Indian Health Service spending was USD 4,078 per user in FY2023 and its 2026 appropriation was set from fiscal 2025 enacted levels

What is happening?

The Indian Health Service is the federal agency responsible for health care for members of 574 federally recognized tribes in 37 states. On its own profile page, marked as based on 2022 data, the population served is 2.8 million American Indians and Alaska Natives. The same page states expenditure per user population of USD 4,078 for FY2023 and sets it beside the US national health expenditure per person of USD 13,493 for CY2022.

The money is set two ways in the same law. Public Law 119-74, enacted 2026-01-23, writes dollar amounts for the discretionary accounts and then appropriates Contract Support Costs and Payments for Tribal Leases under Section 105(l) as such sums as may be necessary, with no dollar figure at all in the statute. Those two accounts are legally mandatory and open ended. The same law writes an advance appropriation of USD 4,789,731,000 for Indian Health Services and USD 516,600,000 for Indian Health Facilities becoming available on 2026-10-01, a total of USD 5,306,331,000 for FY2027.

The rule that produced the discretionary side is written in one sentence in the Senate Appropriations Committee report of 2025-07-24, which states that all programs, projects and activities are maintained at fiscal year 2025 enacted levels unless otherwise stated.

Because two accounts are estimates rather than ceilings, three different totals for FY2026 are printed in opened documents and a fourth follows from adding account lines, and each is defensible on its own scope. A tribal budget table and a tribal organization statement both give the discretionary services and facilities appropriation as USD 5.86 billion. One news report gives a program level of USD 8.05 billion. The same tribal budget table gives a total for the agency of USD 8.25 billion, which is that program level with the Special Diabetes Program for Indians added. The four agency account lines in the FY2027 budget appendix sum to about USD 7.99 billion, and no opened document prints that total. It is arrived at here by adding them. In February 2026 a national tribal health organization corrected its own earlier statement that FY2026 was a decrease, saying the error came from the estimates for those two indefinite accounts having moved.

Against that, a workgroup of tribal representatives that produces an annual need based budget recommendation for the agency put the requirement at USD 73.0 billion for FY2027 and USD 76.10 billion for FY2028. The budget request for FY2027, published in April 2026, was USD 9.094 billion.

Whose problem is this?

The agency is funded through the Interior bill rather than the Labor and Health bill, so the subcommittee that sets its level is the one that also handles land, water and cultural accounts.

Where does this problem end?

The boundary here separates the size of a number from the rule that produced it. A shortfall against an assessment is one statement, and a baseline that reads the previous year rather than the assessment is another, and only the second explains why the first repeats.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state. No statute, appropriations document or agency document opened here names a funding level for this agency that would be sufficient. The Senate Appropriations Committee report of 2025-07-24 states an instruction pointing the other way, directing the agency to establish measurements for tracking the improvement of patient health rather than defining increased funding alone as the metric for measuring improvements. The one numeric target in circulation is the annual need based estimate produced through tribal budget formulation, at USD 73.0 billion for FY2027 and USD 76.10 billion for FY2028. That estimate is recorded in the attempts block below and is not adopted here as the target state, because it is produced by the side asking for the money rather than by the side that sets it. One real government target does exist in this material and it is not a funding target: the department sets a minimum condition index of 90 or higher for all of its buildings, and 39 percent of the rated agency buildings met it as of September 2022.

How big is it?

The affected population is 2.8 million American Indians and Alaska Natives, the figure the agency publishes as population served on 2022 data. It is the best attested number in this document on that axis: the federal auditor writes over 2.8 million in its November 2023 report and a tribal budget book of April 2025 writes approximately 2.8 million. The population file records that count as both bounds of a zero width interval rather than manufacturing a range out of terms that do not match.

Three different population figures appear in this material and they are not interchangeable. The first is population served, at 2.8 million on 2022 data. The second is user population, given as about 3.2 million in an April 2026 tribal budget book and cited there to agency estimates dated 2026-01-08. A user population counts people who actually used the service and should therefore be the smaller of the two, and here it is the larger. The third is implied rather than published: the denominator behind the USD 4,078 per user figure cannot be recovered from the appropriation and either population, which means that figure divides a subset of expenditure rather than the whole appropriation. The agency document that would settle all three is on a host that did not respond to any request made here.

