Figure 7.

The Shapley additive explanations (SHAP) value analysis of 4 wrong prediction cases. This figure shows the SHAP value analysis of 4 wrong prediction cases. The upper 2 cases demonstrate the substantial impact of the average heart rate on the prediction, resulting in false-negative or false-positive outcomes. By contrast, the lower 2 cases exhibit the minimal or negligible effects of the average heart rate, with other factors dominating the prediction results. Note: features with name ending in “_to_m” represent the calculated difference between the original feature and past 4 week average in clinical assessment and between the original feature and past 7 day average in wearable device data. HR: heart rate.