CAUSEWAY

Causal Audit & Uncertainty-propagating Estimand WAY – Causal ML Platform with Multi-Objective Optimization for Decision Support in Health Policy and Discovery of Personalized Prevention Strategies

CAUSEWAY: End-to-End Causal ML Framework with Multi-Objective Optimization for Decision Support in Environmental Health Policy, Healthy Aging and Discovery of Multi-Factor Personalized Prevention Strategies at Any Age

Three gaps at the prediction-to-prescription frontier

  • Gap 1 — No end-to-end causal-ML pipeline in the field unifying DML, TMLE, ML-mediation, and policy MOO under one reproducible orchestration — on SHARE or any comparable harmonised panel.
  • Gap 2 — Exposome-ML on aging panels stops short of policy MOO: HELIX, AMSEN, EXPANSE/HEAP mature on prediction and single-estimand causal inference, but none chains causal estimates → mediation → multi-objective optimisation (HRS, ELSA, or SHARE).
  • Gap 3 — Causal estimates rarely reach MOO: NSGA-II appears in policy papers but almost never receives its objective vector from upstream DML/TMLE estimators with uncertainty propagated.

CAUSEWAY and its SHARE cohort instantiation AETAS close all three.

CAUSEWAY in one picture

Five Core Innovations: What CAUSEWAY Adds

  • 01 Reproducible End-to-End Causal-ML Pipeline: Orchestrating DML, TMLE, ML-mediation, and NSGA-II under a single deterministic Nextflow DAG.
  • 02 The Causal Firewall: A runtime data contract that actively refuses to compute when identification assumptions fail.
  • 03 Four-Slot Inter-Stage Contract: Propagating per-observation EIF to enable downstream variance-aware simulation without retraining.
  • 04 Monte-Carlo Pareto Envelopes: Translating causal uncertainty into policy optimization through non-smooth NSGA-II sorting.
  • 05 Composite Compliance Framework: A unified STROBE × TRIPOD+AI × MI-CLAIM × CHEERS checklist for Causal ML pipelines.

CAUSEWAY: From Core Logic to Analytics Platform

  • The AETAS Project: Our first full-scale empirical deployment using German SHARE panels, SHARE-ENV, and Wave-6 biomarkers.
  • Coming Soon: The complete interactive policy simulator.
  • Contributions: Firewall · Four-slot contract · Monte-Carlo Pareto · Composite checklist
  • Next: 21-analyte DBS expansion · cross-cohort federation (HRS, ELSA) · health-induced retirement coding · pollution natural experiments · Flipping the Causal DAG

More details about CAUSEWAY framework are provided in our recent publication:
CAUSEWAY: An End-to-End Causal Machine Learning and Multi-Objective Optimization Framework for Environmental Health Policy (2026)

© 2019 Medpobs. Powered by VOOV.