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
[Preprint]. 2024 May 6:2024.05.06.24306938.
doi: 10.1101/2024.05.06.24306938.

Forecasting High-Risk Behavioral and Medical Events in Children with Autism through Analysis of Digital Behavioral Records

Yashar Kiarashi  1 , Johanna Lantz  2 , Matthew A Reyna  1 , Conor Anderson  2 , Ali Bahrami Rad  1 , Jenny Foster  2 , Tania Villavicencio  2 , Theresa Hamlin  2 , Gari D Clifford  1   3
Affiliations

Forecasting High-Risk Behavioral and Medical Events in Children with Autism through Analysis of Digital Behavioral Records

Yashar Kiarashi et al. medRxiv. .

Abstract

Individuals with Autism Spectrum Disorder may display interfering behaviors that limit their inclusion in educational and community settings, negatively impacting their quality of life. These behaviors may also signal potential medical conditions or indicate upcoming high-risk behaviors. This study explores behavior patterns that precede high-risk, challenging behaviors or seizures the following day. We analyzed an existing dataset of behavior and seizure data from 331 children with profound ASD over nine years. We developed a deep learning-based algorithm designed to predict the likelihood of aggression, elopement, and self-injurious behavior (SIB) as three high-risk behavioral events, as well as seizure episodes as a high-risk medical event occurring the next day. The proposed model attained accuracies of 78.4%, 80.68%, 85.43%, and 69.95% for predicting the next-day occurrence of aggression, SIB, elopement, and seizure episodes, respectively. The results were proven significant for more than 95% of the population for all high-risk event predictions using permutation-based statistical tests. Our findings emphasize the potential of leveraging historical behavior data for the early detection of high-risk behavioral and medical events, paving the way for behavioral interventions and improved support in both social and educational environments.

PubMed Disclaimer

Figures

Figure 1:
Figure 1:
Schematic of the predictive analysis workflow. a) The feature vector is composed of seven selected behaviors, determined to be most prevalent across the study population, along with any other observed behaviors, each represented in binary form. This information forms a two-dimensional feature vector covering a timespan of the past seven/fourteen days, with the label indicating the occurrence of challenging behaviors and seizure episode on the subsequent day. Cells with lower opacity represent records from previous days, the feature vector consists of the cells enclosed by the black rectangle, and the yellow cells represent the upcoming days for which we aim to predict the presence of high-risk events. b) The diagram illustrates the deep learning model designed for the binary prediction of the occurrence of high-risk behaviorial and medical events on the following day. The model architecture includes two-dimensional convolutional layers, batch normalization, max pooling, dense layers with ReLU activation, a dropout layer for regularization, and a final dense layer with a sigmoid activation function for outputting the likelihood for an individual displaying a given behavior (e.g., self-injurious behavior, elopement, or seizure episode) the following day.
Figure 2:
Figure 2:
Confusion matrices for predicting aggression, self-injurious behavior (SIB), elopement, and seizure. Sub-tables (a), (b), (c), and (d) illustrate confusion matrices predict these behaviors based on a 7-day historical period. Sub-tables (e), (f),(g), and (h) predict these behaviors based on a 14-day period.
Figure 3:
Figure 3:
Representation of features importance using GradCAM for predicting high-risk events. Day -j refers to the jth day preceding the target day for which the prediction is being made.(a), (b), (c), and (d) illustrate the feature importance rankings for predicting aggression, SIB, elopement, and seizures, respectively, based on a 7-day window. (e), (f), (g), and (h) extend the analysis to a 14-day window for the same outcomes. The comparison highlights how the predictive value of specific features shifts with the extension of the historical data period.
Figure 4:
Figure 4:
Representation of prediction results using 7-day and 14-day spans of prior historical data for predicting SIB, elopement, and seizure, presenting AUROC, AUPRC, accuracy, and F1 score. Blue and yellow circles represent cases where we achieved, and could not achieve, statistical significance, respectively. Subfigures (a), (c), (e), and (g) depict the results for aggresion, SIB, elopement, and seizure using 7-day data, while (b), (d), (f), and (h) illustrate the results for SIB, elopement, and seizure using 14-day data.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Abbasi B. and Goldenholz D.. Machine learning applications in epilepsy. Epilepsia, 60:2037–2047, 2019. doi:10.1111/epi.16333. - DOI - PMC - PubMed
    1. Allik H., Larsson J., and Smedje H.. sleep patterns in school-age children with asperger syndrome or high-functioning autism: a follow-up study. Journal of autism and Developmental Disorders, 38:1625–1633, 2008. doi:10.1007/s10803-008-0543-0. - DOI - PubMed
    1. Anders T., Iosif A., Schwichtenberg A., Tang K., and Goodlin-Jones B.. six-month sleep–wake organization and stability in preschool-age children with autism, developmental delay, and typical development. Behavioral sleep Medicine, 9:92–106, 2011. doi:10.1080/15402002.2011.557991. - DOI - PMC - PubMed
    1. Beniczky S., Conradsen I., Henning O., Fabricius M., and Wolf P.. Automated real-time detection of tonic-clonic seizures using a wearable emg device. Neurology, 90, 2018. doi:10.1212/wnl.0000000000004893. - DOI - PMC - PubMed
    1. Buschbacher P. and Fox L.. understanding and intervening with the challenging behavior of young children with autism spectrum disorder. Language Speech and Hearing Services in Schools, 34:217–227, 2003. doi:10.1044/0161-1461(2003/018). - DOI - PubMed

Publication types