Reproduce “Counterfactual Phenotyping with Censored Time-to-Events” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Chirag Nagpal, Mononito Goswami, Keith Dufendach, Artur Dubrawski

Researchers from Carnegie Mellon University developed Cox Mixtures with Heterogeneous Effects (CMHE), a deep latent variable model, to identify patient subgroups exhibiting differential responses to clinical interventions for censored time-to-event outcomes. The method successfully discovered clinically actionable phenotypes in large randomized clinical trials, revealing groups with enhanced or diminished treatment effects.

Key insights

  • Decoupling baseline survival from treatment effect through distinct latent variables enables more accurate identification of counterfactual phenotypes with censored time-to-event data.
  • Counterfactual time-to-event distributions are identifiable under assumptions of independent censoring, conditional proportional hazards, latent independence, and ignorability.
  • Deep learning can capture complex, non-linear relationships in heterogeneous treatment effects, surpassing traditional clustering and 'Virtual Twins' methods for such patterns.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.