research\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(3), 167\u0026ndash;176. US: John Wiley \u0026amp; Sons.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBirmaher B, Khetarpal S, Brent D, Cully M, Balach L, Kaufman J, Neer SM. The Screen for Child Anxiety Related Emotional Disorders (SCARED): Scale construction and psychometric characteristics. J Am Acad Child Adolesc Psychiatry. 1997;36(4):545\u0026ndash;53.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBone D, Bishop SL, Black MP, Goodwin MS, Lord C, Narayanan SS. Use of machine learning to improve autism screening and diagnostic instruments: Effectiveness, efficiency, and multi-instrument fusion. J Child Psychol Psychiatry. 2016;57(8):927\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBone D, Goodwin MS, Black MP, Lee C-C, Audhkhasi K, Narayanan S. Applying machine learning to facilitate autism diagnostics: Pitfalls and promises. J Autism Dev Disord. 2015;45(5):1121\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrodey BB, Girgis RR, Favorov OV, Addington J, Perkins DO, Bearden CE, Woods SW, et al. The Early Psychosis Screener (EPS): Quantitative validation against the SIPS using machine learning. Schizophr Res. 2018;197:516\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarpenter KLH, Sprechmann P, Calderbank R, Sapiro G, Egger HL. (2016). Quantifying Risk for Anxiety Disorders in Preschool Children: A Machine Learning Approach. PLoS ONE, 11(11), e0165524.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConners K. Conners 3rd edition manual. New York: Multi-Health Systems. Inc.[Google Scholar]; 2008.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConstantino JN, Davis SA, Todd RD, Schindler MK, Gross MM, Brophy SL, Metzger LM, et al. Validation of a brief quantitative measure of autistic traits: Comparison of the social responsiveness scale with the autism diagnostic interview-revised. J Autism Dev Disord. 2003;33(4):427\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCostello EJ, Angold A. Scales to assess child and adolescent depression: Checklists, screens, and nets. J Am Acad Child Adolesc Psychiatry. 1988;27(6):726\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003edrivendata. (n.d.). drivendata/cookiecutter-data-science: A logical, reasonably standardized, but flexible project structure for doing and sharing data science work. Retrieved November 16. 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/drivendata/cookiecutter-data-science\u003c/span\u003e\u003cspan address=\"https://github.com/drivendata/cookiecutter-data-science\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEhlers S, Gillberg C. The epidemiology of Asperger syndrome. A total population study. J Child Psychol Psychiatry Allied Discip. 1993;34(8):1327\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEssau CA, Sasagawa S, Frick PJ. Callous-unemotional traits in a community sample of adolescents. Assessment. 2006a;13(4):454\u0026ndash;69.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEssau CA, Sasagawa S, Frick PJ. (2006b). Psychometric Properties of the Alabama Parenting Questionnaire. \u003cem\u003eJournal of Child and Family Studies\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(5), 597\u0026ndash;616. Germany: Springer.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGibbons RD, Chattopadhyay I, Meltzer H, Kane JM, Guinart D. Development of a Computerized Adaptive Diagnostic Screening Tool for Psychosis. Schizophr Res. 2022;245:116\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGlavin D, Grua EM, Nakamura CA, Scazufca M, dos Santos ER, Wong GHY, Hollingworth W, et al. Patient Health Questionnaire-9 Item Pairing Predictiveness for Prescreening Depressive Symptomatology: Machine Learning Analysis. JMIR Mental Health. 2023;10(1):e48444.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGold LH. DSM-5 and the assessment of functioning: The World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0). J Am Acad Psychiatry Law. 2014;42(2):173\u0026ndash;81.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoodman R. The Strengths and Difficulties Questionnaire: A research note. J Child Psychol Psychiatry Allied Discip. 1997;38(5):581\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayes AM, Dombrowski E, Shefcyk A, Bulat J. (2018). \u003cem\u003eLearning Disabilities Screening and Evaluation Guide for Low- and Middle-Income Countries\u003c/em\u003e. RTI Press Occasional Papers. Research Triangle Park (NC): RTI Press. Retrieved November 14, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ncbi.nlm.nih.gov/books/NBK545498/\u003c/span\u003e\u003cspan address=\"http://www.ncbi.nlm.nih.gov/books/NBK545498/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaynos AF, Wang SB, Lipson S, Peterson CB, Mitchell JE, Halmi KA, Agras WS, et al. Machine learning enhances prediction of illness course: A longitudinal study in eating disorders. Psychol Med. 2021;51(8):1392\u0026ndash;402.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHetland J, Braatveit KJ, Hagen E, Lundervold AJ, Erga AH. Prevalence and Characteristics of Borderline Intellectual Functioning in a Cohort of Patients With Polysubstance Use Disorder. Front Psychiatry. 2021;12:651028.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHetrick SE, Parker AG, Hickie IB, Purcell R, Yung AR, McGorry PD. Early identification and intervention in depressive disorders: Towards a clinical staging model. Psychother Psychosom. 2008;77(5):263\u0026ndash;70.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaufman J, Birmaher B, Brent D, Rao U, Flynn C, Moreci P, Williamson D, et al. Initial Reliability and Validity Data. J Am Acad Child Adolesc Psychiatry. 1997;36(7):980\u0026ndash;8. Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKonishcheva K. (2023, October 1). Diagnosis-predictor. Python,. Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/charlie42/diagnosis-predictor\u003c/span\u003e\u003cspan address=\"https://github.com/charlie42/diagnosis-predictor\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKramer E, Koo B, Restrepo A, Koyama M, Neuhaus R, Pugh K, Andreotti C et al. (2020). Diagnostic Associations of Processing Speed in a Transdiagnostic, Pediatric Sample. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1), 10114. Nature Publishing Group.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLam KSL, Aman MG. The Repetitive Behavior Scale-Revised: Independent validation in individuals with autism spectrum disorders. J Autism Dev Disord. 2007;37(5):855\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLange SM. (2006). Early Identification And Interventions For Children At Risk For Learning Disabilities. Int J Special Educ, \u003cem\u003e21\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiddle EB, Batty MJ, Goodman R. The Social Aptitudes Scale: An initial validation. Soc Psychiatry Psychiatr Epidemiol. 2009;44(6):508\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Cao B, Chokka PR. Screening for Adulthood ADHD and Comorbidities in a Tertiary Mental Health Center Using EarlyDetect: A Machine Learning-Based Pilot Study. J Atten Disord. 2023;27(3):324\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Chokka S, Cao B, Chokka PR. Screening for bipolar disorder in a tertiary mental health centre using EarlyDetect: A machine learning-based pilot study. J Affect Disorders Rep. 2021;6:100215.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Hankey J, Cao B, Chokka P. Screening for major depressive disorder in a tertiary mental health centre using EarlyDetect: A machine learning-based pilot study. J Affect Disorders Rep. 2021;3:100062.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu F, Petkova E. A comparative study of variable selection methods in the context of developing psychiatric screening instruments. Stat Med. 2014;33(3):401\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMargolis AE, Broitman J, Davis JM, Alexander L, Hamilton A, Liao Z, Banker S et al. (2020). Estimated Prevalence of Nonverbal Learning Disability Among North American Children and Adolescents. JAMA Netw open, 3(4), e202551.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMujahid MS, Roux D, Morenoff AV, J. D., Raghunathan T. Assessing the measurement properties of neighborhood scales: From psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;12(85):2825\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetterson B, Bourke J, Leonard H, Jacoby P, Bower C. Co-occurrence of birth defects and intellectual disability. Paediatr Perinat Epidemiol. 2007;21(1):65\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaschka S. MLxtend: Providing machine learning and data science utilities and extensions to Python\u0026rsquo;s scientific computing stack. J Open Source Softw. 2018;3(24):638.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRutter M, Bailey A, Lord C. (2003). (SCQ) Social Communication Questionnaire. Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.wpspublish.com/scq-social-communication-questionnaire.html\u003c/span\u003e\u003cspan address=\"https://www.wpspublish.com/scq-social-communication-questionnaire.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchultebraucks K, Stevens JS, Michopoulos V, Maples-Keller J, Lyu J, Smith RN, Rothbaum BO, et al. Development and validation of a brief screener for posttraumatic stress disorder risk in emergency medical settings. Gen Hosp Psychiatry. 2023;81:46\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimons JS, Gaher RM. The Distress Tolerance Scale: Development and Validation of a Self-Report Measure. Motivation Emot. 2005;29(2):83\u0026ndash;102.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteele MM. Making the Case for Early Identification and Intervention for Young Children at Risk for Learning Disabilities. Early Childhood Educ J. 2004;32(2):75\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStringaris A, Goodman R, Ferdinando S, Razdan V, Muhrer E, Leibenluft E, Brotman MA. The Affective Reactivity Index: A concise irritability scale for clinical and research settings. J Child Psychol Psychiatry Allied Discip. 2012;53(11):1109\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSwanson JM, Kraemer HC, Hinshaw SP, Arnold LE, Conners CK, Abikoff HB, Clevenger W, et al. Clinical relevance of the primary findings of the MTA: Success rates based on severity of ADHD and ODD symptoms at the end of treatment. J Am Acad Child Adolesc Psychiatry. 2001;40(2):168\u0026ndash;79.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSymptom Checker. (n.d.). Child Mind Institute. Retrieved November 14. 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://childmind.org/symptomchecker/\u003c/span\u003e\u003cspan address=\"https://childmind.org/symptomchecker/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTartarisco G, Cicceri G, Di Pietro D, Leonardi E, Aiello S, Marino F, Chiarotti F, et al. Use of Machine Learning to Investigate the Quantitative Checklist for Autism in Toddlers (Q-CHAT) towards Early Autism Screening. Diagnostics (Basel Switzerland). 2021;11(3):574.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTutun S, Johnson ME, Ahmed A, Albizri A, Irgil S, Yesilkaya I, Ucar EN, et al. An AI-based Decision Support System for Predicting Mental Health Disorders. Inform Syst Front. 2023;25(3):1261\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWechsler D. (2009). Wechsler Individual Achievement Test | Third Edition. Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Academic-Learning/Reading/Wechsler-Individual-Achievement-Test-%7C-Third-Edition/p/100000463.html\u003c/span\u003e\u003cspan address=\"https://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Academic-Learning/Reading/Wechsler-Individual-Achievement-Test-%7C-Third-Edition/p/100000463.