Weeks 1–15 • All Tracks • Three-Track Architecture

The instructor materials for the complete 15-week graduate course on Indian Health Financing.
These notes and scripts provide the weekly teaching framework, analytical questions, discussion points, and activities across the Policy & History, Research & Methods, and Operations & Administration tracks.

INSTRUCTOR LECTURE NOTES AND SCRIPTS

COURSE LECTURES: Weeks 1–2 | Weeks 3–4 | Weeks 5–7 | Weeks 8–10 | Weeks 11–12 | Weeks 13–14 | Week 15

Weeks 1–15 • All Tracks • Three-Track Architecture.

Complete instructor lecture notes, session structure, discussion guidance, activities, and teaching script for the 15-week graduate course on Indian Health Financing.

INDIAN HEALTH FINANCING SERIES

VI. Graduate Course Materials

Indian Health Financing

Instructor Lecture Notes and Script

Weeks 1–15 — All Tracks

Compact Edition (v4.1)

Edward J. Fox, PhD

Indian Health Financing Graduate Course Materials

Three-Track Architecture: Policy & History • Research & Methods • Operations & Administration

March 2026

Carved mask, gift to the author from the Port Gamble S’Klallam Tribe

INDIAN HEALTH FINANCING — GRADUATE COURSE

Introduction to the Lecture Notes

Compact Edition (v4.1) — Edward J. Fox, PhD — September 2026

Course Purpose and Design Rationale

Indian Health Financing is a 15-week graduate course organized around a single empirical question: how is health care for American Indian and Alaska Native people actually financed, and is that financing adequate? The course approaches this question not primarily as a policy argument but as an analytical problem — one that requires understanding the legal structure that creates the federal obligation, the institutional architecture through which that obligation is discharged, and the data infrastructure that has only recently made the Medicaid piece of the financing picture measurable with precision.

The analytical spine of the course is the evolution from survey-based estimation to administrative data linkage: from American Community Survey-based estimates of AI/AN coverage rates, through the 2012 CMS Medicaid Analytic eXtract analysis using the AIR proxy, to the 2023 CMS–IHS Administrative Data Match — which for the first time directly linked IHS patient registration records to CMS Medicaid enrollment files and produced a verified count of 941,168 IHS-access Medicaid enrollees with $6.4–$6.6 billion in associated annual expenditures. Every methodological innovation in that arc is treated as a response to specific analytical limitations of the prior approach, not as a simple improvement. Students who understand why each tool was insufficient are far better equipped to read the 2023 Data Match findings critically than those who simply treat it as the most authoritative available number.

The policy stakes are introduced in Week 1 and held open until Weeks 11–12: what happens to Indian health programs if Medicaid contracts? That question can only be answered analytically once students understand the trust responsibility, the IHS funding gap, the 100% FMAP mechanism, the PRC substitution dynamic, and the population denominators that determine the fiscal exposure. The course is structured so that each unit builds the analytical capacity needed to answer that question with precision.

Three-Track Architecture

The course is organized in three parallel tracks that run through every unit. All students complete the same sessions, readings, and graded assignments. The tracks differentiate discussion questions, in-class exercises, and analytical emphasis within each unit — allowing the same empirical material to speak to students with different professional orientations.

Track	Policy & History	Research & Methods	Operations & Administration
Primary audience	Policy analysts, advocates, Tribal leaders	Researchers, data analysts, evaluators	Program directors, billing staff, administrators
Central question	What does this mean for Indian health policy and sovereignty?	How do we know this, and how reliable is the evidence?	How does this work in practice for a Tribal health program?
Weeks 1–2 emphasis	Trust responsibility; Snyder Act vs. entitlement architecture	IHS funding gap as measurement problem	I/T/U system mapping; eligibility in practice
Weeks 3–10 emphasis	Policy implications of each methodological era	ACS → MAX → Data Match: what each revealed and why	How Medicaid billing works through IHS and 638 facilities
Weeks 11–14 emphasis	Fiscal risk; mandatory funding argument; trust responsibility	Population A/B; PRC substitution; data sovereignty	100% FMAP mechanics; state savings; Tribal billing strategy
Signature assignment	Midterm: Data Match governance timeline + sovereignty essay	Data Memo: 3-state Population A/B comparison	Final Research Paper (all tracks)

Track differentiation is marked throughout the lecture notes with explicit labels. Where a segment is appropriate for all tracks, it is marked “All tracks.” Where an exercise or discussion is track-specific, the relevant track is named. Instructors should feel free to emphasize track-relevant dimensions without omitting core analytical content that all students need.

15-Week Course Arc

The course is structured in six unit groups, each building directly on the preceding one. The “Leads Into” note at the top of each unit identifies the specific analytical dependency. Instructors teaching a subset of units should read the “Leads Into” notes carefully to understand what prior context students need.

Unit	Title	Core Question	Track Lead
Weeks 1–2	Foundations	What is the IHS, why does it exist, how is it financed, and why does that financing structure produce a chronic funding gap that Medicaid has increasingly been called upon to fill?	All tracks
Weeks 3–4	The Survey Era and Its Limits	Why was survey data insufficient for Indian health financing analysis — and what did each successive methodological innovation reveal that the previous approach could not?	Research & Methods
Weeks 5–7	Building the Data Match	How was the CMS–IHS Data Match built, and why did the governance structures matter as much as the technical work?	Research & Methods
Weeks 8–10	What the Data Shows	What did the administrative data actually show — and what does the Population A/B distinction mean for how we read those findings?	Research & Methods
Weeks 11–12	Policy Implications	What do the Data Match findings mean for Indian health policy — specifically, what is the fiscal risk to Indian health programs if Medicaid contracts, and how should that risk be communicated to policymakers?	All tracks
Weeks 13–14	The Tribal-Led Research Model	What did the NIHB-led model produce that a federal-agency-led project would not — and what are the honest limitations and unresolved tensions for the next generation of Tribal health data work?	Research & Methods
Week 15	Synthesis	What does a student who has completed this course know — analytically and practically — that they did not know 15 weeks ago? And what does the course leave unresolved?	All tracks

The course arc is intentionally asymmetric: the methodological units (Weeks 3–10) are more technically demanding than the policy units (Weeks 11–12), and the synthesis unit (Week 15) requires more from students than any prior session. Instructors should resist the temptation to lighten Week 15 — it is not a review session but a synthesis session, and the distinction matters.

Graded Assignments

The course has three graded assignments, each launched during class with structured scaffolding. The compact edition retains all three assignments from the full v5 edition.

Assignment	Launch / Due	Format	Purpose
Midterm	Launched Week 7; due end of Week 7	Data Match governance timeline + sovereignty essay	Tests whether students can read the Data Match not just as a dataset but as a governance artifact — who designed it, under what constraints, and for whose benefit
Data Memo	Launched Week 10; due end of Week 10	4–6 page analytical memo: 3-state Population A/B comparison	Applies the Population A/B framework to real state-level data; develops the ability to explain methodological scope limits to a policy audience
Final Research Paper	Due end of Week 14	Original policy or research paper (all tracks)	Integrates the full analytical arc of the course; students choose their own question within the Indian health financing domain

How to Use These Notes

Instructor script. Passages marked “READ ALOUD / ADAPT” are suggested instructor script. They are written to be spoken, not read verbatim — adapt pacing and emphasis to your classroom. The goal is not performance but precision: the framing language in these passages carries specific analytical content that lecture improvisation tends to soften.

Discussion protocols. Each unit includes track-differentiated discussion questions with “What to Listen For” guidance. These are not assessment rubrics — they are calibration tools for the instructor, identifying which student responses indicate genuine analytical understanding versus surface familiarity.

Timing. Each session includes a timing overview table. These are targets, not constraints. The most common timing failure is spending too long on the legal foundations in Weeks 1–2 and not enough time on the I/T/U mapping exercise — which is the most pedagogically important activity in the opening unit.

Course arc notes. Each unit opens with a Course Arc Note explaining how that unit fits into the 15-week sequence. Read these before teaching the unit, not after. They contain the most important instructor-facing content in the notes — the analytical rationale for why the unit is structured as it is and what the common pedagogical failure modes look like.

Course-internal documents. Each unit’s Primary Sources list includes references to course-internal analytical documents that are distributed to students separately. These documents are not reproduced in the lecture notes. The table below identifies each document and its role in the course.

Course-Internal Analytical Documents

The following documents are authored by the course instructor and distributed to students as core reading. They are cited by name in the Primary Sources lists throughout these notes but are not reproduced here.

Document	Role in the Course
CMS–IHS Data Match Lecture Paper (Sections I–VII)	The primary analytical spine of the course. Assigned progressively: Sections I–III in early units for historical and methodological context; Sections IV–VII in later units for findings, policy analysis, and data sovereignty.
Revised Analysis 1999–2023	Longitudinal data document reconciling the historical CRIHB/NIHB expenditure series (1999–2012) with the 2023 Data Match findings. Full document assigned in Weeks 8–10; Section III in Weeks 11–12.
Lecture Series Parts 1–3: Counting Who We Serve	Extended methods companion for the Research & Methods track. Part 2 (Weeks 5–7) covers the political history of the Data Match. Part 3 (Weeks 8–14) covers findings, sovereignty, and the Tribal-led research model evaluation.
CMS Investment Estimates Paper (ACS PUMS 2024 × NHE 2024 × Data Match 2023)	Quantitative engine for the total federal investment framework in Weeks 11–12. Tables 1, 3, and 6 are specifically assigned for coverage distribution, spending scenarios, and the total investment framework.

INDIAN HEALTH FINANCING

All Tracks

Weeks 1–2 Instructor Lecture Notes & Script

Foundations: The Trust Responsibility, the IHS System, and Why Financing Matters

Track	All tracks (differentiated discussion in Session 2)
Sessions	Two 90-minute sessions spanning Weeks 1 and 2
Core Question	What is the Indian Health Service, why does it exist, how is it financed, and why does that financing structure produce a chronic funding gap that Medicaid has increasingly been called upon to fill?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. I); Snyder Act (1921); IHCIA (1976, 2010 reauthorization); IHS Budget Justifications (current year); NIHB Tribal Data Project reports; NIHB 2022 Budget Formulation (mandatory funding brief)
Leads Into	Weeks 3–4: Survey-era data methods and ACS limitations

Course Arc Note

Weeks 1–2 establish the structural context that every subsequent analytical unit depends on. Students who do not understand the trust responsibility, the I/T/U system (or non-system) architecture, and the IHS funding gap will not understand why the Medicaid data question matters. These sessions are not background — they are the analytical foundation. The policy stakes introduced here (what happens if Medicaid is cut?) should be held open throughout the course and answered only in Weeks 11–12.

Two-Session Architecture

Session	Core Argument	Key Activity
Session 1 — Week 1	The Indian Health Service exists because of a legal obligation, not a charitable choice. The Snyder Act and the IHCIA established a federal duty to provide health care to eligible AI/AN people. That duty has never been fully funded — and the gap between obligation and appropriation is the structural fact that motivates everything else in this course.	I/T/U system (or non-system) mapping exercise
Session 2 — Week 2	The IHS funding gap is not a budgetary accident. It reflects a political economy in which AI/AN health has historically been underprioritized in federal appropriations. Medicaid entered the picture not as a solution to the gap but as a general entitlement that AI/AN people are eligible for — and that the IHS system has progressively built its financial model around.	Funding gap calculation exercise; Medicaid entry point discussion

SESSION 1  ·  90 minutes — Week 1

The Trust Responsibility and the Architecture of Indian Health

Legal foundations · Snyder Act and IHCIA · I/T/U system (or non-system) structure · eligibility · the funding gap concept

Session 1 Timing Overview

Time	Segment	Format	Notes
0–10	Opening: the question this course answers	Mini-lecture	Establish stakes from day one
10–30	Section 1.1: Legal foundations of Indian health	Lecture	Trust responsibility; Snyder Act; IHCIA
30–50	Section 1.2: The I/T/U system (or non-system)	Lecture + table	Three delivery modes; eligibility; scope
50–68	Section 1.3: I/T/U system (or non-system) mapping exercise	Small groups	15 min work, 8 min debrief
68–80	Section 1.4: What the legal structure does and does not guarantee	Discussion	Obligation vs. appropriation
80–90	Bridge and preview	Mini-lecture	Introduce the funding gap concept

1.0 — Opening Hook (0–10 min)

READ ALOUD / ADAPT

“Here is the question this course is designed to answer: the federal government has a legal obligation to provide health care to American Indian and Alaska Native people. How is that obligation financed — and is it adequately financed? The short answer is: through the Indian Health Service, and no. The IHS has been chronically appropriated at below-need levels for most of its history. The gap between what the law obligates and what Congress funds has been partially, and increasingly, filled by Medicaid — a federal-state entitlement program that AI/AN people access as eligible low-income individuals, not as a trust-responsibility instrument. To understand that sentence fully — to understand what it means for policy, for Tribal sovereignty, for individual patients, and for the financial stability of Indian health programs — you need to understand three things: the legal foundation that creates the obligation, the institutional architecture through which it is (imperfectly) discharged, and the data infrastructure that has only recently made the Medicaid piece measurable. This course covers all three. We start today with the legal foundation.”

1.1 — Legal Foundations of Indian Health (10–30 min)

The federal obligation to provide health care to American Indians and Alaska Natives is rooted in two intersecting legal traditions: the treaty obligation arising from the government-to-government relationship between the United States and Tribal nations, and the plenary power doctrine under which Congress has exercised broad authority over Indian affairs since the nineteenth century.

Key Concept

The trust responsibility: The federal trust responsibility is the legal and moral obligation of the United States to protect the interests of Tribal nations and their members, arising from treaties in which Tribes ceded land and sovereignty in exchange for federal protections including health care. The Supreme Court has affirmed the trust responsibility in numerous decisions. It is not a policy preference — it is a legally binding commitment.

Legal Authority	What It Established
Snyder Act (1921)	Authorized federal expenditures ‘for the benefit, care, and assistance of the Indians throughout the United States’ including ‘relief of distress and conservation of health’ — the statutory basis for IHS appropriations
Transfer Act (1954)	Transferred Indian health functions from the Bureau of Indian Affairs to the Public Health Service, creating the institutional predecessor to IHS
Indian Health Care Improvement Act (1976)	Established IHS as a permanent federal agency; defined service population; created the legal framework for Tribal self-determination in health service delivery; set goals for health status parity
Indian Self-Determination and Education Assistance Act (1975)	Authorized Tribes to contract with the federal government to operate IHS programs directly, creating the Tribal (638) contract and compact system
IHCIA Reauthorization (2010)	Modernized IHS authorities; strengthened Medicaid and Medicare provisions; expanded behavioral health and other service categories; permanently authorized the Tribal self-governance program
Snyder Act appropriations structure	Congress appropriates IHS funds annually through the Interior, Environment, and Related Agencies appropriations bill — subject to the same political pressures as all discretionary spending, with no guaranteed funding floor

Tribal Sovereignty Note

Government-to-government relationship: The trust responsibility is grounded in the government-to-government relationship between the United States and federally recognized Tribal nations. This means IHS is not a social services agency providing benefits to a disadvantaged population — it is the federal government partially discharging a treaty-based legal obligation to sovereign nations. This distinction has significant implications for how Medicaid integration should be framed: Medicaid revenue flowing to Indian health programs is not charity, it is an entitlement that AI/AN people earn through their tax contributions and Medicaid eligibility, accessed through a trust-responsibility-defined health system.

1.2 — The I/T/U system (or non-system) (30–50 min)

The Indian health system (or non-system) is not a single federal agency. It is a three-component delivery network — IHS-operated facilities, Tribally-operated programs, and Urban Indian health organizations (the I/T/U system (or non-system)) — that together serve approximately two million AI/AN people annually.

Component	Description	Medicaid Billing Authority
IHS-operated (I)	Facilities directly operated by the federal Indian Health Service: hospitals, health centers, and health stations on or near reservations. Staffed by federal employees.	Bills Medicaid as an IHS facility; receives 100% federal medical assistance percentage (FMAP) for most services — the federal government pays 100% of the non-federal share
Tribally-operated (T)	Health programs operated by Tribal nations under self-determination contracts (638 contracts) or self-governance compacts. Range from small clinics to large hospital systems.	Same 100% FMAP as IHS-operated facilities when serving AI/AN patients; Tribes retain Medicaid revenue directly rather than returning it to IHS
Urban Indian Organizations (U)	Nonprofit organizations providing health care to AI/AN people in urban areas, funded through IHCIA Title V grants. Serve an AI/AN population that is disproportionately not IHS-registered.	Do not receive 100% FMAP; bill Medicaid at standard rates; most significant access point for Population B (non-IHS-registered) AI/AN Medicaid patients

Key Concept

The 100% FMAP provision: Normally, states pay a share of Medicaid costs (the non-federal share, typically 30–50%). For services provided at IHS and tribally operated facilities to AI/AN patients, the federal government pays 100% of the Medicaid costs — eliminating the state financial stake. This provision is a major driver of state behavior around AI/AN Medicaid enrollment: states have limited financial incentive to invest in AI/AN Medicaid outreach because they bear none of the cost. Understanding this incentive structure is essential context for the enrollment data analysis in later weeks.

1.4 — What the Legal Structure Does and Does Not Guarantee (68–80 min)

READ ALOUD / ADAPT

“The Snyder Act authorizes health expenditures for AI/AN people. The IHCIA defines the IHS mission and service population. The trust responsibility creates a legal obligation. None of these documents specifies how much money Congress must appropriate. The legal obligation and the appropriations process are structurally disconnected. Congress can — and historically has — appropriated IHS at levels that analysts across the political spectrum agree are insufficient to meet the obligation the law establishes. This is not a contested interpretation: IHS budget justifications themselves acknowledge the gap between funding levels and population health needs. This structural disconnection is the starting point for everything that follows in this course. If Congress fully funded the trust responsibility, Medicaid would be a supplementary revenue source for Indian health programs. Because Congress does not, Medicaid has become load-bearing infrastructure — and the question of how much Medicaid flows through the Indian health system (or non-system), and what would happen if it contracted, becomes a question of whether the federal government can meet its legal obligations at all.”

Common Misconception

Common misconception to address directly: Students sometimes assume that the trust responsibility functions like a legal mandate that forces Congress to appropriate sufficient funds. It does not. Courts have generally declined to specify appropriations amounts required to discharge the trust responsibility. The obligation is real and legally recognized; its enforcement through the appropriations process is not.

Bridge and Preview (80–90 min)

Instructor Note

Preview Session 2 by introducing the IHS funding gap concept quantitatively. The IHS Budget Justification uses a ‘current services’ vs. ‘need’ comparison that produces an annual gap figure. Preview question: if IHS is funded at approximately 60–70% of assessed need, and Medicaid increasingly fills the remaining gap, what do we need to know about Medicaid flows to Indian health to understand the real fiscal picture? That is the question Weeks 3–15 answer.

SESSION 2  ·  90 minutes — Week 2

The IHS Funding Gap and Medicaid’s Entry into Indian Health Finance

IHS appropriations history · the funding gap · Medicaid eligibility for AI/AN · IHCIA Medicaid provisions · why Medicaid became load-bearing

Session 2 Timing Overview

Time	Segment	Format	Notes
0–10	Bridge: from obligation to funding	Mini-lecture	Connect legal framework to fiscal reality
10–32	Section 2.1: The IHS funding gap	Lecture + table	Appropriations history; gap quantification
32–52	Section 2.2: Medicaid’s entry into Indian health	Lecture	Eligibility; 100% FMAP; IHCIA provisions
52–68	Section 2.3: Funding gap calculation exercise	Small groups	Apply the gap framework; Medicaid offset analysis
68–82	Section 2.4: Track-differentiated discussion	Groups	Three-track questions
82–90	Bridge to Weeks 3–4	Mini-lecture	The data measurement problem

2.1 — The IHS Funding Gap (10–32 min)

The IHS funding gap refers to the difference between the amount Congress appropriates for the Indian Health Service and the amount analysts estimate would be required to provide health services at a level comparable to the general U.S. population. The gap has been documented by IHS itself, by the Government Accountability Office, and by independent researchers across multiple decades. The NIHB’s 2022 Budget Formulation brief quantifies the gap using per-capita comparisons: combined federal spending (IHS appropriations plus an estimated $3 billion from CMS programs) reaches approximately $5,625 per active user — roughly 50% of the 2018 national health expenditure benchmark of $11,172 per capita. The brief estimates full funding at $29.05 billion annually (FY 2022 base) for the 2.6 million service population, plus a one-time $10 billion infrastructure investment, and proposes a phased glide path reaching that level by FY 2027.

Year / Period	Key Financing Fact
1955–1970s	IHS transferred to DHHS predecessor; chronic appropriations below assessed need established as structural pattern from first decade
1976	IHCIA establishes health status parity as an explicit goal; Congress acknowledges gap between funding and goal at time of passage
1999	Total AI/AN Medicaid payments approximately $1.45 billion; IHS/Tribal facility billing share approximately 13%
2000s	IHS appropriations grow slowly in nominal terms; health care cost inflation and population growth widen real gap annually
2008	AI/AN Medicaid payments approximately $3.86 billion; IHS/Tribal billing share approximately 21%; Medicaid already a dominant revenue stream
2010	IHCIA reauthorization strengthens Medicaid and Medicare provisions; acknowledges Medicaid as central to Indian health system (or non-system) sustainability
FY2023	IHS appropriations approximately $8.6 billion; CMS–IHS Data Match estimates $6.4–6.6 billion in IHS-access Medicaid expenditures; Medicaid approaches IHS appropriations in magnitude
Funding gap estimate	IHS Budget Justifications and independent analyses consistently estimate IHS is funded at 55–70% of comparable-care need; the gap is $3–6 billion annually depending on methodology

Key Concept

The PRC funding gap as the sharpest indicator: The Purchased/Referred Care program — which funds care that IHS cannot provide directly (specialty care, hospital services beyond IHS capacity) — is where the funding gap is most visible. When PRC funds run out, patients are told their referrals cannot be approved. Medicaid increasingly absorbs PRC-equivalent costs for enrolled patients. The 2023 Data Match estimates $2.6 to $4.0 billion in Medicaid expenditures that substitute for what would otherwise be PRC costs. This substitution relationship is the analytical core of the fiscal risk argument in Weeks 11–12.

2.2 — Medicaid’s Entry into Indian Health Finance (32–52 min)

Medicaid was enacted in 1965 as Title XIX of the Social Security Act. AI/AN people were not a primary design consideration — but as low-income Americans, they were eligible for Medicaid from its inception. The integration of Medicaid into Indian health finance evolved gradually through three mechanisms: direct patient eligibility, the 100% FMAP provision, and the IHCIA Medicaid-specific authorities.

Mechanism	How It Works	Policy Significance
AI/AN Medicaid eligibility	AI/AN people who meet income and residency requirements are eligible for Medicaid like all other Americans; IHS service is not required for Medicaid eligibility	Creates a large potential enrollment population; the question is whether eligible individuals are actually enrolled and whether their IHS care is billed to Medicaid
100% FMAP	For services at IHS and tribally operated facilities, the federal government pays 100% of Medicaid costs; states pay nothing The state Medicaid program pays and the federal government reimburses eligible services at 100% of Federal All-inclusive IHS/CMS Rate.	Eliminates state financial stake in AI/AN Medicaid enrollment; reduces state incentive to invest in outreach; makes federal government sole payer
IHS Medicaid billing authority	IHS facilities can bill Medicaid directly for covered services provided to enrolled patients; Medicaid becomes a revenue source for IHS programs	Transformed Medicaid from a patient benefit into an IHS institutional revenue stream; the 1999–2008 billing share growth reflects expanding use of this authority
IHCIA special provisions	2010 reauthorization added exemptions from Medicaid cost-sharing for AI/AN patients; strengthened protections for AI/AN managed care access to I/T/U providers	Protected AI/AN patients from cost-sharing barriers and managed care network restrictions that could limit IHS access

2.3 — Funding Gap Calculation Exercise (52–68 min)

EXERCISE: FUNDING GAP ANALYSIS  |  16 minutes

Format: Individual or pairs. Using the data points introduced in Session 2, work through the following calculations and questions. IHS appropriations for FY2023 are approximately $8.6 billion. What is the implied annual funding gap given uncertainty about the level of need ($15 billion to $32 billion)The CMS–IHS Data Match estimates $6.4–6.6 billion in Medicaid expenditures for IHS-access enrollees (Population A). If Population B adds approximately $1.88 billion, what is total estimated AI/AN Medicaid spending?Total federal investment in AI/AN health = IHS appropriations + AI/AN Medicaid + Medicare + other. Using $8.6B IHS + $8.45B Medicaid (total), what is the approximate combined federal investment? How does Medicaid compare to IHS as a share?If Medicaid were reduced by 15%, what is the estimated dollar impact on IHS-access Medicaid spending alone? What share of the IHS annual appropriation does this represent?   Debrief answers: (1) Gap = ~$4.6B at 65% funding. (2) Total AI/AN Medicaid ~$8.45B (Pop. A $6.4–6.6B IHS-access + Pop. B ~$1.88B non-IHS-registered). (3) Combined ~$19.3–19.9B (midpoint $19.6B): IHS ~$8.6B + Medicaid ~$8.45B + Medicare ~$2.52–2.88B midpoint $2.64B (ACS-derived: 260,131 IHS-access Medicare enrollees × 2024 NHE PMPY) + other (CHIP, VA, BIA not estimated). Note: total CMS investment across all coverage groups in the IHS-access population is $9.23–$12.49B (midpoint $10.86B), with 51.1% of the IHS-access population (778,438 individuals) having no CMS coverage at all. Medicaid approximately equal to IHS. (4) 15% reduction = ~$960M–1.27B; approximately 11–15% of IHS annual appropriation. These figures will return in Weeks 11–12.

2.4 — Track-Differentiated Discussion (68–82 min)

Track	Discussion Question	What to Listen For
Policy & History	The trust responsibility creates a legal obligation but does not specify a funding amount. How has Congress historically interpreted its appropriations duty under the trust responsibility, and what political factors explain the persistent funding gap?	Students should identify: the discretionary appropriations structure; the relative political power of AI/AN constituencies; the role of advocacy organizations like NIHB; the historical context of federal Indian policy cycles
Research & Methods	IHS funding gap estimates range from $3 billion to $6 billion annually depending on methodology. What methodological choices produce this range? What would a well-designed study of the IHS funding gap need to measure?	Students should identify: the comparator population choice (what does ‘comparable care’ mean?); the cost vs. utilization gap distinction; the role of unmet need vs. foregone care; data limitations
Operations & Administration	You are the director of a tribal health program that operates under a 638 compact. Your program receives both IHS base funding and Medicaid revenue. If Medicaid payments declined by 20%, what operational decisions would you face, and in what order?	Students should identify: staffing and capacity decisions; PRC prioritization; billing and collections investment; relationship with state Medicaid agency; tribal government advocacy

Bridge to Weeks 3–4 (82–90 min)

Instructor Note

Close Session 2 by naming the data measurement problem explicitly. Students now understand what IHS is, why it is underfunded, and why Medicaid has become load-bearing. The next question is: how do we know how much Medicaid actually flows through the Indian health system (or non-system)? That turns out to be surprisingly hard to measure — and the effort to measure it accurately is the methodological story of the course. Weeks 3 and 4 begin that story with the survey era and its limitations.

