Showcase · Data layers · CMS, HHS and U.S. Census open data

If Oklahoma’s weakest rural hospitals closed

How much longer would people drive to the nearest hospital, and how full would the hospitals that stay open get? A scenario, not a forecast: the rural hospitals that meet one written rule from their own cost reports are taken off the map, and everything else stays as it is.

16 of Oklahoma’s 74 rural hospitals meet the rule. If all of them closed, the people living more than 30 minutes’ drive from a hospital would rise from 132,970 to 348,649, and those more than 45 minutes away from 11,902 to 91,235. 369,390 Oklahomans would drive 23 minutes longer on average. Of the 34 hospitals that would take their patients, 12 would be 85% full or more, against 4 today.

What is changing

The 2025 budget reconciliation law, Public Law 119-21 (H.R. 1, approved 4 July 2025), changes Medicaid in steps over the next six years, and sets up a fund for rural health care beside it. In its own words and dates:

• Work or “community engagement” (section 71119). For adults aged 19 to 64 covered by the Medicaid expansion, a state must make it a condition of eligibility that they show 80 hours a month of work, community service, a work program or half-time education (or an income of at least 80 hours at the federal minimum wage), “beginning not later than the first day of the first quarter that begins after December 31, 2026”. The Secretary may exempt a state making a good faith effort, but an exemption “shall expire not later than December 31, 2028”. Among those excluded: American Indians, parents and carers of a child of 13 or under, veterans with a total disability rating, and people who are medically frail.
• Eligibility checked every 6 months for the same adults, instead of every 12 (section 71107), for redeterminations scheduled from the first quarter that begins after 31 December 2026.
• Provider taxes (section 71115). The limit under which a state’s tax on hospitals and other providers is treated as not holding them harmless falls, in expansion states, from 6% to 5.5% of net patient revenue in fiscal year 2028, then 5%, 4.5%, 4%, and 3.5% from fiscal year 2032 (taxes on nursing homes and intermediate care facilities keep their rate). States that did not expand Medicaid keep their current taxes and may not add new ones.
• State directed payments (section 71116). New payments through Medicaid managed care plans are capped at 100% of the Medicare rate in expansion states and 110% elsewhere. Payments approved earlier, including those for rural hospitals approved by the date of the law, are “reduced by 10 percentage points each year” from rating periods beginning on or after 1 January 2028, until they reach the cap.
• Cost sharing (section 71120). From 1 October 2028, expansion adults with incomes above the poverty line pay a share of some care, no more than $35 a service and no more than 5% of family income.
• The Rural Health Transformation Program (section 71401). $10 billion a year for fiscal years 2026 to 2030, $50 billion in all, for states with an approved plan: half shared equally among them, half as CMS decides. A state must use it for at least three of the uses the law lists, among them payments to health care providers, recruiting clinicians to rural areas for at least 5 years, technology, and helping rural communities “right size their health care delivery systems”. Oklahoma’s first-year award was about $223.5 million (Office of the Governor, 29 December 2025).

Oklahoma expanded Medicaid to adults aged 19 to 64 with incomes up to 138% of the poverty line from 1 July 2021, after voters approved State Question 802 in June 2020 (Oklahoma Health Care Authority), so the expansion-state rules apply to it. This page does not estimate what the law will do to any hospital’s income. It asks a narrower question that the data can answer: if the hospitals already in the weakest position closed, what would change for the people they serve and the hospitals round them?

Why Oklahoma

We applied one rule to every rural hospital in the United States with cost reports for 2021 to 2024. Oklahoma has more hospitals that meet it than any other state: 20 of 79 (25.3%; the 79 include hospitals that have filed cost reports since 2021 and have closed or converted since). Next come New York 13 of 56, Tennessee 12 of 54, California 12 of 64. Oklahoma is also an expansion state, it has lost 20 rural hospitals’ inpatient care since 2005 (second only to Texas, 39), and a state-wide road network and its hospitals are small enough to time every block group.

Which hospitals: one written rule

We do not use anyone’s list of “hospitals at risk”. The rule is ours and it is public: a rural hospital meets it when its total margin was below zero in at least 3 of its cost reports for fiscal years 2021 to 2024, and its Medicaid share of charges is at or above the median for rural hospitals in the United States (15.1%). The total margin is net income over net patient revenue plus other income (local taxes, grants, gifts), so a hospital its county keeps afloat does not meet it. Rural means a critical access hospital, a hospital CMS places outside any metro area, or one whose ZIP code is rural or small town by the USDA’s 2020 commuting codes. Hospitals run by tribal nations or the Indian Health Service are left out of the rule: much of their funding is federal appropriations outside the cost report.

