The Evidence Is Real
These aren't weak correlations. Wealth and sleep are among the strongest predictors of health outcomes ever documented. The question is whether knowing them improves prediction beyond what we already know.
Wealth & Mortality
Lowest income quintile: 44% die within follow-up. Highest: just 10%. Data from 32,230 Americans tracked over 30 years (HRS).
If wealth predicts death this strongly, shouldn't it improve our models?
Sleep & Health Decline
People with zero sleep problems have a 69% positive health outcome rate. One or more problems: 39–41%. Data from 24,155 people (HRS).
If sleep quality separates outcomes this clearly, shouldn't it help?
The ROC Curves Tell the Truth
ROC curves reveal how well a model distinguishes positive from negative cases across every possible threshold. If adding data helps, the curve should shift upward. Watch what actually happens.
Adding Wealth & Income Data
Q15: 32,230 people · 8,321 deaths · 30-year follow-up
Adding Sleep Data
Q16: 24,155 people · 6-year health decline outcome
The Reclassification Shell Game
Net Reclassification Index (NRI) counts how many patients move to a more appropriate risk category. For every person the new data helps, another is hurt.
SDOH Reclassification
32,230 patients across 4 risk categories
Sleep Reclassification
24,155 patients across 4 risk categories
The Signal Is Already Captured
Feature importance reveals why: the new variables rank near the bottom. Age, self-rated health, and existing conditions already carry the signal that wealth and sleep correlate with.
Health + SDOH Model Features
GradientBoosting importance — SDOH features highlighted
Health + Sleep Model Features
GradientBoosting importance — sleep feature highlighted
Why Wealth Doesn't Help
Wealth doesn't kill directly — it works through the health conditions it causes: diabetes, heart disease, obesity, smoking. Those conditions are already in the model. Adding wealth is adding a proxy for information the model already has.
Why Sleep Doesn't Help
Poor sleep doesn't cause health decline independently — it's a symptom of the conditions that do. Self-rated health (the #1 feature at 19.4%) already captures what sleep problems signal. Adding sleep is adding a downstream indicator.
Two Paradoxes, One Lesson
The longevity study surfaced two symmetrical findings. Together, they define the boundaries of the right-fidelity principle.
The Biology Paradox
More computation doesn't always help. ML alone (r = 0.478) was worse than a textbook formula (r = 0.527) for biological age.
The Data Paradox
More data doesn't always help. Adding wealth (+0.002 AUC) and sleep (+0.001 AUC) to mortality models changed virtually nothing.
The ADM Principle
The right model at the right fidelity. Not the most complex, not the most data-rich — the one matched to the question and the decision it supports.