Update Georgia parent/caretaker Medicaid limit for 2025; verify Medicare IRMAA (#8787) and KS CHIP (#8792) accuracy

[MaxGhenis]

Summary

Accuracy review of three health-program issues (#8787 Medicare IRMAA, #8834 GA parent/caretaker Medicaid, #8792 Kansas CHIP per-capita denominator). All three had already been addressed by recently merged PRs; this PR verifies each against primary sources and lands the one remaining accuracy correction.

Net code change: one dated Georgia Medicaid parameter update (2025). The Medicare and Kansas CHIP issues are verified correct as-is and should be closed with the evidence below.

#8834 — Georgia parent/caretaker Medicaid (fix in this PR)

PR #8913 set income_limit.GA = 0.33 at 2024-01-01 (28% FPL + 5% MAGI disregard) from MACPAC Exhibit 36 (July 2024), and its own note flagged that the frozen-dollar standard should be re-verified as the FPL rises.

Verified against MACPAC Exhibit 36 (July 2025 edition, data as of January 2025): Georgia's "Parents and caretaker relatives of dependent children" level is now 26% FPL. Exhibit 36's own note states the effective limit is 5 percentage points higher than the table value (general MAGI disregard), and that states using dollar standards have those dollars converted to an FPL percentage — exactly Georgia's frozen-dollar case. So the effective 2025 limit is 0.31 (26% + 5%).

Cross-check on the +5% convention: MACPAC lists Kansas at 33%, and PE already stores KS = 0.38 (33% + 5%) — confirming the parameter encodes the with-disregard value.

Note on the DFCS "247% FPL" figure: Georgia DFCS Appendix A2 lists a "Parent/Caretaker with Children" row at 247% FPL. That is the AU-level test for a unit that contains children (governed by the children's higher limit), not the Section 1931 parent floor this parameter models. The parent floor is MACPAC's 26% (frozen dollar ≈ $551/month for a family of three in 2024–early 2025).

Changes:

• Add GA: 2025-01-01: 0.31 with the July 2025 MACPAC reference.
• Add regression tests pinning GA at 0.31 for 2025 and 2026 (frozen-dollar carry-forward).

#8787 — Medicare Part B/D IRMAA (verified correct; recommend close)

Already fixed by merged PR #8788. Independently verified every 2026 value against SSA POMS HI 01101.020 and the CMS 2026 fact sheet figures:

• Part B base premium 2026 = $202.90 ✓
• Part B IRMAA surcharges 2026 (single/HoH/surviving spouse/joint) = 81.20 / 202.90 / 324.60 / 446.30 / 487.00 ✓ (POMS total premiums 284.10 / 405.80 / 527.50 / 649.20 / 689.90 minus $202.90 base)
• Part B MFS = base+446.30 up to $391,000, then +487.00 ✓; separate.yaml correctly carries type: single_amount (as do all five filing statuses)
• Part D IRMAA surcharges 2026 = 14.50 / 37.50 / 60.40 / 83.30 / 91.00 ✓
• All MAGI thresholds match ($109k/$137k/$171k/$205k/$500k single; $218k/$274k/$342k/$410k/$750k joint; $391k MFS top)
• Lagged-MAGI reads (period.offset(-2, "year") on AGI + tax-exempt interest) are present and correct in both gross_medicare_part_b_premium and income_adjusted_part_d_premium_surcharge
• Merged PR Fix Medicare IRMAA schedules and MFS lookup #8788 already added 2026 boundary tests across filing statuses including MFS

The Medicare test tree passes (120 tests). The only item in #8787 not addressed is uprating metadata for projecting thresholds past 2026 — deliberately left out here rather than guessed, since the statutory rule (CPI-U over the 12 months ending August, rounded to $1,000, with special top-bracket handling) needs an exact encoding; approximating it would reduce accuracy versus holding the last official year. Recommend tracking uprating as a separate enhancement.

#8792 — Kansas CHIP per-capita denominator (verified correct; recommend close)

Already fixed by merged PR #8798. per_capita_chip now explicitly divides separate-CHIP spending by separate-CHIP enrollment (the exact concept the issue requested), replacing the total-CHIP denominator (~89,803) the issue flagged.

Verified the Kansas inputs against MACPAC MACStats (Feb 2026 Data Book, FY2024 data):

• Exhibit 32 (Child Enrollment): Kansas separate-CHIP enrollment = 72 thousand → parameter 72_018 ✓
• Exhibit 33 (CHIP Spending): Kansas "Separate CHIP programs and coverage of pregnant women" total = $136.6M → parameter 136_581_734 ✓
• Per-capita = 136,581,734 / 72,018 = $1,896.49/yr ✓ (matches the test)

The issue's ~61,100 KLRD figure is superseded by MACPAC, which is the model's cited national calibration source (and the source PE uses for every other state's denominator). No change needed.

Tests

• Medicare tree: 120 passed
• CHIP tree: 98 passed
• Kansas state tree: 331 passed
• Georgia state tree: 391 passed
• Medicaid tree: 370 passed
• medicaid_parent_income_limit: 9 passed (2 new GA cases)

🤖 Generated with Claude Code