Generalizability is the practical question behind every deployment decision: will this work here. It is established through external validation, ideally across several sites that differ meaningfully from the development setting.

Where This Gets Misread

Generalizability is not a property a model has or lacks in general ... it is always relative to a specific target setting. A model validated across five academic centers may still fail in a community hospital with older equipment and a different population, and that result would not contradict the published validation.