Introduction and purpose: Overtraining syndrome (OTS) might occur in athletes experiencing extreme physical and mental stress over a longer period of time without adequate recovery (Meeusen et al., 2013). A decrease in sports performance and chronic fatigue are the most frequent symptoms (Carrard et al., 2021; Meeusen et al., 2013). Reliable diagnostic and monitoring tools are lacking but are strongly needed due to the high prevalence of OTS of 5 to 64 % (depending on definition and sample) and its potential reducing risk of injury (Meeusen et al., 2013). We aimed to develop novel sex-specific, non-invasive and multiparametric recovery monitoring models. Methods: Seventy-three youth and young adult elite athletes (51 females, age 19.7 ± 4.0 years) from mainly team and speed/power-oriented sports, e.g., handball and athletics, participated. Weekly measurements were conducted over 16 weeks to assess the athletes` recovery state, resulting in 663 measurement timepoints. Forty parameters - including sleep, training load, occupational load, social load, menstrual cycle, heart rate and heart rate variability (HRV), core body temperature, grip strength, and single and double leg jump performance - served as predictors of the athletes` subjective rating of recovery and stress (Short Recovery and Stress Scale, SRSS, Kellmann & Kölling, 2020). Lasso, Ridge, and Elastic Net regularized regression was applied for automated parameter selection, training, and cross-validation of the binomial prediction models. Results: For the female athletes` model AUC = 0.819 was calculated (sensitivity = 79.8%, specificity = 72.9%). Thereby, the parameters social load, single and double leg jump performance, sleep quality, training load, grip strength, and occupational load were ranked within the top ten highest predictive parameters (Figure 1). The male athletes` model demonstrated similar predictive performance with AUC = 0.797 (sensitivity = 74.3%, specificity = 71.4%). Thereby, grip strength, HRV, single leg jump performance, and social load were among the top ten most predictive parameters. Discussion: A broad and novel combination of non-invasive parameters was analysed to capture a holistic picture of the athletes` recovery and stress state. The resulting sex-specific models showed good predictive performance. The development of sex-specific recovery monitoring prediction models seemed crucial due to the observed differences in parameter importance. Conclusion: This study provides a deeper understanding of the relevance of specific parameters for recovery and stress monitoring in female and male youth and young adult elite athletes.
© Copyright 2026 Current Issues in Sport Science. Österreichische Sportwissenschaftliche Gesellschaft. Alle Rechte vorbehalten.