Dynamic Prediction of Alternating Recurrent Events via Neural Network

2026-06-29

AI Summary

This study addresses the statistical challenges inherent in dynamic prediction of alternating recurrent events—such as cyclical depressive episodes—including observation-induced correlation, recurrent outcomes, and censoring. It proposes the first online dynamic prediction framework that integrates neural networks with inverse probability weighted pseudo-observations. By incorporating deep learning into the modeling of alternating recurrent events and leveraging pseudo-observation techniques from survival analysis, the method constructs an interpretable architecture tailored for a statistical audience. The framework enables real-time updating and individualized prediction of event-free time. Extensive simulations demonstrate its superior performance, and its practical utility is evidenced by strong predictive capability in forecasting low-mood episodes among medical students during their first clinical internship year.

Abstract

Alternating recurrent events -- event-times of a specific nature that trigger a secondary refractory period -- occur in a wide-range of fields, including behavioral science, criminal justice, and biostatistics. Analysis of these events requires careful attention to the statistical nuance, including correlated observations and repeated outcomes subject to potential censoring. We develop an online dynamic prediction framework appropriate for predicting subsequent alternating recurrent events, by developing neural network theory for a statistical audiences and applying inverse probability weighted pseudo-observations. The proposed model is applied to dynamically predict alternating recurrent event-free time, showing good performance in simulation, and outstanding capability in application to predicting periods of low mood for first-year medical residents. We close with a discussion.

Problem

• alternating recurrent events
• dynamic prediction
• censoring
• correlated observations
• recurrent event-free time

Innovation

• alternating recurrent events
• neural network
• dynamic prediction
• inverse probability weighting
• pseudo-observations