Appendix: Glossary of Key Terms

Term	Definition
Trust responsibility	The legal and moral obligation of the United States to protect the interests of Tribal nations and their members, arising from treaties; the foundational basis for federal Indian health care obligations
Snyder Act (1921)	Federal statute authorizing appropriations for ‘the benefit, care, and assistance of the Indians’ including health; the statutory basis for IHS funding
Indian Health Care Improvement Act (IHCIA)	Federal statute (1976, reauthorized 2010) establishing IHS as a permanent agency, defining the service population, and authorizing the full range of Indian health programs including Medicaid-specific provisions
I/T/U system (or non-system)	The three-component Indian health delivery network: IHS-operated facilities (I), Tribally-operated programs (T), and Urban Indian Organizations (U)
638 contract/compact	Mechanism under the Indian Self-Determination Act by which Tribal nations assume operation of IHS programs; 638 contractors and self-governance compactors bill Medicaid directly and retain revenue
100% FMAP	The federal medical assistance percentage applied to Medicaid services at IHS and tribally operated facilities; the federal government pays 100% of Medicaid costs, states pay nothing
IHS funding gap	The difference between IHS appropriations and the amount estimated to be required to provide health services at a level comparable to the general U.S. population; estimated at $3–6 billion annually
Purchased/Referred Care (PRC)	The IHS program that funds health services that cannot be provided directly by IHS or tribal facilities; the program where IHS funding shortfalls are most visibly felt
Population A	AI/AN individuals with IHS registration who are Medicaid-enrolled and generated at least one paid IHS service; the population measured by the CMS–IHS Data Match
Population B	AI/AN Medicaid enrollees without IHS registration; primarily urban populations; not captured by IHS-registration-based data systems

Indian Health Financing  ·  All Tracks  ·  Weeks 1–2 Lecture Notes  ·  March 2026

 Weeks 3–4 

INDIAN HEALTH FINANCING

Research & Methods Track

Weeks 3–4 Instructor Lecture Notes & Script

Before the Match: Survey Data, Administrative Claims, and the Road to the CMS–IHS Data Match

Track	Research & Methods (appropriate for all tracks; methods emphasis marked throughout)
Sessions	Two 90-minute sessions spanning Weeks 3 and 4
Core Question	Why was survey data insufficient for Indian health financing analysis — and what did each successive methodological innovation reveal that the previous approach could not?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. I–III); ACS documentation; CMS MAX technical documentation; NWTEC/NPAIHB institutional materials; TTAG (2015) Data Symposium Report
Builds On	Weeks 1–2: IHS financing foundations; trust responsibility; I/T/U system (or non-system) structure
Leads Into	Weeks 5–7: CMS–IHS Data Match construction, governance, and Tribal data sovereignty

Course Arc Note

These two sessions establish the methodological arc that runs through the entire course. Students who understand WHY each data system was inadequate — not just that it was inadequate — will be far better equipped to read the 2023 Data Match findings critically. The progression from ACS to MAX to Data Match is not a story of simple improvement; it is a story of different tools answering different questions, each with its own blind spots. That nuance is what distinguishes advanced policy analysts from advocates who simply cite the largest available number.

Two-Session Architecture

Session	Core Argument	Key Activity
Session 1 — Week 3	The American Community Survey was the primary tool for estimating AI/AN Medicaid enrollment for two decades — but its design was never built for financing analysis. Understanding exactly what it could and could not show is the foundation for reading every subsequent data development critically.	ACS vs. administrative data mapping exercise
Session 2 — Week 4	The 2012 MAX administrative analysis marked the first reliable national count of IHS-access Medicaid enrollment and established the structural boundary — the IHS enrollment limit — that still shapes every AI/AN Medicaid dataset today. NWTEC’s regional linkage model illustrates the same boundary at sub-national scale, using a public-health-oriented data infrastructure that previews the technique the federal Data Match would scale nationally.	Methodological evolution timeline construction

SESSION 1  ·  90 minutes — Week 3

Survey Data and Its Limits: What the ACS Could and Could Not Show

ACS design and AI/AN coverage · what self-report reveals · the Medicaid enrollment estimate problem · what the survey era accomplished

Session 1 Timing Overview

Time	Segment	Format	Notes
0–10	Opening: the question no one could answer	Mini-lecture	Establish policy stakes
10–30	Section 1.1: What the ACS is and what it was built for	Lecture + table	Design and AI/AN application
30–50	Section 1.2: What the ACS could and could not show	Lecture + table	Strengths and structural limits
50–68	Section 1.3: ACS vs. administrative data exercise	Small groups	15 min work, 8 min debrief
68–80	Section 1.4: What the survey era got right	Full group	Avoid presentism; set up Session 2
80–90	Bridge and preview	Mini-lecture	Transition to MAX and administrative data

1.0 — Opening Hook (0–10 min)

READ ALOUD / ADAPT

“Imagine you are the chief financial officer of a tribal health program in 2015. A state legislator asks you: how many of your patients are currently enrolled in Medicaid? You know roughly how many patients walk through your door. You know roughly what your state Medicaid payments were last year. But you cannot tell her with precision how many individuals on your active patient list are currently Medicaid-enrolled, because you do not have a direct link between your patient records and the state Medicaid enrollment file. Now imagine a researcher asks a different version of the same question at the national level: how many American Indians and Alaska Natives are enrolled in Medicaid? And of those, how many are actually using IHS or tribally operated facilities? How much is Medicaid actually paying for their care? For most of the period between 1990 and 2020, the best available answer to those national questions came from a household survey — the American Community Survey. Today we are going to understand exactly what that means: what the ACS could measure, what it could not, and why a field of analysts spent two decades trying to build something better.”

Instructor Note

This is a good moment to connect back to Weeks 1–2: what does chronic IHS underfunding have to do with needing better Medicaid data? The connection is direct — if policymakers cannot accurately quantify how much Medicaid flows through Indian health programs, they cannot make the case that cutting Medicaid would devastate the IHS system. The data gap is not just a research problem. It is a policy advocacy problem.

1.1 — What the ACS Is and What It Was Built For (10–30 min)

The American Community Survey replaced the Census long form in 2005. It is the largest ongoing household survey in the United States, reaching approximately 3.5 million addresses annually. Its primary purpose is to measure the demographic, social, economic, and housing characteristics of the American population — not to track health insurance program enrollment.

ACS Design Element	What It Means for AI/AN Health Coverage Analysis
Sample size	~3.5 million addresses/year nationally; AI/AN oversample insufficient for tribal-level analysis in most states
Coverage question format	Self-reported: respondents answer whether they currently have health coverage and what type; no administrative verification
IHS/Indian health category	Respondents can report ‘Indian Health Service’ as a coverage type; this is a self-identification category, not linked to IHS patient records
Medicaid coverage question	Single yes/no; does not distinguish between IHS-facility users and mainstream-provider users among AI/AN Medicaid enrollees
Geographic granularity	Reliable state-level estimates for large AI/AN populations; unreliable for most individual tribal service areas due to small cell sizes
Race identification	Self-reported; AI/AN race coding varies; multiracial individuals may or may not identify as AI/AN
Frequency	Annual rolling sample; 5-year estimates used for small populations; 3–5 year lag in data availability

Instructor Note

A useful analogy: asking the ACS to measure AI/AN Medicaid enrollment in specific tribal programs is like asking a population census to track whether individuals have active prescriptions. The survey was designed to measure population-level coverage rates, not to verify who is enrolled in which program. That is not a design flaw — it is a scope mismatch between the tool and the question analysts needed to answer.

What the ACS Measured for AI/AN Populations

Despite its limitations, the ACS was the primary data source for AI/AN health coverage research from the mid-2000s through the early 2020s. The NIHB Tribal Data Project produced annual reports using ACS data that were central to federal advocacy. Understanding what the ACS could show is essential context before evaluating its limits.

What the ACS Could Show	Why It Mattered
National AI/AN uninsured rates over time	Documented the disproportionately high uninsured burden; provided baseline for ACA impact analysis
State-level coverage rate comparisons	Enabled cross-state comparisons of Medicaid vs. private coverage vs. uninsured for AI/AN populations
Post-ACA enrollment trends	Documented Medicaid expansion impact on AI/AN coverage rates in expansion vs. non-expansion states
IHS as a coverage type	Captured self-reported IHS coverage as a distinct category; enabled estimates of IHS-access populations in survey form
Rapid-turnaround estimates	ACS data available annually; faster than building administrative linkages
Urban vs. reservation coverage differentials	Documented coverage disparities between reservation-resident and urban AI/AN populations

1.2 — What the ACS Could Not Show (30–50 min)

READ ALOUD / ADAPT

“Let me be precise about what the ACS limitations actually were — because there are two kinds of limitations, and conflating them leads to sloppy analysis. The first is a measurement limitation: the ACS could not verify enrollment, could not link to payment records, and could not tell you what Medicaid actually paid for AI/AN health care. The second is a structural limitation: the ACS was not designed to distinguish between AI/AN Medicaid enrollees who use IHS facilities and those who use mainstream providers. Both limitations matter, but they matter differently — and only the second is truly unsolvable by improving survey design.”

Common Misuse / Warning

The ACS AI/AN Medicaid enrollment estimates and the CMS–IHS Data Match figure of 941,168 are not measuring the same population. The ACS estimated approximately 1.1–1.3 million AI/AN Medicaid enrollees nationally (varies by year). The Data Match found 941,168 IHS-access Medicaid enrollees. The difference is not primarily a data quality problem — it is a population scope difference. The ACS includes Population B (AI/AN Medicaid enrollees who do not use IHS facilities); the Data Match captures only Population A (IHS-registered individuals). Students who conflate these figures will produce analytically invalid arguments.

What the ACS Could Not Show	Why It Mattered for Financing Analysis
Verified Medicaid enrollment	Self-report has known inaccuracies; no administrative confirmation
Medicaid payment flows	Coverage data is not claims data; enrollment tells you nothing about what Medicaid paid
IHS vs. non-IHS Medicaid use	No way to distinguish AI/AN Medicaid enrollees using IHS from those using mainstream providers — the distinction at the core of the Data Match
Tribal-level estimates	Sample size insufficient for most individual tribal service areas
Expenditure magnitudes	The ACS contains no spending data; ‘how much does Medicaid spend on AI/AN health?’ is entirely unanswerable from ACS alone
PRC substitution effects	Whether Medicaid was absorbing costs that would otherwise fall on Purchased/Referred Care cannot be measured from survey data
Enrollment duration	ACS point-in-time snapshot; cannot distinguish year-round from intermittent enrollment

Key Concept

The financing analysis gap: The core question for Indian health financing advocacy is not ‘what percentage of AI/AN people have Medicaid?’ but rather ‘how many dollars flow through Indian health programs via Medicaid, and what would happen to those programs if that flow were disrupted?’ The ACS can help with the first question and is essentially useless for the second. Administrative data — enrollment records linked to claims records — is the only way to answer the second question.

1.3 — ACS vs. Administrative Data Mapping Exercise (50–68 min)

EXERCISE: DATA SOURCE MAPPING  |  15 minutes

Format: Groups of 3–4 students. Each group receives the following policy questions. For each, answer: (1) Can the ACS address this? If yes, how reliably? (2) Would administrative data answer it better, and what would that require? (3) Is the question answerable at all with current data? Policy questions: How many AI/AN individuals are currently enrolled in Medicaid nationally?How much did Medicaid spend on AI/AN health care in FY2023?What is the Medicaid uninsured rate among AI/AN adults in New Mexico?What share of patients at Navajo Nation health programs are Medicaid-enrolled?Did Medicaid expansion under the ACA increase AI/AN enrollment in reservation states?How much of the IHS Purchased/Referred Care budget is effectively substituted by Medicaid?   Debrief: Questions 1, 3, and 5 are partially answerable from the ACS with caveats. Questions 2, 4, and 6 require administrative data — and question 6 requires not just enrollment data but claims data linked to PRC records. Emphasize that the question you are trying to answer should determine the data source, not the other way around.

1.4 — What the Survey Era Got Right (68–80 min)

READ ALOUD / ADAPT

“Before we move to the administrative data era, I want to name something the survey era accomplished that is easy to undervalue in retrospect. The NIHB Tribal Data Project, using ACS data, produced a decade of consistent, publicly available estimates of AI/AN health coverage rates. Those reports landed on congressional desks. They were cited in floor debates. They informed ACA Indian health provisions. You cannot retroactively credit the ACS with precision it did not have. But you also should not dismiss what it achieved: a regular, credible, publicly accessible series of coverage estimates that kept AI/AN health equity visible in federal policy conversations during a period when it could easily have been invisible. The 2023 Data Match findings are more precise and more policy-powerful. But they did not appear from nowhere. They appeared because two decades of ACS-based advocacy established that this was a population policymakers needed to think about. The lesson is not ‘the ACS was bad.’ The lesson is that every data system has a specific purpose and a specific set of questions it can answer. The ACS was directionally right about the scale of AI/AN Medicaid enrollment. It was not precise enough for financing analysis. Understanding both things is what makes you a good analyst rather than just an advocate.”

Bridge and Preview (80–90 min)

Instructor Note

Close Session 1 by previewing Session 2. Core transition: if survey data cannot link enrollment to payment, the obvious next question is — what if we go directly to the Medicaid claims data? That is exactly what the 2012 MAX analysis did. But accessing Medicaid administrative data for AI/AN populations introduced a new problem: how do you identify AI/AN patients in Medicaid records when race coding is unreliable? That identification challenge is the methodological core of Session 2. Preview question for students: The IHS serves AI/AN patients at IHS and tribal facilities. If you had access to state Medicaid claims files, how would you identify which claims went to AI/AN patients? What identifier would you use? What would be missing?

SESSION 2  ·  90 minutes — Week 4

Administrative Data and Its Architecture: MAX, NWTEC, and the Structural Boundary

The 2012 MAX analysis · the AIR proxy method · what MAX revealed · NWTEC as a linkage technique preview · the IHS enrollment boundary · why a federal data match became necessary

Session 2 Timing Overview

Time	Segment	Format	Notes
0–10	Bridge: the administrative data question	Mini-lecture	Connect to Session 1
10–32	Section 2.1: The 2012 MAX analysis	Lecture + tables	Core advance over survey era
32–52	Section 2.2: What MAX revealed — and what it missed	Lecture	AIR proxy and its structural limits
52–65	Section 2.3: NWTEC as a regional linkage model	Lecture (reduced emphasis)	Technique preview; public health orientation
65–78	Section 2.4: The IHS enrollment boundary	Discussion	Structural ceiling across all eras
78–90	Timeline exercise and bridge to Weeks 5–7	Small groups	Methodological arc construction

2.0 — Bridge Opening (0–10 min)

READ ALOUD / ADAPT

“Last week we established what the ACS could and could not do. The core limitation for financing analysis was this: survey data can tell you roughly how many people report having Medicaid coverage. It cannot tell you what Medicaid actually paid, or which of those people used IHS facilities, or how spending is distributed across states. The obvious next step was to go directly to administrative data. If you want to know what Medicaid paid for AI/AN health care, go to the Medicaid records. That sounds straightforward. But it introduces an immediate methodological problem: Medicaid enrollment files do not contain a reliable AI/AN race identifier. To find AI/AN Medicaid enrollees in the claims data, you need a proxy. And the quality of that proxy determines the quality of everything that follows.”

2.1 — The 2012 MAX Analysis: The First Administrative Advance (10–32 min)

The Medicaid Analytic eXtract (MAX) was the CMS data system that preceded T-MSIS. MAX files contained state-level Medicaid enrollment and claims records. The 2012 MAX-based analysis was the first serious attempt to quantify AI/AN Medicaid enrollment and expenditures from administrative data rather than survey estimates. It was a major step forward and a revealing object lesson in the limits of proxy-based identification.

MAX System Element	What It Provided
Data type	Administrative: actual Medicaid enrollment records and paid claims, not self-reports
Coverage	All 50 states + DC; federally maintained
AI/AN identification method	The AIR proxy: individuals with at least one encounter at an IHS-affiliated facility during the study year (see below)
Expenditure data	Actual paid claims — the first time IHS-access Medicaid spending could be measured directly
Geographic detail	State-level totals; tribal-level data not produced due to small-cell suppression
Key limitation	AIR proxy missed AI/AN Medicaid enrollees with no IHS encounter in the study year

Core Finding

What the 2012 MAX analysis found: Approximately 1.1 million AI/AN individuals were identified in Medicaid records using the AIR proxy. Associated Medicaid expenditures totaled approximately $2.8 billion in 2012 in paid claims. This was the first empirically grounded national estimate of IHS-access Medicaid spending derived from administrative records — substantially larger than most prior survey-based estimates had suggested.

The AIR Proxy: The Methodological Core

Methods Note

How the AIR proxy worked: An individual was classified as AI/AN for the 2012 MAX analysis if they had at least one paid Medicaid claim at an IHS-operated or tribally operated facility (I/T/U facility) during the study year. This approach used IHS facility encounter as a proxy for patient race identification — logical because virtually all IHS users are AI/AN — but it structurally excluded any AI/AN Medicaid enrollee who had no IHS encounter in the measurement year.

AIR Proxy Strength	AIR Proxy Limitation
Avoids unreliable self-reported race coding in Medicaid files	Misses AI/AN Medicaid enrollees with no IHS encounter in study year (Population B)
Directly links to IHS system use — the policy-relevant population	Misses IHS-registered individuals who used only external providers in the study year
Administrative verification rather than self-report	Urban AI/AN populations with no IHS facility access are systematically excluded
Consistent across states (unlike race coding quality)	Annual snap: enrolled all year but with no IHS encounter appears unidentified
Enables expenditure analysis tied to IHS system use	Cannot distinguish AI/AN from non-AI/AN at some Tribal facilities serving mixed populations

Methodological Caution

The AIR proxy vs. the CMS–IHS Data Match: The 2012 MAX analysis and the 2023 Data Match both identify IHS-access Medicaid enrollees, but with different methods and different populations. The AIR proxy uses facility encounters as the identifier; the Data Match uses direct person-level linkage between IHS patient registration files and CMS enrollment records. The Data Match is more precise because it identifies all IHS-registered individuals regardless of whether they had an IHS encounter in the study year. The 2012 figure (~1.1M) and the 2023 figure (941,168) are not directly comparable — different methods, different years, different population definitions.

2.2 — What MAX Revealed — and What It Missed (32–52 min)

The 2012 MAX analysis produced three analytically significant findings that shaped the subsequent decade of research:

1. The Medicaid-IHS revenue relationship is substantially larger than survey estimates suggested. ACS-based analyses had estimated approximately $1–2 billion in Medicaid flows to Indian health. The 2012 MAX analysis found $2.8 billion in actual paid claims — substantially higher, and derived from records rather than extrapolated from self-reported enrollment rates.
2. Geographic concentration is extreme. Five states — Arizona, Oklahoma, Alaska, New Mexico, and Minnesota — accounted for more than 50% of total AI/AN Medicaid expenditures in 2012. This pattern reflects the geographic distribution of IHS facilities and reservation-based AI/AN populations, and would remain stable through the 2023 Data Match.
3. The IHS/Tribal facility billing share was rising. In 2012, approximately 21% of total AI/AN Medicaid expenditures flowed through IHS and tribally operated facilities — up from approximately 13% in 1999. This structural shift in how Medicaid dollars moved through the Indian health system (or non-system) was the precursor to the 2023 Data Match findings.

What MAX Could Show	What MAX Could Not Show
Actual paid claims for IHS-access enrollees	AI/AN Medicaid enrollees with no IHS encounter (Population B)
State-level expenditure totals	Tribal-level data (suppressed)
IHS vs. non-IHS billing share	Urban AI/AN Medicaid populations (largely excluded by AIR proxy)
Year-over-year trend from 1999 historical files	Verified IHS-registration status (uses facility encounter as proxy)
Service type breakdown	2013–2022 data (T-MSIS transition; gap in validated series)

2.3 — NWTEC as a Regional Data Linkage Model (52–65 min)

Instructor Note

NWTEC is covered here with reduced emphasis relative to the standalone NWTEC/CRIHB module. For Weeks 3–4 purposes, NWTEC serves a specific pedagogical function: it is the clearest pre-Data Match example of what direct administrative data linkage — not a proxy method, not a survey — looks like in practice. Instructors who use the standalone NWTEC module should treat this section as a brief preview. Instructors not using that module should use this as the primary treatment. Do not expand this section at the expense of Section 2.4 — the IHS enrollment boundary is the conceptual priority.

While national researchers were building toward the 2012 MAX analysis, a parallel methodological tradition was developing at the regional level. The Northwest Tribal Epidemiology Center (NWTEC), housed within the Northwest Portland Area Indian Health Board (NPAIHB), had been building a direct administrative data linkage infrastructure since the late 1990s.

NWTEC’s Northwest Tribal Registry connects IHS patient registration records directly to state Medicaid enrollment and claims data for Oregon, Washington, and Idaho using probabilistic record matching — doing at the regional scale what the CMS–IHS Data Match would eventually do nationally. This is methodologically significant because it demonstrates that person-level data linkage was feasible before the federal infrastructure existed, and because it reveals the structural limitations of linkage-based approaches that would also appear in the national Data Match.

Instructor Note

Governance note: This Washington State data match also illustrates a governance problem that became central to the design of the 2023 national Data Match. After initial presentations, consensus could not be reached among Washington Tribes on whether to repeat the effort — some Tribes did not want to participate, and no follow-up data matches were conducted. The data existed and the analytical capacity was present, but durable Tribal consent was not. This is the institutional problem that NIHB spent the following decade solving through the Tribal Advisory Committee structure, data suppression rules, and negotiated governance protocols that made the national Data Match possible. When students encounter the governance architecture of the 2023 Data Match in Weeks 5-7, this Washington episode is the counterfactual: what happens when Tribal data sovereignty is not built into the design from the outset.

A parallel state-level effort emerged in Washington around the same period. Working directly with Washington State Medicaid, Ed Fox of the Skokomish Tribe developed a data match between state Medicaid files and all 29 Washington Tribal health clinics — both IHS-operated and Tribally operated — to produce actual paid claims expenditures by Tribe and by encounter type. The 2013 reference-year analysis covered 25 Tribes and documented $49.5 million in total Medicaid payments for 21,332 patients ($2,322 per person). After removing two outlier Tribes with high non-Indian patient shares, the 23-Tribe analytic sample showed $37.7 million for 18,379 patients ($2,050 per person). A dramatic 38% increase followed in 2014 — rising to $52 million — driven by ACA Medicaid expansion and restoration of adult dental coverage. The encounter-type breakdown was analytically notable: medical services accounted for 41% of payments, mental health 34%, chemical dependency 13%, and dental 12%, with behavioral health combined (47%) exceeding the medical share at the aggregate level for several Tribes. Medicaid as a share of active IHS users ranged from 15% to 62% across 19 reporting Tribes, with an average of 32% — a wide variation that directly anticipates the fiscal risk typology students will analyze in Weeks 11-12. (Fox, E. 2016. Affordable Care Act, Medicaid, and Washington Tribes. Presentation, Skokomish Tribe Health Clinic, February 4.)

Instructor Note

NWTEC’s public health orientation: NWTEC is primarily a public health institution whose research portfolio covers health status, mortality, cancer incidence, and environmental health. Its Medicaid data work is a component of that broader mission — not its central focus. The analytical framework for financing — expenditures, PMPY modeling, PRC substitution analysis — came from the national administrative data work, not from NWTEC. Instructors should be clear that NWTEC’s value for this course is methodological (it demonstrates direct linkage) and structural (it documents the IHS enrollment boundary), not as a financing analysis model.

Key Concept

The NWTEC structural parallel: NWTEC’s Northwest Tribal Registry captures approximately 75–80% of the AI/AN population in Oregon, Washington, and Idaho. The remaining 20–25% are not captured because they do not use IHS or tribally operated facilities — primarily urban AI/AN populations and those with no IHS registration. The CMS–IHS Data Match’s 941,168 IHS-access enrollees represent approximately 78% of total AI/AN Medicaid spending nationally. A near-identical coverage rate — reflecting the same underlying structural limit: IHS-registration-based data systems cannot capture AI/AN individuals who are not registered in the IHS system. NWTEC documented this ceiling at the regional scale before national data confirmed it.

2.4 — The IHS Enrollment Boundary (65–78 min)

KEY CONCEPT — THE DENOMINATOR PROBLEM

A population estimate is meaningful only in relation to the population definition used to construct it. Census/ACS estimates, IHS registration records, and CMS–IHS linked administrative data answer different questions. The course therefore treats population definition as a methodological issue, not merely a data-quality issue.

READ ALOUD / ADAPT

“By 2015, researchers working on AI/AN Medicaid data had converged on a fundamental insight that the ACS era had obscured and the MAX and NWTEC work had revealed: the central analytical challenge in AI/AN Medicaid research is not data quality. It is population definition. The IHS serves a specific, bounded population: individuals who are registered in the IHS system. That population is large — nearly a million people are Medicaid-enrolled within it. But it is not the full AI/AN population. Depending on the state, between 20% and 70% of AI/AN Medicaid enrollees have no IHS registration. They receive care through mainstream providers, urban Indian health organizations, or FQHCs that are not IHS-affiliated. Every data system we have discussed hits this same ceiling. ACS estimates are higher than administrative counts partly because they capture this non-IHS population. The 2012 MAX AIR proxy captured roughly 1.1 million by using facility encounters — and still missed individuals with no IHS encounter. NWTEC’s Registry captures 75–80%. The 2023 Data Match captures approximately 78% of total AI/AN Medicaid spending. This is not a solvable data quality problem. It is a structural feature of any IHS-registration-based data system. When you encounter this boundary in any dataset for the rest of this course, you will know exactly what it means and where it comes from.”

Population	Who They Are	Captured By
Population A	IHS-registered AI/AN individuals who are Medicaid-enrolled and had at least one paid IHS service	CMS–IHS Data Match (2023); 2012 MAX (AIR proxy, partial); NWTEC Registry (Northwest only)
Population B	AI/AN Medicaid enrollees with no IHS registration — primarily urban, mainstream-provider users	ACS (partial, self-reported); historical all-AIAN payment files; not captured by IHS-registration systems
Uninsured IHS users	IHS-registered individuals not Medicaid-enrolled (Medicare-only, uninsured, privately insured)	IHS patient registration data; not in Medicaid datasets
Non-IHS AI/AN	AI/AN individuals with no IHS registration and no IHS access	ACS (partial); no administrative linkage currently possible

Instructor Note

Bridge to Weeks 5–7: Close by previewing the next unit. Students now understand what problem the Data Match was built to solve. They have seen what each predecessor approach got right and wrong. Weeks 5–7 will ask: how was the Data Match actually built? Who was involved in the governance decisions, and why did the governance matter as much as the technical work? Preview question: the Data Match required linking two separate federal databases — IHS patient registration files and CMS Medicaid enrollment records. What legal, technical, and governance obstacles would you predict arose in that process?

