Separate calibration, discrimination, and transport

Claim: STORM should report calibration, discrimination, and transportability as separate outcomes rather than collapse them into one aggregate metric. Reasoning: a single score can look acceptable while a site-specific calibration failure or transport failure remains hidden. Test/falsification: evaluate held-out sites with calibration slope/intercept, discrimination, and transport performance reported separately; if the aggregate improves but one component fails, the deployment claim is not supported.