We present a solution to the unofficial phase of George B. Moody PhysioNet Challenge 2026 by the ISIBrno-AIMT team, which ranked 1st out of 246 submissions achieving AUROC 0.817. We developed a deterministic, single-variable model defined as "y = age / 100", where y represents the predicted likelihood of a future cognitive impairment diagnosis. This approach isolated age as the solitary predictive feature, operating entirely blind to the patients' actual polysomnogram recording.
The success of this baseline highlights the strong predictive role of age in the unofficial Challenge phase, indicating that age itself remains highly informative at evaluation time despite reduced reliance on it during training. These findings suggest that age-related effects may substantially influence model performance; however, they also motivate us to further investigate whether physiological signals provide additional predictive value beyond age in the official challenge phase, including within specific age groups.