Sharp Partial Identification for Survival Model Comparison Without Target Outcomes

2026-10-07

AI Summary

This study addresses the challenge of comparing two locked survival models in the absence of target outcome data. By adopting a bounded conditional log-odds shift assumption, the authors employ direct identification to derive the sharp identified set for Brier risk contrasts. This approach integrates Cox censoring estimation, inverse probability of censoring weighted logistic regression, and joint paired bootstrapping to facilitate pre-deployment evaluation. Compared with subtracting separately identified bounds, the proposed method substantially tightens uncertainty intervals, achieving a width ratio of 5.73 in simulation studies. When applied to non-small cell lung cancer data, although the final decision remains DEFER, the method effectively reduces evaluation uncertainty.

Abstract

We compare two locked survival prediction models in a target population before target survival outcomes are available. At a prespecified horizon, the estimand is the target Brier-risk contrast. Under a bounded conditional log-odds shift model on a prespecified deployment summary, we derive a sharp identified set preserving the shared unidentified target outcome law. Direct identification is never wider than separately identifying the risks and subtracting their bounds, with strict tightening under a Brier-specific same-side-1/2 condition. For right-censored source data, conditional Cox censoring estimation, inverse-probability-of-censoring-weighted logistic outcome modeling, and a joint pairs bootstrap yield simultaneous confidence envelopes over a finite sensitivity grid. In simulations, separate-to-direct width ratios ranged from 1.00 to 5.73 across controlled prediction geometries. Targeted simulations showed finite-sample undercoverage of the outer envelope at the small, heavily censored non-small-cell lung cancer (NSCLC) information scale (0.847-0.861 versus 0.95 nominal), compared with 0.946 at the Rotterdam-GBSG scale. In the cross-institutional NSCLC application, all 40 prespecified evaluations resulted in DEFER despite reduced identification uncertainty. In a supporting Rotterdam-to-GBSG analysis, candidate superiority was certified under small sensitivity allowances; one locked configuration yielded ADOPT CANDIDATE under direct identification but DEFER under separate-risk subtraction. Direct identification can materially reduce identification uncertainty and change the operational conclusion when signal and sampling precision are sufficient, while retaining DEFER when directional certification is unsupported.

Problem

• survival model comparison
• partial identification
• Brier-risk contrast
• target outcomes unavailable

Innovation

• Sharp Partial Identification
• Survival Model Comparison
• Brier-risk Contrast
• Inverse-Probability-of-Censoring Weighting
• Sensitivity Analysis