Correct Missouri Medicaid child income limit to 153% of FPL

[daphnehanse11]

Fixes #9126.

Change

Corrects Missouri's Medicaid income limit for children ages 1-18 from 1.55 to 1.53 in both the young_child and older_child parameters.

Missouri's MO HealthNet for Kids base standard is 148% of FPL for all children ages 1-18 (DSS Manual § 1830.010.05), and the conditional 5% MAGI disregard (§ 1805.030.20.20.05) produces an effective 153% ceiling. The state's published dollar limits equal exactly 148% of the FPL ($23,621 for one person at the 04/2026 guidelines).

The prior 1.55 traces to KFF's figure, which reflects the pre-MAGI-conversion 150% standard (+5). Since the 148% MAGI-converted standard predates the parameter's 2018 start date, the value is corrected at the existing date rather than dated as a policy change. KFF reports 155% for January 2018, 2021, and 2025 alike, so this is a source discrepancy, not a recent change — a comment in each parameter file documents this.

Impact

• Children in the 153-155% FPL band were incorrectly shown Medicaid-eligible (confirmed by partner testing for one- and two-person households).
• The same children were misrouted away from CHIP, since is_chip_eligible_child requires non-Medicaid-eligibility before applying the CHIP ceiling (MO: 305%).

Tests

• Adds four boundary regression tests (152% eligible / 154% ineligible, for both age groups).
• Full Medicaid suite (380), CHIP + Missouri state suites (626), and partner contract suite (630) all pass; partner tests untouched.

🤖 Generated with Claude Code

[DTrim99]

Program Review — PR #9133 (Correct Missouri Medicaid child income limit to 153% of FPL)

Author: daphnehanse11 · Draft · Fixes #9126

Source Documents (all HTML, verified via WebFetch)

• MO DSS Manual § 1830.010.05 — MO HealthNet for Kids base standard: 148% FPL, ages 1-18 (and a separate 196% standard for under-1)
• MO DSS Manual § 1805.030.20.20.05 — conditional 5%-of-FPL MAGI disregard
• MyDSS published income limits — $23,621 (HH1) = exactly 148% FPL
• KFF — reports 155% for MO (Jan 2018/2021/2025): the legacy pre-MAGI 150% + 5, i.e. the prior value's source

Branch Status

⚠ PR branch is 132 commits behind main (1 ahead). Recommend rebasing before merge. Review was scoped strictly to the PR's merge-base diff, so the staleness did not cause false-positive findings. (Also: this is a draft PR.)

Summary

Correct, well-sourced value fix. The MO entry in both the young_child and older_child Medicaid income-limit params moves 1.55 → 1.53 at the existing 2018-01-01 key. Independently verified: 148% base + 5% MAGI disregard = effective 153%, confirmed by two MO DSS manual sections and the MyDSS dollar figure. The prior 1.55 is the legacy pre-MAGI 150% + 5 = 155% (KFF's figure) and is genuinely wrong for the MAGI era. 0 critical. CI green (33/33).

The main correctness risk is cleared: young_child = ages 1-5, older_child = ages 6-18 — together exactly the ages 1-18 group MO sets at 148%(→153%). The under-1 infant limit (196%, encoded as 201%) lives in a separate param and is not touched, so this PR does not wrongly lower any infant limit. In-place 2018-01-01 dating is correct (the value was wrong since inception, not a policy change); MO-only (every other state byte-for-byte unchanged); folding the 5% disregard into the ceiling is the right encoding (medicaid_income_level carries no disregard of its own, so no double-count).

Critical (Must Fix)

None.

Should Address (non-blocking)

1. Pin the exact 153% boundary + the strict-< operator (test gap, highest-value). The added tests straddle 1.53 with 1.52 (eligible) and 1.54 (ineligible) but never assert at 1.53. Eligibility uses strict <, so medicaid_income_level: 1.53 → ineligible; a future refactor to <= would silently flip it and no test would catch it. Add one case per file at exactly 1.53 → false. (This is the boundary the PR exists to set.)
2. Add an integration test for the Medicaid → CHIP re-routing (the headline behavioral claim). The changelog says the 153-155% band was misrouted away from CHIP; is_chip_eligible_child gates on ~is_medicaid_eligible (MO CHIP ceiling 305%), so a MO child at ~154% FPL should now be is_medicaid_eligible: false and is_chip_eligible_child: true. No test pins this end-to-end — the 4 unit tests only take scalar medicaid_income_level and can't reach CHIP. (No existing medicaid/chip child integration file was found — worth adding a full-household case rather than forcing it into the unit files.)
3. Tighten the param comment wording (regulatory + reference agents both flagged). The comment says "…not the 150% KFF reports", but KFF's published figure was 155% (150% + the 5% disregard already baked in), and KFF's current live snapshot now shows 148%. Reword to attribute 155% to the historical KFF snapshot, e.g. "KFF historically reported 155% (a 150% legacy base + 5% disregard); the MAGI-converted base is 148%, so the effective ceiling is 148 + 5 = 153%."