KEY CONCEPT — THE DATA MATCH ANSWERS ONE QUESTION VERY WELL — NOT EVERY QUESTION

The CMS–IHS Data Match provides a verified national measure of IHS-access Medicaid enrollment and an associated expenditure estimate. It does not measure the entire American Indian and Alaska Native Medicaid population, total Indian health financing, or all Indian health needs. Students should ask of every statistic: Who is included? Who is excluded? What is observed? What is estimated?

Cross-Era Comparison: ACS, MAX, NWTEC, and the CMS–IHS Data Match

Instructor Note

Distribute this table to students at the end of Week 4. It is a reference document for the rest of the course. Every analytical claim about AI/AN Medicaid data in Weeks 5–14 should be checkable against this table.

Dimension	ACS (2005–present)	2012 MAX Analysis	NWTEC / NW Registry	CMS–IHS Data Match (2023)
Data type	Survey (self-reported)	Administrative claims (paid)	Administrative linkage (registry + state claims)	Administrative linkage (IHS registrant file + CMS enrollment records)
AI/AN identification	Self-reported race; no verification	AIR proxy: IHS facility encounter in study year	Direct IHS RPMS linkage + probabilistic record match	Direct person-level CMS enrollment + IHS registrant file linkage
Geographic scope	National	National (50 states)	Regional (OR, WA, ID)	National (50 states)
Population covered	~1.1–1.3M estimated (includes Pops. A+B)	~1.1M (AIR proxy; partial Pop. B exclusion)	75–80% of NW AI/AN (Pop. A focus)	941,168 IHS-access (Pop. A only; ~78% of total AI/AN Medicaid spending)
Expenditure data	None	$2.8B in paid claims (2012)	State-level claims for NW tribes	$6.4–6.6B (PMPY modeled)
Tribal-level detail	Unreliable (small N)	State-level only (suppressed)	Yes, for many NW tribal programs	State-level only (suppressed)
Primary research orientation	Coverage rates; uninsured burden; trends	Financing: national expenditure baseline	Public health: mortality, cancer, utilization; not financing-primary	Financing: national IHS-Medicaid fiscal flows; PRC substitution; budget advocacy
Best for asking…	What % of AI/AN have Medicaid? ACA impact?	How much did Medicaid spend on IHS-access care in 2012?	Which NW tribal programs are Medicaid-dependent and to what degree?	How much does Medicaid spend on IHS-access care nationally? What is fiscal risk of Medicaid contraction?
Primary limitation	Cannot link to payment records; no IHS facility identification	AIR proxy misses non-IHS-encounter AI/AN enrollees; 2013–2022 data gap	Regional only; public health orientation; cannot generalize to non-NW tribes	Population A only; no tribal-level data; no 2013–2022 trend data; PMPY modeled (not summed from claims)

Track-Differentiated Discussion Protocol

Full-Group Discussion (All Tracks, 15 minutes)

Discussion Questions

Q1. The ACS estimated approximately 1.1–1.3 million AI/AN Medicaid enrollees nationally. The 2023 Data Match found 941,168 IHS-access enrollees. A critic argues: ‘The Data Match shows AI/AN Medicaid enrollment is lower than we thought.’ How do you respond? What is this critic misunderstanding about population definitions? Q2. The 2012 MAX analysis used the AIR proxy to identify AI/AN patients — facility encounter rather than race coding or registration. Under what conditions does this proxy produce the most accurate results? The least accurate? Where would it perform worst geographically? Q3. NWTEC built a regional data linkage infrastructure over 25 years that preceded the federal Data Match by more than a decade — but NWTEC’s primary orientation is public health, not financing. Why does that institutional orientation matter for how we use NWTEC’s data in financing arguments?

Track-Differentiated Breakout (20 minutes)

Track	Policy & History	Research & Methods
Task	Construct a one-page annotated timeline of AI/AN Medicaid data development from 2000 to 2023. For each major development, annotate: what policy argument became possible that had not been possible before?	You are designing a study of AI/AN Medicaid trends from 2005 to 2023. Write a methods paragraph explaining how you would handle the transition between the ACS, the 2012 MAX analysis, and the 2023 Data Match — specifically addressing the Population A/B scope difference and the 2013–2022 data gap.
Deliverable	Annotated timeline; share with full group	Methods paragraph (8–12 sentences); share with full group
Discussion prompt	Which data development produced the single most policy-important advance, and why?	Which source would you use for a longitudinal analysis of IHS-access Medicaid spending, and what adjustment is required to maintain comparability across eras?

Instructor Note

Operations & Administration track: Assign the Cross-Era Comparison table as the primary reference. Ask students to identify the one data source they would rely on as a tribal health program administrator making the case to state legislators for protecting AI/AN Medicaid reimbursement rates. They must explain their choice using specific table entries and identify what they cannot show from that source alone.

Key Data Points for Student Reference

Fact	Value / Note
ACS annual sample size	~3.5 million addresses
ACS AI/AN Medicaid enrollment estimate (pre-Data Match)	~1.1–1.3 million (varies by year; includes Populations A+B)
2012 MAX: AI/AN Medicaid enrollees identified	~1.1 million (AIR proxy)
2012 MAX: Medicaid expenditures	~$2.8 billion in paid claims
WA State data match: Medicaid payments, 25 Tribes (2013; Fox, 2016)	$49.5M total (23-Tribe sample: $37.7M; 18,379 patients; $2,050 PMPY); 2014: $52M (+38%); Medicaid avg 32% of active IHS users (range 15-62%); encounter mix: Medical 41%, MH 34%, CD 13%, Dental 12%
IHS/Tribal billing share 1999	~13% of total AI/AN Medicaid payments
IHS/Tribal billing share 2008	~20% of total AI/AN Medicaid payments
NWTEC Northwest Tribal Registry coverage	~75–80% of Northwest AI/AN population; primary public health orientation
2023 Data Match: IHS-access Medicaid enrollees	941,168
2023 Data Match: IHS-access Medicaid expenditures	~$6.4–6.6 billion (PMPY modeled)
2023 Data Match as share of total AI/AN Medicaid	~78%; Population A only
IHS enrollment boundary (structural gap)	~20–25% of AI/AN population not IHS-registered; not captured in any IHS-based data system

Appendix: Glossary of Key Terms

Term	Definition
American Community Survey (ACS)	Annual U.S. household survey; primary source of AI/AN health coverage estimates during the survey era; not designed to measure Medicaid enrollment verification or payment flows.
Medicaid Analytic eXtract (MAX)	CMS administrative data system (replaced by T-MSIS) containing state Medicaid enrollment and claims records; used for the 2012 AI/AN Medicaid expenditure analysis.
AIR proxy	American Indian Roster proxy: approach used in 2012 MAX analysis to identify AI/AN Medicaid enrollees by linking to records of patients with at least one encounter at an IHS or tribally operated facility.
T-MSIS	Transformed Medicaid Statistical Information System: replaced MAX as the CMS national Medicaid data system; source of Medicaid enrollment data used in the CMS–IHS Data Match.
IHS enrollment boundary	The structural ceiling of any IHS-registration-based data system: approximately 20–25% of the AI/AN population is not IHS-registered and is systematically excluded from IHS-linked datasets.
Population A	AI/AN individuals with verified IHS access who are Medicaid-enrolled and generated at least one paid IHS service; the population captured by the CMS–IHS Data Match.
Population B	AI/AN Medicaid enrollees without IHS registration, primarily urban and mainstream-provider users; captured partially by ACS and historical all-AIAN payment files but not by IHS-registration-based administrative systems.
NWTEC	Northwest Tribal Epidemiology Center: housed within NPAIHB; built the Northwest Tribal Registry linking IHS RPMS records to state Medicaid files for OR, WA, and ID; primary public health (not financing) orientation.
Northwest Tribal Registry	NWTEC’s master roster of AI/AN individuals in the Pacific Northwest, constructed from IHS RPMS records; captures ~75–80% of Northwest AI/AN population.
Probabilistic matching	Statistical method for linking records across datasets without a common unique identifier; uses name, date of birth, and other fields; core technique in both NWTEC and the CMS–IHS Data Match.
PMPY	Per-member-per-year: standard Medicaid expenditure modeling framework used in the 2023 Data Match to estimate total expenditures from verified enrollment counts.
All-AIAN payment files	Historical CMS Medicaid payment files (1999–2012) covering the full AI/AN Medicaid population (Populations A+B); the only national series capturing Population B spending.

Indian Health Financing  ·  Research & Methods Track  ·  Weeks 3–4 Lecture Notes  ·  March 2026

  Weeks 5–7 

INDIAN HEALTH FINANCING

Research & Methods Track

Weeks 5–7 Instructor Lecture Notes & Script

REQUIRED READING — Paper II

A History of CMS–IHS Tribal Data Improvement: The Role of the CMS Tribal Technical Advisory Group

Discussion: Why did the Data Match take so long? The CMS–IHS Data Match was the product of a long institutional and technical history. Students should distinguish three dimensions: technical data linkage, federal administrative structures, and Tribal governance/data sovereignty. CMS–TTAG should be treated as a distinct data-development workstream beginning in the early 2000s, even though some individuals participated in both CMS–TTAG and broader IHCIA-related work.

Building the Match: History, Governance, and Tribal Data Sovereignty

Element	Detail
Track	Research & Methods (appropriate for all tracks)
Sessions	Three 90-minute sessions spanning Weeks 5, 6, and 7
Core Question	How was the CMS–IHS Data Match built, and why did the governance structures matter as much as the technical work?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. III–VII); Lecture Series Part 2; Professor Introductions Weeks 5–7
Builds On	Weeks 3–4 ACS and MAX limitations framework
Leads Into	Weeks 8–10 findings, PMPY modeling, Population A vs. B analysis
Midterm	Timeline + sovereignty essay launched in Session 3 (Week 7), due end of week

These notes cover three sessions in full. Each session has its own character: Session 1 (Week 5) is the institutional and political story of why the project was built. Session 2 (Week 6) is the technical architecture — how matching works, what the VRDC is, what data use agreements do. Session 3 (Week 7) is the sovereignty and governance deep-dive, and serves as the launch pad for the midterm assignment.

Shaded script boxes are written for read-aloud or free adaptation. Colored boxes mark governance (purple), technical (dark blue), sovereignty (brown), instructor notes (amber), and warnings (orange).

Three-Session Architecture

Session	Week	Core Argument	Key Activity
Session 1	Week 5	The Data Match was not primarily a technical project — it was a political and institutional achievement built on years of relationship-building between NIHB, CMS, and Tribal communities	Actor mapping exercise (governance network diagram)
Session 2	Week 6	Deterministic and probabilistic matching are methodologically distinct choices with real consequences; the VRDC and DUAs are not bureaucratic formalities but structural protections that shaped what the project could and could not do	Matching strategy decision exercise
Session 3	Week 7	Tribal data sovereignty is not a constraint on good research — it is a design principle that produced better research; the decisions made about small-cell suppression and state-level-only reporting reflect a specific theory of who the research serves	Midterm launch: timeline construction and sovereignty essay scaffolding

WEEKS 5-7 IN THE COURSE ARC

These three weeks are the methodological and institutional heart of the course. Students arrive knowing what was wrong with the ACS (Weeks 3–4). They leave knowing how the field built something better, why it took the shape it did, and what principles governed its design. The midterm at the end of Week 7 tests both the factual knowledge (timeline) and the analytical understanding (sovereignty essay). Do not rush Session 3 to make room for more technical content — the sovereignty discussion is what distinguishes this course from a general Medicaid data course.

SESSION 1  ·  90 minutes — Week 5 Why It Got Built: The Institutional and Political Story Phase 1 (2017-2019) · the diagnosis · actors and interests · why NIHB led, not CMS or IHS

Session 1 Timing Overview

0-10 min

Opening Hook

From diagnosis to design Script: the hospital that still can’t answer the question

10-35 min

Section 1.1

The five phases: a political history of the project Lecture with timeline table; emphasis on institutional dynamics, not just chronology

35-55 min

Section 1.2

Why NIHB, not CMS or IHS? Core analytical question; NIHB’s structural advantages and what they produced

55-75 min

Section 1.3

Actor mapping exercise Small groups construct governance network diagram; 15 min work, 5 min report-out

75-90 min

Bridge

What governance means for research design Transition framing for Session 2 technical content

1.0 — Opening Hook (0-10 min)

READ ALOUD  /  ADAPT

“Two weeks ago, we established what the ACS could and could not do. We saw that survey data gave us estimates but not verification — approximate enrollment counts but no payment records, access questions but not registration confirmation. And we saw the MAX data gave us something more precise but had its own blind spots: it missed urban Indian patients, people without IHS encounters in a given year, and states with poor billing data. By 2016, NIHB and the CMS–IHS Tribal Advisory Committee had internalized that diagnosis. They knew what they needed: a direct linkage between the IHS registrant database and the CMS Medicaid enrollment files. Person by person. Verified. With payment data. That sounds like a data engineering project. And in one sense it is. But what we are going to spend the next three weeks understanding is that it was primarily a governance and institutional project — and that the governance work was harder, slower, and more important than the technical work. Think about what it actually takes to link two federal databases. CMS has Medicaid enrollment records. IHS has registrant files. They are separate agencies under the same Department. They serve the same people. And yet linking their databases requires years of legal negotiation, formal data use agreements, secure computing environments, and continuous oversight from Tribal communities who have every historical reason to be suspicious of federal agencies combining data about them without consent. Today we begin the story of how that got done — and why it matters as much for what it tells us about Tribal-federal relationships as for what it tells us about Medicaid enrollment numbers.”

INSTRUCTOR NOTE

This is a good moment to surface the historical context of data and Indigenous communities without derailing into a full history lesson. The point is compact but important: federal data collection has historically been used against Indigenous communities — in enrollment disputes, in termination-era policies, in blood quantum determinations. That history is why governance is not an afterthought. Students with Indigenous studies backgrounds will recognize this immediately. Students coming from health policy or public health may need it surfaced explicitly. One sentence is enough: ‘Tribes have specific historical reasons to be cautious about federal agencies collecting and linking data about their members — and those reasons shaped every governance decision in this project.’

1.1 — The Five Phases: A Political History (10-35 min)

This is lecture content with a timeline table as the visual anchor. Write the phases on the board as you proceed through them, or display the table as a slide. The goal is not chronological memorization — it is understanding what was at stake at each phase and who drove each transition.

The Phases as a Decision Sequence

Present the following framing before showing the timeline:

READ ALOUD  /  ADAPT

“The five phases of this project are not just a chronology of what happened when. They are a sequence of decisions — decisions about what the research question was, who should lead the work, how data should be protected, and what counts as an acceptable result. Each phase transition represents a moment where someone made a choice that shaped everything that followed. As we go through the timeline, I want you to notice: at each phase, who is doing the deciding?”

Phase	Period	What Happened	Key Decision Made	Who Drove It
1: Recognizing the Need	2017-2019	NIHB’s Tribal Data Project used ACS data and hit its analytical ceiling. The 2012 MAX analysis showed what administrative data could reveal but also its blind spots. Research teams and Tribal advisory partners concluded survey data was insufficient for the policy questions they needed to answer.	Shift from survey-based to administrative-data strategy	NIHB Tribal Data Project team; CMS–IHS Tribal Technical Advisory Group (TTAG)
2: Designing the Match	2020-2021	NIHB began formally planning linkage of IHS registrant files with CMS enrollment data. Matching logic was designed (deterministic primary, probabilistic secondary). By October 2021, NIHB formally presented the project at the National Tribal Health Conference as a concrete architecture.	Deterministic vs. probabilistic matching approach; IHS registrant + Medicaid enrollment as the two source files	NIHB research staff; CMS data partners; technical consultants
3: Legal and Infrastructure Negotiation	2019-2022	Data Use Agreements (DUAs) between NIHB, CMS, and IHS negotiated and executed. NIHB team gained access to the Virtual Research Data Center (VRDC). Federal privacy law requirements (HIPAA, Privacy Act, IHS data authorities) navigated.	VRDC as the secure analysis environment; DUA terms governing data use, storage, and reporting restrictions	Federal legal counsel; NIHB leadership; CMS data governance staff
4: First Linkage Runs and Validation	2021-2023	Initial matching runs executed in VRDC. Data cleaning and validation performed. T-MSIS transition managed as federal data infrastructure shifted. Small-cell suppression decisions made. Tribal Advisory Committee reviewed findings before any publication.	State-level only reporting (suppress tribal-level data); T-MSIS as the Medicaid data source replacing MAX	Tribal Advisory Committee; NIHB analysts; CMS statisticians
5: Publication and the 2023 Results	2023-2025	Validated results released: 941,168 AI/AN individuals with verified IHS access enrolled in Medicaid. PMPY expenditure modeling applied to produce $6.42-6.56B estimate. October 2025 NIHB report published with full state-level tables.	PMPY modeling framework for expenditure estimation; state-level aggregation as the primary output unit	NIHB; October 2025 report team; Tribal Advisory Committee final review

Pause and Discuss: The 2019-2022 Legal Negotiation Phase

This phase deserves special attention because students often underestimate the legal complexity. Surface these specific points:

1. The Privacy Act of 1974 restricts how federal agencies share personally identifiable information between systems. Linking CMS and IHS records required specific legal authority and data agreements.
2. HIPAA’s protections for health information apply differently to federal agencies than to private health systems — but the principles of minimum necessary use and purpose limitation had to be codified in the DUAs.
3. IHS data has its own legal protections under IHS data governance frameworks. IHS cannot simply share registrant files with another agency without specific authorization.
4. Negotiating all three sets of requirements simultaneously, while also satisfying Tribal Advisory Committee conditions, is why this phase alone took two to three years.

INSTRUCTOR NOTE

Practitioner connection: The legal negotiation phase is where most ambitious inter-agency data projects stall or die. In your experience as a health director, you have likely navigated similar multi-party data sharing negotiations at the state or regional level. This is an opportunity to describe what it looks like in practice — the multiple review cycles, the legal back-and-forth, the moments where a single agency’s data counsel can hold everything up. Frame this as normal, not exceptional: ‘This is what good data governance actually looks like. It is slow because it is being done right.’

1.2 — Why NIHB, Not CMS or IHS? (35-55 min)

This is the analytical core of Session 1 for the Research & Methods track. The question of who leads a research project shapes the questions asked, the interpretations offered, and the credibility of findings. Work through this as a structured argument.

READ ALOUD  /  ADAPT

“Here is a question I want you to sit with for a moment. Two federal agencies — CMS and IHS — both had direct access to the data needed for this project. CMS had the Medicaid enrollment files. IHS had the registrant database. Either agency could theoretically have initiated and led this data match in-house. Why didn’t they? Why did NIHB — a nonprofit Tribal advocacy organization — end up as the lead research entity? The answer is not simple. It is not that federal agencies were incompetent or uninterested. It is that NIHB’s role produced something that neither CMS nor IHS acting alone could have produced: research with both technical credibility and Tribal legitimacy. Those two things are both necessary for this work to matter in policy debates. Let me unpack why.”

Dimension	CMS Leading Alone	IHS Leading Alone	NIHB Leading (What Actually Happened)
Research agenda	Driven by CMS program interests — likely focused on cost, fraud, enrollment accuracy	Driven by IHS operational interests — budget justification, utilization patterns	Driven by Tribal health policy priorities — funding advocacy, disparity documentation, trust responsibility
Tribal legitimacy	Low — federal agencies have complicated trust history with Tribal communities; data could be used for adverse purposes	Medium — IHS has a trust relationship but is also a federal agency subject to shifting political priorities	High — NIHB is accountable to Tribal nations; its mission is explicitly Tribal health advocacy
Interpretation authority	Federal framing of findings; Tribes would receive results, not shape them	Same concern — federal framing	Tribal framing throughout; Tribal Advisory Committee reviewed findings before any external release
Political exposure	High — federal agencies are subject to FOIA, political direction changes, congressional scrutiny in real time	Same concern	Lower — nonprofit has more analytical independence from political interference
Data ownership narrative	Federal agencies ‘own’ the findings; Tribes are research subjects	Same concern	NIHB and Tribal communities have a claim to ownership of findings that shaped the policy narrative

GOVERNANCE PRINCIPLE: LEAD ORGANIZATION SHAPES RESEARCH OUTCOMES

In health policy research involving historically marginalized communities, the identity of the lead research organization is not a procedural detail — it is a substantive methodological choice. Who leads determines: (1) what questions get asked, (2) how findings get framed, (3) who sees results first, and (4) how credible the findings are to the communities they describe. The NIHB model established a precedent that Tribal-led administrative data research at federal scale is both possible and preferable to federal-agency-led approaches for these populations.

The Tribal Technical Advisory Group (TTAG)

Students should also understand the role of the CMS Tribal Technical Advisory Group (TTAG) — a standing committee of Tribal representatives that advises CMS on Medicaid policy affecting Tribal communities. TTAG provided the formal CMS-side channel through which NIHB could initiate and sustain the Data Match partnership. Without TTAG, NIHB would have had to negotiate with CMS ad hoc, without an established governance relationship. TTAG’s existence reflects the broader principle of Tribal consultation requirements under federal policy — but also shows how consultation structures can become real partnership structures when led with purpose.

1.3 — Actor Mapping Exercise (55-75 min)

EXERCISE: GOVERNANCE NETWORK DIAGRAM  |  20 minutes

Format: Groups of 3-4 students. Each group draws a network diagram on paper or whiteboard showing all the actors in the CMS–IHS Data Match Project and the relationships between them. Actors to include: NIHB — CMS — IHS — Tribal Advisory Committee — TTAG — Individual Tribal Nations — State Medicaid Agencies — VRDC (as infrastructure) — Congress (as appropriator and oversight body) For each relationship, annotate: What does each party provide to the relationship?What does each party need from the others?Where might conflicts of interest or competing priorities arise? Debrief (5 min): Groups share one relationship they found most complex. Instructor synthesis: The most important relationship to understand is NIHB-to-Tribal Advisory Committee, because it is where Tribal legitimacy and research independence intersect.

INSTRUCTOR DEBRIEF NOTE

Watch for two common patterns in the diagrams: (1) Students often put CMS and IHS at the center with NIHB as a peripheral actor — challenge this directly, since NIHB was the lead. (2) Students often draw ‘individual Tribal nations’ as a single node rather than recognizing the diversity of 574 federally recognized Tribes with different relationships to the I/T/U system (or non-system), different political interests, and different levels of engagement with the advisory process. Surface this complexity: the Tribal Advisory Committee did not represent a monolithic Tribal perspective — it navigated genuine disagreements about how data should be protected and used.

1.4 — Bridge to Session 2 (75-90 min)

READ ALOUD  /  ADAPT

“We now have the political and institutional map. We know who the actors were, what they needed from each other, and why the project took the shape it did. Next week we go inside the machine. What does it actually mean to match two federal databases? What is deterministic matching versus probabilistic matching — and why does it matter which you use? What is the VRDC, and why does doing analysis inside a secure federal computing environment change what research looks like? I want you to come to Session 2 having thought about this question: If you are the NIHB analyst who has just signed a Data Use Agreement and received VRDC access, what is the first thing you actually do? What does the data look like? What problems do you expect to find? Think like a researcher about to start working with two imperfect federal databases that have never spoken to each other before.”

SESSION 2  ·  90 minutes — Week 6 Inside the Machine: Technical Architecture and Data Infrastructure Deterministic vs. probabilistic matching · VRDC · DUAs · data cleaning · T-MSIS transition

Session 2 Timing Overview

0-10 min

Bridge

What does the analyst actually see? Brief discussion; surface technical intuitions

10-35 min

Section 2.1

How matching works: deterministic vs. probabilistic Lecture with matching logic table; worked example

35-55 min

Section 2.2

VRDC and DUAs: what secure analysis environments mean for research Lecture with implications table; a non-trivial research constraint

55-75 min

Section 2.3

Data quality problems: cleaning, T-MSIS transition, what goes wrong Active discussion; instructor draws on practitioner experience

75-90 min

Section 2.4

Matching strategy decision exercise Small-group decision scenario; 10 min work, 5 min debrief

2.0 — Bridge Opening (0-10 min)

READ ALOUD  /  ADAPT

“At the end of last week I asked you to think like an NIHB analyst who has just gotten VRDC access for the first time. You have two databases in front of you — conceptually. The IHS registrant file: millions of records, including name, date of birth, tribal affiliation, registration status, and a registrant ID. The CMS Medicaid enrollment file: tens of millions of records, with name, date of birth, Social Security Number or Medicaid ID, state of enrollment, and coverage dates. Your job is to find the people who appear in both. Simple idea. Hard execution. What problems do you expect to find? [Take 2-3 responses from students.] Good. Let me tell you what the actual problems were — and how the team decided to handle them. Those decisions are where methods and policy intersect most directly.”

2.1 — How Matching Works: Deterministic vs. Probabilistic (10-35 min)

The Core Concept: What Is Record Linkage?

Record linkage is the process of identifying records in two (or more) databases that refer to the same real-world entity — in this case, the same individual person. When the databases were built independently, without a common unique identifier, linkage requires using shared attributes to make the connection.

TECHNICAL CONCEPT: THE TWO MATCHING STRATEGIES

Deterministic Matching Requires exact agreement across specified identifiers. If all specified fields match exactly — same SSN, same date of birth, same name spelling — it is declared a match. If any field disagrees, it is not a match. Strengths: High precision (very low false positive rate). Every match found is almost certainly a true match. Weaknesses: Low recall (misses real matches due to typos, name changes, transposed digits, inconsistent SSN recording). In messy real-world data, exact matching misses a substantial share of true matches. Probabilistic Matching Uses a scoring algorithm to assess the probability that two records refer to the same person. Each matching field contributes a weighted score. Records above a threshold are accepted as matches; records in a middle range are reviewed manually or flagged as uncertain. Strengths: Higher recall — captures real matches that deterministic would miss (e.g., ‘Mary’ vs. ‘Maria’, or a transposed digit in a SSN). Weaknesses: Requires threshold decisions (where to set the acceptance cutoff); lower precision than deterministic; introduces some false positives that must be validated or accepted as noise.

What the CMS–IHS Data Match Actually Used

The project used deterministic matching as the primary approach. This means the verified 941,168 enrollment count is a conservative figure — it represents people who could be matched with high confidence, not the full universe of people who might belong to both databases.

CRITICAL METHODOLOGICAL IMPLICATION — ADDRESS DIRECTLY

Because the project used deterministic matching as the primary method, the 941,168 figure is likely an undercount of the true IHS-access Medicaid population. Real individuals who appear in both databases but whose records contain typos, name variations (including use of traditional names in one system and English names in another), or inconsistent SSN recording would not be captured as confirmed matches. Students must understand this when interpreting the findings in Weeks 8–10: the administrative figure is verified but conservative, while the ACS estimate was higher but unverified. Neither number is simply ‘the answer’ — they triangulate toward a true range.

