Hydration Status and Fluid Replacement Strategies of High-Performance Adolescent Athletes: An Application of Machine Learning to Distinguish Hydration Characteristics

TL;DR

This field study characterized hydration status and fluid balance in 105 heat-acclimatized high-performance adolescent athletes across 11 sports, assessing first-morning and pre-training hydration and ad libitum fluid intake during low- and high-intensity sessions. Hypohydration was common but few athletes exceeded 2% body-mass loss, and k-means clustering identified three hydration-profile clusters (heavy sweaters with sufficient/insufficient compensatory habits and light sweaters with sufficient habits).

Key findings

• First-morning urine specific gravity (USG) classified 20% (USG > 1.025) and 44% (USG > 1.020) of athletes as hypohydrated; before low-intensity training, 21% and 44% began hypohydrated, and before high-intensity training, 15% and 34% began hypohydrated. Post-training, 2% of athletes in the low-intensity session and 0% in the high-intensity session showed excessive dehydration (>2% body-mass loss).
• Athletes who arrived hypohydrated consumed significantly more fluid during training than euhydrated athletes (main effect: F(1, 183.85) = 5.91, p = 0.016; mean difference 82.1 mL/h, 95% CI 15.5–148.7, p < 0.05).
• Linear mixed models found sport × session interactions for sweat rate (F(10, 94) = 2.901, p = 0.003) and fluid consumption per hour (F(10, 94) = 2.297, p = 0.018), plus a main effect of sport on percentage body-mass change (F(10, 94) = 3.558, p = 0.001). Sex main effects were found for sweat rate (F(1, 103) = 4.64, p = 0.034) and percentage body-mass change (F(1, 103) = 6.347, p = 0.013).
• K-means cluster analyses converged on three clusters: “Heavy sweaters with sufficient compensatory hydration habits” (HSSH), “Heavy sweaters with insufficient compensatory hydration habits” (HSIH), and “Light sweaters with sufficient compensatory hydration habits” (LSSH).

Methods

• Descriptive field study. A total of 199 Singaporean adolescent athletes from 11 sports were recruited by convenience sampling from a high-performance sports academy; 105 (male: 66, female: 39; age 14.1 ± 1.0 y) completed all protocols and were analyzed, with 94 unable to complete sessions due to academic commitments.
• Hydration status was assessed on three occasions: first-morning urine, and 30 min before low- and high-intensity training sessions held 1 week apart between 1500 and 1800 h. Hypohydration was defined by USG > 1.020 and, secondarily, USG > 1.025; post-training dehydration was defined as ≥2.0% body-mass change. Fluid consumption was measured by bottle weighing with ad libitum drinking; sweat rate was calculated from body mass and urine output.
• Training intensity was measured by session RPE (CR10 scale), training load as RPE × duration, and thirst on a 9-point scale. Analyses included linear mixed models with participant as random effect, Pearson correlations, and K-means clustering (optimal n = 3 by elbow method) with UMAP projection.

Limitations

• Paper-stated: hydration status was assessed using USG rather than plasma osmolality; USG is recognized to produce higher false positives for dehydration, so a secondary USG > 1.025 threshold was reported.
• Paper-stated: low- and high-intensity training sessions were classified according to coaches’ interpretation and execution of training plans, not by a standardized research-imposed intensity.
• Evident from the text: the study used convenience sampling from one high-performance academy in Singapore, had substantial attrition (199 recruited, 105 analyzed), and was an observational field study, limiting generalizability and causal inference.