Towards Clinical AI Fairness: Filling Gaps in the Puzzle