Repository navigation
Apply each state's rule for a Medicaid applicant whose spouse is not aged, blind or disabled - #10640
Conversation
…aged, blind or disabled Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…it-ineligible-spouse
|
Rulings by the project lead on the source conflicts in this PR, 2026-11-02. General rule: when a state's own rule differs from SSI's deeming rule, the model follows the state's rule. SSI's rule applies where the state follows it or is silent.
Scope: this PR covers SSI's rule, budgeting together and the couple limit without deeming. A follow-up issue tracks the states with rules of their own. |
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
The follow-up for the states with rules of their own is #10641. |
Give CO, MA and NE candidates a second source, which clears the failing CI check. A couple split by a facility counts no spouse income (42 USC 1396r-5), read at the same instant as the income it gates. Medically needy budget-together uses one method at every income. Hawaii budgets a couple together. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…it-ineligible-spouse
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PR Review: Apply each state's rule for a Medicaid applicant whose spouse is not aged, blind or disabledPR #10640 (author: hua7450), head Path prefixes used below:
SummaryWhat works. The federal default is right, and the facility split rests on the right federal law:
What needs to change. Two state settings give wrong results against the sources the PR itself cites, and I reproduced both on the head:
Should address. There are three items:
The axiom line is valid for this head. Critical (must fix)C1. Oklahoma is marked "budget together", but its cited rule adds an own-income test.
C2. Budgeting together drops the ineligible-child allocation, even in states whose sources subtract it.
Should addressS1. In non-209(b) states, budgeting together is sometimes more restrictive than SSI's rule, and that choice isn't recorded.
S2. The seven candidates are incomplete under CONTRIBUTING.md step 1.
S3. Ruled readings and most per-state flags are unpinned (mutation rule).
Suggestions
Verified correct
Axiom line assessment
CI statusAll 34 checks pass at head I ran one probe of four households on the head, using policyengine-core 3.33. It confirmed the C1, C2 and S1 figures above, with no unexpected values for any applicant. Branch statusThe branch is 95 commits behind
Sibling notes (not findings):
Verdict: REQUEST_CHANGES |
DTrim99
left a comment
There was a problem hiding this comment.
Requesting changes per the review above: 2 critical (OK own-income test; budget-together drops the ineligible-child allocation in AZ/AR), 3 should, 6 suggestions.
…geting together OAC 317:35-7-40(1) tests the applicant's own income against the individual standard as well as the couple's against the couple standard, so Oklahoma follows SSI's rule, the closer method, not budgeting together. A state that budgets a couple together now keeps SSI's allocations for ineligible children (20 CFR 416.1163(b)) and drops only the deeming threshold, as Arizona's MA611 A and Arkansas's MS O-531 do. Missouri alone counts the couple's gross income (MHABD 0805.015.05). Tests: Oklahoma, Arizona and Arkansas cases; Massachusetts and Kansas Cases 42 to 45 re-expected; one household per state pins every couple-rule flag; a 209(b) case pins the facility split in the expense deduction. source_conflicts/candidates.yaml: the seven candidates gain readings, prs and the deciding parameter files, and a new ruled candidate records 42 CFR 435.601(d)(2)-(3) against the state rules outside 209(b) states. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…into medicaid-couple-limit-ineligible-spouse Main now totals a couple's spend-down budget through the SSI spouse test (PR 10141). In the 209(b) and medically needy formulas, keep this branch's couple-standard switch, facility rule and spouse-income adjustment, and total a couple with main's helpers through couple_budget_total. In the spouse-deeming formula, use main's couple totals for the spouse's gross income and keep this branch's budget-together and facility rules. Main's new cases put a couple and an unrelated adult in one default marital unit. The couple-standard helper read a couple only from a two-person marital unit, so those spouses were each tested alone. It now also reads a pair that the SSI spouse test finds inside a larger unit. Two-person units are read as before. Tests: keep main's medically needy Cases 37 to 39 and renumber this branch's cases 40 to 53. Three comments in main's cases now name the couple limit, which this branch applies to those couples. Case 44 drops its ssi_countable_income assertion: main computes it month by month, so the year no longer equals the December-status figure. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
|
Ruling by the project lead, 2026-11-10. The general rule of 2026-11-02 applies outside the 209(b) states too: where a state's own rule differs from SSI's deeming rule, the model follows the state's rule, including where that is stricter than SSI's rule in a narrow band (42 CFR 435.601(d)(2)-(3)). |
|
Ready for another look. Your change requests of 5 November are covered under "Changes since review, 2026-11-11" at the top of the description:
A law check on 11 November corrected the description with no code change. It adds the missing own-income test as a limit and groups the budget-together states by how their own texts treat children. #10641 tracks both. This PR and #10629 still conflict with each other, so whichever merges second takes |
PR Review (round 2): Apply each state's rule for a Medicaid applicant whose spouse is not aged, blind or disabledPR #10640 (author: hua7450), head Path prefixes used below:
How each claim below was checked is stated with it: run (executed on the head with policyengine-core 3.33), hand trace (arithmetic or code read by hand) or reading (the tests or the source text). SummaryBoth criticals and all three should items from round 1 are fixed, and each fix is pinned.