Worked Example: Why Names Are Hard

Walk through this concrete scenario to make the matching challenge tangible:

Field	IHS Registrant File Entry	CMS Medicaid File Entry	Match?
Last Name	RUNNINGWATER	RUNNING WATER	Deterministic: NO — space difference Probabilistic: Likely YES
First Name	SARAH	SARA	Deterministic: NO Probabilistic: Likely YES
Date of Birth	1978-04-12	04/12/1978	After normalization: YES
SSN	XXX-XX-4521	XXX-XX-4512	Deterministic: NO — transposed digits Probabilistic: Moderate probability
Overall Verdict	Likely the same person	Cannot confirm deterministically	Missed by deterministic; recovered by probabilistic

READ ALOUD  /  ADAPT

“This is not a hypothetical edge case. Name formatting differences between federal systems are extremely common in AI/AN data because: tribal naming conventions do not always translate consistently to Anglo name formats; hyphenated and compound names get split or merged differently across systems; and data entry in rural IHS facilities decades ago was not always consistent. Any analysis that uses deterministic matching on AI/AN federal records is accepting that it will miss some real individuals — and that the miss rate is not randomly distributed. It is more likely to miss individuals from communities with traditional naming practices or with older, less-complete registrant records. That is a form of systematic bias that the research team had to acknowledge.”

2.2 — VRDC and Data Use Agreements: What Secure Analysis Means (35-55 min)

The Virtual Research Data Center (VRDC)

The VRDC is a CMS-maintained secure computing environment that allows approved external researchers to analyze sensitive federal health data without the data ever leaving CMS infrastructure. It is not a dataset — it is an environment. Researchers log in remotely, run analyses inside the environment, and export only aggregate results (never individual-level records) after CMS review.

WHAT THE VRDC MEANS FOR RESEARCH PRACTICE

The VRDC shapes research practice in ways students should understand concretely: No local data copies: Analysts cannot download the matched dataset to their own computers. All analysis is done inside the VRDC environment, which means every coding and analytical decision is made under monitoring.Output review: Results exported from the VRDC are reviewed by CMS before release. This adds time to every analytical cycle but protects against accidental disclosure of small-cell data.Approved project scope: The DUA defines exactly what research questions can be analyzed. A researcher cannot simply pivot to a new question without amending the agreement — a process that can take months.Access limitations: Only named individuals on the approved project team can access the VRDC. Team member changes require formal amendments. This affects the project’s ability to bring on new analysts or collaborators quickly. Research implication: The VRDC is not a constraint on good research — it is what makes this research trustworthy enough to be policy-relevant. Any result that came from a dataset that could be copied and circulated would face legitimate questions about data security and individual privacy. The VRDC is what allowed NIHB to say: this analysis was conducted under CMS-monitored conditions with no individual-level data ever leaving the secure environment.

Data Use Agreements: What They Actually Govern

A Data Use Agreement is a legal contract — not a formality. In the context of the CMS–IHS Data Match, the DUAs specified:

• Permitted purposes: The data can only be used for the specific research questions defined in the agreement — health coverage analysis for AI/AN IHS-access populations. Not for cost control, not for enforcement, not for other federal programs.
• Prohibited uses: Data cannot be used in ways that could harm Tribal communities, including any use for enrollment audits, eligibility challenges, or identification of individuals for any administrative purpose.
• Output restrictions: Only aggregate, state-level results can be published. No tribal-level breakdowns, no individual-level data, no data for groups below the small-cell suppression threshold.
• Retention and destruction: Matched data files have defined retention periods and must be destroyed on schedule.
• Publication review: Tribal Advisory Committee has pre-publication review rights — findings must be reviewed and approved before external release.

WHY DUA NEGOTIATION TOOK YEARS — NOT A COMPLAINT, AN EXPLANATION

Students sometimes read the multi-year timeline as evidence of bureaucratic dysfunction. It is not. Each DUA clause required review by: CMS Office of General Counsel; IHS legal counsel; NIHB legal counsel; Tribal Advisory Committee members (who represent sovereign governments with their own legal interests). Reaching agreement across those five parties on every provision — including prohibited uses, small-cell suppression thresholds, tribal review rights, and data destruction schedules — took time because it was being done with care. A rushed DUA that later generated a tribal data misuse controversy would have damaged trust in Tribal-federal data partnerships for a generation.

2.3 — Data Quality Problems: What Actually Goes Wrong (55-75 min)

The Reality of Working with Two Imperfect Federal Databases

This section is primarily discussion-based. Use the following points to frame an open exchange, drawing on your own experience with administrative data where relevant.

Problem Category	Specific Issue	How It Affected the Match	What the Team Did
Name formatting	Inconsistent capitalization, spacing, hyphenation; traditional vs. legal names across systems	Reduced deterministic match rate; required normalization before matching	Pre-match data cleaning; probabilistic supplement for near-matches
SSN completeness	IHS registrant files for older patients sometimes had incomplete or missing SSNs; SSN quality varied by IHS Area	Some older/rural patients could not be matched on SSN; date of birth + name used as secondary identifiers	Tiered matching strategy: SSN-confirmed matches first, then DOB+name matches with manual validation
Race coding	IHS uses tribal enrollment-based identification; CMS uses self-reported race on enrollment forms. Not all AI/AN Medicaid enrollees are coded as AI/AN in CMS records	Individuals in CMS with non-AI/AN race coding but verified IHS registration would be missed in any race-based filter	Project did not rely on CMS race coding — used IHS registrant status as the AI/AN identifier, sidestepping CMS race data quality issues
T-MSIS transition	CMS replaced the MAX system with T-MSIS during the project. Different data structure, different quality standards, different state reporting timelines	Required methodological revalidation mid-project; some state data in early T-MSIS was incomplete	Phased approach: continued MAX-based analysis while T-MSIS data was validated; documented methodology transitions explicitly
IHS Area variation	Data quality and completeness of IHS registrant files varied significantly across the 12 IHS Areas	Some Areas had more complete records than others; affected geographic reliability of results	Area-by-area quality review; flagged states where registrant data quality was lower; noted limitations in findings

INSTRUCTOR NOTE

This is the most concrete connection to practitioner experience in Weeks 5–7. As a health director, you have almost certainly encountered one or more of these problems: patients who appear differently in different systems, SSN inconsistencies in older records, race coding problems in state Medicaid data that undercount your AI/AN population. Name one or two specific examples from your experience — without identifying individuals. The goal is to make these ‘data problems’ feel real rather than abstract. Students who have never worked with federal administrative data often assume it is clean and consistent. It is not.

SESSION 3  ·  90 minutes — Week 7 Tribal Data Sovereignty as Research Design Sovereignty principles · small-cell suppression · state-level only reporting · tradeoffs · midterm launch

Session 3 Timing Overview

0-10 min

Opening

Why sovereignty is a design principle, not a constraint Script framing

10-35 min

Section 3.1

Five sovereignty principles and how each shaped the project Lecture with principles table; move through each systematically

35-55 min

Section 3.2

The small-cell suppression decision: who it protected and who it limited Analytical deep-dive; the hardest tradeoff in the project

55-70 min

Section 3.3

Mock Tribal council statement exercise Students draft a 1-paragraph statement from a Tribal council’s perspective

70-90 min

Midterm Launch

Timeline + sovereignty essay: scaffolding and Q&A Walk through the midterm assignment; answer questions; assign timeline construction as first step

3.0 — Opening (0-10 min)

READ ALOUD  /  ADAPT

“Over the past two weeks we have covered the institutional history of the Data Match project and its technical architecture. Today I want to reframe what we have learned by asking a different question. We have been treating Tribal data sovereignty as a series of governance requirements that the project had to satisfy — small-cell suppression, state-level-only reporting, tribal review before publication. Things the team had to do. I want to argue today that this is the wrong frame. Tribal data sovereignty is not a compliance checklist. It is a design principle — a set of affirmative values that, when taken seriously, produce different and better research than you would get without them. Consider: NIHB as lead organization rather than CMS is a sovereignty principle. It is also what produced research that Tribal communities trust. State-level suppression of tribal-specific data is a sovereignty principle. It is also what made the data release politically sustainable — no Tribe was exposed to data it had not consented to make public. The sovereign governance of this project is not what made it slow. It is what made it possible. And understanding why requires thinking carefully about what data sovereignty actually means for communities that have experienced data being used against them.”

3.1 — Five Sovereignty Principles and How Each Shaped the Project (10-35 min)

Work through each principle as a lecture point, pausing after each to ask: ‘What would the project have looked like if this principle had been violated or ignored?’

Sovereignty Principle	How It Appeared in the Project	What the Project Would Have Looked Like Without It
1. Self-determination in research leadership	NIHB — a Tribal advocacy organization — led the research, not CMS or IHS. This was a deliberate structural choice, not a default.	A federal-agency-led project would have had different research questions, framed findings in federal program terms, and lacked the Tribal legitimacy that made findings credible to the communities they described.
2. Meaningful consent and consultation	The Tribal Advisory Committee was not advisory in name only — it had real governance authority, including pre-publication review rights and the ability to halt or redirect the project.	A nominal advisory process would have produced legitimate grievance from Tribal communities about being research subjects without research partners. The findings could have been repudiated by Tribal leaders.
3. Data minimization and purpose limitation	DUAs prohibited use of the matched data for any purpose other than health coverage analysis. Specifically prohibited: enrollment audits, eligibility challenges, enforcement actions.	Without explicit prohibition, matched data could theoretically have been used to challenge Medicaid enrollment accuracy, audit individual tribal programs, or support eligibility restriction arguments. The prohibition transformed the data from a potential threat into a research asset.
4. Protection of individual and community privacy	Small-cell suppression rules prevented publication of data for groups below threshold size. State-level-only reporting prevented tribal-level disclosure.	Tribal-level data publication — even in aggregate — could have identified individuals in small communities, exposed tribal health utilization patterns to outside scrutiny, and undermined future participation in similar projects.
5. Community ownership of findings	Tribal Advisory Committee reviewed results before any external release. Tribes were positioned as co-owners of the findings, not recipients of federal data products.	Federal-agency-first publication would have positioned Tribal communities as subjects of research rather than participants. The political and narrative authority over what the numbers mean would have resided with federal agencies.

SOVEREIGNTY IN RESEARCH DESIGN: THE DEEPER PRINCIPLE

Each of these five principles reflects a theory of who the research serves. When data sovereignty is treated as compliance, the implicit answer is: the research serves the researchers or the funding agencies, subject to certain community protections. When data sovereignty is treated as a design principle, the answer is: the research serves the communities it describes, and every methodological and governance decision should be evaluated against that standard. This is the lens through which students should read the Data Match design choices. Not ‘what did NIHB have to do to satisfy Tribal requirements?’ but ‘what did NIHB build so that this research would actually serve Tribal interests?’ Those are different questions and they produce different designs.

3.2 — The Small-Cell Suppression Decision: The Hardest Tradeoff (35-55 min)

What Small-Cell Suppression Is

Small-cell suppression is the practice of withholding data cells in published tables when the underlying count is below a defined threshold — typically fewer than 11 individuals. It exists to prevent what is called the identification problem: when a cell contains very few people, outside observers can sometimes identify specific individuals by cross-referencing known demographic information.

In the context of the Data Match, suppression meant: no published data for any state-level subgroup (by age, sex, coverage type) where the AI/AN Medicaid enrollee count was below the threshold. And more significantly: no tribal-level data published at all, regardless of tribal size.

The Specific Tradeoff

The decision to suppress tribal-level data was the single most consequential governance choice in the project. It protected individual privacy and prevented tribal-specific data from being misused — but it created a significant limitation: individual Tribal nations could not see their own tribe-specific Medicaid enrollment and expenditure data in published reports.

Perspective	What Small-Cell Suppression Enabled	What Small-Cell Suppression Prevented
Individual Tribal members	Protected: a community member in a small tribe could not be identified through the data	No direct loss to individuals — aggregate figures don’t affect individual privacy either way
Tribal nation governments	Protected: no external actor could use tribe-specific data for adverse purposes (enrollment challenges, sovereignty disputes, funding arguments against specific tribes)	Lost: tribal governments cannot see their own verified Medicaid enrollment figures in public reports — must rely on internal billing data or special data requests
NIHB and Tribal researchers	Enabled: state-level findings are publishable with high credibility and low privacy risk	Limited: cannot produce tribe-level analyses from public data; future research requiring tribal-level granularity requires separate data agreements
State Medicaid agencies	Protected (unintentionally): states cannot see which specific tribes drive their AI/AN Medicaid enrollment — limiting potentially adverse state-level policy targeting	Neutral — states can still see their total AI/AN Medicaid figures, which is sufficient for state policy planning
Federal advocates and policy analysts	Enabled: state-level tables are sufficient for national funding arguments, PRC modeling, and broad equity advocacy	Limited: cannot make tribe-specific arguments about per-capita funding gaps or disparities without separate data agreements

READ ALOUD  /  ADAPT

“Here is the core tension I want you to sit with. A Tribal nation — a sovereign government — may have a legitimate interest in knowing exactly how many of its citizens are enrolled in Medicaid and how much Medicaid spends on their care. That is not a trivial interest. That is budget planning information. Advocacy data. Evidence for negotiations with state and federal agencies. And yet the project decision was: this data will not be published at the tribal level. Individual tribes cannot see their own numbers in public reports. The Tribal Advisory Committee endorsed this decision. They made a collective sovereignty judgment that protecting all communities from data exposure was more important than giving any individual community access to tribe-specific public data. That is a coherent decision. It is also a decision that some individual Tribal governments would not have made for themselves if asked independently. This is what real data sovereignty looks like in practice: not a clean answer, but a negotiated one. The Tribal Advisory Committee represented collective Tribal interests, not individual Tribal nation interests. Those are not always the same thing. And the decision they made — suppress tribal-level data — reflects a particular theory of how data can harm communities that deserves your critical engagement.”

DISCUSSION CAUTION: AVOID FALSE BINARY

Students sometimes frame this as ‘suppression was wrong’ vs. ‘suppression was right.’ Push back on the binary. The more productive frame is: what theory of harm was being protected against, and was the protection proportionate to the limitation it imposed? The project designers had specific historical precedents in mind — cases where tribal health data was used in enrollment disputes or in arguments for tribal termination. Understanding those precedents is necessary to evaluate the decision fairly. Assign the discussion question below to surface this.

MIDTERM ASSIGNMENT  ·  Due End of Week 7 Timeline + Sovereignty Essay Launched in Session 3 (Week 7) — launched and scaffolded in class

3.4 — Midterm Assignment: Launch and Scaffolding (70-90 min)

The midterm has two components. Walk through both in class. Allow 15 minutes for launch, leaving 5 minutes for questions.

Component	Description	Length	Due
Part A: Project Timeline	Construct a detailed annotated timeline of the CMS–IHS Data Match Project from 2017 to 2023. For each phase, identify: (1) what happened, (2) who led it, (3) why it mattered, and (4) one key decision made that shaped subsequent phases. Must be formatted as a visual timeline or structured table — not prose.	1-2 pages (visual or table format)	End of Week 7
Part B: Sovereignty Essay	In 5-7 pages, analyze: How did the principle of Tribal data sovereignty shape the design of the CMS–IHS Data Match Project? Your essay must engage specifically with at least three of the five sovereignty principles from Session 3, analyze the small-cell suppression decision in depth, and evaluate whether the governance model produced research that served Tribal interests. Conclude with a recommendation: what, if anything, should be done differently in the next phase of the project?	5-7 pages	End of Week 7

In-Class Scaffolding: Timeline Construction (15 min)

Give students 10 minutes to begin populating their timeline using their notes and the source materials distributed. This converts the launch into an active exercise rather than a passive assignment introduction. Then spend 5 minutes in whole-class Q&A.

Prompt students to ask: ‘What is the single most important transition point in the timeline?’ Listen for: the Phase 1-to-Phase 2 shift from ACS to administrative data design. This is the key inflection — everything before it is diagnosis, everything after it is construction.

INSTRUCTOR NOTE: MIDTERM GRADING CRITERIA

The timeline will be graded primarily on accuracy and completeness — students should be able to populate all five phases with correct actors, key decisions, and sequencing. The essay will be graded on: (1) accurate representation of the three chosen sovereignty principles, (2) analytical depth on the small-cell suppression decision (not just description of what it was, but analysis of the tradeoff), and (3) quality of the concluding recommendation — which should reflect genuine engagement with the tensions, not just a summary. A strong essay will acknowledge that the governance model involved real costs as well as real benefits.

Indian Health Financing  ·  Research & Methods Track  ·  Weeks 5–7 Lecture Notes  ·  February 2026

  Weeks 8–10 

INDIAN HEALTH FINANCING

Research & Methods Track

Weeks 8–10 Instructor Lecture Notes & Script

What the Data Match Found: Findings, PMPY Modeling, and the Population A/B Reconciliation

Element	Detail
Track	Research & Methods (appropriate for all tracks; methods emphasis marked throughout)
Sessions	Three 90-minute sessions spanning Weeks 8, 9, and 10
Core Question	What did the administrative data actually show — and what does the Population A/B distinction mean for how we read those findings?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. IV–V); Revised Analysis 1999-2023 (all sections); Lecture Series Part 3; Professor Introductions Weeks 8–10; TTAG (2015) Data Symposium Report
Builds On	Weeks 5–7 Data Match construction, governance, and sovereignty framework
Leads Into	Weeks 11–12 policy implications, Medicaid as structural pillar, fiscal risk scenarios
Data Memo	Major analytical assignment launched in Session 3 (Week 10), due end of week

These three sessions deliver the payoff the course has been building toward. Students who have followed the ACS-to-MAX-to-Data Match arc now encounter the verified findings and learn how to read them critically — which requires understanding the Population A vs. Population B distinction in full. Session 3 introduces the 1999-2023 longitudinal reconciliation and launches the data memo assignment.

Box colors: findings (green), PMPY/modeling (dark blue), Population A/B analysis (purple), fiscal risk (dark red), instructor notes (amber), warnings (orange), exercises (teal).

Three-Session Architecture

Session	Week	Core Argument	Key Deliverable
Session 1	Week 8	The 941,168 enrollment figure and the $6.4-6.6B expenditure estimate are robust within their defined scope — but that scope (Population A only) is a specific, bounded population that omits a significant share of AI/AN Medicaid spending	ACS vs. administrative cross-era comparison; PMPY walkthrough
Session 2	Week 9	The 1999-2008 historical baseline establishes the structural trajectory that produced the 2023 findings; the IHS/Tribal billing share doubling from 13% to 21% is the precursor to everything the Data Match revealed	Historical longitudinal analysis; state growth pattern exercise
Session 3	Week 10	The Population A vs. Population B distinction is not a footnote — it is the central analytical challenge for any researcher working with AIAN Medicaid data across time periods or across states with different IHS penetration	Data memo launch: 3-state Population A/B comparison assignment

WEEKS 8-10 IN THE COURSE ARC

This is the course’s data-heaviest block. Students are simultaneously reading quantitative findings, evaluating methodological scope, and building toward a major assignment. Three pacing principles: (1) Do not bury the lede — the $6.4-6.6B finding and the Population A/B distinction are the two most important things students need to understand, and both should be introduced early in Week 8, not saved for Week 10. (2) Use the state case studies as the primary analytical vehicle — Oklahoma and Minnesota are more pedagogically powerful than any lecture abstraction. (3) The data memo assignment is designed to be partially scaffolded in class during Week 10. Budget 20 minutes for that scaffolding; students who leave Week 10 with a partially constructed memo do substantially better than those who start cold.

SESSION 1  ·  90 minutes — Week 8 What the Data Match Found: The Core Numbers 941,168 enrollees · $6.4-6.6B expenditure estimate · PMPY mechanics · sensitivity analysis table · ACS vs. administrative final verdict · Population A scope

Session 1 Timing Overview

0-10 min

Opening

From five years of construction to a number Script: the payoff moment framing

10-30 min

Section 1.1

The enrollment finding: 941,168 verified What the number means, how it was produced, what it is not

30-55 min

Section 1.2

The expenditure estimate: PMPY modeling unpacked Walk through per-member-per-year mechanics; where the $6.4-6.6B comes from

55-75 min

Section 1.3

ACS vs. administrative: the final comparison table Side-by-side across all three eras; what changed and what each source is still useful for

75-90 min

Section 1.4

Introducing Population A — what the 941,168 includes and excludes Critical scope framing; preview of Sessions 2 and 3

1.0 — Opening (0-10 min)

READ ALOUD  /  ADAPT

“For seven weeks we have been building toward this moment. We traced the trust responsibility and its chronic underfunding. We diagnosed what survey data could and could not show. We spent three weeks inside the construction of the CMS–IHS Data Match — the governance negotiations, the VRDC, the matching algorithms, the sovereignty framework. And now we get to ask: after all of that, what did the administrative data actually find? The headline is this: 941,168 American Indian and Alaska Native individuals with verified IHS access were enrolled in Medicaid in 2023. Associated Medicaid expenditures for that population are estimated at $6.4 to $6.6 billion annually. Medicaid, as a funding stream for Indian health, is now roughly comparable in scale to IHS direct appropriations. But this course has taught you to ask: what does that number actually measure? Who is in it, and who is not? What assumptions does the $6.4-6.6 billion rest on? And what does the number look like if you account for the AI/AN Medicaid population that the Data Match, by design, does not capture? Those questions are what Weeks 8 through 10 are about. We are going to read these findings with the same critical precision we applied to the ACS. The numbers are better — substantially better. But they are not the last word. They are a defined answer to a defined question. Understanding exactly what that definition is will determine how well you can use these findings in research, advocacy, and policy analysis.”

1.1 — The Enrollment Finding: 941,168 Verified (10-30 min)

What the Number Represents

The 941,168 figure is the count of AI/AN individuals who satisfy all three of the Data Match’s inclusion criteria simultaneously during the 2023 study year:

1. Verified IHS registrant status — confirmed through direct linkage of CMS Medicaid enrollment files with IHS patient registration records
2. Medicaid enrollment — confirmed active enrollment in Medicaid during the study year (not just self-reported coverage)
3. At least one paid service delivered through an Indian health program — an IHS, Tribal, or Urban Indian facility that generated a paid Medicaid claim during the year

All three conditions must be met. A person who is IHS-registered and Medicaid-enrolled but whose care in 2023 was delivered entirely through external providers does not appear in this count. A person who received IHS care but was not Medicaid-enrolled does not appear. This is the definition of Population A.

CORE FINDING: 941,168 IHS-ACCESS MEDICAID ENROLLEES (2023)

This figure represents the most precise, verified count of the IHS-access Medicaid population ever produced. It is not an estimate from a sample — it is derived from administrative records. Its precision is its primary advantage over the ACS and MAX estimates that preceded it. Geographic concentration: Approximately 95% of enrollment is concentrated in 25 states. Four states — Arizona (~17%), Oklahoma (~13%), New Mexico, and Alaska — account for a disproportionate share. This geographic pattern reflects the distribution of IHS facilities and reservation-resident AI/AN populations. Scale context: The total AI/AN population of the United States is estimated at 4-8 million (depending on race identification methodology). Finding that over 940,000 individuals satisfy all three IHS-access-Medicaid criteria confirms that Medicaid is a primary health coverage pathway for the IHS-served population — not a fringe program.

What the 941,168 Is Not

This framing matters for how students use the number in advocacy and research contexts:

It is not…	Because…	What it is instead
The total AI/AN Medicaid population	Excludes Population B: AI/AN Medicaid enrollees with no IHS registration (urban, mainstream-provider users)	The IHS-access Medicaid population — a specific, policy-relevant subset
A complete count of IHS users	IHS users who are not Medicaid-enrolled (Medicare-only, uninsured, privately insured) are outside the Data Match scope entirely	The intersection of IHS registration and Medicaid enrollment
A count of tribal members	Tribal membership is not required for IHS access; some IHS users are descendants without formal enrollment; some tribal members have no IHS registration	A count of IHS registrants who meet the Medicaid enrollment and service criteria
A headcount of all people who used IHS in 2023	Requires at least one paid Medicaid service at an I/T/U facility; IHS users whose Medicaid claims were for non-IHS providers are excluded	The subset of IHS users who generated at least one paid IHS Medicaid claim

MOST COMMON MISUSE OF THE 941,168 FIGURE

Students and advocates sometimes cite 941,168 as the ‘AI/AN Medicaid population’ without qualification. This understates the total AI/AN Medicaid population (which includes Population B) and mischaracterizes what the Data Match measured. The correct framing is: ‘941,168 AI/AN individuals with verified IHS access were enrolled in Medicaid in 2023.’ The distinction matters most in policy contexts where urban AI/AN communities are arguing for Medicaid protections — those communities are Population B, largely invisible in the 941,168 figure.

1.1b — CMS Coverage Distribution Within the IHS-Access Population (ACS 2024)

A cross-tabulation of Medicare (HINS3) × Medicaid (HINS4) × IHS access (HINS7) from the 2024 ACS 5-Year PUMS (weighted N = 1,523,017 IHS-access population) produces four mutually exclusive coverage groups. This is the first nationally representative breakdown of CMS coverage status within the IHS-access population. It provides context the Data Match cannot — the Data Match captures only the Medicaid-enrolled subset (Population A); the ACS reveals the full coverage landscape including Medicare and the uninsured majority.

CMS Coverage Group	IHS-Access Pop.	% of IHS Pop.	Data Match Scope	Spending Est. (midpoint)
Dual Eligible (Medicare + Medicaid)	102,865	6.8%	Included (Medicaid side)	$2.86B
Medicare Only (no Medicaid)	157,266	10.3%	Not in scope	$2.64B
Medicaid Only (no Medicare)	484,448	31.8%	Included (core Population A)	$5.35B
No CMS Coverage (uninsured / other)	778,438	51.1%	Entirely invisible to Data Match	— (IHS/PRC only)
Total IHS-Access Population (ACS)	1,523,017	100%	Data Match captures 587,313 (38.6%)	$10.86B (CMS-covered only; 744,579 individuals)

Source: ACS 5-Year PUMS 2024, U.S. Census Bureau. Weighted cross-tabulation of HINS3 (Medicare) × HINS4 (Medicaid) × HINS7 (IHS/Indian health access), weighted by PWGTP. Total IHS-access population N = 1,523,017. The 51.1% without CMS coverage is entirely outside the Data Match scope — this population is visible only through survey data. Medicare total (any Medicaid): 260,131 (17.1%); Medicaid total (any Medicare): 587,313 (38.6%). Spending estimates (midpoint scenario) from CMS Investment Estimates Paper, Table 1 (ACS 5-Year PUMS 2024 × 2024 NHE PMPY rates): Dual $2.86B, Medicare-only $2.64B, Medicaid-only $5.35B, total CMS-covered $10.86B. No CMS spending accrues to the uninsured group; their care costs fall on IHS appropriations and PRC.

INSTRUCTOR NOTE: The 51.1% figure is the most important number in this table for policy purposes. The majority of the IHS-access population has no CMS coverage — they are entirely dependent on IHS appropriations and PRC for care. This is the population whose access deteriorates first when PRC budgets are exhausted. The Data Match tells us everything about the 38.6% who have Medicaid, but nothing about the 51.1% who do not. When students hear the $10.86B total CMS investment estimate (midpoint; dual + Medicare-only + Medicaid-only), they should immediately ask: what about the other half? That question drives the funding gap analysis in Weeks 11–12.

