Falls represent a significant threat to the motor health of older adults and individuals with neuromuscular disorders. Early identification of those at risk plays a crucial role in fall prevention and the design of rehabilitation interventions. Considering the importance of the transition phase in generating gait instability, the objective of this research was to develop and evaluate a composite index, termed the “Gait Stability Index” (GSI), to differentiate between healthy individuals and those at high risk of falls. This cross-sectional analytical study involved 60 participants, comprising 30 healthy individuals and 30 individuals at high risk of falls. Gait data were collected using inertial sensors and a 3D motion analysis system. The GSI was calculated based on dynamic features including center of mass (CoM) oscillations, gait asymmetry, speed instability, and Lyapunov exponents. To assess the diagnostic performance of this index, ROC analysis and machine learning models including Random Forest, Logistic Regression, and Support Vector Machine (SVM) were employed. The results indicated that the mean GSI was significantly higher in the high-risk group compared to the healthy group. Furthermore, a very large effect size (d = 2.90) demonstrated a high discriminative power. ROC analysis yielded an AUC of 0.90, with a sensitivity of 0.87 and specificity of 0.83. The Random Forest model also provided the highest classification accuracy compared to other models. Based on these findings, the Gait Stability Index can serve as an effective, cost-efficient, and practical tool for fall risk screening and the development of intelligent motion assessment systems.