Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jul;48(7):2418-2433.
doi: 10.1007/s10803-018-3509-x.

Prediction of Autism at 3 Years from Behavioural and Developmental Measures in High-Risk Infants: A Longitudinal Cross-Domain Classifier Analysis

G Bussu  1 , E J H Jones  2 , T Charman  3 , M H Johnson  2 , J K Buitelaar  4 ; BASIS Team
Collaborators, Affiliations

Prediction of Autism at 3 Years from Behavioural and Developmental Measures in High-Risk Infants: A Longitudinal Cross-Domain Classifier Analysis

G Bussu et al. J Autism Dev Disord. 2018 Jul.

Abstract

We integrated multiple behavioural and developmental measures from multiple time-points using machine learning to improve early prediction of individual Autism Spectrum Disorder (ASD) outcome. We examined Mullen Scales of Early Learning, Vineland Adaptive Behavior Scales, and early ASD symptoms between 8 and 36 months in high-risk siblings (HR; n = 161) and low-risk controls (LR; n = 71). Longitudinally, LR and HR-Typical showed higher developmental level and functioning, and fewer ASD symptoms than HR-Atypical and HR-ASD. At 8 months, machine learning classified HR-ASD at chance level, and broader atypical development with 69.2% Area Under the Curve (AUC). At 14 months, ASD and broader atypical development were classified with approximately 71% AUC. Thus, prediction of ASD was only possible with moderate accuracy at 14 months.

Keywords: Autism; Data integration; Early prediction; High-risk; Individual prediction; Longitudinal study; Machine learning.

PubMed Disclaimer

Conflict of interest statement

Conflict of interest

JKB has been a consultant to/ member of advisory board of/and/or speaker for Janssen Cilag BV, Eli Lilly, Lundbeck, Shire, Roche, Novartis, Medice and Servier. He is neither an employee nor a stock shareholder of any of these companies. The present work is unrelated to these relationships. The other authors declare not to have competing interests.

Ethical Approval

Ethical approval and informed consent were made available for the current study through the BASIS. Ethical approval for BASIS Phase 1 and Phase 2 was given by the London Central NREC (06/MRE02/73) on 28 August 2007, covering the collection of phenotypic data and saliva samples from the infants.

Figures

Fig. 1
Fig. 1
Developmental trajectories of estimated means for MSEL measures by clinical outcome groups. This figure shows the longitudinal trajectory of scores per outcome groups (LR, HR-Typical, HR-Atypical, HR-ASD) obtained through multilevel mixed modelling for each scale of the MSEL. The developmental trajectories are built on four time-points, one for each visit, which are approximately: 8; 14; 24; 36 months. 95% bootstrap confidence interval on group trajectories is shown as shaded area. Individual scores are also shown (points) with different colours by outcome group. The average normative score is shown by the red line. MSEL Mullen Scales of Early Learning, LR low-risk controls, HR high-risk siblings
Fig. 2
Fig. 2
Developmental trajectories of estimated means for VABS measures by clinical outcome groups. This figure shows the longitudinal trajectory of scores per outcome groups (LR, HR-Typical, HR-Atypical, HR-ASD) obtained through multilevel mixed modelling for each scale of the VABS. The developmental trajectories are built on four time-points, one for each visit, which are approximately: 8; 14; 24; 36 months. 95% bootstrap confidence interval on group trajectories is shown as shaded area. Individual scores are also shown (points) with different colours by outcome group. The average normative score is shown by the red line. VABS Vineland Adaptive Behavior Scales, LR low-risk controls, HR high-risk siblings
Fig. 3
Fig. 3
Prediction of ASD clinical outcome at 36m: AUC. In this figure the area under the curve (AUC) is reported for different classifiers based on behavioural measures (MSEL, VABS and AOSI) and their combination at different time-points (8 months, 8 months + change factor, 14 months). The classification made is between high-risk infants who are going to develop ASD at 36 m, and high-risk infants with typical and atypical (but not ASD) outcome at 36 m. The change factor is computed as the difference between measures at 14 and 8 months over the age difference between the two visits. The 95% confidence interval is also reported for each classifier. AUC area under the curve, MSEL Mullen Scales of Early Learning (5 scores), VABS Vineland Adaptive Behavior Scales (4 scores); AOSI Autism Observation Scale for Infants, in this study we considered the total score
Fig. 4
Fig. 4
Prediction of atypical clinical outcome (including ASD) at 36 m: AUC. In this figure the AUC is reported for different classifiers based on behavioural measures (MSEL, VABS and AOSI) and their combination at different time-points (8 months, 8 months + change factor, 14 months). The classification made is between high-risk infants with atypical development (including an ASD diagnosis at 36 m), and high-risk infants with typical outcome at 36 m. The change factor is computed as the difference between measures at 14 and 8 months over the age difference between the two visits. The 95% confidence interval is also reported for each classifier. AUC area under the curve, MSEL Mullen Scales of Early Learning (5 scores), VABS Vineland Adaptive Behavior Scales (4 scores); AOSI Autism Observation Scale for Infants, in this study we considered the total score
See this image and copyright information in PMC

References

    1. American Psychiatric Association . Diagnostic and Statistical Manual of Mental Disorders. 5. Washington, DC: American Psychiatric Publishing; 2013.
    1. Arbabshirani MR, Plis S, Sui J, Calhoun VD. Single subject prediction of brain disorders in neuroimaging: Promises and pitfalls. Neuroimage. 2017;145:137–165. doi: 10.1016/j.neuroimage.2016.02.079. - DOI - PMC - PubMed
    1. Barbaro J, Dissanayake C. Developmental profiles of infants and toddlers with autism spectrum disorders identified prospectively in a community-based setting. Journal of Autism and Developmental Disorders. 2012;42(9):1939–1948. doi: 10.1007/s10803-012-1441-z. - DOI - PubMed
    1. Bates D, Machler M, Bolker B, Walker S. Fitting linear mixed-effects models using {lme4} Journal of Statistical Software. 2015;67:1–48. doi: 10.18637/jss.v067.i01. - DOI
    1. Brian J, Bryson SE, Garon N, Roberts W, Smith IM, Szatmari P, Zwaigenbaum L. Clinical assessment of autism in high-risk 18-month-olds. Autism. 2008;12(5):433–456. doi: 10.1177/1362361308094500. - DOI - PubMed