Two new should items remain, both small.
The merge is sound. All seven hand-resolved files combine both parents correctly. Dropping Case 44's Prior findings
Oklahoma's two remaining differencesBoth are disclosed accurately, in "Changes since review", under "Own-rule states" and "Own-income test", and in the Oklahoma row of #10641. Neither is new (hand trace):
See suggestion G10 for the wording. New findingsCritical (must fix)None. Should addressS4. Minnesota's setting follows the manual against a statute, and no candidate records it.
S5. The default-marital-unit code added in the merge is half pinned and half applied.
Suggestions
Verified
Axiom
CIAll 36 checks pass at Local runs on the head, with policyengine-core 3.33: one probe of 15 households, one in-memory mutation pass over the four setting files, and one pass for the 209(b) split, the Missouri cases and the default-unit helper. Nothing was run on BranchThe branch contains
Sibling notes (not findings):
Verdict: REQUEST_CHANGES |
DTrim99
left a comment
There was a problem hiding this comment.
Requesting two small changes. Both round-1 criticals and all three should-address items are fixed and pinned: Oklahoma follows SSI's rule (Case 16), child allocations are kept in budget-together states except Missouri, the 2026-11-10 ruling is recorded and followed, the seven earlier candidates are complete, and every true state setting fails a test when flipped.
- Minnesota's setting follows the manual against a statute, with no candidate. Minn. Stat. 256B.056, subd. 1a(a)(1) requires SSI's methodologies unless state law or rule says otherwise; the budget-together rule is in the manual only. #10641 already says this "has no candidate entry". Please add one;
state: openis accepted, so it needn't wait for a ruling (Should Address S4). - The default-marital-unit code added in the merge is half pinned and half applied.
spouses_sum's branch for a marital unit of three or more people changes no asserted output when replaced by the plain sum, and the pathway's income test still sums the whole unit while the 209(b) and medically needy budgets total only the couple. Two cases and a one-line change are in the report above (S5).
All seven hand-resolved merge files combine both parents correctly. Dropping Case 44's ssi_countable_income assertion is right for now, since the wrong value comes from main (#10818, fixed by #10820). CI is green at 27ca507758; the branch still conflicts with #10629 in two files.
The aged, blind and disabled pathway's income test summed the whole marital unit under the couple standard, so a default unit of more than two people added another adult's income. Use spouses_sum, as the 209(b) and medically needy budgets do, and pin it with a West Virginia household (the third adult is not the applicant's spouse) and an Arizona one (the applicant passes on the couple's own income). One of main's California cases now finds the second couple eligible on its own income. Add the open candidate us-mn:abd-spouse-income-statute-vs-manual: Minn. Stat. 256B.056, subd. 1a(a)(1) requires SSI's methodologies, while the manual the model follows deems a co-resident spouse's income. Link the 2026-11-10 ruling comment from the budgeting-together candidate, and add a case showing the stricter outcome that ruling covers. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Main now holds PR 10629. In is_209b_ssi_recipient_income_eligible_for_medicaid, the parental medical expense share comes off inside the same couple total as the couple's own expenses, as is_medically_needy_for_medicaid already does. In is_medically_needy_for_medicaid.yaml, main's Illinois Case 40 keeps its number and this branch's cases move to 41-54, with their cross-references; Case 45 now names issue 10818 and PR 10820 for the omitted ssi_countable_income. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Ready for another look at
On the merged code, the nine changed test files and three of 10629's pass locally (321 cases). |
PR Review (round 3): Apply each state's rule for a Medicaid applicant whose spouse is not aged, blind or disabledPR #10640 (author: hua7450), head Path prefixes used below:
Nothing was run locally this round. Each claim below says what it rests on: hand trace (arithmetic or code read by hand), reading (the code, tests or source text) or CI log (the run at this head). Where I say what a changed line would do to a test, that is by my reading, not a run. SummaryBoth should items from round 2 are fixed, and each fix is pinned.
