Benchmarking multi-step methods for the dynamic prediction of survival with numerous longitudinal predictors

2024-03-21

AI Summary

Existing dynamic survival prediction methods lack standardized benchmarking, particularly for high-dimensional longitudinal biomedical data. Method: This study systematically evaluates multi-step dynamic survival prediction approaches—including mixed-effects models, multivariate functional principal component analysis, Cox regression, random survival forests, and landmark analysis—across multiple real-world datasets. We rigorously control key factors (sample size, covariate dimensionality, and follow-up duration) to assess predictive accuracy, robustness, and computational efficiency. Contribution/Results: The analysis reveals critical trade-offs among modeling choices, identifies context-specific adaptability requirements, and delineates practical performance limits. It provides the first empirically grounded, methodological guideline for selecting appropriate dynamic risk prediction strategies in clinical settings, thereby advancing evidence-based decision support for time-varying prognostic modeling.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Search and Optimization: Metareasoning and Metaheuristics
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

In recent years, the growing availability of biomedical datasets featuring numerous longitudinal covariates has motivated the development of several multi-step methods for the dynamic prediction of time-to-event ("survival") outcomes. These methods employ either mixed-effects models or multivariate functional principal component analysis to model and summarize the longitudinal covariates' evolution over time. Then, they use Cox models or random survival forests to predict survival probabilities, using as covariates both baseline variables and the summaries of the longitudinal variables obtained in the previous modelling step. Because these multi-step methods are still quite new, to date little is known about their applicability, limitations, and predictive performance when applied to real-world data. To gain a better understanding of these aspects, we performed a benchmarking of the aforementioned multi-step methods (and two simpler prediction approaches) based on three datasets that differ in sample size, number of longitudinal covariates and length of follow-up. We discuss the different modelling choices made by these methods, and some adjustments that one may need to do in order to be able to apply them to real-world data. Furthermore, we compare their predictive performance using multiple performance measures and landmark times, and assess their computing time.

Problem

• Dynamic prediction of survival with numerous longitudinal predictors
• Benchmarking multi-step methods for real-world applicability
• Comparing predictive performance and computing time of methods

Innovation

• Uses mixed-effects models for longitudinal covariates
• Applies Cox models for survival probability prediction
• Benchmarks methods on diverse real-world datasets