Author(s): EGGENBERGER, P., BUFFAT, N., WEBER, T., GUBLER, R., BRUNNER, E., Institution: OST - EASTERN SWITZERLAND UNIVERSITY OF APPLIED SCIENCES, Country: SWITZERLAND, Abstract-ID: 2354

INTRODUCTION:
Youth and adult athletes in high-level sports experience an elevated risk for injury and illness during phases of higher training and competitive load. When recovery is neglected under these circumstances, psychological and physical health problems might arise. These are referred to as nonfunctional overreaching or overtraining syndrome, OTS (1,2,3). OTS is characterized by athletic performance beeing reduced for more than 3-4 weeks up to months. Concurrent symptoms include mood and sleep disturbances, feelings of depression, respiratory tract infections, and weight loss, among others. This condition is highly prevalent, with 10-20% of young adult and about 29% of young athletes from various sports beeing affected (4). To our knowledge, no validated and reliable measurement system currently exists, that would allow for preventive, early diagnosis of overreaching states that might lead to OTS (5). The aim of this study is to develop and evaluate a multiparameter prediction model to assess the recovery and stress state of athletes.
METHODS:
Twenty-five male ice hockey players from the highest level Swiss leagues, at their respective age groups, participated (i.e., National League, n = 11, age = 24.8+-4.1 years and U20, n = 14, age = 18.5+-1.5 years). Over 5 - 10 weeks during the in-season (i.e., competition phase) measurements were performed on 10 separate days, either after 1 day of recovery or after a day with match/intensive training. The test battery comprised 40 predictor variables from counter movement jump (CMJ), heart rate variability (HRV), executive function, tympanic temperature, weekly rating of perceived exertion, and sleep measurements. The Stress Recovery Short Scale (SRSS) served as reference variable. Least Absolute Shrinkage and Selection Operator (LASSO) regularized regression analysis was performed for variable selection, training, and cross-validation of a binomial prediction model.
RESULTS:
Based on 163 timepoints of measurement, our developed LASSO regression model predicted SRSS scores < 5 (i.e., highly stressed state) with very good performance (area under curve, AUC = 0.921, sensitivity = 0.889, specificity = 0.843). The prediction model retained variables from all applied measurement methods, with HRV, CMJ, and sleep beeing represented among the 10 most important predictive variables.
CONCLUSION:
Due to the multisystemic nature of overreaching and OTS states, we conclude that a multiparameter prediction model, containing relatively easily measurable parameters, might be most reliable and practicable for long-term monitoring in athletes.
REFERENCES:
1) Daly, E., et al. (2022). Front Sports Act Living, 4, 1058326.
2) Jones, C. M., et al. (2017). Sports Med, 47(5), 943-974.
3) Kiely, J. (2018). Sports Med, 48(4), 753-764.
4) Matos, N. F., et al. (2011). Med Sci Sports Exerc, 43(7), 1287-1294.
5) Weakley, J., et al. (2022). Int J Sports Physiol Perform, 17(5), 675-681.