A comparative study of methods for dynamic survival analysis
IntroductionDynamic survival analysis has become an effective approach for predicting time-to-event outcomes based on longitudinal data in neurology, cognitive health, and other health-related domains. With advancements in machine learning, several new methods have been introduced, often using a two...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-02-01
|
| Series: | Frontiers in Neurology |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fneur.2025.1504535/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850024382586421248 |
|---|---|
| author | Wieske K. de Swart Marco Loog Jesse H. Krijthe Jesse H. Krijthe |
| author_facet | Wieske K. de Swart Marco Loog Jesse H. Krijthe Jesse H. Krijthe |
| author_sort | Wieske K. de Swart |
| collection | DOAJ |
| description | IntroductionDynamic survival analysis has become an effective approach for predicting time-to-event outcomes based on longitudinal data in neurology, cognitive health, and other health-related domains. With advancements in machine learning, several new methods have been introduced, often using a two-stage approach: first extracting features from longitudinal trajectories and then using these to predict survival probabilities.MethodsThis work compares several combinations of longitudinal and survival models, assessing their predictive performance across different training strategies. Using synthetic and real-world cognitive health data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), we explore the strengths and limitations of each model.ResultsAmong the considered survival models, the Random Survival Forest consistently delivered strong results across different datasets, longitudinal models, and training strategies. On the ADNI dataset the best performing method was Random Survival Forest with the last visit benchmark and super landmarking with an average tdAUC of 0.96 and brier score of 0.07. Several other methods, including Cox Proportional Hazards and the Recurrent Neural Network, achieve similar scores. While the tested longitudinal models often struggled to outperform simple benchmarks, neural network models showed some improvement in simulated scenarios with sufficiently informative longitudinal trajectories.DiscussionOur findings underscore the importance of aligning model selection and training strategies with the specific characteristics of the data and the target application, providing valuable insights that can inform future developments in dynamic survival analysis. |
| format | Article |
| id | doaj-art-b514b269f55f4bb8a6abb6857dcc7900 |
| institution | DOAJ |
| issn | 1664-2295 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neurology |
| spelling | doaj-art-b514b269f55f4bb8a6abb6857dcc79002025-08-20T03:01:07ZengFrontiers Media S.A.Frontiers in Neurology1664-22952025-02-011610.3389/fneur.2025.15045351504535A comparative study of methods for dynamic survival analysisWieske K. de Swart0Marco Loog1Jesse H. Krijthe2Jesse H. Krijthe3Institute for Computing and Information Sciences, Radboud University, Nijmegen, NetherlandsInstitute for Computing and Information Sciences, Radboud University, Nijmegen, NetherlandsInstitute for Computing and Information Sciences, Radboud University, Nijmegen, NetherlandsPattern Recognition Laboratory, Delft University of Technology, Delft, NetherlandsIntroductionDynamic survival analysis has become an effective approach for predicting time-to-event outcomes based on longitudinal data in neurology, cognitive health, and other health-related domains. With advancements in machine learning, several new methods have been introduced, often using a two-stage approach: first extracting features from longitudinal trajectories and then using these to predict survival probabilities.MethodsThis work compares several combinations of longitudinal and survival models, assessing their predictive performance across different training strategies. Using synthetic and real-world cognitive health data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), we explore the strengths and limitations of each model.ResultsAmong the considered survival models, the Random Survival Forest consistently delivered strong results across different datasets, longitudinal models, and training strategies. On the ADNI dataset the best performing method was Random Survival Forest with the last visit benchmark and super landmarking with an average tdAUC of 0.96 and brier score of 0.07. Several other methods, including Cox Proportional Hazards and the Recurrent Neural Network, achieve similar scores. While the tested longitudinal models often struggled to outperform simple benchmarks, neural network models showed some improvement in simulated scenarios with sufficiently informative longitudinal trajectories.DiscussionOur findings underscore the importance of aligning model selection and training strategies with the specific characteristics of the data and the target application, providing valuable insights that can inform future developments in dynamic survival analysis.https://www.frontiersin.org/articles/10.3389/fneur.2025.1504535/fullsurvival analysisdynamic predictionlongitudinal datalandmarkingmachine learningADNI |
| spellingShingle | Wieske K. de Swart Marco Loog Jesse H. Krijthe Jesse H. Krijthe A comparative study of methods for dynamic survival analysis Frontiers in Neurology survival analysis dynamic prediction longitudinal data landmarking machine learning ADNI |
| title | A comparative study of methods for dynamic survival analysis |
| title_full | A comparative study of methods for dynamic survival analysis |
| title_fullStr | A comparative study of methods for dynamic survival analysis |
| title_full_unstemmed | A comparative study of methods for dynamic survival analysis |
| title_short | A comparative study of methods for dynamic survival analysis |
| title_sort | comparative study of methods for dynamic survival analysis |
| topic | survival analysis dynamic prediction longitudinal data landmarking machine learning ADNI |
| url | https://www.frontiersin.org/articles/10.3389/fneur.2025.1504535/full |
| work_keys_str_mv | AT wieskekdeswart acomparativestudyofmethodsfordynamicsurvivalanalysis AT marcoloog acomparativestudyofmethodsfordynamicsurvivalanalysis AT jessehkrijthe acomparativestudyofmethodsfordynamicsurvivalanalysis AT jessehkrijthe acomparativestudyofmethodsfordynamicsurvivalanalysis AT wieskekdeswart comparativestudyofmethodsfordynamicsurvivalanalysis AT marcoloog comparativestudyofmethodsfordynamicsurvivalanalysis AT jessehkrijthe comparativestudyofmethodsfordynamicsurvivalanalysis AT jessehkrijthe comparativestudyofmethodsfordynamicsurvivalanalysis |