What PMPY Means

Per-member-per-year (PMPY) analysis is the standard Medicaid expenditure modeling framework. It answers the question: if we know how many people are enrolled and for how long, how much should Medicaid spending be? The logic is:

PMPY FORMULA AND ITS COMPONENTS

Total Annual Expenditure = PMPY Rate x Enrolled Months / 12 x Adjustment Factor Where: PMPY Rate: The average annual Medicaid cost per enrolled individual. The Data Match analysis used approximately $6,975 adjusted, derived from national Medicaid PMPY benchmarks calibrated for the AI/AN IHS-access population’s known utilization patterns (higher acute care, lower long-term care relative to national average). Enrolled Months / 12: Not all 941,168 enrollees were covered for the full 12 months of 2023. Some enrolled mid-year; some lost coverage. The analysis adjusts by dividing total enrolled member-months by 12, equivalent to counting full-year enrollees. Adjustment Factor (0.90): The paper applies a 0.90 adjustment factor to account for expected variation in utilization between the national PMPY benchmark and the specific AI/AN IHS-access population. This produces the $6.42-6.56B range depending on the PMPY rate used.

Walking Through the Calculation

READ ALOUD  /  ADAPT

“Let me walk through this so the math is transparent. We have 941,168 verified enrollees. But some of them were enrolled for less than 12 months — say an average of 10.8 months. That means full-year equivalent enrollees are roughly 941,168 times 10.8 divided by 12, which is about 848,000 full-year equivalent enrollees. Multiply that by the adjusted PMPY rate of about $6,975 and you get $5.9 billion before the adjustment factor. Apply the 0.90 factor: about $5.3 billion. But the paper’s PMPY is calibrated higher — around $7,750 for the IHS-access population — reflecting the known higher acute care burden. At $7,750 times 848,000 full-year equivalents times 0.90, you arrive at approximately $5.9 billion at the low end, $6.56 billion at the high end. What I want you to notice is this: the $6.4-6.6 billion is a modeled figure, not a summed-from-claims figure. The paper explicitly acknowledges that no 2023 primary state-level claims data was available at time of writing. The PMPY approach applies national benchmarks to a verified enrollment count. It is a sound methodology — it is how Medicaid actuaries work — but it produces a range, not a precise claim, and that range rests on the PMPY assumptions. When you cite this number, you are citing a model output, not an audited claims total.”

Why the Range Matters for Advocacy

The $6.42-6.56B range is tight enough to be compelling in policy arguments. Students should understand what drives variation within the range and what could move the estimate outside it:

Assumption	Low-End Effect	High-End Effect	Analytical Implication
PMPY rate calibration	If IHS-access population has lower utilization than national AI/AN benchmark	If IHS-access population has higher utilization due to unmet need	The PMPY is the single largest driver of the range; future claims data will validate or revise it
Enrolled months per enrollee	If average enrollment duration is shorter than assumed	If more enrollees have full-year continuous coverage	Medicaid continuous enrollment policy (2020-2023) likely pushed this upward; post-unwinding period may reduce it
Adjustment factor (0.90)	Lower adjustment = lower total (0.85 yields ~$5.6B)	Higher adjustment = higher total (0.95 yields ~$5.9B at same PMPY)	The 0.90 factor is a modeling convention; its sensitivity should be acknowledged when using the figure in advocacy

INSTRUCTOR NOTE: CLAIMS DATA LIMITATION

The absence of 2023 primary claims data is worth a brief discussion. Future phases of the Data Match project are expected to incorporate actual claims records rather than PMPY modeling. When that happens, the expenditure figure will shift from a model output to a direct measurement. Whether it goes up or down relative to the PMPY estimate is itself an empirical question — and an important one for tracking whether the IHS-access population is over- or under-utilizing relative to national benchmarks. Students interested in health services research careers should flag this as an open research question.

1.2b — Sensitivity Analysis: Five Estimates Across Data Sources and Reliance Assumptions

Once students have worked through the PMPY mechanics, introduce the sensitivity analysis table as the analytic tool that makes the range of estimates legible. The table presents five distinct estimates of the AI/AN population with both Medicaid coverage and IHS access, organized by data source and conceptual scope. It is the bridge between the specific $6.4–6.6B finding and the broader question of what different modeling assumptions produce.

Distribute the table before this section or display it on screen.

Estimate Type	Data Source	Concept Measured	Population	Est. Medicaid Spending
Survey estimate (lower bound)	American Community Survey (ACS)	Self-reported Medicaid coverage and IHS access	587,313 (Medicaid-enrolled with IHS access; ACS cross-tab 2024)	~$5.35B (midpoint; 2024 NHE Medicaid PMPY × ACS enrollment)
Administrative overlap	CMS–IHS Data Match	Verified enrollment in both Medicaid and IHS patient registry	≈ 941,000	$6.4–$6.6B
Reliance-adjusted (80%)	Sensitivity analysis	Estimated population primarily using IHS/Tribal providers	≈ 753,000	$5.1–$5.3B
Reliance-adjusted (70%)	Sensitivity analysis	Conservative reliance estimate	≈ 659,000	$4.5–$4.6B
Reliance-adjusted (60%)	Sensitivity analysis	Lower-bound estimate of reliance on I/T/U system (or non-system)	≈ 565,000	$3.8–$4.0B

Note: ACS Medicaid count (587,313) is the verified cross-tabulation of HINS4 (Medicaid) × HINS7 (IHS access) from the 2024 ACS 5-Year PUMS, replacing the prior ≈798,000 survey estimate. The Data Match count (941,168) is 60% larger than the ACS figure, consistent with the known ACS IHS-coverage undercount documented in Census Bureau linkage studies. Critical enrollment-basis distinction: the Data Match likely reflects “ever enrolled” (anyone enrolled ≥1 month in 2023), while the ACS captures point-in-time coverage. Full-year-equivalent (FYE) adjustment of the Data Match denominator yields ≈847,000–899,000 FYE enrollees, narrowing the gap. The CMS–IHS Data Match provides the empirically strongest enrollment estimate and should anchor the discussion. Reliance-adjusted estimates apply alternative assumptions about the share of the matched population whose care primarily flows through IHS, Tribal, or Urban Indian (I/T/U) providers.

HOW TO TEACH THIS TABLE

The administrative overlap row (CMS–IHS Data Match, ≈941,000, $6.4–$6.6B) is the anchor. Every other row is a comparison point, not a competing claim. The reliance-adjusted rows all start from the same 941,000 denominator — what changes is the assumed share of that population who rely primarily on I/T/U providers for their care. The table is not saying the population is smaller; it is saying that if only 80% (or 70%, or 60%) of matched enrollees depend primarily on IHS/Tribal care, the policy-relevant spending estimate adjusts accordingly.

Key insight to surface: even at the most conservative reliance assumption (60%), Medicaid spending tied to the IHS-access population is $3.8–$4.0B annually — roughly 44–47% of IHS appropriations ($8.6B, FY2023). Medicaid is structurally significant at every point in the sensitivity range.

Discussion prompt: A tribal health director is presenting to a state Medicaid committee. Which row of this table should she lead with, and how should she characterize the others? What changes if the audience is a Congressional appropriations subcommittee vs. a state budget office that is skeptical of the $6.4B figure?

1.3 — ACS vs. Administrative: The Final Comparison (55-75 min)

Present this as the definitive answer to the question the course has been building toward since Week 3: what changed when we moved from survey to administrative data, and what is each source still good for?

Data Source	Year	AI/AN Medicaid Enrollees	Expenditure Estimate	What It Could Show	What It Could Not Show
ACS Survey	2013	~1.3M (self-reported)	None	Coverage prevalence, uninsured rates, broad trends, state comparisons	Verification, claims data, payment flows, IHS-specific enrollment, spending magnitude
MAX Administrative	2012	~1.1M (verified with caveats)	$2.8B (actual claims)	Enrollment counts, payment totals, utilization patterns, IHS facility share	Urban Indian patients (missed by AIR proxy), non-encounter-year registrants, state-level AI/AN race code quality
CMS–IHS Data Match	2023	941,168 (verified IHS-access)	$6.42-6.56B (PMPY modeled)	Verified IHS-access enrollment, state distribution, fiscal magnitude, trend 2019-2023	Population B (non-IHS-access AI/AN Medicaid), actual 2023 claims, tribal-level data

What Is Each Source Still Useful For?

A key takeaway for Research & Methods students: administrative linkage does not replace survey data. It answers different questions. The ACS remains the best available tool for:

1. Monitoring national AI/AN uninsured rates and coverage trends over time
2. Rapid-turnaround state-level coverage estimates where VRDC access is not available
3. Capturing Population B — AI/AN Medicaid enrollees with no IHS registration, who appear in ACS self-report but not in the Data Match
4. Post-ACA longitudinal coverage analysis where the question is about coverage rates, not payment magnitudes

The Data Match is essential — and the ACS is insufficient alone — for:

• Verified enrollment counts tied to IHS system use
• Fiscal magnitude: how many dollars are actually flowing through Indian health programs
• State-level distribution of IHS-access Medicaid spending for budget advocacy
• PRC substitution modeling: estimating how much Medicaid absorbs costs that would otherwise fall on Purchased/Referred Care
• Trend analysis within the IHS-access population across the 2019-2023 study window

THE CENTRAL METHODS LESSON OF WEEKS 8-10

Survey data and administrative data are complements, not substitutes. The ACS tells you about coverage prevalence for the full AI/AN population. The Data Match tells you about payment magnitude for the IHS-registered subset. A complete picture of AI/AN Medicaid requires both — and explicitly accounting for the population scope difference between them. That reconciliation is what Weeks 9 and 10 are about.

1.4 — Introducing Population A: What the 941,168 Includes and Excludes (75-90 min)

READ ALOUD  /  ADAPT

“I want to close today by introducing a distinction that will be at the center of the next two weeks. The Data Match produces data about what we are calling Population A: AI/AN individuals with verified IHS access who are Medicaid enrolled and generated at least one paid IHS service. That is a precisely defined, policy-relevant population. It is the population that the IHS system directly touches. But there is a second population — Population B — consisting of AI/AN Medicaid enrollees who use mainstream providers and have no IHS registration. They are in the ACS. They were in the 1999-2012 historical payment files. They are not in the 2023 Data Match. And for some states, they represent a majority of total AI/AN Medicaid spending. If you connect the historical payment files — which capture both populations — to the 2023 Data Match data — which captures only Population A — without correcting for this scope difference, you will produce trend lines that are analytically invalid. You will conclude that Medicaid spending fell in Minnesota when it did not. You will miss New York entirely. You will overstate IHS-access as a share of total AI/AN Medicaid in urban-dominant states. Next week we go back to the 1999-2008 baseline data. The week after that, we build the full reconciliation. Come prepared to work with numbers.”

SESSION 2  ·  90 minutes — Week 9 The Historical Baseline: What 1999-2008 Tells Us About 2023 All-AIAN spending trajectory · IHS/Tribal billing share shift · geographic concentration · state growth patterns · bridging to the Data Match

Session 2 Timing Overview

0-10 min

Bridge

Recap Population A framing; introduce the historical question Brief verbal recap; pose the central question for Session 2

10-35 min

Section 2.1

National trajectory 1999-2008: $1.45B to $3.86B Lecture with year-by-year table; 11.5% CAGR; the IHP billing share doubling

35-55 min

Section 2.2

Geographic concentration and state growth patterns State table; the top-5 concentration; Minnesota’s structural position

55-75 min

Section 2.3

Bridging 1999-2023: valid comparisons and the two approaches Cross-dataset analysis; why direct comparison fails; how to correct

75-90 min

Section 2.4

State growth pattern exercise Groups assigned states; analyze growth vs. scope shift

2.0 — Bridge Opening (0-10 min)

READ ALOUD  /  ADAPT

“Last week we established what the 2023 Data Match found and what Population A means. Today I want to go backward — to 1999 — and ask: how did we get here? The $6.4-6.6 billion in Medicaid spending for IHS-access enrollees in 2023 did not appear suddenly. It was the product of a 25-year structural transition. The historical data shows us that transition in detail. And understanding the trajectory matters for one specific analytical reason: the 2023 paper and the historical payment files measure different populations. If you try to draw a trend line from 1999 to 2023 without correcting for that difference, you will get the wrong answer. Today we build the foundation for making that correction correctly.”

2.1 — National Trajectory 1999-2008: The Structural Transition (10-35 min)

The Growth Numbers

Present this table as the anchor for the session. Spend time on it — students should be able to read this data and identify the key patterns before you name them.

Year	National Total (All AIAN)	YOY Growth	IHP Billing (IHS+Tribal)	IHP Share of Total	Cumulative Growth
1999	$1.451B	—	$193M	13.3%	baseline
2000	$1.628B	+12.2%	$230M	14.1%	+12.2%
2001	$2.012B	+23.6%	$266M	13.2%	+38.7%
2002	$2.455B	+22.0%	$319M	13.0%	+69.2%
2003	$2.816B	+14.7%	$376M	13.3%	+94.1%
2004	$3.192B	+13.4%	$531M	16.7%	+120.1%
2005	$3.243B	+1.6%	$672M	20.7%	+123.6%
2006	$3.164B	-2.4%	$610M	19.3%	+118.1%
2007	$3.387B	+7.0%	$667M	19.7%	+133.5%
2008	$3.857B	+13.9%	$767M	19.9%	+165.9%

The Two Patterns Students Must Identify

READ ALOUD  /  ADAPT

“Before I tell you what to see in this table, I want you to look at it for two minutes and tell me: what are the two most significant patterns you notice? [Pause for responses.] The first pattern is the growth rate itself: $1.45 billion to $3.86 billion in nine years. That is 166% growth at a compound annual rate of 11.5%. This substantially outpaced general Medicaid growth in the same period. Why? A combination of enrollment expansion post-welfare reform, increasing Tribal Medicaid billing capacity, and rising per-unit costs in AI/AN communities. The second pattern is more important for our analytical purposes: the IHP billing share. Look at 1999-2003: it holds steady at about 13%. Then look at 2004-2005: it jumps from 16.7% to 20.7% in a single year. This is a structural break. The IHS and Tribal facilities did not suddenly start treating more patients. What happened is that they got dramatically better at billing Medicaid for the patients they were already treating. Indian Health Care Improvement Act provisions expanded Tribal billing rights; Tribes invested in revenue cycle capacity. This 2004-2005 shift is the precursor to the 2023 finding. The Tribal billing infrastructure that now generates $6+ billion in annual Medicaid revenue was being built in those years. The 2023 number did not emerge from nothing — it was the culmination of a deliberate, decade-long investment in Tribal billing capacity.”

THE IHP BILLING SHARE DOUBLING: WHY IT MATTERS FOR 2023

The IHP share doubling from ~13% (1999-2003) to ~20% (2005-2008) means that an increasing portion of total AIAN Medicaid payments was flowing through IHS and Tribal facilities — making Population A spending progressively more visible in administrative records. By 2008, the billing infrastructure that would make the Data Match analytically possible was largely in place. The 2023 paper’s $6.4-6.6B figure is not a sudden revelation — it is the documentation of a trajectory that was already visible in 2008.

The 2005-2006 Anomaly

Students should also notice the 2005-2006 plateau: near-flat growth in 2005 (+1.6%) and nominal contraction in 2006 (-2.4%). This anomaly warrants attention because it appears across multiple large states simultaneously — a pattern inconsistent with genuine spending reduction in those years, since the underlying IHS-registered population did not contract. The most likely explanation is changes in CMS reporting methodology or state Medicaid reconciliation practices. The IHP billing share actually peaked in 2005, suggesting the anomaly reflects the all-AIAN total component (Population B) more than the IHS-specific component. Students should understand this as a potential data artifact when drawing trend lines — and as a reminder that even administrative data is not immune to reporting inconsistencies.

2.2 — Geographic Concentration and State Growth Patterns (35-55 min)

The Top-5 Concentration

Five states — Arizona, Alaska, Oklahoma, New Mexico, and Minnesota — consistently accounted for more than 50% of national AIAN Medicaid payments from 1999 to 2008. This is not a policy artifact; it reflects the geographic distribution of both IHS facilities and AI/AN population density. Present the state growth table:

State	1999	2004	2008	% of 2008 National	1999-2008 CAGR	Population type
Arizona	$292M	$585M	$950M	24.6%	14.0%	Reservation-dominant (~70% IHS penetration)
Alaska	$143M	$358M	$373M	9.7%	11.3%	Reservation-dominant (~75% IHS penetration)
Oklahoma	$114M	$232M	$332M	8.6%	12.6%	Mixed (~55% IHS penetration)
New Mexico	$167M	$318M	$390M	10.1%	9.9%	Reservation-dominant (~65% IHS penetration)
Minnesota	$89M	$147M	$234M	6.1%	11.4%	Urban-dominant (~30% IHS penetration)
California	$41M	$131M	$178M	4.6%	17.7%	Urban-dominant (~28% IHS penetration)
Washington	$61M	$120M	$164M	4.2%	11.6%	Mixed (~50% IHS penetration)
North Carolina	$52M	$99M	$156M	4.1%	13.0%	Mixed (~40% IHS penetration)
Montana	$55M	$102M	$124M	3.2%	9.5%	Reservation-dominant (~68% IHS penetration)

Minnesota’s Structural Position in the Historical Data

Minnesota’s historical position — 5th by payment volume in 2008, $234M — warrants extended attention because it is the state where the Population B scope problem is most starkly visible in the cross-dataset comparison. Minnesota’s large urban AI/AN population (concentrated in Minneapolis-Saint Paul) had relatively low IHS registration rates (~30% penetration). This means approximately 70% of Minnesota’s 2008 AIAN Medicaid payments reflect Population B spending — individuals who are AI/AN Medicaid enrollees using mainstream providers with no IHS registration.

When the 2023 Data Match — which captures only Population A — shows Minnesota with $190 million (down from $329 million in 2012 actual), the apparent 42% decline is almost entirely a measurement artifact. The Population B component that drove Minnesota’s historical ranking simply does not appear in the 2023 paper. Students must internalize this before attempting any cross-dataset analysis.

2.3 — Bridging 1999-2023: Two Valid Approaches to Cross-Dataset Analysis (55-75 min)

Present this as the core methodological challenge of the module — and the reason the Revised Analysis exists. The 1999-2012 historical files and the 2023 Data Match paper measure different populations. Direct year-over-year comparison is analytically invalid for the reasons we have established. Two correction approaches exist:

THE TWO VALID CROSS-DATASET APPROACHES

Approach A: Adjust 2023 to all-AIAN scope (add Population B back) Add estimated Population B spending (~$1.88B for the 15-state analysis) to the 2023 paper’s $6.56B IHS-access figure. Estimated total: ~$8.45B. This is directly comparable to the historical all-AIAN payment totals. Implied CAGR from 2008 ($3.86B) to 2023 estimated all-AIAN (~$8.45B): approximately 5.4% per year. This represents a marked deceleration from the 11.5% CAGR of 1999-2008, consistent with maturing enrollment and moderating Medicaid growth. Approach B: Estimate Population A’s share of historical spending (adjust history down) Using national IHS penetration rates, Population A represented approximately 63-70% of total AIAN Medicaid payments historically. This implies a Population A 2008 baseline of roughly $2.4-2.7B. The CAGR from that range to the 2023 paper’s $6.56B is approximately 6.3-7.2% — higher than the all-AIAN rate, reflecting real expansion of IHS-access enrollment and Tribal billing infrastructure.

READ ALOUD  /  ADAPT

“Both approaches produce defensible trend lines. They tell slightly different stories. Approach A — adding Population B back to 2023 — gives you the broadest federal Medicaid commitment to AI/AN health. That is the right frame for arguing that total federal investment should be increased. Approach B — estimating Population A historically — gives you the IHS-system-specific trajectory. That is the right frame for arguing about IHS funding specifically. Neither approach fills the 2013-2022 data gap — the eleven years between the last reliable historical data point and the 2023 paper. For those years, you can interpolate, but you should be explicit that you are interpolating, not reporting actual data. The gap is real, and filling it requires T-MSIS data extraction that has not yet been done for this population. That is the single highest-value data infrastructure investment for anyone doing longitudinal AI/AN Medicaid analysis.”

What the Two Approaches Produce	Approach A: Adjust 2023 to all-AIAN	Approach B: Estimate Pop. A historically
2023 endpoint	~$8.45B (IHS-access + Population B estimate)	$6.56B (paper figure, IHS-access only)
Comparable historical baseline	1999-2012 all-AIAN actual data	Estimated 63-70% of 1999-2012 totals
Implied 1999/2008 to 2023 CAGR	~7.3% (1999 to 2023) / ~5.4% (2008 to 2023)	~6.3-7.2% (2008 IHS-access estimate to 2023)
Best for arguing	Total federal Medicaid commitment to AI/AN health	IHS-system-specific Medicaid trajectory
Key limitation	Population B estimates are illustrative, not validated claims	IHS penetration rates are approximations; historical IHS-access share requires estimation

2.4 — State Growth Pattern Exercise (75-90 min)

EXERCISE: REAL GROWTH VS. SCOPE SHIFT  |  15 minutes

Format: Groups of 2-3 students. Each group is assigned one of the following state pairs. Using the historical table (Session 2.2) and the 2023 reconciliation table (from the Revised Analysis), groups answer the three questions below. State pair assignments: (1) Arizona + Alaska | (2) Oklahoma + Minnesota | (3) California + North Carolina | (4) New Mexico + New York Questions: For each state: is the apparent change from 2012 actual to 2023 paper primarily real growth, primarily a scope artifact, or a mixture? What evidence supports your classification?For your most complex state: write one sentence you would include in a research paper to correctly characterize the cross-dataset trend for that state.Which state in your pair would be most misleading if cited without the Population B correction, and why? Debrief: Oklahoma/Minnesota pair always produces the clearest contrast. Oklahoma’s growth is real (2021 expansion). Minnesota’s apparent decline is entirely scope. If any group got this wrong, walk through it explicitly — this is the core analytical skill the data memo will test.

SESSION 3  ·  90 minutes — Week 10 Full Reconciliation: Population A/B, Fiscal Risk, and the Data Memo Reconciled 2023 state table · PRC substitution · fiscal risk scenarios · data infrastructure gap · data memo launch

Session 3 Timing Overview

0-10 min

Bridge

Synthesis: what the full reconciliation requires Script: bringing Approaches A and B together

10-35 min

Section 3.1

The reconciled 2023 state table: reading it correctly Full 15-state table; state category framework; analytical flags

35-55 min

Section 3.2

PRC substitution and what it means for Population A and B separately Fiscal mechanism deep-dive; the $2.6-4B estimate and its Population A specificity

55-70 min

Section 3.3

Fiscal risk scenarios and the measurement gap What remains unknown; the 11-year data gap; policy implications of the gap

70-90 min

Data Memo Launch

Assignment scaffolding: 3-state Population A/B comparison In-class start; walk through structure; Q&A

3.0 — Bridge Opening (0-10 min)

READ ALOUD  /  ADAPT

“Over the past two weeks we have built all the pieces: the 2023 Data Match findings and their Population A definition; the 1999-2008 historical baseline and its all-AIAN scope; the two valid approaches to cross-dataset comparison; and the state cases that illustrate when scope distortion is most severe. Today we put it together. We will look at the most complete available picture of total AIAN Medicaid spending by state in 2023 — combining the Data Match findings with Population B estimates. We will analyze the PRC substitution mechanism and what it does and does not cover. And we will be honest about what we still do not know and why that matters. Then we will launch the data memo assignment. By the end of class today, each of you should have a framework and a partial draft for three states. This assignment asks you to do exactly what we have been doing together — read the numbers, identify the scope issues, and explain what the difference between Population A and Population B means for how those numbers should be used in advocacy.”

3.1 — The Reconciled 2023 State Table (10-35 min)

Present this as the analytical centerpiece of the three-week block. Every number in this table has been discussed; now students see them all at once. Walk through it systematically, using the three state categories as your organizing framework.

State	Pop. Type	2023 IHS-Access (Paper)	Est. Non-IHS (Pop. B)	Est. Total A+B	Pop. B Share	Analytical Flag
Arizona	Res.-dominant	$1,110M	$207M	$1,317M	16%	Trend valid — scope gap small
Oklahoma	Mixed	$810M	$330M	$1,140M	29%	Real growth from 2021 expansion
New Mexico	Res.-dominant	$720M	$157M	$877M	18%	Trend valid — reservation-dominant
Alaska	Res.-dominant	$580M	$64M	$644M	10%	Trend valid — scope gap minimal
California	Urban-dominant	$470M	$271M	$741M	37%	Pop. B significant; caution on trends
Washington	Mixed	$360M	$98M	$458M	21%	Mixed; use with caveat
Oregon	Urban-dominant	$300M	$64M	$364M	18%	IHS-access growth real (billing infra)
South Dakota	Res.-dominant	$210M	$49M	$259M	19%	Trend valid — reservation-dominant
Montana	Res.-dominant	$230M	$50M	$280M	18%	Trend valid — reservation-dominant
Minnesota	Urban-dominant	$190M	$106M	$296M	36%	Apparent decline is scope artifact
Wisconsin	Mixed	$170M	$63M	$233M	27%	Moderate Pop. B effect
Michigan	Mixed	$180M	$79M	$259M	30%	Real IHS growth + Pop. B component
North Carolina	Mixed	$130M	$179M	$309M	58%	Pop. B dominates — high distortion
North Dakota	Res.-dominant	$140M	$27M	$167M	16%	Trend valid — reservation-dominant
New York	Urban-dominant	Not in paper	$141M+	$141M+	~100%	Absent from paper; Pop. B only
TOTAL (15 states)		$4.59B	$1.88B	~$6.47B	~29%	Paper understates by ~$1.88B for these states

Reading the Three State Categories

Walk through the three categories explicitly. Students need a heuristic for which states’ Data Match figures can be used as trend indicators and which need Population B correction:

Category	States	Can you use 2023 paper figures as trend indicators?	Key analytical move required
Reservation-dominant (IHS penetration >65%)	AZ, AK, SD, MT, ND, NM	Yes — Population B is small; cross-dataset trends are largely valid	Cite paper figures directly; note they exclude a modest Population B component
Mixed (IHS penetration 40-60%)	OK, WA, WI, MI, NC	With caution — interpret growth and decline signals carefully	Distinguish real policy-driven change (OK expansion) from scope-driven patterns; note Population B share explicitly
Urban-dominant (IHS penetration <40%)	MN, CA, OR, NY	No — cross-dataset comparisons are most misleading here	Add Population B estimate before citing trends; note that paper figures significantly understate total AIAN Medicaid for these states

3.2 — PRC Substitution: What It Covers and What It Does Not (35-55 min)

The PRC Substitution Mechanism

The Purchased/Referred Care (PRC) program is the IHS budget line that pays for care that IHS facilities cannot provide in-house — specialist referrals, hospital admissions, and services outside the geographic reach of IHS clinics. PRC is chronically underfunded; many IHS areas have exhausted PRC funds before the end of the fiscal year, at which point referrals for non-emergency services are effectively rationed.

