Results
What the check found across the 28 demonstration notes. SIMULATED synthetic patients; not a measured rate.
7 notes leave out something the results show.
3 notes say more than the results support.
8 notes could not be checked, and the screen says why.
Agreement with a coder
SIMULATED Label set synthetic-demo: synthetic labels written for this demonstration; replace with a coder's labels. 33 checks on 28 notes.
- Alerts the coder would also query (precision): 12/13 = 92% [67%–99%]
- Coder queries the tool also raised (recall): 12/13 = 92% [67%–99%]
- False alerts per 100 notes: 1 in 28 notes = 3.6
- Does the note state it? Agreement on presence: kappa 0.94, AC1 0.94, positive 97%, negative 97% (n = 33)
Every miss (1)
- KE-P10.kidney: coder would query; tool said “Not assessable: the data cannot decide” · coder: “stated 3b but the latest results are mixed; I would ask”
Every false alert (1)
- KE-P13.kidney: tool said “Definitions disagree; the tool does not choose”; coder would not query · “outside lab report is enough for me; would not query”
Import a coder's labels (CSV)
Both directions, by who drafted the note
| Notes | n | Met, not stated | Stated, not supported | Stage differs (higher / lower) | Definitions disagree | Not assessable | Read it yourself | Net |
|---|---|---|---|---|---|---|---|---|
| All notes | 28 | 7 | 1 | 2 / 2 | 1 | 10 | 0 | +6 |
| AI-drafted | 6 | 2 | 0 | 2 / 0 | 0 | 1 | 0 | +2 |
| Clinician-typed | 22 | 5 | 1 | 0 / 2 | 1 | 9 | 0 | +4 |
Alert budget
SIMULATED <built-in method items of dict object at 0x7925fc2104c0> things to look at across 28 notes = 46 per 100 notes against a budget of 25 per 100 (an assumed planning figure; set per site). Over budget.
Decisions recorded: 0.