That is why no per capita figure in this document is computed rather than quoted. Dividing an appropriation by a population would produce a number that reads as measured and is not, and it would sit next to a published per user figure built on a different base, so the two would be compared as if they answered the same question. Details are in population/us-indian-health-service-funding-shortfall.json.

Under what conditions does it arise?

1. Two accounts in the bill have no number. Contract Support Costs and Payments for Tribal Leases under Section 105(l) are appropriated as such sums as may be necessary. They are legally mandatory, they are open ended, and the statute sets no ceiling for either. What the agency and the budget office publish for them are estimates, and those estimates move.

2. Everything else starts from the previous year. The Senate Appropriations Committee report of 2025-07-24 states that all programs, projects and activities are maintained at fiscal year 2025 enacted levels unless otherwise stated. A baseline rule of that shape reads the previous appropriation and it does not read any measure of need.

3. Growth in the open ended accounts is absorbed inside the same topline. Testimony to the Senate Committee on Indian Affairs on 2025-10-29 states that in FY2024 agency accounts were reduced to make room for growing Contract Support Costs and Section 105(l) lease payments, and that the accounts that bore the brunt were the facilities and electronic health record line items. That statement is tribal testimony rather than a committee finding, and no document opened here from either appropriations subcommittee confirms or contests it.

4. The estimating body is not the appropriating body. The need based estimate is produced through tribal budget formulation and submitted upward. The appropriation is written by two subcommittees working from the prior year. Nothing in the process opened here obliges the second to answer the first.

5. Advance appropriations cover part of the agency and not all of it. Testimony of 2025-10-29 puts roughly USD 1.3 billion across six accounts outside that protection, with the agency covering those costs as a stop gap. The advance for FY2027 of USD 5.31 billion is itself below the FY2026 enacted services and facilities total of USD 5.86 billion.

6. The workforce moved during the same period. The same testimony states that a voluntary early retirement authority cost the agency over 1,000 employees, that about 500 positions were refilled, and that the offer acceptance rate was the lowest in the history of the agency. No source opened here gives a post measurement of the vacancy rate against a published denominator.

7. Replacement runs slower than aging. Nine facilities remain on the replacement priority list of 1993, and the agency told the federal auditor that at roughly USD 250 million a year for medical facilities construction it would take about ten years to fund them. Tribal testimony puts agency hospitals at an average of 39 years against 11.5 years for US not for profit hospitals, and states that at the existing replacement rate a facility built in 2026 would not be replaced for 290 years.

What has been tried?

Two of these are audit recommendations with recorded outcomes, and the outcomes point in opposite directions. The data recommendations of 2011 were all implemented. The allocation recommendation of 1991 was closed as Not Implemented, which is a stronger fact about why the pattern persists than the original finding was.

What was found?

Why is it still unsolved?

Cost structure — the appropriation is built from the previous year, and two mandatory open ended accounts absorb growth inside that same topline.

The first part is that the baseline is last year rather than an assessment. The Senate committee report states the rule in a single sentence, and a rule of that shape cannot register a shortfall, because the quantity it reads is the previous appropriation. The federal auditor said the same thing about distribution among areas in 1991, in a report named for the finding that funding is based on historical patterns and not on need, and that recommendation was closed as Not Implemented after the relevant committees declined to alter the allocation formula. The 1991 finding concerned how the total is divided. The same shape now operates on the total itself, and the record shows the mechanism was put to Congress once and was declined.

The second part is that two accounts inside the bill have no ceiling. Contract Support Costs and lease payments under Section 105(l) are legal obligations that must be paid at whatever level they reach, and they are appropriated as such sums as may be necessary. Tribal testimony describes what follows when they grow inside a flat topline, which is that other accounts are reduced to make room, and it names the facilities and electronic health record lines as the ones that bore it in FY2024. That is testimony rather than a committee finding, and it is the only mechanism in this material that explains why the discretionary program accounts move as they do.

The third part is that the size of the thing is contested at the first step. Three totals for a single fiscal year are printed in opened documents and a fourth follows from adding account lines, all of them correct on their own scope, and the spread between the smallest and the largest is larger than the annual increase being argued about. A tribal organization reported a decrease for FY2026 and corrected it two weeks later, saying the estimates for the two indefinite accounts had moved. A quantity that needs a scope note before it can be read does not travel well, and every statement of the gap inherits that.