In Oklahoma 27 rural hospitals lost money in 3 or more of the 4 years, 46 have a high Medicaid share, and 16 have both. Four more met the rule and have since stopped inpatient care, becoming Rural Emergency Hospitals: Perry and Blackwell in 2023, Stilwell and Nowata in 2026. Meeting the rule does not mean a hospital will close. Many hospitals run losses for years.

Beds and patients a day: HHS, mean of the weeks reported, May 2023 to April 2024, or the latest cost report. Years with a loss and Medicaid share: CMS cost reports, fiscal years 2021 to 2024.

Closures that already happened

CMS’s Provider of Services file keeps every hospital whose Medicare agreement ended. A rural hospital’s agreement often ends for reasons that are not a closure: a new owner, or a change to a critical access hospital, each with a new provider number. So each ending since 2005 is classed by what the file shows at the same address or in the same town afterwards. By that rule 324 rural hospitals in the United States stopped inpatient care: 236 closed, 34 closed and later reopened, and 54 became Rural Emergency Hospitals, a Medicare type that began in 2023 and keeps emergency and outpatient care without inpatient beds. In Oklahoma: 20 (9 closed, 5 later reopened, 6 Rural Emergency Hospitals).

The UNC Sheps Center’s list of rural hospital closures, which reads each case, counts 198 closures and conversions since January 2005 and does not count Rural Emergency Hospitals. For Oklahoma it lists 10; 9 of them are in our list, and the tenth (Valley Community Hospital, 2025) is still active in the third-quarter 2026 file. Its data is not licensed for republishing, so this page compares counts only.

The map

Each block group is shaded by the drive from where its people live to the nearest hospital open to everyone: today, if the 16 hospitals closed, or the minutes it adds. White dots ringed in black are the 16; the other dots are the hospitals that stay open, sized by beds. The coloured rings are the rural hospitals that stopped inpatient care since 2005: press Play to step from 2005 to 2026. Hover over anything, or tap it on a phone, for its numbers.

How much longer the drive

The drives run on the Census Bureau’s TIGER/Line roads of Oklahoma and the 155 counties round it in Texas, Kansas, Arkansas, Missouri, Colorado, New Mexico and Louisiana (the TIGER/Line 2024 roads of all 232 counties), from each of Oklahoma’s 3,374 block groups to the nearest of 285 hospitals open to everyone, so a family near the state line can go to a hospital across it. Today the average Oklahoman lives 10.7 minutes’ drive from a hospital, and the average rural Oklahoman 12.6. If the 16 closed, those would be 12.8 and 18.3 minutes.

In the app, the Drive times row ran the same question live on OpenStreetMap roads for the 104 block groups of seven counties round McAlester (108,640 people) and the 14 hospitals nearest them: an average drive of 20.2 minutes today and 39.2 with the 5 that meet the rule left out. The state run on TIGER/Line roads gives 17.6 and 37.5 for the same block groups: the two maps of the roads differ a little, and the live run knew only the hospitals nearest.

The change falls on a few counties. The biggest rises in the average drive: Pittsburg County 14 to 52 minutes; Pushmataha County 21 to 58 minutes; Harmon County 4 to 38 minutes; Woodward County 9 to 41 minutes. In Pittsburg County nearly everyone would be more than 30 minutes from a hospital: McAlester Regional Health Center, one of the 16, is its only hospital.

Beds at the hospitals that stay open

The 16 hospitals have 526 inpatient beds and held 215 patients a day on average in the last year hospitals reported to HHS. In the scenario each closing hospital’s patients are shared among the block groups nearest to it today, in proportion to their people, and each share goes to the hospital nearest that block group without it. Nothing else changes: the number of patients is the same, no beds are added, and the receiving hospitals’ own patients stay as they were.

By that rule 35 hospitals would take patients, 11 of them in a neighbouring state (52 patients a day). Of the 34 whose beds are known, 12 would be 85% full or more (today 4) and 6 over 100% (today 1). The fullest: Coalgate (critical access, 25 beds): 15% to 133%; Mangum (critical access, 22 beds): 49% to 139%; Stigler (critical access, 25 beds): 11% to 78%; Clinton (short-term, 22 beds): 7% to 82%.

