Doctors sometimes use a binary “yes or no” classification to predict a long-term diagnosis using patient records, but this model doesn’t work well when patients stop coming in for visits, leaving incomplete data. A newer machine learning model called a Discrete-Time Neural Network (DTNN) is more effective when working with incomplete patient records, particularly for individuals with less access to care, helping doctors make predictions about conditions like autism and ADHD that are more accurate for all children.
Read more about this analysis from researchers at the Duke-NUS Medical School in Singapore and Duke Center for Autism and Brain Development in the Journal of Medical Interest Research AI: https://ai.jmir.org/2025/1/e62985/
Loh D, Hill E, Liu N, Dawson G, Engelhard M
Limitations of Binary Classification for Long-Horizon Diagnosis Prediction and Advantages of a Discrete-Time Time-to-Event Approach: Empirical Analysis
JMIR AI 2025;4:e62985
URL: https://ai.jmir.org/2025/1/e62985
DOI: 10.2196/62985