A model with a good average score can still miss sepsis more often in older patients, or produce alerts that nurses cannot act on. Evaluation therefore looks at several things at once: accuracy, performance in each patient group, performance at other sites, and whether people can use the tool in daily work.
Prerequisites¶
| Prerequisite | Importance | Level |
|---|---|---|
| Completed development phase checklist | Necessary | Not applicable |
| Held-out test set (data kept apart and not used during development) | Necessary | Not applicable |
| Understanding of statistical validation | Necessary | Intermediate |
| Access to representative users for usability testing | Helpful | Not applicable |