How is the final test-set ranking computed (2026 Challenge)?

[indie Cha]

Dear Challenge Organizers,

Thank you for running the 2026 Challenge. 

The scoring page and leaderboard show two metrics — the age-conditioned AUROC and the prevalence-based reward — but I could not find how they combine into the final test-set ranking.

Could you clarify whether the final ranking is based on the age-conditioned AUROC, the prevalence-based reward, or some combination of the two? If it is a combination, what is the formula?

Since the two metrics can favor different operating points, this would help teams target the correct objective.

Thank you for your time.

[PhysioNet Challenge]

Dear Indie,

We will use the age-conditioned AUROC metric to determine the final rankings for the 2026 Challenge.

However, we are reporting both metrics on the leaderboard because, like you noted, different metrics can provide different context to model performance.

In particular, the age-conditioned AUROC metric evaluates a model across operating points from the models' real-valued outputs, the prevalence-based reward metric evaluates a model at a specific operating point from the models' binary outputs, and both metrics evaluate the ability of a model to predict future cognitive impairment diagnoses while addressing age as a known covariate for cognitive impairment.

Best,
Matt
(On behalf of the Challenge team.)

Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address. This email is maintained by a group. Please do not email us individually.