Methodology and conflict of interest
Short version: every time figure comes from a cited study or a named vendor's own page, every tool fact from the vendor's site on the date shown, and the worked example is synthetic. Medrecords AI publishes this site and sells 1 of the tools on it, so we follow 6 written rules, listed below.
Limits: We have not timed the listed tools on a shared test file: time figures come from the cited studies and vendor pages, and the worked example is synthetic.
Where the numbers come from
| Source | Used for | Read |
|---|---|---|
| Peer-reviewed studies: Brysbaert 2019 (reading rate), BMC Med Inform Decis Mak 2022 (abstraction time), Ann Intern Med 2020 via Healio (EHR chart review time) | Time benchmarks | 2026-11-12 |
| Peer-reviewed studies of AI summaries: npj Digital Medicine 2025, JAMA Network Open 2026, Nature Medicine 2024 | Error types and rates | 2026-11-12 |
| Vendor pages: medicalrecordsreview.com, medchronosai.com, botminds.ai, assemblysoftware.com | Vendor-stated reading paces, labeled as vendor estimates | 2026-11-12 |
| 20 tool vendors' own websites | The tool table: price, trial, BAA, built for | 2026-11-12, linked per row |
| California Labor Code 4628, 8 CCR 9795, 8 CCR 10682; 42 CFR 482.43 | Rules on who summarizes, what reports list, fee schedule pay, transfers | 2026-11-12 |
| Google results for 4 searches | What existing pages cover, and what this site adds | 2026-11-12 |
If a vendor doesn't publish a fact, we write "not published" or "not on the page read" rather than guess. We don't use paid placements or affiliate links. No vendor paid to be listed.
How the time benchmarks were built
We converted each published pace to minutes per 100 pages and kept the source's own range. We label vendor figures as vendor estimates, because those companies sell against manual review. We did not time anyone ourselves. File-level hours (for example 20 to 50 hours for 2,000 pages) are our arithmetic from the 40 to 100 pages an hour range. Dollar figures for physician time use $200 an hour as an example input, inside the range Sermo reports for legal chart review.
How the worked example was built
The 40-page file on the worked example page is synthetic: we wrote every page summary, name, date and finding. No real patient's records were used. We wrote all 3 summaries, including the AI-style draft, which shows the error types the studies report and is not output from any product. The comparison table describes those 3 versions; it is not a measured comparison of methods.
Our conflict of interest, and the rules we follow
Medrecords AI publishes this site and sells Medrecords AI, AI medical record review software that appears in the tool table. That is a conflict. These are the rules we hold ourselves to:
- A disclosure appears at the top and bottom of every page.
- Our product is labeled as ours wherever it appears.
- Our product gets the same table columns as every other tool, and our page lists who should pick something else.
- Links to medrecords.ai appear only on the page about our product and in boxed, labeled notes on a few pages.
- We make no claim about another product that its own website doesn't support.
- Vendors can correct their own entries by emailing [email protected]; confirmed fixes go live within 2 business days.
The method on this site works with any tool or none. The templates, the checks and the calculator don't assume Medrecords AI.
How often pages change
We recheck tool prices and the benchmark sources at least every 45 days. The "Updated" date on each page is the last time its facts were checked.
Questions about how this site is made
Did any vendor pay to be listed?
No. The site has no paid placements or affiliate links, and no vendor can pay to be listed or moved.
Why are some time figures from vendors?
Because no professional body publishes a standard pace for medical record review. We show vendor figures next to the peer-reviewed ones, labeled, so you can weigh them.
Is the worked example real?
No. It is synthetic, written for teaching, with no real patient's information.
How do you keep your own product from getting an edge?
The 6 rules above: labeled everywhere, same columns as everyone else, links limited to 1 page and boxed notes, and a list of who should choose another tool.