Medicaid provides a partial solution to this problem. When an IHS-registered AI/AN patient who is Medicaid-enrolled is referred to an external provider, Medicaid can cover that referral cost — sparing PRC from having to pay. The Data Match analysis estimates that $2.6-4.0 billion in IHS-access Medicaid spending absorbs referral costs that would otherwise fall on finite PRC appropriations. This is the PRC substitution mechanism.

PRC SUBSTITUTION: THE $2.6-4.0B ESTIMATE

For IHS-access Medicaid enrollees (Population A), Medicaid’s ability to cover referral and external care costs means that PRC appropriations can be reserved for uninsured patients and care situations not coverable by Medicaid. The $2.6-4.0B estimate represents the portion of Population A Medicaid spending that flows to non-IHS providers — external hospitals, specialists, and long-term care — that would otherwise require PRC funding. Without Medicaid coverage for this population, PRC would face an impossible funding shortfall.

What PRC Substitution Does NOT Cover

READ ALOUD  /  ADAPT

“Here is the part of the PRC substitution story that the 2023 paper does not address — and that the Population B analysis makes visible. PRC substitution is an IHS-system-specific mechanism. It only functions for Population A — people who are IHS-registered and Medicaid-enrolled. When a Population A patient needs a referral, Medicaid can cover it, sparing PRC. When the same patient is uninsured and needs that referral, PRC has to cover it. That is the substitution. But Population B — urban AI/AN Medicaid enrollees with no IHS registration — never touches PRC at all. Their care is fully externalized from the IHS system from day one. Their Medicaid coverage pays for mainstream provider care directly. There is no PRC buffer for them to draw on; there is no substitution mechanism operating on their behalf. They are simply Medicaid enrollees like any other low-income individual. This has a specific policy implication: the PRC substitution mechanism understates the total value of Medicaid to AI/AN communities, because it only measures the fiscal relief that Medicaid provides to the IHS system. It does not account for the direct Medicaid-funded access to care that Population B receives outside the IHS system entirely. A complete measure of Medicaid’s value to AI/AN health would add both.”

FISCAL RISK: WHAT ENROLLMENT CONTRACTION SCENARIOS MEAN

The 2023 paper models two scenarios: 10% Population A enrollment contraction: Shifts ~$650M toward PRC or uncompensated care. Specifically, the portion of current Medicaid spending that covers IHS referrals and external care for those 94,000 enrollees would have to be absorbed by PRC appropriations or left as uncompensated care. 20% Population A enrollment contraction: Shifts ~$1.3B. At current PRC appropriation levels (approximately $1B/year nationally), a $1.3B shift would be catastrophic for IHS referral capacity. Population B adds a separate risk layer: Urban AI/AN Medicaid enrollees (Population B) are generally more vulnerable to eligibility policy changes — ACA rollback, work requirements, redetermination processes — than reservation-resident enrollees, who have more stable eligibility pathways through IHS registration. A policy change targeting urban Medicaid populations could reduce Population B spending significantly without appearing in IHS-access enrollment statistics at all. This risk is invisible to the IHS-only analytical framework.

3.3 — The Measurement Gap and What It Means for Policy (55-70 min)

The 11-Year Hole

The historical AIAN Medicaid payment files end in 2012 (with 2008 as the last fully reliable national total year). The 2023 Data Match paper uses a different population definition and was published in 2023-2025. No source currently provides a validated, all-AIAN Medicaid payment series for 2013-2022.

This 11-year gap means:

• The impact of the ACA Medicaid expansion (2014) on AI/AN enrollment and payments has not been systematically documented in a population-consistent series
• COVID-19’s impact on AIAN Medicaid — including the continuous enrollment provision (2020-2023) and its subsequent unwinding — cannot be quantified against a historical baseline for the same population
• State-level fiscal planning for AI/AN Medicaid lacks a 25-year trend series that policymakers can use to model future exposure
• Any interpolated estimates for those years are visualization aids, not validated data

What Filling the Gap Would Require

The T-MSIS system (which replaced MAX/MSIS) contains Medicaid enrollment and claims data for the gap years. Extracting and reconciling an all-AIAN payment series from T-MSIS would require:

• VRDC access and a data use agreement specifically covering the 2013-2022 T-MSIS files
• Methodology for identifying AI/AN Medicaid enrollees in T-MSIS consistent with both the historical all-AIAN approach (race coding) and the Data Match approach (IHS registration linkage)
• Population B estimation or direct measurement for the gap years
• Validation against the 2012 historical endpoint and the 2023 Data Match figures

This is the single highest-value data infrastructure investment for longitudinal AI/AN Medicaid analysis. It would close the gap between what we know about 2023 and what we know about the trajectory that produced 2023.

INSTRUCTOR NOTE: CONNECTING TO RESEARCH CAREERS

For students considering health services research or health policy analysis careers, this is a direct pointer to a genuinely open research problem. The T-MSIS gap-fill work described here has not been done. A dissertation chapter, a fellowship project, or a collaborative research proposal with NIHB could produce this analysis. Surface this explicitly: ‘This is not a hypothetical future research agenda. This is an identified gap in the data that exists right now and that a motivated researcher with VRDC access could address.’

DATA MEMO ASSIGNMENT  ·  Due End of Week 10 Three-State Population A/B Comparison Launched and scaffolded in Session 3 (Week 10)

3.4 — Data Memo Assignment: Launch and Scaffolding (70-90 min)

The data memo asks students to apply the Population A vs. Population B analytical framework to three states of their choosing. It is a 4-6 page analytical memo written for a policy audience. Walk through the full assignment, then spend 10 minutes of in-class scaffolding to ensure students leave with a working framework.

Component	Description	Length	Key Analytical Move
State selection and classification	Select three states: one reservation-dominant, one urban-dominant, and one mixed. Justify the classification using IHS penetration estimates and population composition.	1 page	Students must demonstrate understanding of the state category framework before analyzing numbers
Population A analysis for each state	Report the 2023 Data Match figure for each state. Explain what it measures and what it excludes. Identify whether the state’s Data Match figure is a reliable trend indicator or requires Population B correction.	1-1.5 pages	Students apply the scope framework learned in Sessions 1-3
Population B estimation and total	Using the methodology from the Revised Analysis (ACS population x 47% Medicaid rate x [1-IHS penetration] x $5,500 PMPY x 0.90), estimate Population B spending for each state. Calculate estimated total AIAN Medicaid for each state.	1-1.5 pages	Students use the estimation formula; must show their work
Policy implications	For each state: what does the Population A/B gap mean for advocacy? Who is missing from the IHS-access data, and what policy arguments are affected by their absence? What would a complete data picture change?	1-1.5 pages	Tests analytical rather than descriptive comprehension
Data infrastructure recommendation	Identify one specific data gap or methodological improvement that would produce a more complete picture for your three states. Connect to the T-MSIS gap and/or small-cell suppression tradeoffs.	0.5 page	Tests understanding of the measurement gap material from Section 3.3

GRADING NOTE FOR INSTRUCTOR

The data memo is graded primarily on analytical precision, not rhetorical sophistication. Strong memos will: (1) correctly classify all three states, (2) accurately apply the Population B formula and show the calculation, (3) correctly identify which states’ trend lines can be read directly from the paper and which require correction, (4) articulate a specific policy consequence of the Population B gap rather than a generic statement about undercounting. Weak memos will describe what Population A and B are without applying the distinction to the specific states chosen. The scaffolding exercise in class is designed to prevent the latter.

Indian Health Financing  ·  Research & Methods Track  ·  Weeks 8–10 Lecture Notes  ·  February 2026

  Weeks 11–12 

INDIAN HEALTH FINANCING

All Tracks (Policy emphasis)

Weeks 11–12 Instructor Lecture Notes & Script

MEASURING ADEQUACY — THE FULL-FUNDING FRAMEWORK

The question is not simply whether federal investment has increased. The analytical question is whether available resources are sufficient to finance the services and infrastructure required to meet the health needs of the population being considered. Use the Full Funding paper as a case study in constructing an adequacy estimate: NEED → AVAILABLE FEDERAL INVESTMENT → ESTIMATED GAP. Students should distinguish total federal investment from full funding. The two are not equivalent concepts.

From Findings to Policy: Medicaid as Structural Pillar and the Fiscal Risk of Contraction

Track	All tracks; policy and operations emphasis
Sessions	Two 90-minute sessions spanning Weeks 11 and 12
Core Question	What do the Data Match findings mean for Indian health policy — specifically, what is the fiscal risk to Indian health programs if Medicaid contracts, and how should that risk be communicated to policymakers?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. V–VI); Revised Analysis 1999–2023 (Sec. III); Lecture Series Part 3 (Sec. 3.4–3.5); IHS Budget Justifications; CMS Investment Estimates Paper (ACS 5-Year PUMS 2024 × NHE 2024 × Data Match 2023): Tables 1, 3, and 6 for coverage distribution, spending scenarios, and total federal investment framework; NIHB 2022 Budget Formulation (mandatory funding brief; full-funding estimate and glide path); TTAG (2015) Data Symposium Report
Builds On	Weeks 8–10: Data Match findings; PMPY modeling; Population A/B reconciliation; data memo
Leads Into	Weeks 13–14: Tribal data sovereignty; NIHB-led research model; implications for next-generation data infrastructure

Course Arc Note

These two sessions are the policy payoff of the quantitative work in Weeks 8–10. Students arrive knowing the Data Match findings in detail. They now need to translate those findings into policy arguments — specifically, arguments about what Medicaid contraction would mean for Indian health programs that have built their financial models around Medicaid revenue. The PRC substitution analysis is the analytical bridge between the enrollment figures and the fiscal risk argument. Instructors should resist the temptation to treat these sessions as advocacy training; the goal is analytical fluency in constructing and evaluating fiscal risk arguments, not producing advocates.

Session	Core Argument	Key Deliverable
Session 1 — Week 11	Medicaid has evolved from a revenue supplement to a structural pillar of Indian health financing. The historical trajectory from 13% IHS/Tribal billing share (1999) to 21% (2008) to the 2023 Data Match findings reflects a structural shift with no policy reversal. The $8.45 billion total federal investment framework reframes the policy conversation.	Total federal investment table; structural pillar argument construction
Session 2 — Week 12	PRC substitution is the mechanism through which Medicaid contraction becomes an IHS program threat, not merely an insurance coverage issue. The fiscal risk scenarios — 10%, 20%, and 30% contraction — quantify the exposure in terms directly comparable to IHS appropriations. The geographic concentration of risk amplifies the impact on specific states and tribal programs.	Fiscal risk scenario exercise; advocacy memo draft (track-differentiated)

SESSION 1  ·  90 minutes — Week 11

Medicaid as Structural Pillar: The Policy Argument the Data Makes Possible

Total federal investment framework · historical trajectory · the structural pillar argument · what $8.45 billion means · advocacy applications and analytical risks

Session 1 Timing Overview

Time	Segment	Format	Notes
0–10	Opening: from measurement to argument	Mini-lecture	Bridge from data to policy
10–32	Section 1.1: The structural pillar argument	Lecture + table	Historical arc; total investment framework
32–52	Section 1.2: What $8.45 billion means — and what it does not	Lecture	Scope, caveats, and analytic risks
52–68	Section 1.3: Advocacy applications exercise	Small groups	Constructing arguments from data
68–80	Section 1.4: Common misuses and how to correct them	Full group	Analytical discipline
80–90	Bridge to Session 2	Mini-lecture	Introduce PRC substitution

1.0 — Opening (0–10 min)

READ ALOUD / ADAPT

“You have spent three weeks learning what the Data Match found. 941,168 IHS-access Medicaid enrollees. $6.4 to $6.6 billion in associated expenditures. A 1999–2008 historical trajectory showing the IHS/Tribal billing share doubling from 13% to 21%. A Population A/B reconciliation showing another $1.88 billion flowing to non-IHS-registered AI/AN Medicaid patients in 15 states. Now the question is: what does all of that mean for policy? In these two sessions we are going to build the policy argument that the data supports — and we are going to be equally rigorous about what the data does not support, because in policy advocacy, overreach destroys credibility faster than silence.

1.1 — The Structural Pillar Argument (10–32 min)

The central policy claim that the Data Match supports is this: Medicaid has evolved from a supplementary revenue source into a structural pillar of Indian health system (or non-system) financing. The word ‘structural’ is doing significant analytical work in that sentence and needs to be unpacked carefully.

Key Concept

What ‘structural’ means in this context: A revenue source is structural when the health system has built its capacity, staffing, facilities, and service commitments around that revenue — when a significant reduction in that revenue would require not just budget cuts but program elimination, facility closure, or workforce reduction. By 2023, Indian health programs have been billing Medicaid for decades, hiring staff whose salaries depend on Medicaid revenue, and expanding service capacity on the assumption that Medicaid revenue will continue. That is what makes Medicaid structural rather than supplemental.

Historical Marker	What It Shows
1999: IHS/Tribal billing share ~13%	Medicaid present but secondary; most AI/AN Medicaid spending flows through non-IHS providers
1999–2008 CAGR of 11.5%	AI/AN Medicaid payments growing faster than IHS appropriations and faster than general Medicaid trend; structural shift underway
2008: IHS/Tribal billing share ~21%	IHS and tribal facilities now account for more than one-fifth of all AI/AN Medicaid spending; Medicaid central to program finance
2023: $6.4–6.6B IHS-access spending	IHS-access Medicaid spending approaches total IHS appropriations ($8.6B); structural parity reached
2023: $8.45B total AI/AN Medicaid (est.)	Combined IHS + Medicaid + Medicare federal investment approximately $19.3–19.9B (midpoint $19.6B); Medicaid now comparable in magnitude to IHS appropriations; Medicare adds an estimated $2.64B (ACS-derived, 260,131 IHS-access enrollees)

Core Finding

The total federal investment framework: The 2023 Data Match enables a complete accounting of federal investment in AI/AN health for the first time. IHS appropriations (~$8.6B) + AI/AN Medicaid (~$8.45B total; ~$6.4–6.6B IHS-access + ~$1.88B Population B) + Medicare (~$2.64B; midpoint of $2.52–$2.88B range based on 260,131 ACS-estimated IHS-access Medicare enrollees × 2024 NHE Medicare PMPY) + other federal sources = approximately $19.3–19.9B (midpoint $19.6B) in total federal health investment for AI/AN people annually. Medicaid alone is approximately equal to IHS appropriations. This reframes the policy conversation: the federal government’s investment in AI/AN health through Medicaid is as large as its investment through IHS, yet Medicaid’s continuation is not protected by the trust responsibility. New integrated CMS figure: Combining all three coverage groups within the IHS-access population (dual eligible, Medicare-only, Medicaid-only), total CMS investment is estimated at $9.23B–$12.49B annually (low–high scenario), with a midpoint of $10.86B. This $10.86B figure is the first published estimate of combined Medicare and Medicaid investment in the IHS-access population specifically. Note the population it excludes: 778,438 IHS-access individuals (51.1%) have no CMS coverage and generate zero CMS expenditure; their care costs fall entirely on IHS appropriations and PRC. The $10.86B figure captures the 744,579 CMS-covered individuals; the majority of the IHS-access population remains invisible to it.

Instructor Note — ACS vs. Data Match Reconciliation (for advanced discussion or graduate sections): A cross-tabulation of the 2024 ACS 5-Year PUMS yields 587,313 IHS-access Medicaid enrollees — 60% below the Data Match count of 941,168. This gap is expected and methodologically explicable: the ACS undercounts IHS coverage by approximately 41% relative to administrative records (Census Bureau linkage study); the ACS is a point-in-time survey while the Data Match counts anyone ever enrolled during the year; and the ACS pools five survey years (2020–2024) including peak COVID continuous enrollment. The ACS Medicaid-only spending estimate ($5.35B midpoint at 2024 NHE PMPY) is 19% below the Data Match $6.4–6.6B figure, consistent with the enrollment undercount. Key teaching point: neither figure is wrong — they are measuring related but distinct things. The Data Match is the authoritative count for IHS-access Medicaid enrollment; the ACS is the only available source for Medicare enrollment and for the 51.1% with no CMS coverage. Both are needed for a complete investment picture. Source: ACS 5-Year PUMS 2024 × 2024 NHE PMPY; NIHB CMS–IHS Data Match 2023.

1.2 — What $8.45 Billion Means — and What It Does Not (32–52 min)

Analytical Warning

Scope discipline is essential in policy advocacy: The $8.45B figure is an estimate combining the verified 2023 Data Match IHS-access figure ($6.4–6.6B) with a 15-state reconciliation estimate of Population B spending ($1.88B). It is not a CMS-certified national total; it is an analytical estimate. Advocates who cite it as an audited national figure will be challenged and will lose credibility. The correct framing is: ‘The best available evidence suggests total AI/AN Medicaid spending is approximately $8.45 billion, combining the verified 2023 Data Match IHS-access figure with a 15-state reconciliation of non-IHS-registered enrollees.’

What the $8.45B Framework Supports	What It Does Not Support
Medicaid is approximately equal to IHS appropriations in magnitude	That Medicaid was designed as a trust-responsibility instrument (it was not)
A significant Medicaid reduction would have IHS-scale fiscal consequences	That every dollar of Medicaid reduction translates directly to IHS program loss (substitution is partial)
The federal government’s total investment in AI/AN health is approximately $19.3–19.9B (IHS $8.6B + Medicaid ~$8.45B + Medicare ~$2.64B ACS-derived midpoint + other)	That $17–20B is sufficient to meet the trust responsibility (it is not)
IHS programs have built their financial models around Medicaid revenue	That IHS programs can recover from Medicaid contraction through appropriations increases alone
Geographic concentration of AI/AN Medicaid makes certain states especially exposed	That all states face equivalent risk (five states account for >50% of IHS-access spending)

1.3 — Advocacy Applications Exercise (52–68 min)

EXERCISE: POLICY ARGUMENT CONSTRUCTION  |  16 minutes

Format: Groups of 3–4. Each group is assigned one of the following policy contexts. Using the Data Match findings and the structural pillar framework, construct a 3-minute verbal policy argument for the assigned audience. Groups must: (1) cite at least two specific data points with correct scope framing; (2) identify the single most important analytical risk their argument faces; (3) explain how they would respond to a critic who questions their primary figure. Assigned contexts: Group A: Congressional testimony to the Senate Committee on Indian Affairs opposing a block grant proposal for MedicaidGroup B: Briefing to a Tribal council considering whether to invest in expanded Medicaid outreach and enrollmentGroup C: Response to a state Medicaid director proposing to reduce IHS facility reimbursement rates by 10%Group D: Op-ed for a tribal newspaper explaining why proposed federal Medicaid cuts threaten IHS program viability   Debrief: The most common error across all groups is treating the $6.4–6.6B figure as total AI/AN Medicaid spending. Correct framing every time it appears. The second most common error is claiming that Medicaid cuts would directly eliminate IHS services dollar-for-dollar — the substitution relationship is partial, and programs have some (limited) adaptive capacity.

SESSION 2  ·  90 minutes — Week 12

PRC Substitution and Fiscal Risk: Quantifying the Exposure

PRC substitution mechanism · $2.6–4.0B estimate · fiscal risk scenarios · geographic concentration · state typology · policy responses

Session 2 Timing Overview

Time	Segment	Format	Notes
0–10	Bridge: from magnitude to mechanism	Mini-lecture	Connect structural pillar to PRC substitution
10–32	Section 2.1: The PRC substitution mechanism	Lecture + table	How Medicaid absorbs PRC costs
32–52	Section 2.2: Fiscal risk scenarios	Lecture	10/20/30% contraction; geographic concentration
52–68	Section 2.3: State typology and differential exposure	Lecture + table	Reservation-dominant vs. mixed vs. urban-dominant
68–82	Section 2.4: Fiscal risk scenario exercise	Small groups	Quantify exposure for assigned state type
82–90	Bridge to Weeks 13–14	Mini-lecture	Why governance matters for the next generation of data

2.0 — Bridge: From Magnitude to Mechanism (0–10 min)

READ ALOUD / ADAPT

“Last session established the magnitude: Medicaid is approximately equal to IHS appropriations in total federal investment terms. Today we establish the mechanism: specifically, how does a Medicaid reduction translate into an IHS program threat? The answer runs through Purchased/Referred Care. PRC is the IHS program that pays for services IHS cannot provide directly — specialty referrals, hospital care beyond IHS capacity, off-site services. When an IHS-enrolled patient with Medicaid needs a specialist, Medicaid pays. When the same patient does not have Medicaid, the cost falls on PRC — or the service is denied. This substitution relationship is the financial mechanism through which Medicaid contraction becomes an IHS budget crisis, not just a coverage access problem.”

2.1 — The PRC Substitution Mechanism (10–32 min)

Key Concept

How PRC substitution works: IHS-enrolled patients who are Medicaid-enrolled can have their referred/purchased care costs billed to Medicaid rather than drawn from the PRC budget. When Medicaid pays for these services, the PRC budget is not consumed — it is effectively substituted. The 2023 Data Match estimates that $2.6 to $4.0 billion in annual Medicaid expenditures for IHS-access enrollees consists of services that would otherwise require PRC funding. This is the single most policy-consequential finding in the Data Match for IHS program stability analysis.

PRC Substitution Element	Details
Substitution estimate	$2.6–4.0 billion annually (Population A only; range reflects methodological variation)
Scope	Applies only to IHS-registered Medicaid enrollees (Population A); Population B has no PRC relationship
Mechanism	For each IHS-enrolled patient with Medicaid, Medicaid pays for referred/specialty care; without Medicaid, this cost would fall on PRC or be denied
PRC budget context	Annual PRC appropriation is approximately $1.0–1.2 billion; the substitution estimate of $2.6–4.0B exceeds PRC’s entire annual budget by 2x–3x
Implication	If Medicaid enrollment declined significantly, PRC would face demand it could not fund at any plausible appropriation level; services would be denied
Advocacy framing	A 20% Medicaid contraction affecting IHS-access enrollees would shift $520M–$800M in PRC-equivalent costs onto an already-underfunded PRC budget; this is not an insurance coverage problem, it is an IHS budget crisis

2.2 — Fiscal Risk Scenarios (32–52 min)

The fiscal risk scenarios translate the substitution estimate into concrete budget impact numbers at different levels of Medicaid contraction. These scenarios are useful for advocacy precisely because they express the risk in terms comparable to the IHS annual appropriation.

Fiscal Risk Scenario

Baseline: $6.4–6.6B in IHS-access Medicaid expenditures (Population A). Of this, $2.6–4.0B is estimated as PRC-substituting. IHS annual appropriation: ~$8.6B. PRC annual appropriation: ~$1.0–1.2B. 10% contraction: $640M–$660M reduction in IHS-access Medicaid spending. Of this, $260M–$400M shifts to PRC demand. This is 22–33% of PRC’s entire annual budget. As a share of total IHS appropriations: ~7–8%. 20% contraction: $1.28B–1.32B reduction. PRC-shifting portion: $520M–$800M. This equals or exceeds PRC’s entire annual budget. As a share of IHS appropriations: ~15%. 30% contraction: $1.92B–1.98B reduction. PRC-shifting portion: $780M–1.2B. PRC would face more than its entire annual budget in new demand from this source alone. Service denial at scale would be unavoidable. As a share of IHS appropriations: ~22–23%.

Instructor Note

These scenarios apply only to Population A (IHS-access enrollees). Population B contraction would reduce AI/AN Medicaid coverage further but would not add PRC demand, since Population B patients are not IHS-registered. The combined effect of a broad Medicaid contraction on all AI/AN patients — including coverage loss, reduced access, and increased uncompensated care — is larger than the PRC substitution analysis captures. But the PRC mechanism is the most concrete link from Medicaid policy to IHS program stability. FYE note for advanced students: The 941,168 Data Match figure is an ‘ever-enrolled’ count — anyone enrolled for at least one month in 2023. The full-year equivalent (FYE) denominator, used in per-enrollee spending rate calculations, is smaller. National Medicaid FYE churn is approximately 4.5% below ever-enrolled; AI/AN churn is likely 10–15% higher due to reservation economies and redetermination disruptions. FYE-adjusted denominators: approximately 847,000–899,000. The fiscal risk scenarios use the reported 941,168 figure, which is appropriate for policy contexts. Analysts computing per-enrollee rates should use the FYE-adjusted denominator and cite the implied PMPY (~$6,906) as evidence of partial-year enrollment compression, not as evidence of lower utilization.

2.3 — State Typology and Differential Exposure (52–68 min)

Geographic concentration of IHS-access Medicaid spending means that Medicaid contraction risk is distributed unequally across states and tribal programs. The three-state typology developed for cross-dataset comparability also describes differential fiscal risk exposure.

State Type	Characteristics	Fiscal Risk Profile
Reservation-dominant (e.g., AZ, NM, SD, MT)	Large IHS-registered AI/AN population; high Medicaid penetration among IHS patients; IHS and tribal facilities are primary providers	Highest direct exposure: large PRC substitution volumes; tribal health programs with high Medicaid revenue dependence; 10–20% contraction translates to immediate program-level budget crises
Mixed (e.g., OK, MN, WA)	Significant reservation and urban populations; both IHS facility use and mainstream provider use common; varies by tribe and region	Moderate exposure: IHS programs face substitution risk; urban Indian organizations face coverage access risk; differential impact by program type within state
Urban-dominant (e.g., CA, TX, IL)	AI/AN population primarily urban; limited IHS facility presence; Urban Indian Organizations as primary I/T/U component; most AI/AN Medicaid patients are Population B	Lower PRC substitution risk (Population B has no PRC relationship) but significant coverage access risk; Urban Indian Organization revenue at risk from standard Medicaid rate reductions

Bridge to Weeks 13–14 (82–90 min)

Instructor Note

Close by connecting fiscal risk analysis back to data governance. The fiscal risk scenarios in these two sessions are only possible because the 2023 Data Match produced verified, policy-grade enrollment and expenditure data for the first time. The next two weeks ask: what governance decisions made that data trustworthy? And what does the NIHB-led research model mean for the next generation of data infrastructure that will be needed to track the policy changes these two sessions have been analyzing?

Track-Differentiated Discussion Protocol

Discussion Questions

Q1 (All tracks): The PRC substitution estimate suggests that $2.6–4.0 billion in Medicaid spending effectively funds services that would otherwise cost IHS. But IHS’s PRC budget is only $1.0–1.2 billion annually. What does this arithmetic mean for the IHS’s ability to absorb a significant Medicaid reduction? What would happen to patients? Q2 (Policy & History): The trust responsibility creates a federal obligation to provide health care to AI/AN people. If Medicaid contraction reduces Indian health program capacity, does this create a trust responsibility claim against the federal government? What would such a claim look like legally? Q3 (Research & Methods): The fiscal risk scenarios use a simple proportional reduction model — a 15% Medicaid cut = a 15% reduction in IHS-access spending. What are the limitations of this model? What factors would cause the actual impact to be larger or smaller than the proportional estimate? Q4 (Operations & Administration): You are the administrator of a tribal health program that derives 40% of its operating revenue from Medicaid. Design a 3-year contingency plan for a 20% Medicaid reduction. What are your first three actions in year one?

