/claim #16

Summary

• Adds a distinct survival/time-to-event peer-review assistant for the AI-Powered Research Assistant Suite.
• Flags immortal-time bias, unsupported proportional-hazards assumptions, arm-level censoring imbalance, low events-per-variable support, competing-risk model mismatch, short follow-up windows, missing censoring tables, and overstrong survival-benefit claims.
• Emits deterministic JSON, Markdown, SVG, and MP4 reviewer artifacts from synthetic survival-analysis packets.

Non-overlap

This is separate from existing broad assistant, external-validity, Bayesian prior sensitivity, structured abstract, unit/dimension, randomization/blinding, literature freshness, assay calibration, image integrity, reporting-guideline, prompt-safety, and uncertainty-calibration submissions. It focuses specifically on time-to-event endpoint validity, censoring evidence, exposure-time modeling, and survival-model assumptions.

Demo Artifact

• ai-research-assistant/survival-endpoint-review-assistant/artifacts/demo.mp4
• Reproducible with node ai-research-assistant/survival-endpoint-review-assistant/make-demo-video.js

Validation

• node ai-research-assistant/survival-endpoint-review-assistant/test.js
• node ai-research-assistant/survival-endpoint-review-assistant/demo.js
• node ai-research-assistant/survival-endpoint-review-assistant/make-demo-video.js
• ffprobe ai-research-assistant/survival-endpoint-review-assistant/artifacts/demo.mp4 -> H.264, 640x360, 5s, 60 frames
• git diff --check

Synthetic data only. No credentials, private manuscripts, protected health information, external APIs, or network calls are used by the module.