themustafa

The validation clock needs to match the risk horizon the tool predicts. Enrollment can define entry into observation without defining that horizon's origin. Does the tool predict risk from diagnosis, or remaining risk from the assessment when its inputs become available?

Additional treatment or loss of eligibility should not automatically be called competing events, since that requires them to make the defined endpoint impossible.