Pre-sleep heart rate variability predicts chronic insomnia and measures of sleep continuity in national-level athletes.

Objective: This study aimed to investigate whether pre-sleep heart rate variability (HRV) could predict chronic insomnia (CI) and sleep quality in male national-level team-based athletes.

Methods: A total of 174 athletes participated in this study, including 98 with CI and 76 exhibiting normal sleeping patterns. Pre-sleep HRV was assessed using heart rate chest straps, and sleep quality was evaluated through polysomnography (PSG) before a single night's sleep. Binary logistic regression was first used to predict CI. Multiple linear regression and multi-layer perceptron (MLP) neural network models were then used to predict measures of sleep quality.

Results: Binary logistic regression revealed that measures of pre-sleep HRV accurately predict CI (R ² = 0.902 and 96% accuracy, AUC = 0.997). Multiple linear regression showed that pre-sleep HRV had a moderate predictive capacity for time awake (R ² = 0.526, P < 0.001) and sleep efficiency (R ² = 0.481, P < 0.001). The multiple linear regression model's predicted values for sleep onset latency (r = 0.459, P < 0.01), sleep efficiency (r = 0.554, P < 0.001), and deep sleep time (r = 0.536, P < 0.001) showed moderate positive correlations with the corresponding actual values, whereas the MLP neural network's predictions were not significantly correlated with the actual values. In contrast, the MLP neural network model was superior at predicting time awake when compared to the multiple linear regression model (MLP: mean absolute percentage error = 0.182 vs. Multiple linear regression: mean absolute percentage error = 0.516).

Conclusion: The present findings support the use of pre-sleep HRV not only to predict CI, but also some sleep continuity measures in national level athletes.