Checklist

Clinical Decision Support & AI Safety Checklist (Interactive)

An interactive pre-deployment and post-deployment checklist that helps teams confirm clinical validation, usability, data integrity, equity, monitoring, and governance for CDS and clinical AI. The form records evidence, responsible owners, monitoring metrics, and approval decisions for auditability.

Interactive Tool

Clinical Decision Support & AI Safety Checklist

Use this checklist to confirm clinical validation, usability, equity, data integrity, monitoring, and governance steps before and during deployment of clinical decision support (CDS) or clinical AI. Record owners, evidence links, monitoring metrics, and approval decisions so the deployment can be audited and followed up.

Complete the required fields and attach or link supporting documentation in Evidence fields. Use the stage selector to indicate whether this is a pre-deployment review or a post-deployment monitoring check.

Name of the person completing this checklist.
e.g., Clinical Lead, CDS Engineer, QI Manager.
Where the CDS/AI will be used (department, clinic, ward).
YYYY-MM-DD.
Includes retrospective validation, simulated cases, and prospective pilot testing where possible.
Link to validation report, datasets, test-cases, and key performance numbers.
Human factors review, workflow mapping, and alert volume analysis completed.
Summarize UI wording changes, alert suppression rules, or workflow adjustments made to reduce fatigue.
Sources, transforms, refresh cadence, and feature dependencies are recorded and accessible.
How will upstream schema or source changes be detected, tested, and mitigated?
Includes subgroup performance checks and documentation of known limitations and mitigations.
Summarize differential performance and corrective steps or restrictions on use.
Plan should specify what is monitored, acceptable ranges, owners, and escalation rules.
Examples: AUC, sensitivity, specificity, predictive values, alert volumes, override rates, time-to-action, equity subgroup metrics.
Describe concrete thresholds that will trigger review or escalation (organization-specific).
Enter percent or leave blank if unknown.
Enter percent during monitoring reviews.
Who to contact when thresholds are exceeded or harm suspected?
Provide names, roles, pager/email, and expected response times.
Clear steps to disable or revert the tool quickly if harm is detected.
Link to runbook or describe short playbook steps and responsibilities.
Mechanisms for users to report false positives, false negatives, workflow problems, or suggestions.
Evidence of training, quick-reference guides, or in-system tips.
Logs should capture inputs, model/version, timestamp, and user overrides for investigations.
DPIA or equivalent, and security review completed where required.
Internal legal or external regulator reviews as applicable.
Person granting approval (name).
e.g., Medical Director, CIO, Chief Quality Officer.
YYYY-MM-DD.
Any outstanding items, owners, and target dates for remediation.
You can explore this tool now. Sign in or create an account to save your responses and return to them later.
Sign In to Save Create an Account
Make this tool part of your work

Save a personal copy, bring it to your team, or tailor the questions and workflow to fit what you are hungry to improve.

Member customization and team collaboration are coming soon.
Discuss Ask the Librarian

Discussion

Comments and conversation will live here.