Subtract modeled Medicare Part A premium and Part D IRMAA in the other-premium residual

Type

Pure data / dataset contract.

Context

policyengine-us PR #8532 adds two modeled components to spm_unit_health_insurance_premiums (and therefore SPM MOOP):

• medicare_part_a_premium (person-level; base Part A premium net of QMB Medicare Savings Program coverage), and
• income_adjusted_part_d_premium_surcharge (person-level Part D IRMAA computed from prior-prior-year MAGI).

The MOOP decomposition contract (docs) expects data construction to subtract baseline modeled premiums from CPS-reported non-Medicare premiums when building other_health_insurance_premiums, so reforms move the modeled pieces without double counting. The legacy pipeline implements this in archived policyengine-us-data (derive_other_health_insurance_premiums in policyengine_us_data/datasets/cps/cps.py), subtracting baseline chip_premium, marketplace_net_premium, and medicaid_premium from health_insurance_premiums_without_medicare_part_b (PHIP_VAL, which includes Part A/D amounts).

Populace currently:

• carries health_insurance_premiums_without_medicare_part_b raw from PHIP_VAL (cps_carried.py), and
• does not construct other_health_insurance_premiums at all — it is an eCPS parity known gap pointing at #32.

What needs to happen

1. When Populace ports the other-premium residual (the #32 MOOP family), the baseline subtraction list must include the two new Medicare components in addition to the legacy CHIP/Marketplace/Medicaid three. The Part D IRMAA subtraction is live immediately: the surcharge is computed from MAGI, so IRMAA-band enrollees whose PHIP_VAL includes Part D payments would otherwise double count.
2. medicare_quarters_of_coverage is set nowhere in Populace; policyengine-us now defaults it to 40 (premium-free, matching 99% of beneficiaries), so modeled Part A is $0 in baseline microdata. Record this column as intentionally defaulted in the input-readiness contract (#98), and do not populate it before the Part A subtraction in (1) exists — populating quarters first would double count full Part A premiums ($6,780/yr per affected enrollee in 2026) against reported premiums.

Acceptance criteria

• The other_health_insurance_premiums residual construction subtracts baseline chip_premium, marketplace_net_premium, medicaid_premium, medicare_part_a_premium, and income_adjusted_part_d_premium_surcharge, with a test.
• medicare_quarters_of_coverage is recorded as observed, imputed, unavailable, or intentionally defaulted, and any future population of it is sequenced after the residual subtraction.

Related: #32 (MOOP input family umbrella), #98 (health input readiness, incl. lagged IRMAA MAGI), #204 (us-data replacement tracking), #169 (Part B premium payer-count diagnosis).

🤖 Generated with Claude Code

Start with the archived derive_other_health_insurance_premiums in policyengine_us_data/datasets/cps/cps.py and Populace's cps_carried.py, then read the MOOP decomposition documentation and related issues #32 and #98. Implement the residual contract with a test, and document medicare_quarters_of_coverage as intentionally defaulted with future population sequenced afterward.