The California re-expectation is the correct result of the fix, not a regression. The second couple has no income. The merge is sound. Both hand-resolved blocks combine their parents correctly. No expectation from #10629 changed, and a child alone in a marital unit still gets their own budget. Nothing new blocks. Three new suggestions, one of which corrects a figure from my round-2 report: after #10820 (merged into Prior findings
New findingsCritical (must fix)None. Should addressNone. Suggestions
Verified
Axiom
CIAll 36 checks pass at From the log of the Baseline (contrib-hhs) job, every changed test file ran with no failure:
I couldn't reconcile the body's local count of 321 with these (the nine changed files alone hold 231). CI is the record. BranchThe branch contains
Sibling notes (not findings):
Verdict: APPROVE |
DTrim99
left a comment
There was a problem hiding this comment.
Approving. Both round-2 items are fixed and pinned: the pathway's income test now totals only the couple (West Virginia Case 18, the new Arizona case and main's California case all turn on it), and Minnesota is recorded as the open candidate us-mn:abd-spouse-income-statute-vs-manual. The California re-expectation is the right consequence of the fix, both hand-resolved merge blocks combine their parents correctly, and no #10629 expectation changed.
One correction to my round-2 report, in G12: the value to restore in Case 45 once main is merged again is 14,320 and 27,320 (by my reading of main's formula after #10820), not 20,880 each. No asserted output changes. The other suggestions are optional. CI is green at 25c45f3214.
Fixes #10459
Fixes #10507
Changes since review, round 2 (2026-11-12)
is_optional_senior_or_disabled_income_eligiblenow usesspouses_sum, as the 209(b) and medically needy budgets do. When a household gives no marital units, an unrelated adult in the default marital unit no longer adds income to a couple's test. Two cases pin the default-unit code:is_optional_senior_or_disabled_income_eligible.yaml, pins the income test. The applicant's $9,760 is within the couple limit of $21,640. Adding the 67-year-old's $19,760 would give $29,520 and fail the applicant.maincase is re-expected for that fix. In "California leaves another couple's dependent's premiums out of a couple budget" (medicaid_optional_senior_or_disabled_countable_income.yaml), the second couple has no income and now passes the pathway: $0 is within the couple limit of $29,863.20. The case's old comment put their failure down to the whole-unit sum. Its inputs are unchanged, and its first-couple comment now names the couple limit.us-mn:abd-spouse-income-statute-vs-manual. The setting follows the state manual, which deems a spouse's income whenever the spouses live together. Minn. Stat. 256B.056, subd. 1a(a)(1) adopts SSI's methodologies instead. "Minnesota" under Limits gives an example, and a new "North Dakota" limit covers that state's pathway setting.us:medicaid-abd-budgeting-together-stricter-than-ssinow links the ruling comment. Uses-couple-standard Case 19 runs the candidate's example. The spouse's $5,964 of interest equals SSI's threshold, so SSI deems nothing. Arizona budgets the couple together and denies the applicant ($21,924 against $21,640). Pennsylvania follows SSI's rule and finds the applicant eligible ($15,960 against $15,960).ssi_countable_incomeassertion stays out until Combine an SSI couple's income by each month's facility status #10820 merges. Until then SSI gives each spouse $31,320, where $20,880 is right.24939d2701. Main now holds Let Missouri aged, blind and disabled people who don't receive SSI spend down #10629, so the two PRs no longer conflict. Inis_209b_ssi_recipient_income_eligible_for_medicaid, the parental medical expense share comes off inside the couple total, as inis_medically_needy_for_medicaid. Inis_medically_needy_for_medicaid.yaml, main's Illinois Case 40 keeps its number, and this PR's cases move from 40-53 to 41-54. Every case number in this description is the new one.Changes since review, 2026-11-11
medicaid_optional_senior_or_disabled_uses_couple_standard.yamlCase 16.is_medically_needy_for_medicaid.yamlCases 46 to 49), and two cases are new (deemed-income Case 7, medically needy Case 54).us:medicaid-abd-budgeting-together-stricter-than-ssi). The seven earlier candidates now carryreadings,prs, the deciding parameter files andyears.from: 2015.23a7c9a8f4. Main now totals a couple's spend-down budget through the SSI spouse test (Apply 416.1163(e)(2) to deemed SSI and identify SSI couples by marriage, not tax roles #10141), and this PR's formulas use those totals. The couple-standard helper also reads a couple that the spouse test finds inside a default marital unit of more than two people, which main's new cases need; two-person units are read as before. This PR's medically needy cases moved after main's new Cases 37 to 39. Case 45 drops itsssi_countable_incomeassertion, because main computes that figure month by month.What and why
The model gave every married Medicaid applicant the couple limit, including someone whose spouse is not aged, blind or disabled and whose income SSI wouldn't deem. #10459 and #10507 proposed SSI's rule for these applicants. Research across 42 states found most states don't follow it, so this PR applies each state's own rule. The model uses a method other than SSI's rule in 19 of the 38 states in the aged, blind and disabled column below, and in 17 of the 32 states in the medically needy column.