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWechsler D. (2014). Wechsler Intelligence Scale for Children | Fifth Edition. Retrieved November 16, 2023, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Cognition-%26-Neuro/Gifted-%26-Talented/Wechsler-Intelligence-Scale-for-Children-%7C-Fifth-Edition-/p/100000771.html\u003c/span\u003e\u003cspan address=\"https://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Cognition-%26-Neuro/Gifted-%26-Talented/Wechsler-Intelligence-Scale-for-Children-%7C-Fifth-Edition-/p/100000771.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWillcutt EG, Boada R, Riddle MW, Chhabildas N, DeFries JC, Pennington BF. Colorado Learning Difficulties Questionnaire:Validation of a parent-report screening measure. Psychol Assess. 2011;23(3):778\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. (n.d.). Mental disorders. World Health Organization. Retrieved March 27. 2024, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/news-room/fact-sheets/detail/mental-disorders\u003c/span\u003e\u003cspan address=\"https://www.who.int/news-room/fact-sheets/detail/mental-disorders\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu T-K, Huang S-C, Meng Y-R. (2006). Effects of Feature Selection on the Identification of Students with Learning Disabilities Using ANN. In L. Jiao, L. Wang, X. Gao, J. Liu, \u0026amp; F. Wu, editors, \u003cem\u003eAdvances in Natural Computation\u003c/em\u003e, Lecture Notes in Computer Science (Vol. 4221, pp. 565\u0026ndash;574). Berlin, Heidelberg: Springer Berlin Heidelberg. Retrieved March 27, 2024, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://link.springer.com/10.1007/11881070_77\u003c/span\u003e\u003cspan address=\"http://link.springer.com/10.1007/11881070_77\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoung KS. (1998). Internet addiction: The emergence of a new clinical disorder. \u003cem\u003eCyberPsychology \u0026amp; Behavior\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(3), 237\u0026ndash;244. Belgium: Virtual Reality Medical Institute BVBA.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"machine learning, mental health assessment, psychiatry, learning disabilities, learning disability assessment, feature selection, Healthy Brain Network study","lastPublishedDoi":"10.21203/rs.3.rs-6270695/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6270695/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eAccurate assessment of mental disorders and learning disabilities is essential for timely intervention. Machine learning and feature selection techniques have potential in improving the accuracy and efficiency of mental health assessments. However, limited research has explored the use of large transdiagnostic datasets, as well as the application of these techniques in developing quick, question-based assessments. The goals of this study are to apply machine learning and feature selection techniques to a large transdiagnostic dataset featuring a high number of assessment items, and to create a tool for the streamlined creation of efficient and effective assessment using existing datasets.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study uses the Healthy Brain Network (HBN) dataset to develop a tool for creation of efficient and effective machine learning-based assessment of mental disorders and learning disabilities. Feature selection algorithms were applied to identify parsimonious item subsets. Modular architecture ensures straightforward application to other datasets.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eMachine learning models trained on the HBN data exhibited improved performance over existing assessments. Using only non-proprietary assessments did not significantly impact model performance.\u003c/p\u003e\u003ch2\u003eDiscussion\u003c/h2\u003e \u003cp\u003eThis study demonstrates the feasibility of using existing large-scale datasets for creating accurate and efficient assessments for mental disorders and learning disabilities. The performance of the machine learning models provide estimates of the performance of the new assessments in a population similar to HBN. The modular architecture of the developed tool ensures seamless application to diverse clinical and research contexts.\u003c/p\u003e","manuscriptTitle":"Accurate and efficient data-driven psychiatric assessment using machine learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-01 20:17:20","doi":"10.21203/rs.3.rs-6270695/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-05-12T20:16:25+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-09T09:00:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"237682347518624240700227288100118273635","date":"2025-05-01T05:21:58+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-07T10:15:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"279951969135245033448465833478998268145","date":"2025-03-26T06:06:43+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-03-25T20:53:33+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-03-24T05:39:41+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-03-23T16:09:27+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-03-23T16:07:18+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-03-20T14:39:57+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8c1251a1-776c-48de-90b2-05764491e75e","owner":[],"postedDate":"April 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-01-19T17:04:59+00:00","versionOfRecord":{"articleIdentity":"rs-6270695","link":"https://doi.org/10.1186/s12911-025-03329-5","journal":{"identity":"bmc-medical-informatics-and-decision-making","isVorOnly":false,"title":"BMC Medical Informatics and Decision Making"},"publishedOn":"2026-01-13 16:29:35","publishedOnDateReadable":"January 13th, 2026"},"versionCreatedAt":"2025-04-01 20:17:20","video":"","vorDoi":"10.1186/s12911-025-03329-5","vorDoiUrl":"https://doi.org/10.1186/s12911-025-03329-5","workflowStages":[]},"version":"v1","identity":"rs-6270695","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6270695","identity":"rs-6270695","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}