Newest papers

Doubly Robust Learning of Heterogeneous Survival Treatment Effects under Time-Varying Treatments

Tomàs Grau Garriga, Munib Mesinovic, Luke De Bretton-Gordon, Gerard Sanz, Axel Brando, Tingting Zhu · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26710 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Estimar os efeitos heterogêneos do tratamento de sobrevivência (HSTEs) a partir de dados de tempo a evento sob tratamentos de tempo variável é importante em campos como a medicina (por exemplo, para decidir qual sequência de tratamentos é melhor em termos de probabilidade de sobrevivência). No entanto, esta tarefa é desafiadora porque as decisões de tratamento, as características do paciente e a probabilidade de censura muitas vezes evoluem juntas, exigindo ajuste sequencial para ambos os tratamentos e censura. Os aprendizes de sobrevivência existentes consideram principalmente uma única tarefa de tratamento de base, enquanto os aprendizes de tratamento de efeito longitudinal geralmente assumem resultados não censurados. Neste quadro, apresentamos um novo leitor duplo robusto para efeitos de tratamento heterogêneo durante tratamentos de tempo variável e censura dependente

Machine translation (Meta M2M100); the English original follows.

Abstract

Estimating heterogeneous survival treatment effects (HSTEs) from time-to-event data under time-varying treatments is important in fields such as medicine (e.g., to decide which sequence of treatments is better in terms of survival probability). However, this task is challenging because treatment decisions, patient characteristics, and censoring probability often evolve together, requiring sequential adjustment for both treatments and censoring. Existing survival learners mostly consider a single baseline treatment assignment, while longitudinal treatment-effect learners typically assume uncensored outcomes. In this paper, we introduce a novel doubly robust learner for heterogeneous survival treatment effects under time-varying treatments and history-dependent censoring. Our framework separates the intervention window from the survival horizon, with treatment prescribed for a finite period, after which care follows its natural course. We further establish the favorable theoretical properties of Neyman-orthogonality, rate double robustness and quasi-oracle efficiency. The framework accommodates flexible machine learning models for nuisance estimation and treatment-effect regression. We evaluate our learners through several numerical experiments, demonstrating their effectiveness.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Censoring (clinical trials), Robustness (evolution), Computer science, Survival analysis, Artificial intelligence, Machine learning

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange