Track health input readiness for upcoming PolicyEngine-US model improvements

Type

Pure data / dataset contract.

Context

Several of the PE-US health-model improvements identified from the Populace integration depend on data inputs or source diagnostics that Populace may be able to provide. Existing open Populace work already covers ACA take-up and selected-plan inputs (#70, #74, #75, PR #95), base population scale (#94, PR #93), and target-period Medicaid facts (#91, PR #92).

This issue tracks the remaining data-readiness questions so they do not get lost while the PE-US model-side issues are worked.

Candidate inputs / diagnostics

• Claimant/dependent links for Medicaid MAGI household construction, so PE-US can distinguish known claimant relationships from missing-link fallback.
• Lagged Medicare IRMAA MAGI or components needed to compute MAGI from two years prior in a single-year dataset.
• Rating-area-ready geography, such as county FIPS, ZIP/ZCTA quality diagnostics, or explicit ACA rating area where source support exists.
• ESI offer/affordability support, beyond current ESI coverage and MEPS-derived premium assignment.
• Missing-vs-observed-false flags or diagnostics for eligibility-sensitive health statuses, such as medically frail/special medical needs, recent incarceration, emergency medical condition, and long-term care service receipt.

Why this matters

The PE-US side can add clearer rules and optional inputs, but Populace needs to know which columns can be observed, imputed, diagnosed, or intentionally left unavailable. Without that contract, PE-US defaults can become invisible data assumptions.

Acceptance criteria

• For each candidate input, Populace records whether it is observed, imputed, unavailable, or intentionally deferred.
• Dataset diagnostics expose high-impact missingness/defaulting where possible.
• The export contract can distinguish source inputs from formula-owned PE-US outputs.
• The issue links to the corresponding PE-US model issues once those are implemented or scoped.

Start by reviewing the linked Populace issues and pull requests, then inventory the candidate inputs and current export contract. For each input, document whether it is observed, imputed, unavailable, or deferred; add high-impact missingness diagnostics where possible and distinguish source inputs from formula-owned PE-US outputs.