This site is published by Medrecords AI, which makes 1 of the products it covers. We say which one wherever it appears. How we handle that.

Medical timeline checklist: 12 checks before you trust the picture

By Ahmed Jemaa, Co-Founder and CEO of Medrecords AI. Updated 2026-11-14.

Short version: check the rows first (dates, duplicates, undated pages), then the drawing (true-date axis, same-day events, span ends, color), then the trail back to the record (a page cite on every mark, exports that keep it). Tick the boxes below while you look at a tool's output; the score updates as you go. Under 10 of 12, keep the table as your working copy.

Limits: Tool facts come from each vendor's own website, read 2026-11-12; we have not used those products. The case on this site is synthetic, with no real patient.

Why a timeline needs its own checks

A table that's wrong looks wrong when you read it closely. A timeline that's wrong still looks clean. The eye trusts spacing, color and length before it reads a label, which is the strength of the format and its risk. These checks are about the ways a drawn timeline misleads. They apply to a vendor's view, to a figure you export, and to 1 you draw yourself.

Run them on a closed file you already know, or on the synthetic case, where the answers are known. Tool-by-tool output types are on the tools page.

The 12 checks

The rows behind the picture

1. Every mark matches a row with a page cite
Click 5 marks at random. Each should open a record and a page. A mark with no row behind it is decoration.
2. The date used is the date of service
Find 1 note signed days after the visit. It should sit on the visit date, not the signing or fax date.
3. Duplicates are drawn once
Find a note that appears twice in the file. It should be 1 mark, not 2 dots a day apart.
4. Undated records are listed, not dropped or guessed
The tool should show where undated pages went: a tray, a list or a flag. Count them against your own count.

The drawing

5. The axis uses true dates
A 4-month gap must be wider than a 1-week gap. If events are evenly spaced, the gaps are gone. Any axis break needs a visible break mark.
6. Same-day events are countable
Pick a day with 3 records. The chart should show 3, or a mark that says 3, not 1 dot.
7. Spans with no recorded end look different
A drug with a start and no stop should be marked "end unknown," not drawn to today or to the next visit.
8. Dose and status changes split the bar
A dose increase or a move from modified to full duty should show as a change, not vanish inside 1 bar.
9. Meaning survives without color
Print it in grayscale. Lanes, shapes or labels should still tell you which mark is which.
10. The key date is marked
Date of injury, admission or surgery should be a visible line, so before and after read at a glance.

The trail and the data

11. The export keeps the citations
A PNG or a slide has no links. The export you put in a report should carry page cites, or travel with the cited table.
12. Patient data stays covered
The tool that holds the records has a signed BAA, or the file runs offline. No real records in public sheets or consumer chat accounts.

Score: tick the checks that pass.

How to score a tool in 30 minutes

Pick a closed file you know. 300 to 1,000 pages, with at least 1 known gap, 1 duplicate, 1 late-signed note and 1 undated page. Note where each is before you start.
Run it and open the timeline view. If the tool has no drawn view, export the table and draw it with the template, then score the export.
Go down the 12 checks. Use your known items for checks 2, 3, 4 and 6. Print a page for check 9.
Write down the misses. A miss on checks 1, 4 or 5 is a reason not to use the picture at all. Other misses are workable if you know about them.

What the score means

Score	What to do
12 of 12	Use the timeline for orientation and the cited table for every fact you write down.
10 or 11	Usable if the misses aren't checks 1, 4 or 5. Note the known weakness in your working file.
Under 10	Work from the table. Draw your own figure from it if you need one.
Fails 1, 4 or 5	Don't rely on the picture, whatever the total: it can't be traced, it hides pages, or it hides time.

This checklist covers the picture. Whether the AI read the records correctly in the first place (missed findings, wrong laterality, merged patients) is a separate check on the rows; a timeline only redraws what the rows say.