Ankle And Knee Joint Work Differentiates Drop Jump Performance

In sport science, studying multi-joint coordination patterns and their relationships to performance is important because athletes may coordinate movement differently while striving for the same outcome. Unsupervised machine learning methods, such as cluster analysis, can help identify athletes with similar coordination profiles, providing an avenue to link biomechanics to performance. PURPOSE: To determine if clustering multi-joint coordination of work at the ankle, knee, and hip can differentiate drop jump height (JH), ground contact time (GCT), and reactive strength index (RSI = [JH/GCT]). METHODS: Forty-five NCAA Division I and collegiate club level athletes (23 males, 22 females) from basketball, volleyball, and track & field performed drop jump trials off a 12-inch box. Athletes were instructed to maximize JH and minimize GCT. Marker-based 3D motion and ground reaction forces were recorded. Positive, negative, and net ankle, knee, and hip joint work during stance phase were calculated, normalized to body mass, standardized, and entered into a random forest clustering algorithm. RESULTS: A three-cluster solution was optimal based on Bayesian information criterion. One-way ANOVA and post hoc testing revealed differences between clusters in anthropometrics, joint work, and all jump performance variables (Table 1). Athletes in Cluster 1 achieved the best RSI (Hedges’ g ≥ 1.87) and exhibited greater positive work (g ≥ 1.96), negative work (g ≥ 1.22), and net work (g ≥ 0.96) at the ankle and knee joints than athletes in the other two clusters. CONCLUSION: Clustering of lower extremity joint work identified three groups of athletes with distinct coordination patterns and jump performance profiles. Specifically, athletes who generated more ankle and knee joint work exhibited better RSI. Joint-specific biomechanical variables, used in combination with random forest clustering, provide insight into coordination-based determinants of athletic performance. Table 1. - Between-cluster comparisons in anthropometrics, jump performance (JH: jump height; GCT: ground contact time; RSI: reactive strength index), and lower extremity joint work (J/kg). Cluster 1 Cluster 2 Cluster 3 ANOVA Sig. η 2 Anthropometrics N(M/F) 15 (11/4) 9 (2/7) 21 (5/16) - - Age (years) 19.9 ± 1.3 19.2 ± 1.0 19.7 ± 1.6 0.480 0.03 Height (cm)bc 186.8 ± 7.5 192.7 ± 12.5 179.2 ± 9.2 0.002 0.25 Body Mass (kg)c 83.0 ± 14.6 85.9 ± 9.7 74.4 ± 13.5 0.054 0.13 Jump Performance JH (m)bc 0.356 ± 0.056 0.359 ± 0.077 0.245 ± 0.050 0.001 0.46 GCT(s)ab 0.294 ± 0.045 0.464 ± 0.098 0.408 ± 0.063 0.001 0.51 RSI (m/s)ab 1.24 ± 0.25 0.80 ± 0.18 0.64 ± 0.16 0.001 0.65 Ankle Joint Work Positiveab 448.2 ± 108.5 249.0 ± 48.4 244.4 ± 51.6 0.001 0.63 Negativeab -288.4 ± 86.3 -152.6 ± 72.8 -179.3 ± 46.5 0.001 0.43 Netab 159.8 ± 73.7 96.4 ± 42.2 65.1 ± 31.6 0.001 0.42 Knee Joint Work Positiveab 531.0 ± 96.9 342.3 ± 86.3 327.6 ± 62.5 0.001 0.60 Negativeab -368.3 ± 54.6 -273.5 ± 43.2 -304.7 ± 48.1 0.001 0.37 Netab 162.7 ± 77.8 68.8 ± 59.3 22.9 ± 42.2 0.001 0.54 Hip Joint Work Positiveac 240.6 ± 69.0 375.4 ± 58.9 203.4 ± 56.9 0.001 0.54 Negativec -172.6 ± 62.7 -269.4 ± 58.1 -162.6 ± 34.3 0.001 0.42 Netc 68.0 ± 69.5 105.9 ± 54.6 40.8 ± 34.1 0.011 0.19 η2: Eta-squared effect size from ANOVA a: significant difference (p < 0.05) between Clusters 1 and 2 b: significant difference (p < 0.05) between Clusters 1 and 3 c: significant difference (p < 0.05) between Clusters 2 and 3