North Dakota's 209(b) SSI recipients read the pathway's setting even though the state has no group for people not on SSI.
Changes:
budgets_couple_togetherandcouple_limit_without_deeming.medicaid_optional_senior_or_disabled_uses_couple_standard,medicaid_medically_needy_uses_couple_standardandmedicaid_medically_needy_income_from_spouse.medicaid_optional_senior_or_disabled_income_deemed_from_ineligible_spousenow readsbudgets_couple_together. A budget-together state counts the spouse's income with no deeming threshold, after SSI's allocations for ineligible children (20 CFR 416.1163(b)). Missouri alone counts the spouse's gross income, because MHABD 0805.015.05 says "Do not consider the income and expenses of minor children".medicaid_medically_needy_income_from_spouseadjustsssi_countable_income. In a budget-together state it replaces SSI's deemed amount with the spouse's income after SSI's allocations for ineligible children, counted with no threshold, so one method applies at every income.main, SSI's countable income is now computed month by month (Fix Missouri TANF grants, territory SNAP limits, Hawaii CCAP income and SSI food timing #10332), so for a part-year stay the year's income mixes the two (SSI pools a married couple's income by December's facility status, so a part-year facility stay gives wrong SSI in the other months #10818).A 70-year-old with $1,400 a month of countable income, and a 50-year-old spouse with no income, in 2026:
Sources: each state's citation is in the parameter files. The quotes and what wasn't checked are in the research behind this PR. State plans weren't read, so this rests on state regulations and manuals.
Rulings: Ruled by the project lead, 2026-11-02. When a state's own rule differs from SSI's deeming rule, the model follows the state's rule. SSI's rule applies where the state follows it or is silent. On 2026-11-10 the project lead confirmed that this holds outside the 209(b) states too, where 42 CFR 435.601(d)(2) lets a state's methods be "less restrictive, but no more restrictive" than SSI's. Budgeting together is stricter than SSI's rule only in a narrow band, for example a spouse's income between $473 and $497 a month against poverty-level standards. Candidate
us:medicaid-abd-budgeting-together-stricter-than-ssi. The six earlier rulings cover every state below except Hawaii.us-co:abd-spouse-income-ssi-defined-groupus-ct:abd-spouse-income-hearing-decisionus-hi:abd-spouse-income-household-sizeus-ma:abd-couple-income-standardus-md:medically-needy-non-abd-spouse-unitus-mt:abd-spouse-deeming-thresholdus-ne:medically-needy-unit-sizeLimits:
Own-rule states. AK, NY, OK and CA's aged, blind and disabled groups, and GA, KY, NC, NY and WA medically needy, have rules of their own. This PR gives them the closest method, and Medicaid couple rules: encode the states whose rule is neither SSI's nor budgeting together #10641 tracks the differences. For Oklahoma the closest method is SSI's rule; its two-part test is not modeled. WA keeps its one-person medically needy limit.
Own-income test. SSI's rule has a first step that the model doesn't apply. POMS SI 01320.400 A says "For spouse-to-spouse deeming to apply, the SSI applicant or recipient must be eligible based on their own income", and 20 CFR 416.1163(e)(2) says the result under deeming "cannot be higher than it would be if deeming did not apply". N.J.A.C. 10:72-4.4(d)1 states it for this pathway: "If the countable income (before income deeming) of the aged, blind, or disabled individual exceeds the poverty income guideline for one person he or she is ineligible for benefits and income deeming does not apply." In the states on SSI's rule, once a spouse's income is deemed the model tests only the couple's combined income against the couple limit. A New Jersey 70-year-old with $1,500 a month of Social Security, whose 50-year-old spouse earns $700, has $1,480 of own countable income, above the one-person $1,330, so that rule gives not eligible. The model counts $1,480 + ($700 - $65) / 2 = $1,797.50 against the two-person $1,803.33 and passes the income test (by code trace, not run).
