Adaptive data selection improves wearable prediction under low baseline performance

In three linesStudy on adaptive data selection for wearable prediction across 3 modalities (heart rate, activity, EMA) in longitudinal dataset. AUROC gains up to 0.7 for low-baseline participants, strong inverse correlation (r=-0.67) between improvement and baseline performance. Limited benefit for strong baselines.

Summary generated by Claude — human-verified