Step

Critical Question

Evidence Required to Proceed

1

Is the tool technically valid for my population?

External validation in ≥ 1 independent dataset; performance metrics stratified by age, sex, severity; evaluation in community practice settings; imaging protocol compatibility documented

2

Does it improve patient outcomes relevant to conservative care?

Evidence linking AI outputs to functional outcomes (pain, disability); impact on treatment decisions documented; evaluation in non-surgical populations; relationship between detected findings and clinical presentation established

3

Will it reduce or exacerbate health disparities?

Training data demographic composition reported; performance analyzed across race/ethnicity, socioeconomic status; validation in safety-net/community settings; accessibility barriers identified and addressed

4

Does it support or replace clinical reasoning?

Explainability features present; positioned as decision support not autonomous diagnosis; practitioner override mechanisms clear; training resources available; automation bias risks assessed

5

Are safety governance mechanisms adequate?

Regulatory clearance appropriate for intended use; post-market surveillance plans documented; adverse event reporting mechanisms exist; version control and update procedures clear; liability framework defined

6

Does it support patient-centered care?

Patient communication supports available; informed consent processes defined; integration with shared decision-making demonstrated; psychological impacts (anxiety, medicalization) considered