How the numbers are made — and what they can’t tell you

Where the data comes from

Every figure is parsed directly from IRS Form 990 e-file XML — the hospitals’ own signed filings — primarily Schedule H (Hospitals). Charity care means financial assistance at cost (Part I, line 7a, net of offsetting revenue) — never “at charges,” which inflates it several-fold. Bad debt (Part III) and Medicaid shortfall (line 7b) are always shown as separate figures; they are not charity and we never blend them. Every hospital-level figure links to the exact filing it came from. Aggregate figures — counters, medians, ranges — link instead to a page showing how each was calculated and over what universe, because an aggregate has no single filing behind it.

What the percentage actually measures

Charity care ÷ total expenses. It measures effort relative to size — how much of what this hospital spent went to care it knew it would never be paid for. It is not a measure of quality, of community benefit overall, or of how kind the staff are.

Its five real weaknesses

1. Demographics move it — legitimately

A hospital in a wealthy, fully insured suburb has fewer uninsured patients to forgive. A hospital serving a poor county has more. Some of the spread between hospitals is the population, not the policy. This is the single biggest caveat, and it cuts both ways: it can understate a suburban hospital that gives freely to the few who need it, and overstate a hospital in a poor county that is simply surrounded by need.

2. Medicaid expansion changes the denominator of need

In states that expanded Medicaid, patients who would once have been charity cases are now insured — so charity care falls while the Medicaid shortfall rises. That is why we always show Medicaid shortfall as its own figure beside charity, never merged into it.

3. The line between “charity” and “bad debt” is a policy choice

The same unpaid bill can be recorded as charity care (if the hospital screens the patient and forgives it) or as bad debt (if it doesn’t, and sends it to collections). A hospital that screens patients well converts bad debt into charity care. One that doesn’t screen looks like it gives less. This is the most manipulable seam in the whole measure — which is exactly why we publish bad debt on every page. Read the two together: a low charity figure sitting beside a large bad-debt figure is a different story than a low charity figure with almost no bad debt.

4. Costing methods vary

“At cost” relies on each hospital’s own cost-to-charge ratio. Two hospitals doing identical work can report different numbers because their accountants converted charges to cost differently.

5. Filer boundaries blur the picture

A big system may file one 990 covering many hospitals, so the percentage is a system-wide average that can hide a hospital giving well above the rate and one giving well below it inside the same number. Fiscal years also differ — we always label the year, never mix them silently.

So how should the number be read?

Never on one number. Four signals are harder to explain away together than any one of them alone:

The benchmark

The 2.3%-of-expenses average comes from Bai et al., Health Affairs — roughly $1 of every $43 spent. One honest note about basis: that literature draws partly on Medicare cost-report data, which defines charity care slightly differently than the 990 does. Measured with our own ruler — the same Schedule H line, same parser, every filer nationally — the sector median is about 1.0% (roughly $1 in $97), with half of all hospitals between 0.5% and 2.2%. We show the 2.3% benchmark because it is the published standard, and we tell you where our own measurement lands so you can judge both.

Eligibility for free care

Each hospital reports the income cutoff for free and discounted care in its financial assistance policy, as a percentage of the federal poverty level. We publish it as filed, and convert it to dollars using the HHS 2026 poverty guidelines ($15,960 for one person, $33,000 for a family of four in the 48 contiguous states). Those lines run from 40% to 600% of the poverty level, with a median of 200% — hospitals choose this number, the law does not set it. Income is also not the only test. A peer-reviewed study of financial assistance policies found 64% of hospitals also weigh assets, 8.4% require citizenship or local residency, and a significant minority restrict eligibility for insured patients; policies generally cover medically necessary care rather than elective procedures, and the physician’s bill is often separate from the hospital’s. Schedule H does not capture any of that, which is a real limit on what our eligibility figure can tell you. Policies also change between filing and today — always confirm the current one. How to ask →

The statewide counter

• For each nonprofit filer operating hospital facilities in Ohio, at its latest filed year: shortfall = max(0, 2.3% × total expenses − charity care).
• The counter is the sum of those shortfalls. Hospitals above the average do not offset hospitals below it — it answers “how much more free care if every below-average hospital rose to average,” not a net figure.
• Filers reporting $0 or nothing at all contribute zero — we never impute. The counter is a floor.
• Fiscal years vary by hospital (“latest filed year per hospital”).
• This is a benchmark-relative aggregate. It is not individually claimable dollars, and it implies no individual entitlement.

What we never do

• Zeros are never ranked. A legally filed $0 (offsets exceeding gross) appears labeled, in its own section — never in a ranking at 0.0%.
• Nothing is imputed. “Not reported” renders as not reported.
• No single composite score. Any weighting of these signals is an editorial judgment; we hand you the separate numbers instead.
• Attestations are not court records. Collection-action answers on Schedule H are the filer’s own attestation and are labeled that way — in either direction.
• Patients are never named. When the court-records layer lands, defendants appear only as counts and ranges.
• Questions, not conclusions. The filings raise the questions; readers draw the conclusions.

Corrections

Auto-generated pages carry a verification banner. If a figure doesn’t match its linked filing, we want to know: cashpaymed.org.

A public-interest project of CashPayMed. Informational only — not legal, tax, or medical advice.