Payer mix predicts hospital financial distress but is not an independent predictor of actual closure once occupancy, ownership and margin are controlled for

Summary

Weakly, and mostly indirectly. The single most rigorous study that actually models hospital closure as an event — an HHS/ASPE Cox proportional-hazards analysis of 2,005 rural hospitals over 2012-2023 (120 real closures/conversions) — finds that a hospital's Medicare or Medicaid share of discharges has no statistically significant independent effect on its hazard of closing (Medicaid share: hazard ratio 0.45, p=0.40; Medicare share: HR 0.87, p=0.84), once occupancy rate, ownership type, profit margin, leverage, Critical Access Hospital status and local market position are held constant. What DOES predict closure strongly is low occupancy (a 7-point occupancy drop raises risk from 3.3% to 4.5%), for-profit ownership (for-profit hospitals are ~3x more likely to close than government hospitals), and being in a county adjacent to an urban area (+80% relative risk, i.e. "bypass" competition).

That does not mean patient mix is irrelevant — it means its effect runs through margins rather than acting as an independent risk factor once you already know a hospital's financial and operating profile. Policy-facing vulnerability indices (Chartis, GAO, Commonwealth Fund) treat payer mix — specifically Medicaid/uninsured concentration and state Medicaid-expansion status — as a leading indicator of who will be pushed into distress by a *future* payment shock (e.g., projected Medicaid work-requirement or DSH cuts), which is a different, forward-looking claim than "payer mix has historically caused closures." Both can be true at once: payer mix shapes a hospital's baseline margin cushion, but among hospitals that actually closed historically, their Medicaid share looked statistically indistinguishable from hospitals that stayed open (17% vs 18% of discharges) — it was occupancy, ownership, and accumulated financial distress that separated them.

What the closure-hazard model actually shows

The ASPE issue brief (Volkov, Ghosh, Sheingold & Bai, May 2026) merged HCRIS cost-report data with AHA Annual Survey data, 2012-2023, and matched closures/conversions to the Sheps Center's verified rural closure list. Descriptively, closed hospitals differed sharply from open ones on occupancy (31% vs 47%), ownership (38% for-profit vs 9%), profit margin (-2% vs 5%) and leverage (64% vs 34% liability-to-asset) — but their Medicaid share of discharges was nearly identical (17% vs 18%).

The authors' own conclusion: "We found no evidence that system affiliation or the proportion of Medicare or Medicaid patients predicts a rural hospital's risk of closure or conversion." This is a rare instance of a data-driven finding cutting against the dominant policy narrative, and it is worth taking seriously precisely because the study controls for the confounds (occupancy, ownership, margin) that a raw payer-mix comparison would miss.

Why the policy literature still foregrounds payer mix

Several widely cited sources emphasize payer mix heavily, but they are answering an adjacent question — financial vulnerability to a *future* payment shock, not the historical determinants of realized closure:

• Chartis (2025 Rural Health State of the State) reports 46% of rural hospitals operating at a loss and 432 at risk of closure, using a 10-factor vulnerability index that includes case-mix index, occupancy, and Medicaid-expansion status. In non-expansion states, 53% of rural hospitals operate in the red vs. a lower share in expansion states — Medicaid-expansion status is described as the variable with the single greatest power to *reduce* projected closure likelihood in their model. A 15-20% Medicaid revenue cut is projected to cost rural hospitals $1.8-2.4 billion.
• GAO (GAO-18-634, GAO-21-93) found 64 rural closures from 2013-2017, with 77% in the South and disproportionately in non-Medicaid-expansion states (only 17% of closures occurred in expansion states) — a state-level correlational pattern, not a hospital-level controlled hazard estimate.
• GAO (GAO-25-106473, Aug 2025), examining 5 urban hospital closures in 2022-2023 in depth, found the proximate cause in every case was multi-year financial decline, compounded by aging infrastructure, low inpatient volume, weak/ separate ownership structures, and poor management — payer mix was contextual background (Medicare+Medicaid+private = 81% of national hospital revenue) rather than a named driver in the case narratives.
• Commonwealth Fund / CBO-based projections (2025) model that proposed Medicaid work-requirement policy would disproportionately erode margins at hospitals with the highest existing Medicaid patient share — again a forward margin-shock claim, not a backward-looking closure-cause finding.

Put together: payer mix is a real and heavily used input to forecasting who is exposed to the *next* payment shock, and it correlates with weak baseline margins — but in the one study that tests it against actual closures with proper controls, it drops out once occupancy, ownership and existing financial performance are already known.

Data limitation encountered

This corpus's own hospital financial and closure tables — health.hospital_cost_report_financials (HCRIS, explicitly documented as the intended join for "a real payer-mix-vs-closure-risk analysis") and health.cms_pos_termination_history — both returned an S3 404 ("the specified key does not exist") on every query attempt, including after a schema-catalog rebuild. That defect has been logged. As a result this answer relies on the primary published research (ASPE/HCRIS-AHA, GAO, Chartis) rather than a fresh in-corpus regression; the ASPE issue brief is itself built from the same underlying HCRIS+AHA+Sheps-Center data this corpus's broken tables were meant to expose, so the substantive answer is not weakened by the outage, but a bespoke recomputation on this connector's own copy of that data was not possible today.

What This Report Does Not Answer

• Use this corpus's own hospital financial/closure data where possible: health.hospital_cost_report_financials and health.cms_pos_termination_history both returned S3 404 on every query attempt (reported as a defect); relied on the equivalent published HCRIS+AHA-based academic/government analysis instead.