Phase 1 · Comorbidity Cost Layer · prototype

What obesity actually costs a Medicaid plan — decomposed

The standard headline ("~$1,861 excess per adult/yr") hides where the money goes. This breaks it into the comorbidities that drive it — built bottom-up from public data.

$1,980
bottom-up obesity-attributable comorbidity cost / obese adult / yr
≈ matches $1,861
independent published figure (Cawley 2021)

Two unrelated methods — a published top-down study and this bottom-up comorbidity build — land within ~6% of each other. The convergence is the credibility signal; the breakdown below is the planning value.

The decomposition

Obesity-attributable cost per obese adult/yr = (how much more often obese adults carry the condition) × (what the condition costs). Diabetes + hypertension drive ~80%.

Payer
Low-income cutoff
Total obesity-attributable / obese adult / yr

High cholesterol is omitted from the headline: it carries ≈$0 independent cost once diabetes and heart disease are accounted for (its spend flows through them).

Scale to a plan

Illustrative defaults below — swap in your plan's covered lives for a real figure. (Public averages here; the exact number for your members needs your claims.)

PLAN-LEVEL OBESITY-ATTRIBUTABLE COMORBIDITY COST / YR
Methodology & honest limits

How it's built. Obesity→comorbidity gradient from NHANES 2021–23 (measured BMI; diabetes lab-confirmed at A1c ≥ 6.5%; hypertension, cholesterol, heart disease, stroke from interview), among low-income adults. Comorbidity→cost from MEPS 2023 (age/sex-adjusted incremental annual expenditure per condition for Medicaid adults, in a multivariable model so co-occurring conditions aren't double-counted). Published anchor: Cawley et al. 2021 ($1,861).

  • A cross-check, not proof. Comorbidities are obesity's main cost mechanism, so a comorbidity-built figure should approximate the total; a clean match can also mask offsetting errors.
  • Mild apples-to-oranges. This build is low-income; the $1,861 is all-adult.
  • Rare conditions are noisy. Diabetes/hypertension are solid; coronary disease/stroke rest on few dozen cases — wide uncertainty. Survey-design CIs deferred.
  • Associational, not causal; comorbidities self-reported except diabetes.
  • National averages. A specific plan's members differ — the exact number needs that plan's claims (the paid engagement).

Sources: NHANES 2021–23, MEPS 2023 (H251), Cawley et al. PharmacoEconomics 2021. Calibrated estimate on public data — not a predictive model. Prototype; not for billing.