The fourth part is that the body producing the estimate is not the body setting the number, and the deciding body has stated a different metric. The tribal recommendation is an ask. The committee report directs the agency to measure the improvement of patient health rather than treating increased funding alone as the measure of improvement. Those are not two readings of one number. They are two different questions, and no document opened here reconciles them or records either side answering the other.

What observation would mean it is solved?

Candidates — (a) an appropriation whose stated starting point is an assessment of need rather than the prior year enacted level, which would be readable in the committee report language (b) a fall in the provider vacancy rate measured against a published denominator (c) a rise in the share of federally operated medical buildings meeting the condition index goal of 90 or higher, from 39 percent as of September 2022, together with a fall in the 151 buildings that carry no rating at all.

(a) is the direct observation and it can be read on the face of a committee report. It is also weaker than it looks, because the two indefinite accounts would still be estimated rather than capped, so a stated change of baseline could leave the discretionary program accounts moving exactly as before.

(b) is not currently measurable. The last rate published with a denominator and a measurement date is the 25 percent of November 2017. Everything since is a figure without a denominator, so there is no pair of comparable measurements from which a fall could be read. Establishing that measurement is a precondition for using it as a test rather than a consequence of passing it.

(c) is the most concrete and the narrowest. It covers federally operated medical buildings only, and tribes operate most of the health centers in the system, so the measure can move while most of the system it stands for does not. The 151 unrated buildings matter more than the 39 percent for the same reason the vacancy figures do: a share computed on a subset that was chosen by data availability is not yet a measurement of the whole.

What is it connected to?

Fills with researchthe federal trust and treaty responsibility as a legal doctrine; Medicaid and third party revenue as a growing share of facility funding; electronic health record modernization as a separate account with its own estimate; the December 2018 report of the U.S. Commission on Civil Rights on continuing federal funding shortfalls for Native Americans, which is cited in the Senate testimony and in both tribal budget books; and rural health access, which the corpus carries as problem `us-rural-maternity-unit-closure`. Relation type and evidence grade were not confirmed.

What these sources do not say

• What the per user figure divides by. The agency page states expenditure per user population of USD 4,078 for FY2023 and national health expenditure per person of USD 13,493 for CY2022, and states neither the expenditure base nor the population count behind the first. It cannot be reconstructed from the appropriation and either published population, so it evidently uses a subset of expenditure. No opened source states which subset, which means the pairing that the agency page itself makes between the two amounts is not shown to be a comparison of like measures.
• How population served and user population relate. The agency page gives 2.8 million as population served on 2022 data, and a tribal budget book gives about 3.2 million as user population citing agency estimates dated 2026-01-08. The term that ought to be the smaller carries the larger number, and no opened source reconciles them.
• Where the 46 percent area vacancy figure comes from. It is not in the auditor report, whose area range is 13 to 31 percent, not in either tribal budget book, not in the Senate hearing record and not in federal full text search. It appears in two articles by one outlet in an identical sentence with no attribution in either. The nearest published statistic is a different measure entirely, that 43 percent of agency facilities would have to close if they lost one physician level provider.
• What denominator the current vacancy figures use. Rates of 30 percent overall and 36 percent for physicians, of 32 to 37 percent, and of around 35 percent appear in three tribal documents between April 2025 and April 2026. None publishes a denominator or a measurement date, and one footnotes its range to the auditor report that states 25 percent.
• What the need based estimate is built from. The two budget books describe the method at the level of an actuarial engagement begun in 2003, annual updates using current population and per capita cost information, geographic considerations and a health needs assessment, then area worksheets rolled up nationally. Neither publishes a per capita cost model, a population source or an inflation assumption.
• Why the level is set where it is. The Senate committee report states the baseline rule and directs a different metric. No document opened here from either appropriations subcommittee addresses the size of the difference between the appropriation and the tribal estimate.
• The FY2026 tribal estimate at first hand. USD 63.04 billion reaches this document through reporting alone. The book that states it is not among the budget material currently posted, which lists the FY2027 and FY2028 books only.
• Whether building condition has moved since September 2022. The auditor measured it then, and no opened source publishes a later measurement or reports how many of the 151 unrated buildings have since been rated.
• What became of the announced repair funding. A dedication of USD 1.0 billion in existing funds together with unobligated infrastructure balances was announced in March 2026, and no opened source reports an obligation, a project list or an outcome.
• What the civil rights commission found in 2018. That report is cited in the Senate testimony and in both tribal budget books, and nothing in this document rests on it.
• A federal statement of the account totals from outside the appropriations and budget documents. The department budget pages and the congressional research service both refused automated requests, so no third federal source corroborates the account detail used here.