Key Data Points for Student Reference

Fact	Value / Note
IHS/Tribal facility Medicaid billing share (1999)	~13% of total AI/AN Medicaid payments
IHS/Tribal facility Medicaid billing share (2008)	~21% of total AI/AN Medicaid payments
1999–2008 AI/AN Medicaid CAGR	~11.5%
2023 Data Match: IHS-access Medicaid spending	~$6.4–6.6B (Population A; PMPY modeled)
15-state Population B estimate	~$1.88B additional (non-IHS-registered AI/AN Medicaid)
Total AI/AN Medicaid estimate	~$8.45B (combined Population A + B)
IHS FY2023 appropriations	~$8.6B
PRC annual appropriation	~$1.0–1.2B
PRC substitution estimate	$2.6–4.0B annually (Population A only)
10% Medicaid contraction impact	~$640M–$660M IHS-access spending reduction; ~$260M–$400M PRC demand increase
20% contraction impact	~$1.28B–1.32B reduction; equals or exceeds full PRC annual budget in new demand
Total federal AI/AN health investment estimate	~$19.3–19.9B (midpoint $19.6B); components: IHS ~$8.6B + Medicaid ~$8.45B (Pop. A $6.4–6.6B + Pop. B $1.88B) + Medicare ~$2.52–2.88B (midpoint $2.64B, ACS-derived: 260,131 IHS-access enrollees × 2024 NHE PMPY ±15%) + other federal (CHIP, VA, BIA not estimated)

Indian Health Financing  ·  All Tracks  ·  Weeks 11–12 Lecture Notes  ·  March 2026

  Weeks 13–14 

INDIAN HEALTH FINANCING

Research & Methods Track

Weeks 13–14 Instructor Lecture Notes & Script

Tribal-Led Research: What the Model Produced, What It Cost, and What Comes Next

Element	Detail
Track	Research & Methods (capstone content; appropriate for all tracks)
Sessions	Two 90-minute sessions spanning Weeks 13 and 14
Core Question	What did the NIHB-led model produce that a federal-agency-led project would not have — and what are the honest limitations and unresolved tensions that should inform the next generation of Tribal health data work?
Primary Sources	CMS–IHS Data Match Lecture Paper (Sec. V–VII); Lecture Series Part 3 (Sec. 3.6-3.8); Annotated Bibliography (Fox 2003, Fox & Boerner 2009, Fox 2011, NIHB 2008-2019 comparison, NIHB 2024 retrospective); Professor Introductions Weeks 13–14; TTAG (2015) Data Symposium Report
Builds On	Weeks 5–7 governance framework; Weeks 8–10 findings; Weeks 11–12 policy implications
Leads Into	Week 15 synthesis and final paper presentations
Final Paper	Due Week 15; students should be deep in drafting during Weeks 13–14

These final two content sessions close the methodological arc of the course. Session 1 evaluates the NIHB-led research model honestly: what it produced, what it could not have produced under federal-agency leadership, and what its real limitations and costs were. Session 2 turns to the future: what the next phase of Tribal health data infrastructure should look like, how the ACS and administrative data fit together going forward, and what the 20-year arc from Fox (2003) to the 2023 Data Match reveals about the relationship between data, power, and sovereignty in Indian Country.

Box colors: legacy and significance (green), lessons-learned (purple), future directions (dark blue), limitations and costs (dark red), sovereignty tensions (brown), instructor notes (amber), warnings (orange), exercises (teal).

Two-Session Architecture

Session	Week	Core Argument	Central Activity
Session 1	Week 13	The NIHB-led model produced research that was substantively different from what CMS or IHS would have produced alone — not because NIHB was more competent, but because it was accountable to a different principal. Understanding that accountability structure is the key to understanding why the model worked and why it should be the template for what comes next.	Structured comparison exercise: how would the project have differed under three leadership scenarios (NIHB, CMS, IHS)
Session 2	Week 14	The 20-year arc from Fox 2003 to the 2023 Data Match is a story about the gradual development of Tribal analytic capacity — and about the unresolved tension between the precision that administrative data provides and the community autonomy that governs its use. The next phase must grapple with that tension honestly, not dissolve it.	The arc-of-the-story synthesis discussion; future directions mapping exercise; final paper check-in

WEEKS 13-14 IN THE COURSE ARC

Students enter these sessions with three things: ten weeks of analytical framework, a nearly-complete or draft final paper, and a growing ability to hold complexity without resolving it prematurely. The instructor’s job in these sessions is to honor that complexity — not to deliver a triumphalist narrative about the Data Match as a success story, but to give students the tools to evaluate it as a research model with genuine strengths, genuine costs, and genuine unresolved questions. The final activity of Session 2 — mapping what the next generation of Tribal health data infrastructure should look like — is designed to be generative rather than conclusive. Students should leave with open questions they want to answer, not closed answers they have memorized.

SESSION 1  ·  90 minutes — Week 13 What the Tribal-Led Model Produced — and What It Cost NIHB’s structural advantages · what federal-agency leadership would have produced differently · real limitations and costs · the precedent value

Session 1 Timing Overview

0-10 min

Opening

The accountability question Script: who the research is accountable to determines what it produces

10-35 min

Section 1.1

What NIHB’s leadership produced: five specific outputs Systematic comparison of what the project delivered vs. what a federal-agency project would have delivered

35-55 min

Section 1.2

The real limitations and costs of the NIHB model Honest accounting: resource constraints, small-cell suppression tradeoffs, the Tribal-level data gap

55-75 min

Section 1.3

Leadership scenario comparison exercise Groups analyze how the project would have differed under CMS, IHS, and NIHB leadership respectively

75-90 min

Section 1.4

Precedent value and the template question What this model offers — and what it requires — for future Tribal health data initiatives

1.0 — Opening (0-10 min)

READ ALOUD  /  ADAPT

“We have now covered 12 weeks of this course. You know what the Data Match found, how it was built, what the findings mean for policy, and where the system is vulnerable. Today and next week I want to step back and ask a different kind of question — not what the project found, but what the project was. In Week 5, I argued that NIHB’s role as lead research organization was itself a substantive methodological choice — not a governance formality. I want to develop that argument fully now, in light of everything we have learned. Because the question of who leads a research project, and what that leader is accountable to, turns out to be one of the most consequential methodological decisions in the entire enterprise. Research organizations are accountable to their principals — the parties they answer to. A federal agency is accountable to Congress, to the administration in power, to OMB, to its inspector general. A Tribal advocacy organization is accountable to the Tribal nations it represents, to its board, and to the communities whose data it is stewarding. Those accountability structures produce different research — different questions, different interpretations, different safeguards, different relationships between the researcher and the researched. That is the lens for today. Not ‘NIHB did a good job’ but ‘NIHB’s accountability structure produced specific outcomes that a different accountability structure would not have produced.’ And then: what did that model cost, and was it worth the cost?”

1.1 — What NIHB’s Leadership Produced: Five Specific Outputs (10-35 min)

Work through each output as a claim that needs to be defended, not asserted. For each one, ask: could a CMS-led or IHS-led project have produced this? Why or why not?

Output	What it was	Why NIHB’s accountability structure produced it	Could a federal-agency project have produced this?
1. Research agenda driven by Tribal health policy questions	The core questions — how many AI/AN individuals have verified IHS access and Medicaid enrollment, and what does that cost — were Tribal advocacy questions, not federal program administration questions.	NIHB is accountable to Tribal nations whose primary interest is documenting the fiscal case for IHS funding adequacy and Medicaid protection. Those are the questions it asked.	CMS would have asked: are Medicaid enrollment claims accurate? Is fraud and abuse occurring? IHS would have asked: how can we improve billing rates? Neither question produces the population-level enrollment count that Tribal advocacy required.
2. Tribal community legitimacy for findings	The 941,168 enrollment figure and $6.4-6.6B estimate were immediately usable in Tribal advocacy contexts because they came from an organization that Tribal nations trust and that had kept Tribal communities informed throughout.	NIHB’s Tribal Advisory Committee structure meant findings were reviewed by Tribal representatives before any external release. The research was not delivered to communities — it was co-produced with them.	A CMS-published finding with the same numbers would have been received with more skepticism by Tribal advocates and might have been challenged on grounds that federal framing shaped the analysis. The NIHB imprimatur conferred credibility that CMS could not have provided.
3. Pre-publication Tribal review and interpretation authority	Tribal Advisory Committee members reviewed findings before external release and had meaningful input into how results were framed and contextualized.	This right was negotiated as part of the governance structure because NIHB, as a Tribal organization, understood why it mattered and had the leverage and motivation to secure it.	A federal-agency project might have consulted Tribal stakeholders, but pre-publication review rights with genuine authority to delay or redirect findings are rare in federal research processes. Tribes would have been notified, not co-owners.
4. Protection against adversarial data use	DUA terms explicitly prohibited use of matched data for enrollment audits, eligibility challenges, or any enforcement purpose. Data could only be used for health coverage research and planning.	NIHB negotiated these protections because its principals — Tribal nations — have specific historical reasons to fear federal data being used against their interests. NIHB had strong motivation to make the prohibitions explicit and enforceable.	A CMS-led project would have had standard data use restrictions, but the specific prohibition on enrollment audits and eligibility enforcement as prohibited purposes reflects a Tribal-specific concern that CMS had less incentive to prioritize.
5. Sovereign governance precedent	The model — Tribal organization in the lead, federal agencies as data partners, Tribal Advisory Committee with real governance authority — set a documented precedent for how future Tribal health data initiatives should be structured.	Because NIHB led, it could design the governance structure in ways that reflected Tribal self-determination principles. A federal-agency-led project would have been structured around federal governance norms.	A CMS- or IHS-led project might have produced a technically similar data linkage, but the governance precedent it set would have been federal-centric. That precedent has compounding effects on every future initiative that references it.

THE ACCOUNTABILITY ARGUMENT IN ONE SENTENCE

The CMS–IHS Data Match produced what it produced not just because the technical work was well done, but because the organization leading it was accountable to the communities whose data was being used — and that accountability shaped every question asked, every protection negotiated, and every decision about what to publish and how to frame it.

1.2 — The Real Limitations and Costs of the NIHB Model (35-55 min)

An intellectually honest evaluation of the NIHB model requires acknowledging its real limitations and the real costs it imposed. This section is not a criticism of the project — it is a necessary part of using it as a template. A template you cannot criticize is a template you cannot improve.

FOUR GENUINE LIMITATIONS OF THE NIHB-LED MODEL

Limitation 1: Resource constraints on the lead organization NIHB is a nonprofit advocacy organization, not a research institution with large permanent data infrastructure. The Data Match required sustained technical capacity over multiple years — VRDC access, data cleaning, matching algorithm development, validation. This is resource-intensive work that stretches a nonprofit’s capacity in ways it would not stretch a federal agency. The multi-year timeline was not solely a governance challenge; it also reflected the difficulty of maintaining research capacity in an organization whose primary mission is advocacy, not data production. Limitation 2: The Tribal-level data gap — a cost paid by individual Tribes The decision to suppress tribal-level data and publish only state-level aggregates protected individual privacy and prevented misuse — but it imposed a real cost on individual Tribal governments that have a legitimate interest in their own verified Medicaid enrollment and expenditure data. A Tribal health director who wants to present their Tribal council with a verified count of tribal members enrolled in Medicaid and the associated Medicaid revenue cannot find that figure in any public Data Match output. The decision was made collectively by the Tribal Advisory Committee, but individual Tribes did not necessarily endorse it. The Tribal Advisory Committee represents collective Tribal interests, which are not identical to any individual Tribe’s interests. Limitation 3: PMPY modeling rather than actual claims data The $6.4-6.6 billion expenditure estimate is a model output, not a sum of actual claims. The first phase of the Data Match confirmed enrollment with high precision but estimated spending using national per-member-per-year benchmarks. The PMPY approach is standard and defensible, but it means the expenditure figure is an estimate that could be revised substantially when actual claims data is incorporated. A critic who wants to challenge the $6.4B figure has a legitimate methodological target: the PMPY assumptions, not the enrollment count. The sensitivity analysis table introduced in Week 9 demonstrates exactly how PMPY-assumption variation plays out: across five estimates ranging from the ACS lower bound to the 60%–80% reliance-adjusted scenarios, the structural significance of Medicaid holds at every point in the range — but students should be able to name which row is the empirically strongest and why, and to explain what would have to be true for the reliance-adjusted rows to be the more appropriate citation in a given policy context. Limitation 4: Population B remains unmeasured The Data Match, by design, measures Population A — IHS-registered AI/AN Medicaid enrollees. Population B — urban AI/AN Medicaid enrollees with no IHS registration — is estimated but not verified. The $1.88B Population B estimate from the Revised Analysis is illustrative, not validated. This means total AIAN Medicaid spending remains imprecisely known, and urban AI/AN communities remain the most analytically underserved by the current data infrastructure — the very communities that are most vulnerable to the policy risks identified in Weeks 11–12.

THE SOVEREIGNTY TENSION THAT CANNOT BE FULLY RESOLVED

The deepest limitation of the model is structural: the same sovereignty principles that made the project legitimate also constrained its outputs. State-level-only reporting protects privacy but leaves individual Tribes without verified public data about their own members. Tribal Advisory Committee authority to shape findings protects community interests but could theoretically limit the scope of what gets published. NIHB leadership ensures Tribal framing but also means the research capacity depends on a nonprofit that may not have the sustained infrastructure of a federal data program. These are not design flaws to be corrected. They are the expressions of a genuine tension between two legitimate values: the precision that comes from comprehensive data production, and the sovereignty that comes from community control over that production. Future iterations of the model will not dissolve this tension; they will have to negotiate it again, differently, in a different political context. The work of Weeks 5–7 — understanding how the governance was designed — is the work of knowing what that negotiation looks like.

1.4 — Precedent Value and the Template Question (75-90 min)

READ ALOUD  /  ADAPT

“The Data Match is finished — or rather, the first phase is finished, and the next phase is beginning. What matters for the future of Tribal health data is whether this model gets institutionalized or remains a one-time achievement. Institutionalizing it means: annual or biennial matches that produce updated enrollment and spending figures without requiring years of new negotiation each time. It means the DUA framework developed for this project becomes a standing agreement rather than a fresh legal negotiation. It means VRDC access for NIHB becomes a permanent data access arrangement rather than a project-specific authorization. None of that is guaranteed. Data access arrangements depend on the cooperation of federal agencies whose priorities shift with administrations. DUA renewals can be delayed or blocked. VRDC access protocols can change. The governance infrastructure built so carefully between 2017 and 2023 is not self-sustaining — it requires ongoing investment by NIHB, ongoing cooperation from CMS and IHS, and ongoing engagement from Tribal Advisory Committee members who are doing this work in addition to their primary responsibilities as Tribal health leaders. The question for next week — and for your final papers — is: what would it take to make this model durable? What are the institutional and structural changes that would move from ‘the Data Match happened once and produced important findings’ to ‘Tribal-led administrative data research is a permanent feature of the Indian health policy infrastructure’?”

What institutionalization would require	Current status	Key obstacle	Who would need to act
Standing DUA framework between NIHB, CMS, and IHS that renews automatically	Project-specific DUA negotiated for the 2023 match; renewal for next phase requires fresh negotiation	Legal review cycles at multiple agencies; political risk if any agency’s priorities shift	CMS Office of General Counsel; IHS data governance; NIHB leadership; potentially Congress for authorizing language
Permanent VRDC access for NIHB as an approved research organization	Access was established for this project; permanence requires ongoing authorized status	VRDC access is tied to specific authorized projects, not organizations; each new project scope may require new authorization	CMS data governance; OMB approval processes; NIHB’s capacity to maintain approved researcher status
Annual T-MSIS data extraction for AI/AN Medicaid trend tracking	Not yet established; the 2013-2022 data gap remains unfilled	T-MSIS methodology development; state data quality variation; resource requirements for annual processing	NIHB research team; CMS T-MSIS program; potentially new funding from HRSA or IHS for recurring data work
Tribal-level data access for participating Tribal nations (with appropriate privacy protections)	Currently suppressed in all public outputs; tribal-level access would require separate governance agreements	Small-cell suppression standards; individual Tribe data requests would require individual agreements; risk of inconsistent protections across Tribes	Tribal Advisory Committee; NIHB; participating Tribal nations who choose to enter individual data use agreements
Claims-level spending data replacing PMPY modeling	First phase used PMPY; future phases expected to incorporate claims data	Claims data linkage more complex than enrollment linkage; additional governance requirements for claims records	CMS claims data infrastructure; NIHB technical capacity; Tribal Advisory Committee review of new data scope

SESSION 2  ·  90 minutes — Week 14 The 20-Year Arc: From Fox 2003 to 2023, and What Comes Next The annotated bibliography as an arc · ACS and administrative data as complements · what the next generation looks like · synthesis discussion · final paper check-in

Session 2 Timing Overview

0-10 min

Bridge

From the project to the field: the 20-year arc Script framing the Fox 2003 through 2023 trajectory

10-35 min

Section 2.1

The annotated bibliography as a methodological history Six documents; the arc from structural narrative to empirical precision; what each phase required from the next

35-55 min

Section 2.2

ACS and administrative data: complementarity, not competition When to use each; the survey complement that administrative data cannot replace; the Population B gap as a permanent ACS use case

55-70 min

Section 2.3

Next-generation Tribal health data infrastructure: mapping exercise Groups map what Phase 2 of the Data Match should look like and what a fully institutionalized system would require

70-90 min

Section 2.4

Full synthesis discussion and final paper check-in Arc-of-the-story questions; paper workshop: students share thesis sentences and receive peer feedback

2.0 — Bridge Opening (0-10 min)

READ ALOUD  /  ADAPT

“Last week we evaluated the NIHB-led model on its own terms — what it produced, what it cost, and why the accountability structure was the key variable. Today I want to zoom out further and trace the 20-year arc that led to the Data Match — because understanding where the Data Match sits in that arc is essential to understanding where it should go next. In 2003, Edward Fox testified before the U.S. Civil Rights Commission and argued that Medicaid and Medicare were becoming de facto trust responsibility backfill — that the federal government was discharging part of its obligation to AI/AN health through programs that were not designed for that purpose and that could be modified without any tribal consultation. He called this a structural financing gap and warned that relying on third-party billing risked normalizing underfunded IHS appropriations. Twenty years later, the Data Match confirmed exactly the structure described — and quantified it precisely. $6.4-6.6 billion in Medicaid expenditures for IHS-access enrollees. Structural parity with IHS appropriations. PRC substitution at $2.6-4.0 billion. The Data Match did not discover something new; it verified, with administrative precision, what Tribal health policy analysts had been describing narratively for two decades. That arc — from structural narrative to empirical precision — is the story of Tribal analytic capacity development. And the next chapter of that story is what we are mapping today.”

2.1 — The Annotated Bibliography as a Methodological History (10-35 min)

Walk through the six documents in the annotated bibliography as a chronological arc — not a reading list, but a sequence of methodological advances and their limitations, each of which created the conditions for the next document.

Document	Year	Methodological character	Key contribution	Limitation that required the next step
Fox, Civil Rights Commission testimony	2003	Structural narrative analysis; no primary data	Named the structural financing gap; framed Medicaid as de facto trust responsibility backfill; established the conceptual vocabulary	No quantification; relied on per capita comparisons and general policy argument; could not specify the dollar magnitude of Medicaid flows
Fox & Boerner, Medicaid and Indian Health Programs	2009	Multi-state pooled empirical analysis; state reports + IHS user counts; no administrative data linkage	First multi-state empirical quantification of Medicaid flows to IHPs; 12 states, 90%+ of IHS user population; documented Medicaid approaching or exceeding IHS funding in several areas; introduced state-level spending estimates	Estimates based on state reports of variable quality and coverage; no direct IHS-CMS record linkage; could not isolate IHS-access population from broader AI/AN Medicaid census; state fiscal year comparability issues
Alaska Interagency Task Force report	2010	Regional administrative data + policy analysis	Documented Medicaid-Medicare interaction in Alaska Native context; showed geographic complexity; foreshadowed ACA debates	Regional scope only; could not produce national enrollment counts or spending totals; Alaska-specific dynamics not directly generalizable
Fox, Payment Reform report	2011	Policy analysis of ACA payment reform implications	First Tribal analysis of managed care expansion risk; documented 1995-2000 managed care disruption; argued for Tribal carve-out protections	Forward-looking projection rather than measured outcome; no enrollment or spending data for the post-ACA period
NIHB State Overview Sheets	2013	Survey-based (ACS) modeling	50-state modeling of Medicaid expansion eligibility; quantified potential enrollment gains; served as strategic advocacy tool during ACA implementation	Projection, not measurement; participation rate assumptions proved too optimistic; ACS limitations apply (self-report, sampling error, no claims data)
NIHB 2008 vs. 2019 comparison report	~2021	Administrative data (IHS active user definition)	First direct administrative enrollment-spending comparison across a decade; 46.5% enrollment growth, $3.35B to $7.17B spending growth; grounded in administrative rather than survey data	IHS active user definition is narrower than Data Match’s three-criteria Population A; did not link IHS records directly to CMS Medicaid files; PMPY-based spending estimates
NIHB 2024 retrospective	2024	Retrospective validation of 2013 projections against 2012-2021 data	Found realistic participation rates 20-40% rather than full-uptake assumptions; identified administrative barriers, state delays, outreach variability; demonstrated maturation of Tribal analytic capacity	Does not fill the 2013-2022 data gap for all-AIAN Medicaid spending; retrospective analysis cannot recover the missing decade of longitudinal data

READ ALOUD  /  ADAPT

“Notice what this arc looks like methodologically. From 2003 to 2013, the work is primarily narrative and survey-based — important conceptual framing and enrollment estimates, but no verified administrative counts, no expenditure data, no direct linkage of IHS and Medicaid records. Fox and Boerner in 2009 represent the first serious attempt to quantify Medicaid flows empirically — 12 states, state-level spending estimates, the observation that Medicaid was approaching or exceeding IHS funding in several areas. That was important work. But it was built on state reports of variable quality, and it could not isolate the IHS-access population from the broader AI/AN Medicaid census. The 2013 State Overview Sheets represent the apex of what the ACS-based approach could produce — a 50-state projection model that was valuable for advocacy but whose assumptions proved significantly too optimistic. Then there is a methodological rupture. The 2008-2019 comparison report and the Data Match both use administrative data — not surveys, not projections, but actual enrollment records. The difference in what they can say about the system is not incremental. It is categorical. Fox in 2003 could argue that Medicaid had become structurally important. The Data Match in 2023 could say: $6.4-6.6 billion, 941,168 people, 25 states, here is the state-by-state breakdown. The 2024 retrospective closes the loop: it goes back to the 2013 survey-based projections and asks whether they were right. They were right about the direction — Medicaid expansion did increase AI/AN enrollment and reduce uninsured rates — but wrong about the magnitude, because ACS-based participation rate assumptions were too optimistic. The retrospective demonstrates something important: you can only retrospectively validate a projection if you have better data later. The Data Match is what makes future retrospective validation possible.”

THE ARC IN ONE FRAME: FROM NARRATIVE TO PRECISION

2003: Medicaid is structurally important — argued but not quantified. 2013: Medicaid expansion would produce significant AI/AN enrollment gains — projected but not measured. 2021: Medicaid grew from $3.35B to $7.17B for IHS active users over a decade — measured administratively but with a narrower population definition. 2023: 941,168 verified IHS-access Medicaid enrollees, $6.4-6.6B estimated — measured with maximum available precision, PMPY-modeled spending. 2024: 2013 projections were directionally right, quantitatively optimistic — retrospectively validated. Each step made the next step possible.

2.2 — ACS and Administrative Data: Complementarity, Not Competition (35-55 min)

A conclusion students sometimes reach incorrectly: the Data Match means the ACS is no longer useful for AI/AN health research. This is wrong, and it is worth addressing directly before students write their final papers or enter policy careers with this misconception.

Question type	Best data source	Why	What the other source cannot provide
What is the AI/AN uninsured rate nationally and by state?	ACS	ACS covers all AI/AN individuals regardless of IHS registration, including Population B; produces annual estimates with consistent methodology for trend tracking	Data Match only covers IHS-registered Medicaid enrollees; cannot produce uninsured rate estimates for the full AI/AN population
How many AI/AN individuals are enrolled in Medicaid and verified IHS users?	CMS–IHS Data Match	Administrative linkage produces verified counts not dependent on self-report; eliminates sampling error for the IHS-access population	ACS self-report is subject to misreporting (Bhaskar et al. 2018 found both over- and under-reporting of IHS access and Medicaid enrollment)
What are post-ACA coverage trends for AI/AN populations in expansion vs. non-expansion states?	ACS (with administrative supplement)	ACS provides annual trend data across full AI/AN population; Data Match is currently single-year (2023) without a validated multi-year trend series for the full population	Data Match covers IHS-access population only and lacks a comparable pre-2019 baseline in the same administrative framework
How much Medicaid spending flows through IHS and Tribal facilities annually?	CMS–IHS Data Match + T-MSIS	Only administrative linkage can produce verified payment figures; ACS has no claims or spending data	ACS cannot estimate payment flows; no survey instrument captures Medicaid reimbursement amounts
What is the total AI/AN Medicaid population, including non-IHS-registered urban enrollees?	ACS + Data Match reconciliation (Population A + B approach)	ACS provides the denominator for Population B estimation; Data Match provides verified Population A count; neither alone produces the total	Data Match by design excludes Population B; ACS alone cannot distinguish Population A from Population B
Are 2013 Medicaid expansion projections consistent with actual outcomes?	Administrative data retrospective (NIHB 2024)	Only actual administrative enrollment data can validate survey-based projections; the 2024 retrospective demonstrated participation rates were 20-40% rather than the 50-80% projected	Projecting forward from surveys cannot substitute for measuring backward from administrative data

THE POPULATION B PERMANENT USE CASE

The most durable ACS use case in the post-Data Match era is Population B measurement. Because the Data Match by design captures only IHS-registered enrollees, ACS self-identification of AI/AN race among Medicaid enrollees remains the only available tool for estimating the non-IHS urban AI/AN Medicaid population. Until T-MSIS race coding quality improves enough to reliably identify AI/AN enrollees without IHS registration linkage, ACS will remain indispensable for Population B analysis — which is indispensable for understanding the full scale of AIAN Medicaid spending and for tracking urban AI/AN coverage trends. Students whose final papers involve urban-dominant states (CA, MN, NY, OR, NC) should note this explicitly.

2.4 — Full Synthesis Discussion and Final Paper Check-In (70-90 min)

Arc-of-the-Story Synthesis Discussion (15 min)

Use these questions as the structure. The instructor’s role is to push toward specificity — not ‘the Data Match was worth it’ but ‘here is the specific policy decision that could not have been made without it, and here is why.’

