Research map / Probabilistic and scientific modeling / Molecules and proteins

Learning Survival Models with Right-Censored Reporting Delays

NeurIPS 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Survival analysis under censoring · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes survival models with right-censored reporting delays, jointly modeling event and reporting hazards, fit by Monte Carlo EM, and a two-stage transfer-learning procedure that restores consistency under administrative censoring.

Problem

Estimating event hazards when event times are only revealed upon (delayed, right-censored) reporting, and when administrative censoring with a short observation window violates identifiability of standard survival models.

Approach

Jointly model parametric hazards for event occurrence (T1) and reporting delay (T2) with a likelihood for the observed pattern; estimate via Monte Carlo EM with rejection sampling of latent (z,w); under administrative censoring, transfer the shared baseline/later-hazard parameters from a source domain in a two-stage estimator and use an approximate estimating equation for proportional hazards mode

Key results

• On Colon at tau=0.25, Ours achieves integrated Brier score 0.272 vs 0.319 for Standard (Oracle 0.232).
• On Rotterdam at tau=0.25, Ours achieves 0.203 vs 0.247 for Standard (Oracle 0.179).
• Improvement over Standard is largest for small tau and the gap to Oracle remains across all tau values.

Limitations

• Consistency under administrative censoring requires transfer assumptions (shared baseline and later hazard) that may be misspecified.
• Approximate estimating equation accuracy relies on a small-event-probability regime (Assumption on nu_i y_i being small).
• Experiments use two public benchmark datasets adapted to the setting; no real insurance data validation.

Experimental setup

Datasets: Colon Rotterdam

Benchmarks: Colon Rotterdam

Metrics: integrated Brier score

Baselines: Standard Oracle

Builds on: Cox proportional hazards model piecewise-proportional hazards model Monte Carlo EM Jensen's inequality bound

Closest papers by meaning

• Survival Analysis via Density Estimation (ICML 2025 · similarity 0.87)
• Toward Conditional Distribution Calibration in Survival Prediction (NeurIPS 2024 · similarity 0.85)
• SurvivalPFN: Amortizing Survival Prediction via In-Context Bayesian Inference (NeurIPS 2026 · similarity 0.85)
• Deep Copula-Based Survival Analysis for Dependent Censoring with Identifiability Guarantees (AAAI 2024 · similarity 0.85)
• Assessing the robustness of heterogeneous treatment effects in survival analysis under informative censoring (NeurIPS 2026 · similarity 0.85)
• History-Aware Conformal Prediction Sets for Censored Time-to-Event Outcomes (NeurIPS 2026 · similarity 0.85)