Pitfalls of Administrative Censoring in Survival Models with Time-Indexed Inputs

Jul 16, 2026

Survival models predict time-to-event outcomes using partially observed data and are widely applied in clinical prediction, including cancer risk, disease progression, treatment response, and mortality. Many modern approaches train on rich time-indexed inputs, where retrospective data span long calendar periods. Such data may reveal acquisition time, creating a failure mode when follow-up is truncated by a fixed administrative study-end date. This administrative-cutoff leakage can cause models to predict available follow-up instead of true risk, and it is analyzed, distinguished, and detected.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

Abstract

Introduction

Administrative-Cutoff Leakage

Distinguishing from Informative Censoring and Temporal Risk Changes

Experiments and Evaluation

Simulation Studies

Mammography Cohort Results

Design Principle for Survival Prediction

arXiv :2607 . 10466v 1 [ cs .LG] 11 Jul 2026 Pitfalls of Administrative Censoring in Survival Models with Time-Indexed Inputs Yanqi Xu 1 , Hui Dai2 , Carlos Fernandez-Granda 1 , Krzysztof J. Geras3,4 , Yiqiu Shen3 1 Center for Data Science, New York University, 2 University of Chicagoy 3 NYU Grossman School of Medicine, 4 Ataraxis AI July 14, 2026 Abstract Survival models can model time-to-event outcomes using partially observed data. They are widely used in clinical prediction including cancer risk, disease progression, treatment response, and mortality. Recent models often rely on rich inputs collected at a specific clinical encounter, such as medical images, laboratory tests, electronic health record snapshots, or sensor measurements. In large retrospective datasets, these inputs are usually collected over many calendar years. Asa result, they may contain clues about when they were acquired, through changes in devices, protocols, documentation, patient mix, or clinical practice. This creates a potential failure mode when outcomes are observed only up to a fixed study end date. More recent records necessarily have less possible follow-up than older records. A model that can infer record date from the input may therefore learn to predict how much follow-up was available, rather than the patient’s true risk of experiencing the event. We call this failure mode administrative-cutoff leakage. In this paper, we characterize when this leakage can occur, distinguish it from classical informative censoring and genuine temporal changes in risk, and propose practical ways to detect it. In simulations, we show that administrative-cutoff leakage can inflate fixed-horizon AUC and can also affect Harrell’s C-index under realistic follow-up patterns. We then demonstrate the same behavior in a real mammography cohort. These results motivate a simple design principle for survival prediction: for an n-year prediction task, the dataset should provide at least n years of potential follow-up after the latest input date. Otherwise, the models may be subject to bias induced by administrativecutoff leakage. 1 Introduction Survival analysis is the standard statistical framework for modeling time-to-event outcomes [16, 4], where the central question is not only whether an event will occur, but when it will occur. This distinction is central in medicine because many decisions are tied to clinically meaningful time horizons: whether a patient is likely to develop cancer within the next several years, whether a disease is likely to progress before the next follow-up interval, or whether a treatment is expected to delay recurrence, progression, or death. A fundamental practical constraint in these settings is that not all subjects experience the event of interest during the observation period. Some individuals leave follow-up before the event occurs, while others remain event-free when the study ends. These partially observed outcomes are referred to as censored observations. By accounting for censored observations, survival models can use partially observed follow-up rather than discarding individuals whose final event status is unknown [11] . This capability has made time-to-event modeling widely used across clinical prediction tasks, including cancer risk stratification [31, 22], modeling disease progression and recurrence using endpoints such as progression-free or disease-free survival [29], and estimating treatment-associated outcomes for individualized decision-making [17] . Recent advances in deep learning have substantially expanded survival modeling by enabling timeto-event prediction from high-dimensional and heterogeneous clinical data, including medical imaging, radiology reports, and electronic health records (EHRs) [21, 30 , 18 , 27] . However, because clinical events are often rare, training such models typically requires large datasets assembled over long calendar periods. We refer to these inputs as time-indexed d

FAQs

What is administrative-cutoff leakage in survival modeling?

Administrative-cutoff leakage occurs when a model infers the reference/input time from time-indexed signatures and uses it as a shortcut for how much follow-up is available, rather than learning true biological event risk.

How does administrative-cutoff leakage differ from classical informative censoring?

Classical informative censoring reflects dependence between censoring and risk through the censoring mechanism. Administrative-cutoff leakage specifically arises from administrative truncation at a fixed study-end/extraction date combined with time-indexed cues embedded in the inputs.

What dataset design principle does the paper recommend for n-year prediction tasks?

The dataset should provide at least n years of potential follow-up after the latest input date; otherwise, predictions may become biased due to administrative-cutoff leakage.

Jul 16, 2026

Upload by Theodore