Discussion question	What a strong answer looks like	What to push back on
Was the additional precision worth the five-plus years it took to build?	Names a specific policy argument that required the $6.4B figure or the 941,168 count — not just ‘advocacy’ but ‘the PRC substitution model requires a verified enrollment count to produce the $2.6-4.0B range,’ or ‘Congressional testimony on Medicaid protection requires a figure that cannot be challenged on sampling grounds’	Vague yes answers (‘it gave us better data’) without identifying what the better data enabled; also push back on vague no answers that do not reckon with what the survey alternative actually could have produced
What policy decisions could not have been made based on the ACS estimates alone?	At minimum: PMPY-based expenditure modeling requires a verified denominator; the PRC substitution estimate requires knowing what share of Medicaid spending is for referrals, which requires claims-adjacent data; state-by-state fiscal risk modeling requires state-level enrollment counts with margins of error small enough to be policy-meaningful	Students who say ACS was sufficient for all policy purposes need to be shown the specific claim (‘ACS shows 1.3M AI/AN Medicaid enrollees’) and asked: could you use that figure to model PRC exposure, calculate the PMPY, or defend against a challenge that the enrollment count is inflated?
How does NIHB’s leadership change the nature of the research?	Names the accountability structure argument from Session 1 — not just ‘NIHB is more trusted’ but ‘NIHB’s accountability to Tribal nations produces different research questions, different safeguards, and different interpretation authority than CMS accountability to Congress or IHS accountability to HHS’	Generic statements about trust and relationship without engaging the specific structural mechanism (accountability to different principals produces different research designs)
What comes next — what questions remain that this project’s methods could address?	At minimum: Population B verification (replacing ACS-based estimates with administrative-linked data); T-MSIS gap-year series (2013-2022); claims-level spending data replacing PMPY; tribal-level data access for participating Tribes; Medicare dual-eligible analysis for the IHS-access population	Students who treat the Data Match as a finished achievement rather than the first completed phase of an ongoing data infrastructure project

Final Paper Check-In (15 min)

PEER REVIEW: THESIS AND ANALYTICAL FRAMEWORK CHECK

Each student writes their paper’s thesis sentence on a half-sheet of paper. Sheets are passed to a peer reviewer. Peer reviewers have 3 minutes to answer two questions in writing: (1) Does the thesis identify a specific, bounded policy question? (2) Can you identify from the thesis alone which vulnerability vector or structural mechanism the paper will engage? Writers receive their sheets back and have 2 minutes to revise their thesis based on peer feedback. Instructor circulates during the revision period. Instructor note: The most common thesis problem at this stage is scope. Students whose thesis reads ‘this paper analyzes Medicaid’s role in Indian health’ have not yet committed to a specific argument. Push them toward: ‘this paper argues that [specific claim] because [specific mechanism], using [specific state or policy context] as the analytical case.’ That sentence should be writable by end of Week 14.

CLOSING FRAME FOR THE COURSE’S CONTENT WEEKS

The journey from Fox 2003 to the 2023 Data Match is a 20-year story about what it takes to move from structural diagnosis to empirical precision in Tribal health policy. The structural diagnosis was available early — Fox named the financing gap in 2003 with conceptual clarity that the Data Match has only confirmed. What took 20 years was building the data infrastructure, the governance relationships, the legal frameworks, and the analytic capacity to produce verified numbers that a skeptical audience cannot dismiss on methodological grounds. That 20-year investment is what precision costs in contested policy domains. And the work of your final papers is to use that precision — to take verified numbers, apply the analytical framework developed over this semester, and produce an argument that Fox in 2003 could not have made but that you, in 2026, can.

Indian Health Financing  ·  Research & Methods Track  ·  Weeks 13–14 Lecture Notes  ·  February 2026

  Week 15 

INDIAN HEALTH FINANCING

All Tracks

Week 15 Instructor Lecture Notes & Script

The Arc Completed: Synthesis, Application, and What Comes Next

Track	All tracks; integrated capstone session
Sessions	One 90-minute session — Week 15 (final class meeting)
Core Question	What does a student who has completed this course know — analytically and practically — that they did not know 15 weeks ago? And what does the course leave unresolved?
Prerequisites	Final research paper submitted (end of Week 14); all prior lecture units complete
Assignments Due	Final research paper due end of Week 14; no new assignments in Week 15
Builds On	All prior units; with particular emphasis on Weeks 8–10 findings, Weeks 11–12 policy implications, and Weeks 13–14 Tribal data sovereignty framework

Course Arc Note

Week 15 is not a review session. It is a synthesis session. The distinction matters: review restates what students already know; synthesis asks students to do something with what they know. The goal of this session is for students to leave with three things: (1) a clear account of the analytical arc they have traveled, in their own words; (2) a set of unresolved questions they can carry into their professional lives; and (3) a direct encounter with the political stakes of the data infrastructure they have spent 15 weeks learning to read. The final research papers have been submitted, so the session should feel intellectually spacious rather than high-stakes. Use that.

Session Architecture (One 90-minute session)

Time	Segment	Format	Notes
0–10	Opening: the question you can now answer	Mini-lecture	Instructor-led; connect back to Week 1 opening
10–30	Section 1: The arc of the course, reconstructed	Full-group discussion	Student-led reconstruction of the methodological arc
30–50	Section 2: Five claims every graduate can defend	Lecture	Core analytical fluencies; link to final paper arguments
50–68	Section 3: What the course leaves unresolved	Discussion	The honest accounting; the open research questions
68–80	Section 4: Track-differentiated forward-looking exercise	Small groups	Where each track takes this next
80–90	Closing: the trust responsibility and data	Instructor close	The moral frame; final statement

Instructor Note

Final paper logistics: If you are returning papers at Week 15, do so at the very start of class and allow 3–4 minutes for students to skim comments. Then move directly to the opening section. Do not let paper discussion consume the session — direct individual questions about grades to office hours.

Opening: The Question You Can Now Answer (0–10 min)

READ ALOUD / ADAPT

“In the first session of this course, I asked you to imagine a tribal health program CFO who could not answer a simple question from a state legislator: how many of your patients are currently enrolled in Medicaid? I want to come back to that question today — not because the answer has changed, but because you have. Fifteen weeks ago, you heard that question and understood it as a data management problem. A CFO who can’t track enrollment needs better software, or better staff, or a better process. Now you hear it differently. You understand that the inability to answer that question is not primarily an administrative failure. It is the downstream consequence of a seventy-year funding structure in which the Indian Health Service was chronically appropriated at below-need levels, with no systematic federal infrastructure linking IHS patient records to Medicaid enrollment files — a gap that took a national five-year administrative data match project, led by a Tribal organization with a cooperative agreement from CMS, to partially close. That is what you know now that you did not know then. You can trace the gap from its origins in the trust responsibility doctrine through the survey era, through the administrative data transition, through the governance negotiations that shaped the Data Match, through the Population A/B distinction, through the fiscal risk scenarios — all the way to what a tribal health program administrator needs to understand about the $6.4 to $6.6 billion figure and what it does and does not represent. That analytical literacy is what this course was built to develop. Today we are going to take stock of what it means.”

Section 1: The Arc of the Course, Reconstructed (10–30 min)

Instructor Note

This section is student-led. The instructor’s role is to facilitate, correct errors, and add connective tissue — not to lecture. Ask students to reconstruct the course arc from memory in small groups for 8 minutes, then build a single shared arc on the board for 10 minutes. The goal is for students to hear each other articulate the arc, not just to hear it from you.

EXERCISE: THE ARC IN YOUR OWN WORDS  |  8 minutes small groups, 10 minutes full group

Small group task (8 min): In groups of 3–4, reconstruct the methodological arc of the course as a sequence of no more than six steps. For each step, identify: (1) what problem it was solving, (2) what it could show that the previous step could not, and (3) what new limitation it introduced or revealed. Full group (10 min): Build a shared arc on the board. Resolve disagreements about sequencing or characterization as they arise. Instructor adds bridging language between steps. Expected arc (for instructor reference — students should arrive at something close to this independently): Survey era (ACS): Established that AI/AN Medicaid enrollment was substantial and growing; could not link coverage to payment flows or distinguish IHS users from non-IHS users.Administrative data transition (2012 MAX / AIR proxy): First measurement of IHS-access Medicaid expenditures from actual claims; revealed $2.8B in 2012 paid claims; introduced the AIR proxy limitation and the IHS enrollment boundary.Regional data linkage (NWTEC): Demonstrated direct registry-to-Medicaid linkage was feasible; documented the 75–80% coverage ceiling structurally; revealed the public health vs. financing orientation distinction.CMS–IHS Data Match construction (Weeks 5–7): Showed that building a federal data match required governance and Tribal sovereignty infrastructure, not just technical capability; the five-phase project history is itself an argument about institutional trust.Data Match findings (Weeks 8–10): Established 941,168 IHS-access enrollees and $6.4–6.6B in expenditures; the Population A/B distinction is the course’s central analytical challenge; the 11-year data gap (2013–2022) is the course’s central evidentiary limitation.Policy implications and fiscal risk (Weeks 11–12): Medicaid is now a structural pillar of Indian health finance, approaching IHS appropriations in magnitude; total CMS investment in the IHS-access population is $10.86B (midpoint), with 51.1% of that population holding no CMS coverage; total federal investment ~$19.6B (midpoint); a 10–20% Medicaid contraction represents a $640M–1.3B fiscal shock with no IHS backstop.Tribal data sovereignty (Weeks 13–14): The governance decisions embedded in the Data Match — Tribal Advisory Committee structure, state-level-only reporting, small-cell suppression, NIHB as lead — are expressions of data sovereignty; they are not constraints on the research, they are what made the research legitimate.

Section 2: Five Claims Every Graduate Can Defend (30–50 min)

READ ALOUD / ADAPT

“Let me give you five claims. These are not claims you need to take on faith — they are claims this course has equipped you to defend analytically, with specific evidence, against specific counterarguments. If you can defend all five, you have the core analytical fluency this course was designed to build.”

Claim 1: Medicaid is now a structural pillar of Indian health financing, not a revenue supplement

In 1999, total AI/AN Medicaid payments of approximately $1.45 billion were secondary to IHS direct appropriations as a funding source for Indian health programs. By 2023, IHS-access Medicaid expenditures alone — approximately $6.4 to $6.6 billion — approach total IHS appropriations of approximately $8.6 billion. The structural relationship has inverted: the Indian health system (or non-system) is now substantially dependent on a federal entitlement program whose rules, rates, and continuation are determined by processes outside Tribal control.

Key Concept

How to defend this claim: The historical data series (1999–2008 CAGR of 11.5%; IHS/Tribal billing share doubling from 13% to 21%) documents the trajectory. The 2023 Data Match figure quantifies the magnitude. The new integrated CMS investment estimate ($10.86B midpoint across all three coverage groups within the IHS-access population) and the total federal investment framework ($19.6B midpoint: IHS + Medicaid + Medicare) place Medicaid’s structural role in full relief: Medicaid alone is approximately equal to IHS appropriations in scale, and the combined CMS investment exceeds IHS appropriations by a wide margin. The argument is not that Medicaid is bad for Indian health — it is that structural dependence on any single funding stream creates fiscal vulnerability, and that Medicaid’s rules are determined by CMS and Congress, not by Tribal nations.

Claim 2: The $6.4–6.6 billion figure represents approximately 78% of total AI/AN Medicaid spending, not 100%

The CMS–IHS Data Match’s flagship expenditure figure covers Population A — IHS-access enrollees with at least one paid IHS service. The remaining approximately 22% of total AI/AN Medicaid spending flows to Population B: AI/AN Medicaid enrollees without IHS registration, primarily urban populations. The 15-state reconciliation analysis estimates approximately $1.88 billion in additional Population B spending not captured in the Data Match’s primary figures.

Key Concept

How to defend this claim: The Population A/B distinction introduced in Weeks 8–10, the 15-state reconciled table, and the historical all-AIAN payment file series together establish that the Data Match is a partial count — robust within its defined scope, but not a complete accounting of AI/AN Medicaid spending. Any policy argument that treats the $6.4–6.6B figure as total AI/AN Medicaid spending is making a Population A/B error.

Claim 3: The 11-year data gap (2013–2022) is a structural weakness in the evidence base, not just a data availability inconvenience

The T-MSIS transition created a gap in the validated AI/AN administrative Medicaid data series from 2013 through 2022. During this period — which includes the full ACA expansion era, the COVID-19 public health emergency, and the introduction of continuous enrollment protections — the field had no administrative equivalent to the 2012 MAX analysis or the 2023 Data Match. The 2023 Data Match is, in effect, a point-in-time snapshot after an 11-year gap, not the latest entry in an ongoing validated series.

Key Concept

How to defend this claim: The sensitivity analysis table from Week 9 establishes the range of estimates consistent with the 2023 data. The gap means that the 1999–2008 trajectory and the 2023 findings cannot be smoothly connected — the trend line has a missing segment of exactly the period when the most significant coverage changes were occurring. Analysts who cite the 2023 Data Match without acknowledging this gap are presenting a cross-sectional finding as if it were a longitudinal trend.

Claim 4: Tribal data sovereignty is not an external constraint on research quality — it is the governance design that made the research credible

The CMS–IHS Data Match’s findings are credible in Tribal communities precisely because the project was led by NIHB, governed by a Tribal Advisory Committee, and designed with Tribal priorities — not federal administrative convenience — as the organizing principle. The decisions to suppress tribal-level data, to report only at the state level, and to give Tribal communities first review of findings were not compromises of research quality. They were the conditions under which Tribal communities consented to participate in the research.

Tribal Sovereignty Note

How to defend this claim: Compare the governance structure of the Data Match to a hypothetical CMS-led study with no Tribal advisory process. The findings would be the same numbers — but the data would have been extracted rather than produced collaboratively, the findings would not have Tribal endorsement, and the policy arguments built on them would be contestable on sovereignty grounds. The governance structure is not window dressing; it is what makes the data usable in Tribal advocacy contexts.

Claim 5: PRC substitution is the most policy-consequential finding in the Data Match, and the least-reported one

The Purchased/Referred Care substitution analysis — estimating $2.6 to $4.0 billion in Medicaid expenditures that substitute for PRC costs that would otherwise fall on the IHS budget — is the analytical bridge between the Data Match’s enrollment figures and the IHS funding gap argument. If Medicaid were not absorbing these costs, IHS would need to either fund them directly (requiring a massive appropriations increase) or deny services to IHS-eligible patients who cannot afford them. This is the fiscal risk scenario that makes Medicaid contraction a direct IHS program threat, not merely an insurance coverage concern.

Key Concept

How to defend this claim: The PRC substitution estimate applies only to Population A (IHS-access enrollees), making it a conservative lower bound on total substitution. The fiscal risk scenarios from Weeks 11–12 (10% contraction = $640M shock; 20% = $1.3B) are derived from this substitution relationship. The argument is not that Medicaid was designed to fund IHS — it is that the financing system has evolved to function that way, and the IHS budget does not contain the reserves to absorb a large Medicaid shock.

Section 3: What the Course Leaves Unresolved (50–68 min)

Instructor Note

This section is the intellectual honesty section. Its purpose is to prevent students from leaving with false confidence. The five claims they can defend are important and well-grounded. But there are questions this course has circled without fully resolving, and students who enter the field thinking they have complete answers will eventually be embarrassed by the gaps. Better to name the gaps here.

READ ALOUD / ADAPT

“Every course has a limit. Here is where this one ends. I want to be direct with you about what we have not resolved — not because the course has failed, but because intellectual honesty is itself a professional skill in this field.”

Unresolved Question 1: The Population B Measurement Problem

We know that approximately 22% of total AI/AN Medicaid spending flows to Population B — AI/AN Medicaid enrollees with no IHS registration. We have a 15-state estimate of $1.88 billion for this population. But we do not have a validated national series for Population B equivalent to what the Data Match provides for Population A. The historical all-AIAN payment files provide one approach to bridging this gap, but they have their own methodological limitations and are not continuously updated.

Looking Forward

What would it take to resolve this: A national administrative data match that identifies AI/AN Medicaid enrollees by tribal enrollment status or self-reported race — not just by IHS registration — would capture Population B. This requires either improved race coding in T-MSIS (a known CMS priority) or a second match using tribal enrollment records, which raises its own sovereignty and data access challenges. This is an open research and policy infrastructure question that no one in the field has fully solved.

Unresolved Question 2: The 2013–2022 Trend Data Gap

The validated administrative data series has an 11-year gap. We know the 2012 baseline and the 2023 snapshot. We do not have a validated year-by-year trend through the ACA expansion era. This means that the most analytically important period for understanding how Medicaid became a structural pillar of Indian health finance is also the period for which we have the weakest data. The field is operating on a before-and-after comparison with a missing middle.

Looking Forward

What would it take to resolve this: T-MSIS data from 2013 through 2022 exists and is theoretically accessible through the VRDC. A validated retrospective series using the same methodology as the 2023 Data Match would close this gap. This is a research project that could be conducted now — it has not been done yet, as of the time this course was developed.

Unresolved Question 3: Tribal-Level Data and Sovereignty

The Data Match suppressed tribal-level data to protect small-population confidentiality. This was the right governance decision. But it means that individual tribes cannot use the Data Match to understand their own Medicaid dependence — the very question tribal health program administrators most need answered. State-level aggregates are useful for federal advocacy; they are not useful for tribal-specific program planning.

Tribal Sovereignty Note

The tension here is not resolvable by better statistics. The small-cell suppression decision reflects a genuine conflict between two legitimate interests: protecting individual privacy (and, implicitly, protecting tribes from having their specific data used in ways they did not authorize) versus enabling tribes to access data about their own communities. NWTEC’s Northwest Tribal Registry resolves this for 43 Northwest tribes. There is no national equivalent.

Unresolved Question 4: The IHS Enrollment Boundary as a Policy Problem

Every data system in this course hits the same structural ceiling: the 20 to 25% of the AI/AN population that is not IHS-registered and is therefore excluded from IHS-based datasets. This population is disproportionately urban, younger, and covered by Medicaid or marketplace insurance rather than IHS direct care. The Data Match tells us about the IHS-dependent population with precision. It tells us very little about the AI/AN population that has already moved beyond IHS-dependent care.

Looking Forward

This boundary is not just a data problem — it is a policy design question. If the long-term trajectory of AI/AN health coverage involves a growing population without IHS registration, then the analytical framework built around IHS-access Medicaid spending will increasingly understate the full AI/AN Medicaid picture. The course has given students the tools to understand this limitation. Resolving it requires both data infrastructure investment and policy decisions about how the IHS registration system should evolve.

Unresolved Questions 5–6: Full-Spectrum Need and Tribal Own-Source Financing

5. What is the full cost of providing comprehensive Indian health care? The Full Funding research provides a framework for estimating need, but the estimate depends on definitions, benchmarks, service scope, and available expenditure data.

6. How much Tribal own-source revenue contributes to the overall financing of Indian health care? Federal investment is not the entire financial picture. A complete financing framework ultimately needs to account for Tribal resources used to support health services and infrastructure.

Section 4: Track-Differentiated Forward-Looking Exercise (68–80 min)

Instructor Note

This is the last substantive activity. It asks students to connect the course content to their professional trajectories. Groups should be self-selected by track. Allow 8 minutes for group work, 4 minutes per group to share (or sample 2 of 3 tracks if time is tight).

Track	Forward-Looking Task	Share-Out Prompt
Policy & History	You are advising a Tribal nation’s federal relations team. Congress is considering a 15% reduction in federal Medicaid matching rates as part of a budget reconciliation bill. Using the course’s analytical framework, prepare a two-minute verbal briefing for the Tribal council that explains: (1) what $6.4–6.6B actually represents and what it does not; (2) why PRC substitution makes this a direct IHS threat, not just a coverage issue; (3) what data from this course you would cite as the strongest evidence for the impact.	What is the single most important number from this course for a Tribal council member to understand about Medicaid dependence, and why?
Research & Methods	You are designing the next iteration of the CMS–IHS Data Match, targeting a 2026 publication. Using what you know about the current match’s limitations, identify: (1) one Population B gap you would prioritize closing and how; (2) one methodological improvement to the PMPY modeling framework; (3) one governance design choice you would keep exactly as it was and explain why.	What is the most important methodological improvement the field should prioritize in the next data match cycle?
Operations & Administration	You are the director of a tribal health program in a state with a high IHS penetration rate (reservation-dominant). A new state Medicaid director has proposed reducing reimbursement rates for IHS-facility claims by 8%. Using the course framework, prepare a one-page talking points document that explains: (1) what share of your program’s revenue is Medicaid-derived; (2) why an 8% rate reduction is not an 8% budget problem; (3) what population would lose access to services, and why the IHS PRC budget cannot absorb the loss.	What course concept would be most immediately useful to you in a real tribal health program administrator role?

Closing: The Trust Responsibility and Data (80–90 min)

READ ALOUD / ADAPT

“I want to close with something that runs under every technical finding in this course and that I have been careful not to overstate — because overstating it leads to bad analysis, and understating it leads to analysis that misses the point. The trust responsibility is the legal and moral obligation of the federal government to provide for the health of American Indian and Alaska Native people as partial compensation for treaty rights and land cessions. The Indian Health Service exists because of that obligation. The chronic underfunding of IHS — documented in every analysis this course has engaged — is not a budgetary oversight. It is a recurring federal failure to meet a treaty obligation. Medicaid entered this story not as a trust responsibility instrument but as a general entitlement program that AI/AN patients, like all low-income Americans, are entitled to access. Over 50 years, through a combination of advocacy, regulatory evolution, and administrative data matching, Medicaid has become the primary mechanism through which additional federal dollars flow into the Indian health system (or non-system) above the IHS appropriation. The $6.4 to $6.6 billion figure is not charity. It is part of an entitlement that AI/AN people are legally entitled to access and that the Indian health system (or non-system) has built its financial model around. What the Data Match has done — what NIHB spent five years building and what you have spent 15 weeks learning to read — is make visible the full scale of that financial model for the first time. The $8.45 billion total AI/AN Medicaid figure. The $10.86 billion total CMS investment midpoint when Medicare is added in. The $19.6 billion combined federal investment estimate when IHS appropriations are included. The 51.1% of the IHS-access population — 778,438 people — who are invisible to every CMS data system, entirely dependent on IHS appropriations and PRC. The 78% coverage rate that maps exactly onto the IHS enrollment boundary. The PRC substitution relationship. The fiscal risk scenarios. None of this was visible without the data. And none of it will protect Indian health financing without analysts, advocates, administrators, and policymakers who understand what the data actually says and what it does not say.

Instructor Note

Optional closing activity (if time permits, or as an informal close): Ask each student to write one sentence — not to share aloud, just for themselves — completing the following prompt: ‘The most important thing I understand about Indian health financing that I did not understand before this course is…’ Give them 90 seconds. No collection required. This is a private reflective close, not an assessment.

Final Research Paper: Evaluation Guidance (Instructor Reference)

Instructor Note

The final research paper was assigned at the end of Week 13 and due at the end of Week 14. This section provides evaluation guidance for reading the submitted papers. It is not distributed to students.

Assignment Guidance

Assignment parameters (from syllabus): 8–12 pages; applies the course analytical framework to a policy question of the student’s choosing involving AI/AN Medicaid financing, IHS funding, or Tribal health data governance; must demonstrate command of the Population A/B distinction, the Data Match findings, and at least one of: fiscal risk analysis, PRC substitution, or Tribal data sovereignty governance.

Evaluation Criterion	What Strong Papers Do
Population A/B command	Explicitly distinguishes IHS-access spending from total AI/AN Medicaid spending when citing figures; does not treat $6.4–6.6B as total AI/AN Medicaid
Data Match findings fluency	Cites 941,168 enrollment figure with correct scope; understands PMPY as modeled estimate, not summed from claims; acknowledges 11-year data gap when making trend arguments
Policy argument quality	Makes a specific, falsifiable claim about a policy choice or risk; does not simply describe the Data Match findings without an analytical argument
Sovereignty integration	Treats governance decisions (TAC structure, suppression rules, NIHB leadership) as analytically significant, not merely procedural background
Limitations acknowledgment	Identifies the specific limitations of the data sources used in the paper’s argument; does not cite figures beyond their analytical scope
Arc coherence	Situates the paper’s argument within the historical arc from survey era to administrative data era; does not treat the 2023 Data Match as if it appeared without methodological predecessors

Common Weakness	What It Looks Like
Population A/B conflation	Paper treats the Data Match’s $6.4–6.6B as total AI/AN Medicaid spending and builds an argument on that error
Trend overreach	Paper uses the 1999–2008 CAGR to project a 2023 figure and presents this projection as if it were an empirical finding
Sovereignty as decoration	Paper mentions Tribal data sovereignty in the introduction but does not integrate it into the analytical argument
Single-source dependence	Paper relies entirely on the 2023 Data Match without engaging the historical series, the Population B gap, or the 11-year data gap limitation
Advocacy without analysis	Paper asserts that Medicaid cuts would harm Indian health without specifying which population, which expenditure category, and through what causal mechanism

Course Arc Summary — Student Reference

This table summarizes the full analytical arc of the course. It is distributed to students at the end of Week 15 as a synthesis reference.

Unit	Core Argument	Key Analytical Output
Weeks 1–2: Foundations	The IHS exists because of the federal trust responsibility; it is chronically underfunded relative to the need it is obligated to serve; Medicaid entered the picture not by design but through AI/AN entitlement eligibility	I/T/U system (or non-system) structure; trust responsibility legal framework; the funding gap concept
Weeks 3–4: Survey Era and Administrative Data	The ACS established that AI/AN Medicaid enrollment was large and growing, but could not measure payment flows; the 2012 MAX analysis used the AIR proxy to produce the first administrative estimate ($2.8B); NWTEC demonstrated regional direct linkage; the IHS enrollment boundary appeared across all systems	Cross-era comparison table; Population A/B first definition; IHS enrollment boundary concept
Weeks 5–7: Data Match Construction	The CMS–IHS Data Match was a governance achievement as much as a technical one; the five-phase project history reflects the institutional trust required for Tribal communities to participate in federal data matching	Five-phase project timeline; DUA/VRDC architecture; Tribal Advisory Committee model; midterm
Weeks 8–10: Data Match Findings	941,168 IHS-access Medicaid enrollees; $6.4–6.6B in expenditures (PMPY modeled); Population A/B distinction in full; 15-state reconciliation; 11-year data gap; sensitivity analysis table	Population A/B reconciliation table; PMPY walkthrough; sensitivity analysis; data memo assignment
Weeks 11–12: Policy Implications and Fiscal Risk	Medicaid is now a structural pillar approaching IHS appropriations in magnitude; PRC substitution ($2.6–4.0B) is the mechanism linking Medicaid contraction to IHS program loss; fiscal risk scenarios quantify the exposure	Total federal investment framework ($19.3–19.9B midpoint $19.6B; IHS + Medicaid + Medicare); total CMS investment midpoint $10.86B; 51.1% IHS-access population with no CMS coverage; PRC substitution range; 10%/20% contraction scenarios
Weeks 13–14: Tribal Data Sovereignty	The governance decisions embedded in the Data Match are expressions of Tribal data sovereignty; they are not constraints on research quality but the conditions under which research became legitimate; the NIHB data infrastructure model is replicable	Five sovereignty principles; TAC governance analysis; Tribal data infrastructure design framework; final paper
Week 15: Synthesis	The analytical arc is coherent; five core claims are defensible with course-derived evidence; six major questions remain unresolved; the data’s policy utility depends on analysts who understand its scope and limitations	Synthesized arc; five defensible claims; six open research questions; professional application exercise

Indian Health Financing  ·  All Tracks  ·  Week 15 Lecture Notes  ·  March 2026