Nearest by road is not how hospitals share patients. A critical access hospital keeps at most 25 inpatient beds and sends on the patients who need more. So a second version sends each share to the nearest short-term general hospital instead: 20 hospitals take patients, 10 of the 19 known would be 85% full or more (today 5) and 6 over 100%. The fullest: Muskogee (short-term, 165 beds): 62% to 92%; Clinton (short-term, 22 beds): 7% to 164%; Fort Smith, AR (short-term, 198 beds): 93% to 107%; Vernon, TX (short-term, 35 beds): 29% to 86%.

How it was made in Plotlay

1. A map project over Oklahoma on the USGS National Map topo base map (public domain).
2. States: Take all from the Census file, keyed by their postal code, linked to the rule’s count and to a table of closures each year with the year as a time axis.
3. Closures: the 20 Oklahoma hospitals as discs 8 km across (big enough to see across the state), keyed by CMS number, with their status at the end of every year from 2005 to 2026 on the slider.
4. Hospitals: Take all, 293 discs keyed by CMS number, linked to their beds, the rule’s inputs and the scenario, coloured by what happens to them.
5. Block groups: Take all, 3,374 keyed by GEOID, linked to their people and drive times, coloured today, if they closed and by the minutes added. Formula columns work out each one’s people times minutes and its people over 30 minutes.
6. Counties: roll-ups add up the block groups inside each county, and formula columns divide them into the average drive.
7. Drive times: the state-wide run is Plotlay’s own drive-time engine (the code the app and the connector run) over TIGER/Line roads in Node, in about 20 seconds; its travel-time bands are layers of their own. In the app, the Data layers Drive times row ran live on OpenStreetMap roads for seven counties round McAlester: once as today, once with the hospitals that meet the rule left out, each drawing its bands.
8. Charts, map pictures and a one-page brief in the Analysis workspace, exported as PNG; every layer exported as GeoJSON, TopoJSON and CSV with WKT.

More about Data layers · Make a data layer of your own

Power BI

Python

A narrated Jupyter notebook reads the tables and shapes, applies the rule again from the cost reports, and recomputes every number on this page.

Download the notebook (with its outputs) and the data. Both are also in the repository on GitHub.

Read it carefully

• A scenario, not a forecast. Nothing here says any of these hospitals will close. The rule picks hospitals that lost money in most recent years and draw more of their charges from Medicaid than most rural hospitals. It does not weigh reserves, debt, a parent health system, local taxes still to come or the Rural Health Transformation money.
• Closing could mean converting. Six Oklahoma hospitals have become Rural Emergency Hospitals since 2023. They keep an emergency department, so the drive to emergency care would not change as much as the drive to a hospital bed, which is what is timed here.
• Drive times are modelled on TIGER/Line roads at a typical speed for each kind of road: 105 km/h (65 mph) on interstates and other limited-access roads, 80 (50 mph) on US and state highways, 45 (28 mph) on local roads, 60 on ramps, 50 on service drives, and 20 km/h from a block group’s centre of population to the nearest road. Roads count both ways; no traffic, weather or ferries. They were checked against a second, independent routing in plain Python on the same roads: the median block group differs by 0.2 minutes, and 95% have the same nearest hospital.
• Hospitals open to everyone. Tribal and Indian Health Service hospitals are on the map but are not counted as anyone’s nearest hospital: they serve American Indian and Alaska Native patients first. Rural Emergency Hospitals have no inpatient beds and are not destinations either.
• Beds and patients are each hospital’s mean over the weeks it reported to HHS from May 2023 to April 2024, all inpatient beds included, or its latest cost report (for the 16 hospitals in the scenario: 14 from HHS, 2 from cost reports). A hospital that did not report and has no cost report has its certified beds and no patient count; it is left out of the bed counts.
• Patients go to the nearest hospital. In life they go where their doctor, their insurer or the ambulance takes them, and complex care already goes to the cities. Both versions here are simple on purpose, and both are stated.
• Counts of closures depend on the rule. Ours reads the Provider of Services file mechanically; the Sheps Center reads each case. Their totals differ for that reason.
• No patient-level data. Every value is a total per hospital, block group or county from public files.

The other United States showcases

• How full were LA County’s hospitals?: every hospital week by week, June 2020 to April 2024.
• Where are the national parks really crowded?: visits per km² of each park’s visitor core.
• Which buildings face Local Law 97 fines?: every benchmarked New York City building against its limit.