Include Medicare Part A premiums and Part D IRMAA in MOOP

[daphnehanse11]

Summary

• add a net medicare_part_a_premium variable for Part A premiums paid out of pocket
• add msp_part_a_premium_coverage so QMB-covered Part A premiums do not count as household MOOP
• include paid Part A premiums and Part D IRMAA in SPM health insurance premiums and medical out-of-pocket expenses

Fixes #8529.

Tests

uv run python -m policyengine_core.scripts.policyengine_command test -c policyengine_us policyengine_us/tests/policy/baseline/gov/hhs/medicare/eligibility/medicare_part_a_premium.yaml policyengine_us/tests/policy/baseline/gov/hhs/medicare/savings_programs/msp_part_a_premium_coverage.yaml policyengine_us/tests/policy/baseline/household/income/spm_unit/spm_unit_medical_out_of_pocket_expenses_medicare_part_b.yaml

uv run python -m policyengine_core.scripts.policyengine_command test -c policyengine_us policyengine_us/tests/policy/baseline/gov/hhs/medicare policyengine_us/tests/policy/baseline/household/income/spm_unit/spm_unit_medical_out_of_pocket_expenses_medicare_part_b.yaml

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (7f028e9) to head (78b8288).
⚠️ Report is 479 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #8532   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            2         5    +3     
  Lines           30        58   +28     
  Branches         0         1    +1     
=========================================
+ Hits            30        58   +28     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[daphnehanse11]

Scope mismatch: this PR says it includes Medicare Part D premiums in MOOP, but spm_unit_health_insurance_premiums only adds income_adjusted_part_d_premium_surcharge. The surcharge variable documentation correctly says this is paid in addition to the beneficiary's underlying Part D plan premium, and I do not see a modeled base Part D premium or Part D enrollment variable in the PR. As written, ordinary Part D premiums remain omitted from MOOP, and high-income beneficiaries only get the IRMAA surcharge. Please add the base Part D premium path, or narrow the PR/docs/tests to say this only models Part D IRMAA.

[MaxGhenis]

Reviewed in depth (agent-assisted). The Part A amounts check out against the CMS 2026 fact sheet ($565/$311, QMB-months proration correct), but three things block merge as-is:

1. Double-count mechanism is unimplemented. spm_unit_health_insurance_premiums relies on the residual design in docs-quarto/methodology/moop-decomposition.qmd (other_health_insurance_premiums = reported non-Medicare premiums − baseline computed premiums), but nothing in policyengine-us-data writes other_health_insurance_premiums (the CPS build writes health_insurance_premiums_without_medicare_part_b from PHIP_VAL, which includes Part A/D amounts). Adding modeled Part A + Part D IRMAA on top will double-count once that input is populated. The data-side subtraction should land with (or before) this, and the methodology doc's component table should list the two new components.

2. The May 29 self-review is still open: only the Part D IRMAA surcharge enters MOOP, not the base plan premium, so MOOP rises only for IRMAA-band (high-income) beneficiaries. Either add the base premium path or narrow all doc strings to IRMAA-only and note the gap.

3. msp_part_a_premium_coverage gates on is_qmb_eligible, while msp_part_b_premium_coverage deliberately gates on msp_income_eligible & msp_asset_eligible to avoid the Medicaid → medically-needy → MOOP cycle. It appears cycle-safe today, but mirroring the Part B pattern (or documenting why the QMB gate is safe) would prevent a future cycle.

Also: draft #8721 duplicates the two new variables and the changelog fragment from this PR (same author, fewer tests) — they hard-conflict if both land, so one should be closed; this one looks canonical.