Contra Hanson On Medical Effectiveness
Read the original on Astral Codex Ten →
Follow-up reading
↳ Response to Hanson On Health Care — Scott's reply to Hanson · Apr 2024
↳ Highlights From The Comments On Hanson And Health Care — Highlights on the debate · May 2024
Summary
A rigorous rebuttal of Robin Hanson's strong claim that medicine doesn't work. Three prongs. (1) Medicine obviously works -- secondary-prevention survival rates (cancer 5-year, post-heart-attack 30-day, stroke, CHF, MS, type-1 diabetes) have improved dramatically over 50 years in ways nutrition/sanitation can't explain, since they're measured AFTER diagnosis. (2) Hanson's three free-insurance experiments (RAND, Oregon, Karnataka) don't support his thesis -- the centerpiece is a 'funnel' framework: for insurance to show a health effect you need insurance->more visits->more diagnoses->more treatment->treatment works, and significance is lost at DIFFERENT steps in each study due to statistical power, not true nulls (the diabetes/HbA1c math showing the 'null' is exactly consistent with meds working; the blood-pressure subgroup analysis). (3) Bigger, better studies (the 4.5M-person IRS-letter natural experiment; Sommers/Baicker Medicaid-expansion mortality studies) show clear positive mortality effects (p=0.01, p=0.001). Skewers Hanson for dismissing thousands of clinical RCTs as biased while leaning on underpowered insurance studies, then dismissing the one adequately-powered insurance study as 'noise'.
Why this score
Quality 76 · Excellent. Excellent floor. Meticulous and quantitatively sharp -- the insurance-study 'funnel' framework is an original, genuinely clarifying contribution, and the per-study statistical-power dissection is first-rate. A near-definitive rebuttal of a prominent contrarian thesis. Held at the floor as a (rigorous) rebuttal rather than a standalone framework-builder.
Claude’s paradigm shift 52 · Moderate. Moderate. The conclusion (medicine works) isn't novel, but the funnel framing of the insurance-study causal chain and the power-analysis machinery are a fresh, reusable contribution.
Real-world impact 3 · Moderate. Moderate. The definitive response to a widely-cited 'medicine doesn't work' thesis within the rationalist/health-econ discourse; conceptual influence, no direct material change. 3.