maingives the same result, so this PR doesn't change it; Medicaid couple rules: encode the states whose rule is neither SSI's nor budgeting together #10641 tracks the test. In Oklahoma, whose standards are $1,330 and $1,803.33 a month, a working spouse adds only (wages - $65) / 2. At spouse wages just over $497 a month, the model passes own countable income up to about $1,587 a month, about $257 above the individual standard. The gap closes at spouse wages of about $1,012.Iowa, Maine and Idaho. Ruled by the project lead, 2026-11-02: Medicaid couple rules: encode the states whose rule is neither SSI's nor budgeting together #10641 tracks these, and this PR doesn't change them.
Minnesota. The setting follows Minnesota's eligibility manual in both programs: "Except in the situations noted below, the income of a spouse is deemed to a person if the spouses live together" (EPM 2.3.3.3.2.1.1). Minn. Stat. 256B.056, subd. 1a(a)(1) adopts SSI's methodologies for these groups. The open candidate
us-mn:abd-spouse-income-statute-vs-manualrecords the conflict, and the model follows the manual pending a ruling. A 70-year-old with $16,800 of Social Security and a spouse with no income has $16,800 of countable income, because Minnesota deducts no $20 disregard (subd. 1b). That is within the two-person $21,640 budgeted together, and above the one-person $15,960 tested alone.North Dakota. No North Dakota text read gives a couple standard for the pathway setting. The setting reaches only SSI recipients on the 209(b) route, and at SSI's own standard both methods pass them, so no household result is expected to turn on it (arithmetic, not run). Medicaid couple rules: encode the states whose rule is neither SSI's nor budgeting together #10641 tracks it.
Child allocations. The model applies SSI's allocation for ineligible children (20 CFR 416.1163(b)) in every budget-together state except Missouri, whose manual bars it. The states' own texts, re-read on 2026-11-11, are not all silent on children:
So SSI's allocation is the state's own rule only in the first group, and none of the other treatments is modeled. The Massachusetts and Kansas cases rest on 130 CMR 520.002(A)(1) and MKEESM 4312, which count the spouses' total income and don't mention allocations.
Facility splits.
ssi_countable_income, sossiand theSSI_RECIPIENTroute). Split couples still pool assets.ms_hmw_income_eligibletests a split couple's countable income against its own two-person limit.Dates. The settings start in 2015 because the sources don't date each state's rule.
Population effect. It hasn't been estimated.
Test plan
medicaid_optional_senior_or_disabled_uses_couple_standard.yaml, 6 inmedicaid_medically_needy_uses_couple_standard.yaml, 14 inis_medically_needy_for_medicaid.yaml(Cases 41 to 54), 4 inis_209b_ssi_recipient_income_eligible_for_medicaid.yaml, 3 inmedicaid_category.yaml, 2 inmedicaid_optional_senior_or_disabled_income_deemed_from_ineligible_spouse.yamland 1 inis_optional_senior_or_disabled_income_eligible.yaml. They cover part-year facility stays, couples with an ineligible child on both sides of SSI's threshold, Arizona's and Arkansas's child allocations, Oklahoma, every state setting, facility residents with a spouse who isn't aged, blind or disabled, the 209(b) route and its facility split, both Hawaii settings, the final category for the WV, AZ and RI example, an unrelated aged adult in a default marital unit, and the stricter outcome of budgeting together. One existing case's expected value changes: inmain's "California leaves another couple's dependent's premiums out of a couple budget", the second couple now passes the pathway. Four comments in casesmainadded on 2026-11-10 now name the couple limit, which this PR applies to a couple found inside a default marital unit.state: ruledcandidates are insource_conflicts/candidates.yaml(CO, CT, HI, MA, MD, MT and NE, and the general one on budgeting together outside the 209(b) states), and onestate: opencandidate,us-mn:abd-spouse-income-statute-vs-manual. The rulings above state each ruled one, and "Minnesota" under Limits states the open one. There are no register entries or markers yet.make format(ruff format on the changed variables)changelog.d/Before the merge with main at
24939d2701, the tests for the round-2 fixes passed locally (145 passed). On that merge, the PR's nine changed test files and three of #10629's test files were run one process at a time: 321 passed. Other suites were not run locally; CI validates.axiom: TheAxiomFoundation/rulespec-us#2012 queued
🤖 Generated with Claude Code