Additive hazard causal model with a binary instrumental variable

Area of Science:

• Biostatistics
• Survival Analysis
• Causal Inference

Background:

• Estimating causal treatment effects with censored outcomes is crucial in medical research.
• Unmeasured confounding can bias results, necessitating advanced statistical methods like instrumental variables (IV).
• Existing IV methods, like two-stage least squares, are primarily for linear regression and need adaptation for survival data.

Purpose of the Study:

• To develop a novel instrumental variable (IV) framework for binary treatments with censored outcomes.
• To quantify the causal treatment effect using an additive hazards model specifically for compliers.
• To establish a weighted estimator with an explicit form for improved accuracy in survival analysis.

Main Methods:

• Utilized a distinctive binary instrumental variable (IV) framework tailored for censored data.
• Adapted the principle of conditional score to develop a weighted estimator.
• Established asymptotic properties and provided variance estimators for the proposed method.

Main Results:

• A weighted estimator with an explicit form was successfully derived.
• Asymptotic properties of the proposed estimators were theoretically established.
• Extensive simulations demonstrated the finite sample performance of the estimator.

Conclusions:

• The proposed weighted estimator effectively addresses bias in instrumental variable (IV) analysis for censored survival data.
• The method provides a robust tool for quantifying causal treatment effects in the presence of unmeasured confounding.
• Applied to end-stage renal disease patient data, it enables comparison of survival across different dialytic modalities.