AETAS

Ageing & Environmental Trade-off Assessment System – powered by CAUSEWAY causal framework

AETAS: a causal-inference platform, not another regression. It is based on double machine learning, negative-control outcomes, Oster δ, E-values, Holm correction — plus a policy layer: multi-objective optimization and Monte-Carlo.

Healthy ageing: what we can prove — and where care and prevention pay

Causal inference on 28 countries — rejecting the obvious, and what survives.

  • Data: SHARE — ex-ante harmonised panel, 28 countries, adults 50+, waves 4–8 + COVID surveys + environmental exposome SHARE-ENV.
  • A "Causal Firewall" that refuses to estimate when confounding is too severe.
  • A platform that knows when to stay silent is more useful than one that always answers.

Anna: a tangle of plausible causes — which arrows are real?

Most "X harms healthy ageing" claims are correlations in disguise

  • By 2050, nearly 1 in 3 Europeans will be over 65 (1 in 5 today) — every prevention euro will count more than ever.
  • The literature offers many "X harms late-life health" claims — air pollution, lifestyle, deprivation.
  • Which are causal — worth acting on — and which are confounded correlations?
  • Getting it wrong wastes budgets and misdirects care.

Pillar 1

Prediction tells you what will happen. Causation tells you what to do about it.

Pillar 2 — what survives: the social

  • Everyday ageism → later depression: prospective, survives baseline (β ≈ +0.37). A hidden cause would need to be ~4× stronger than everything we measured to explain it away (Oster δ = 4.1, E-value 1.68; survives Holm).
  • Life-course persecution → depression: replicated in the Spanish Civil War cohort (p < 10−5; δ = 13.9 — a hidden cause would need to be ~14× stronger than everything measured); directionally same in a Czech cohort (marginal), so the weight is on Spain.
  • Psychosocial and modifiable — the WHO anti-ageism agenda has a causal target here.

Pillar 3 — what survives: the systemic

  • Pre-pandemic chronic-disease burden → losing care when COVID overloaded the system.
  • Not age or poverty — survives age + sex + education + income (DML), and not a contact artefact.
  • The driver is the system postponing/refusing appointments (OR 1.26).
  • A year on → more hospitalisations (+26%).

The levers we can prove are social & systemic

  • Air pollution, at the scale we can measure → no detectable effect (a small one could still hide).
  • What does survive scrutiny → social adversity (ageism, persecution) and health-system resilience.
  • So rigour redirects us — toward the levers we can actually prove, and act on.

Back to Anna: the levers we can actually pull

The policy engine: crank the levers, see the dividend

  • Move a lever; the engine returns the fiscal impact.
  • Pollution lever → €0 (band through zero) — we proved it's confounded.
  • Anti-ageism → net +€1.8B/yr (the fiscal engine).
  • Care-continuity runs at a net cost — bought for crisis-equity, not euros.
  • Integrated nets a modest +€0.4B [−0.8, 1.8].
  • Worked example: Germany 50+ — German population & costs; the method transfers to any country with its own numbers.

Under the hood ① — propagate: anti-ageism firmly pays; integrated is modest but positive

Under the hood ② — optimise the split: a frontier of best answers

Not one best answer — a map of best answers.

What this means for policy and practice

  • Base budgets on causal evidence (Pollution needs finer data or a natural experiment)
  • Anti-ageism is clinical prevention (A modifiable mental-health lever, not just a virtue)
  • Build shock-resilient care continuity (Protect the highest-need first — an equity choice, not a cost saving)
© 2019 Medpobs. Powered by VOOV.