The rule
What hospitals actually have to publish
The rule requires two things. First, a consumer-facing display of
shoppable services, which most hospitals satisfy with a price estimator
tool. Second, and the part this site is built on, a single
machine-readable file of all standard charges: gross charges,
discounted cash prices, payer-specific negotiated rates, and
de-identified minimums and maximums, updated at least yearly.
Since 2024 hospitals must also advertise where that file lives, in a
standard text file at a fixed address on their own website:
cms-hpt.txt at the domain root. That one requirement is
what makes a site like this possible. Finding a hospital's file is a
lookup, not a research project, as long as the hospital complies.
The reality
What the files look like when you open them
They are not pamphlets. The largest single file we have parsed carries
over 13 million rows, and one hospital system's files run to hundreds
of megabytes zipped. Formats vary between a wide CSV, a tall CSV and
JSON, each with its own quirks, and one parser here handles every
publisher shape we have met. Nothing about them is designed for a person with a
spreadsheet, which is the practical reason so few patients have ever
seen a number from one.
Compliance is uneven in both directions. Of the
551 hospitals in the metros we cover, we could find no
machine-readable file at all for 138,
after checking the mandated address, the system's other domains, and
the rebrands that make an old domain go quiet. Each of those hospitals
gets a page here saying exactly that, because "we looked and found
nothing" is a finding about compliance, not a gap in our coverage.
The catch
A compliant file can still tell a patient nothing
The discounted cash price column is mandatory. Filling it is not, in
practice. We have read files that are millions of rows long, with the
required column present in every row and empty in every row. Other
hospitals leave the column blank and put the cash price in a payer row
labelled "Self Pay" instead, which a naive reader misses entirely; we
read those too, and mark them, because pricing the self-payer like an
insurer is not the same claim as filling in the mandated column.
Then there are the traps inside the numbers themselves: one code
carried at seven different prices with nothing saying which applies, a
single contract figure attached to nearly every service in the file,
rates expressed as percentages that read as sub-dollar prices. Each of
these would have made our numbers look better than the truth, which is
why they are measured and published on
what we found reading these files rather
than quietly worked around.
How to use them
What the files are good for, honestly stated
Used carefully, the files answer questions nothing else can: what a
named hospital says it charges a self-pay patient for a named code,
how that compares across a city, and whether the hospital publishes
anything at all. Those answers are real leverage before a scheduled
test or procedure, and the practical playbook is in
how to find a hospital's
self-pay price before you go.
What the files cannot do: they carry facility charges only, so the
doctor's own fee is missing; they are not quotes; and the same code is
not always the same unit of sale at two hospitals, which is why
blanket "cheapest hospital" claims built on these files deserve
suspicion. Where prices genuinely diverge and why is its own story:
why the same test costs twice as
much across town.