Health, home
For developers: Deedglass for Developers Talk to us
Talk to us about a pilot Deedglass for Developers

What could unnoticed AI mistakes cost you?

Sometimes an AI tool says it saved one thing, and your system saved another. This page helps you put a number on that. You enter your own figures. The answer is your estimate, not a promise.

Your estimate, made at deedglass.com/healthcare/savings.

Your numbers

Currency
a day

How many times a day AI tools save something to your HIS or billing system. Count only writes you plan to route through a check.

%

Your best guess, or a figure from an audit. Our own tests were built to cause mismatches, so they can’t tell you your rate.

What happens after a mismatch? Split it three ways.

Together, the three shares can’t be more than the whole. For context, US figures put claims denied at first at 11.8% (2024) [3] to nearly 15% for private payers [4]. That is a general figure, not a mismatch rate.

%
%
%
What each one costs you
$

Filled in: $43.84, Premier’s US average for one denied claim in 2022, all payers, reported by HIA [2]. A 2016 study of 724 hospitals found about $118 [1]. Use your own figure if you have one.

$

Filled in: $96, from one Texas hospital, published in 2009 [5]. Others give $50 a pair, and $1,099 on average once cleanup and repeat tests are counted [6]. These are old, so check them.

minutes
$
%

Start at 0%: you find out later, or never. Reference rate, for context only: in Deedglass’s FHIR write benchmark of 17 made-up cases on Medplum 5.2.6, it caught 4 of the 11 where the server said success but stored something else (36%). It missed 7. Those cases were built to be hard, so they are not your rate.

Your estimate

Fill in your numbers to see your estimate.

How it is worked out

  1. Mismatches a year = writes a day × 365 × mismatch share
  2. Cost of one mismatch = claim share × claim cost + record share × record cost + staff share × (minutes ÷ 60) × hourly cost
  3. Cost a year = mismatches a year × cost of one mismatch
  4. Part a check would show = cost a year × share caught after the write

A check shows you a mismatch. Your team still has to fix it, so this is not money saved by itself.

Read this first

What this number is, and isn’t.

It is your estimate

Every figure can be changed. The US figures are only a starting point, and some are old. Your own numbers are always better.

It is not a measured saving

No hospital has measured savings with Deedglass yet. We will only quote measured numbers after a pilot measures them.

See how it works in a hospital Deedglass Health

Sources

  1. 1Change Healthcare claims study of 724 hospitals (2016 claims), as reported in 2017: about $118 to rework each denied claim · hcinnovationgroup.com · checked 21 October 2026
  2. 2HIA, citing Premier: an average of $43.84 per denied claim in 2022, all payers · hiacode.com · checked 21 October 2026
  3. 3Kodiak Solutions, 2024 data published May 2025: initial denial rate 11.8% · businesswire.com · checked 21 October 2026
  4. 4TechTarget RevCycle, on a Premier survey of over 500 providers (March 2024): nearly 15% of claims to private payers denied at first · techtarget.com · checked 21 October 2026
  5. 5HFMA (hfm, 2009): a Texas hospital put the cost of one duplicate record at $96 · hfma.org · checked 21 October 2026
  6. 6Becker's Hospital Review: about $50 for each duplicate pair, and $1,099 on average with cleanup and repeat tests (Children's Medical Center Dallas study) · beckershospitalreview.com · checked 21 October 2026

Vendor figures are the vendors’ own. Nothing on this page is legal advice. If something here is wrong, write to hello@deedglass.com.