See the evidence

Eighteen of the twenty-four rows above rest on a document that was opened, and ten of those are federal. The statute settles the FY2027 advance to the dollar and settles the fact that two accounts carry no dollar figure at all. The Senate committee report settles the baseline rule and the metrics instruction. The budget appendix and the auditor reports settle the account tables, the building condition figures, the 2017 vacancy measurement and the disposition of two old recommendations. Where sources overlap they mostly agree: population served appears as 2.8 million in three of them, and the FY2026 services and facilities total appears as USD 5.86 billion in two. Where they disagree the disagreement is left visible rather than resolved: three sources give three different FY2024 totals for the agency, the auditor and tribal testimony give nine and seven for facilities remaining from the priority list of the early 1990s, the two population terms run in the wrong direction relative to each other, and two sources give different comparison figures for life expectancy. Where a figure is weak it is marked and quarantined: the USD 63.04 billion need estimate for FY2026 rests on reporting alone and is graded low, the current vacancy rates are graded medium because none publishes a denominator, and the 46 percent area rate is recorded as not established. Each of those weak figures is carried only where it is graded, and no claim elsewhere in this document rests on one of them. That is a choice made deliberately at one point in particular. The need estimate matching the fiscal year of the appropriation described here is the FY2026 figure of USD 63.04 billion, and the figures that were read in an opened document are the FY2027 and FY2028 estimates. This document uses the opened documents rather than the matching year, and states the year each estimate belongs to wherever it appears. The population file records a zero width interval at 2.8 million rather than assembling a range out of two terms that the sources do not reconcile. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 2,800,000–2,800,000 As of 2022 data, published on the agency profile page captured 2025-03-19

Sensitivity The interval has zero width and is not a confidence interval. Three opened sources give the same figure for the same term, and the one adjacent figure that could have formed an upper bound was deliberately excluded. That figure is about 3.2 million, described in the FY2028 tribal budget book of April 2026 as the people served by the hospitals and health clinics account and cited there to agency user population estimates dated 2026-01-08. It is not used as a bound for two reasons. First, it is a different term: a user population counts people who actually used the service and should therefore be smaller than the population served, and here it is larger, which means the relationship between the two terms is the opposite of what the terms imply and cannot be resolved from any opened source. Second, the underlying agency estimate could not be opened, because the agency host refused every connection attempted, so the larger figure reaches this file through one document that cites it. Assembling an interval from 2.8 million and 3.2 million would present a range built from two definitions as though it were a range printed for one quantity. A third population is implied and never published: the denominator behind the agency figure of USD 4,078 in expenditure per user population for FY2023 cannot be recovered from the appropriation and either published count, which means that figure divides a subset of expenditure rather than the whole appropriation. No per capita figure is computed anywhere in this document for that reason. Corrections to the 2.8 million run in both directions and neither is computable here. Downward, because the count is a population the agency is responsible for rather than a count of people who used a service in a given period. Upward, because the reference year is 2022 and no later count of the same term was published in any opened source.

Regional breakdown No opened source breaks the population served down by IHS area, by state or by tribe. What does exist at area level is a count of buildings rather than of people: the auditor report gives 363 federally operated medical buildings as of September 2022 across ten areas, being Albuquerque 41, Bemidji 16, Billings 40, California 7, Great Plains 64, Nashville 6, Navajo 80, Oklahoma 16, Phoenix 73 and Portland 20, with none federally operated in Alaska or Tucson. Those counts cannot stand in for a population decomposition. They exclude tribally operated facilities entirely, and tribes operate 331 health centers against 52 federally operated, so an area with few federal buildings may hold a large served population. Splitting the national count by building counts, by area office count or by resident population would be proportional allocation, which this repository forbids, and it would be wrong here in a specific way: the agency operates through 12 area offices, 170 service units and 41 urban Indian organizations whose catchments differ in size and in the share of care delivered by tribes rather than by the federal system, so a proportional split would misstate every area at once.