Motor Performance Phenotypes and Balance–Coordination Characteristics Identified by Unsupervised Clustering in Nigerian School-Age Children

Background: Motor competence shapes childhood health, but its heterogeneity in low- and middle-income countries remains unclear. Purpose: This study used unsupervised clustering to identify distinct motor performance profiles among Nigerian school-age children. Research Design: A cross-sectional study. Study Sample: The study included a convenience sample of 200 school-age children aged 6–16 years from five schools in Ebonyi State, Nigeria. Data Collection and Analysis: Motor performance was assessed using 15 motor tasks. K-means and hierarchical clustering were applied to 17 features, including motor tasks, age, and sex, with cluster validity assessed using internal validation indices and between-cluster differences examined using the Mann–Whitney U test or Kruskal–Wallis H test, as appropriate. A Bernoulli mixture model (BMM) was applied as a model-based validation approach specifically designed for binary data. Results: Across clustering approaches, approximately 31–36% of children were classified within a motor-delayed performance phenotype characterized by an age–performance mismatch: despite being chronologically oldest (median age 14.0 years in k-means and 12.0 years in hierarchical clustering), this group demonstrated selectively reduced coordination and balance performance. Team balance on beam consistently emerged as the strongest task-level discriminator, showing the largest performance gap in k-means clustering (54.2% vs. 86.6–86.9%, p < 0.001), and robust separation in hierarchical clustering (35.4% vs. 94.0%, p < 0.001). The BMM corroborated this two-profile structure (BIC-optimal two-class solution; ∼31% lower-performing), with perfect classification stability (ARI = 1.0). The motor-delayed group also exhibited higher prevalence of reported disabilities (16.9–19.4%), prior therapy exposure (15.4–19.4%), and lower sports participation compared with higher-performing groups. Hierarchical clustering demonstrated stronger between-cluster discrimination in total motor score than k-means ( η H 2 = 0.745 vs. 0.059) and identified one small outlier cluster ( n = 2), while cluster composition did not differ significantly by sex. Conclusions: These findings indicate that unsupervised clustering can reveal distinct motor performance phenotypes in school-age children and highlight the value of balance-related tasks for differentiating coordination and motor control profiles during childhood.