Identification of newborns at risk for autism using electronic medical records and machine learning
Rayees Rahman1, Arad Kodesh2,3, Stephen Z Levine3
1Department of Pharmacological Sciences, Icahn School of Medicine at Mount Sinai, New York, New York, USA.
Insights
Machine learning models applied to electronic medical records can predict autism spectrum disorder (ASD) risk early in life. This approach may improve early detection and intervention for ASD in large populations.
Area of Science:
- Computational psychiatry
- Pediatric health informatics
- Machine learning in healthcare
Background:
- Early identification of autism spectrum disorder (ASD) is crucial for improving developmental outcomes, but current methods are limited.
- Most children with ASD are diagnosed after age 4, missing critical early intervention windows.
- Developing effective early screening tools for ASD in the general population remains a significant challenge.
Purpose of the Study:
- To evaluate the efficacy of machine learning (ML) models in predicting ASD risk using electronic medical records (EMRs).
- To assess the potential of ML-driven EMR analysis for early ASD detection in a general population sample.
- To identify novel risk factors for ASD through ML analysis of routinely collected health data.
Main Methods:
- Utilized EMR data from a large cohort of children and their parents in Israel.
- Extracted features from parental sociodemographic information, medical history, and medication data.
- Trained and evaluated various ML algorithms, including logistic regression, neural networks, and random forest, using 10-fold cross-validation.
Main Results:
- Machine learning models demonstrated consistent performance in predicting ASD.
- The models achieved an average C-statistic of 0.709, with a specificity of 98.18% and an accuracy of 95.62%.
- The positive predictive value (PPV) was 43.35%, indicating a substantial rate of correct positive predictions.
Conclusions:
- ML algorithms applied to EMRs can effectively identify early signs of ASD risk.
- This approach reveals previously unrecognized associations between EMR data and ASD risk.
- ML-based EMR analysis holds promise for enhancing the accuracy and efficiency of early ASD detection in pediatric populations.
Background:
Current approaches for early identification of individuals at high risk for autism spectrum disorder (ASD) in the general population are limited, and most ASD patients are not identified until after the age of 4. This is despite substantial evidence suggesting that early diagnosis and intervention improves developmental course and outcome. The aim of the current study was to test the ability of machine learning (ML) models applied to electronic medical records (EMRs) to predict ASD early in life, in a general population sample.
Methods:
We used EMR data from a single Israeli Health Maintenance Organization, including EMR information for parents of 1,397 ASD children (ICD-9/10) and 94,741 non-ASD children born between January 1st, 1997 and December 31st, 2008. Routinely available parental sociodemographic information, parental medical histories, and prescribed medications data were used to generate features to train various ML algorithms, including multivariate logistic regression, artificial neural networks, and random forest. Prediction performance was evaluated with 10-fold cross-validation by computing the area under the receiver operating characteristic curve (AUC; C-statistic), sensitivity, specificity, accuracy, false positive rate, and precision (positive predictive value [PPV]).
Results:
All ML models tested had similar performance. The average performance across all models had C-statistic of 0.709, sensitivity of 29.93%, specificity of 98.18%, accuracy of 95.62%, false positive rate of 1.81%, and PPV of 43.35% for predicting ASD in this dataset.
Conclusions:
We conclude that ML algorithms combined with EMR capture early life ASD risk as well as reveal previously unknown features to be associated with ASD-risk. Such approaches may be able to enhance the ability for accurate and efficient early detection of ASD in large populations of children.