Suggestions

• Add a one-line note that the 5% disregard is folded into the limit because medicaid_income_level carries no disregard of its own (pre-empts a "model the disregard separately?" question). (regulatory S2)
• The 152% "eligible" test is not a regression guard — it passes under both 1.55 and 1.53; only the 154% case does real regression work. Acceptable as a both-sides pair, but the exact-1.53 case (item 1) is the one that matters. (tests M1)
• Pre-existing, out of scope: MN: 2.80 on line 79 of both files carries a trailing zero (repo style → 2.8). Not introduced by this PR. (code)

Validation Summary

Check	Result
Regulatory Accuracy	Correct — 1.53 = 148% base + 5% MAGI disregard for ages 1-18 (both params); infant 196%/201% correctly untouched; 1.55 genuinely wrong for MAGI era; in-place dating correct; MO-only
Reference Quality	0 critical, 0 missing; 3 new MO refs corroborate 148%+5%=153%; KFF retained transparently (not deleted); 2 minor wording/date-anchor notes
Code Patterns	0 critical; right date key, MO-only, both params consistent, changelog present + fixed; 1 pre-existing trailing-zero (out of scope)
Test Coverage	4 boundary tests correct (154% flips → genuine regression guard); gaps: exact-1.53 boundary + < operator unpinned; no Medicaid→CHIP integration test
Source Audit	148% + 5% = 153% confirmed by 2 DSS manual sections + MyDSS $ figure; KFF 155% = legacy 150%+5
CI Status	Passing (33/33)

Review Severity: COMMENT

A correct, cleanly-scoped, well-sourced value correction with the one real risk (infant band) verified clear. No blocking issues. Before marking ready: add the exact-1.53 boundary test + a Medicaid→CHIP integration test (items 1-2), tighten the KFF comment (item 3), and rebase off the 132-commit lag.

Next Steps

To auto-apply the actionable items: /fix-pr 9133

Review generated with Claude Code via /review-program

[daphnehanse11]

Addressed the review items and rebased onto main (branch is now main + 2 commits):

1. Exact-153% boundary test — added to both category test files, but with one deviation from the suggestion: the case can't assert false. YAML inputs are stored as float32, and float32(1.53) ≈ 1.52999997 lands just below the float64 parameter 1.53, so the exact boundary evaluates eligible under both < and <= — the operator is unpinnable at this value. This happens to match the manual's income-maximum (≤) semantics, so the tests pin the actual behavior (true) with a comment explaining the float mechanics.
2. Medicaid → CHIP integration tests — added to is_chip_eligible_child.yaml: a full-household MO child at 154% FPL is is_medicaid_eligible: false and is_chip_eligible_child: true (computed, not input), plus the 152% converse.
3. Comment wording — reworded in both parameter files to attribute KFF's 155% to the legacy pre-MAGI 150% base plus the 5% disregard, and to note the disregard is folded into the limit because medicaid_income_level applies no disregard of its own.

Full Medicaid/CHIP/MO suites (999) and the partner contract suite (626) pass locally.

[DTrim99]

Re-review — all three items addressed ✅

Thanks for the quick turnaround. The follow-up commit resolves every should-address item from the prior review:

1. Exact-153% boundary — pinned, and your analysis corrects mine. Both category files now assert medicaid_income_level: 1.53 → eligible. I'd flagged the strict < as making 1.53 ineligible, but you're right: the input is stored as float32 (1.53 → 1.5299999713897705), which lands just below the float64 parameter 1.53, so it evaluates eligible under both < and <= — matching the manual's income-maximum (<=) semantics. Verified: float32(1.53) < 1.53 (float64) == True. Good catch; the comment documents it clearly.
   • Minor note (non-blocking): this relies on 1.53's nearest float32 rounding down. It's stable for this value, but it's a float-precision coincidence rather than an explicit <=; worth keeping the explanatory comment (which you have) so a future reader doesn't "simplify" it away.
2. Medicaid → CHIP re-routing — pinned both sides. is_chip_eligible_child.yaml now has MO 154% FPL → is_medicaid_eligible: false + is_chip_eligible_child: true (routes to CHIP) and MO 152% → the inverse. This is exactly the end-to-end behavioral claim the PR makes.
3. Comment wording — tightened. Now correctly attributes KFF's 155% to the legacy pre-MAGI 150% base + 5% disregard, and adds the note that the disregard is folded into the limit because medicaid_income_level carries none of its own. 👍

CI is green (31/31). The value itself was already verified correct against the MO DSS manual (148% base + 5% disregard = 153%) with the infant band correctly untouched. This is good to merge once you're ready to take it out of draft / rebase off the (now larger) main lag.