The Geography of Medicare Spending

Medicare is a single national program with a single set of rules, yet what it spends per beneficiary varies dramatically by geography. This post uses CMS standardized spending data to explore why higher spending predicts neither better outcomes nor lower readmission rates.

Medicare standardized spending per beneficiary varies by more than 2.5× between the cheapest and most expensive states — and by far more at the county level. Yet higher spending predicts neither better outcomes nor lower readmission rates. The data consistently shows weak-to-zero correlation between what Medicare spends and how well patients do.

Medicare is a single national insurance program with a single set of rules. Yet what it spends on a beneficiary in Manhattan, Kansas is dramatically different from what it spends in Manhattan, New York. Some of this is expected: prices are higher in urban markets, specialists charge more, and illness is more complex in some populations. But when you standardize for these factors — removing the geographic payment rate differences so you're comparing actual resource use — the variation persists and remains enormous.

The CMS Geographic Variation Public Use File captures this. Every year, CMS publishes per-beneficiary Medicare spending broken down by geography (national, state, county) and service type, standardized to remove local wage and price differences. The result is a dataset that lets you ask: ignoring local prices, how much Medicare care does this area actually consume?

This post explores that question using the 2023 data (latest available), with trend context going back to 2014.

Standardized Medicare spending per fee-for-service beneficiary in 2023, by state. Darker blue = more spending.

A few things jump out immediately:

• The South spends more. States like Louisiana, Mississippi, Alabama, and Florida consistently appear among the highest spenders. The Northeast and Mountain West tend toward the lower end.
• Utah is a persistent outlier on the low end. Utah has among the lowest Medicare spending in the country, a pattern that has held for decades and is associated with a younger average age, lower rates of chronic disease, and a health system historically oriented around the LDS hospital network.
• The range is substantial. The difference between the lowest- and highest-spending states represents thousands of dollars per beneficiary per year — real money, multiplied across millions of patients.

The obvious question is whether higher-spending states achieve better outcomes. The answer, consistently, is no. This scatter plots standardized spending against the 30-day hospital readmission rate — one of Medicare's primary quality metrics.

The correlation is near zero. High-spending states like Louisiana and New York do not have systematically lower readmission rates than low-spending states like Hawaii or Utah. If anything, the relationship trends slightly positive: states that spend more tend to have slightly higher readmission rates, though the relationship is too weak to be meaningful.

This is one of the most replicated findings in health economics, associated most prominently with the Dartmouth Atlas research: more care does not automatically produce better outcomes. Much of the geographic variation in Medicare spending reflects differences in practice patterns, not patient needs.

Medicare per-beneficiary spending has grown over the past decade, but not uniformly across service types. This chart breaks down standardized national spending by category.

Several patterns stand out:

• Inpatient spending has declined as a share of total spending. Hospital stays are expensive, and the secular trend has been toward shorter stays and more outpatient care.
• COVID-19 visibly disrupted 2020 spending. Total spending dipped as elective procedures were deferred and patients avoided hospitals. Post-acute and home health care were particularly affected.
• Recovery was uneven. Some service categories rebounded quickly (outpatient, E&M); others recovered more slowly.
• Total spending resumed growth after 2020. By 2022–2023, total standardized spending exceeded pre-COVID levels.

State averages mask enormous county-level variation. These are the 20 counties with the highest and lowest standardized Medicare spending in 2023, restricted to counties with at least 1,000 fee-for-service beneficiaries to reduce small-sample noise.

The spread is striking. The highest-spending counties can exceed $20,000 per beneficiary per year in standardized spending — more than twice what the lowest-spending counties spend on the same standardized basis. The geographic clustering is notable too: extreme-high-spending counties tend to cluster in Louisiana, south Texas, and urban Florida. Extreme-low-spending counties tend to cluster in Utah, rural Nebraska, and parts of the Pacific Northwest.

If higher spending doesn't produce better outcomes, why does the variation persist? Several factors:

Practice pattern variation is the most heavily studied explanation. Physicians in high-spending areas order more tests, refer more, hospitalize more — not because patients are sicker, but because local norms evolved that way. Once established, these patterns are sticky: physicians trained in high-intensity environments carry those habits forward.

Supplier-induced demand plays a role where supply of specialists and hospitals is high. More available beds tend to be filled; more available specialists tend to generate more referrals. The healthcare market doesn't clear like a normal market because patients don't pay marginal costs and don't comparison-shop.

Chronic disease burden explains some (but not all) variation even after standardization. The South has higher rates of diabetes, obesity, and hypertension — chronic conditions that generate ongoing Medicare spending. But the Dartmouth research suggests this explains only a fraction of the variation; the rest is attributable to intensity of treatment for equivalent patients.

Fee-for-service incentives reward volume, not value. A system that pays per procedure will always face pressure toward more procedures in areas where practice patterns allow it.

All data comes from the CMS Medicare Geographic Variation by National, State & County dataset, 2014–2023 releases.

• Population: Original Medicare fee-for-service (FFS) beneficiaries aged 65+. Medicare Advantage enrollees are excluded because their claims are not reported to CMS in the same format.
• Standardized spending: CMS standardizes payment rates to remove geographic differences in input prices (wage index, rural adjustments, graduate medical education payments). Standardized spending reflects actual resource use — how much care, not how much it costs locally. This is what allows fair geographic comparison.
• Readmission rate: Percentage of inpatient stays followed by an unplanned readmission within 30 days. Includes all-cause readmissions.
• County minimum: The top/bottom chart filters to counties with ≥1,000 FFS beneficiaries to reduce noise from very small populations with erratic per-capita figures.
• Age level: All figures use the "All" age stratum (65 and over combined).