Abstract

Cardiovascular disease (CVD) is a leading cause of global mortality. The accurate prediction of a patient’s time-to-event outcomes, such as heart failure or mortality, is a critical challenge in clinical practice. Traditional survival models, including the Cox Proportional Hazards (CoxPH) model and Multi-Task Logistic Regression (MTLR), often rely on simplifying assumptions that limit their ability to capture complex, non-linear relationships and temporal dynamics inherent in patient health data. This research introduces a novel hybrid survival model that integrates an ELU-activated Deep Surv network with an MTLR ensemble, enhanced by a temporal feature fusion layer. The model is designed to leverage the complementary strengths of both approaches: Deep Surv’s capacity for learning deep, nonlinear representations and MTLR’s ability to estimate time-specific survival probabilities. By fusing static patient attributes (e.g., age, gender) with dynamic temporal features (e.g., blood pressure readings over time), the model provides a more comprehensive view of a patient’s health trajectory. We train the model using a survival-specific loss function and evaluate its performance using metrics such as the C-index, Integrated Brier Score (IBS), and time-dependent AUC. When applied to the Framingham Heart Study dataset, the hybrid model consistently outperformed traditional methods, yielding higher predictive accuracy and more clinically interpretable risk stratification. This approach demonstrates a promising step towards developing personalized, time-aware predictive analytics for cardiovascular care.

References

• D. R. Cox, “Regression models and lifetables,” Journal of the Royal Statistical Society: Series B (Methodological), vol. 34, no. 2, pp. 187–202, 1972.
• C. N. Yu, R. Greiner, H. C. Lin, and V. Baracos, “Learning patient-specific cancer survival distributions as a sequence of dependent regressors,” Advances in Neural Information Processing Systems (NeurIPS), vol. 24, 2011.
• J. L. Katzman, U. Shaham, A. Cloninger, J. Bates, T. Jiang, and Y. Kluger, “Deep Surv: Personalized treatment recommender system using a Cox proportional hazards deep neural network,” BMC Medical Research Methodology, vol. 18, no. 1, p. 24, 2018.
• S. Fotso, “Deep neural networks for survival analysis based on a multi-task framework,” arXiv preprint arXiv:1801.05512, 2018.
• H. Kvamme, Ø. Borgan, and I. Scheel, “Time-to-event prediction with neural networks and Cox regression,” Journal of Machine Learning Research, vol. 20, no. 129, pp. 1–30, 2019.
• C. Lee, W. R. Zami, J. Yoon, and M. van der Schaar, “Deep Hit: A deep learning approach to survival analysis with competing risks,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 32, no. 1, 2018.
• B. A. Goldstein, A. M. Navar, M. J. Pecina, and J. P. Ioannidis, “Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review,” Journal of the American Medical Informatics Association, vol. 24, no. 1, pp. 198–208, 2017.
• A. Kvamme and Ø. Borgan, “Continuous and discrete-time survival prediction with neural networks,” Lifetime Data Analysis, vol. 27, no. 4, pp. 710–736, 2021.
• R. Bellot and M. van der Schaar, “Neural survival machines: Non-parametric survival analysis with neural networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 9, pp. 4824–4836, 2022.
• A. Luck, B. C. Sylvain, H. Lu, and J. Pineau, “Deep learning for survival analysis: A survey,” ACM Computing Surveys, vol. 55, no. 9, pp. 1–38, 2023.
• J. Esteva, B. Kuper, R. A. Novoa, J. Ko, S. M. Swetter, H. M. Blau, and S. Thrun, “Dermatologist-level classification of skin cancer with deep neural networks,” Nature, vol. 542, no. 7639, pp. 115–118, 2017.
• E. Wulczyn, C. Steiner, R. Xu, S. Sadhwani, A. Wang, F. Flament-Auvigne, B. M. Mermel, C. Chen, N. T. Chen, and Y. H. Tan, “Deep learning-based survival prediction for multiple cancer types using histopathology images,” PLoS Medicine, vol. 17, no. 9, e1003332, 2020.