Physical inactivity is a major risk factor for low life expectancy and is a major public health concern in countries worldwide. Subjective life expectancy majorly influences health behaviors, healthcare costs, and economic patterns, therefore, identifying its determinants can help early prevention and policy planning. Especially understanding risk factors in physically inactive older adults is critical, given their higher vulnerability to disease progression. Machine learning models traditionally have been used to enhance prediction accuracy rather than to explain why certain outcomes occur. Unlike classical statistical methods, machine learning systems are designed to detect complex, nonlinear patterns that improve prediction performance. Tools such as SHAP allow researchers to examine which features most heavily influence the model’s decisions. In Yang’s study, SHAP outputs showed that age, perceived health status, physical pain, depression, and life satisfaction were the strongest contributors to subjective life expectancy risk across the inactive group. However, the active groups' greatest contributing factors were perceived health, age, education level, pension insurance, and physical pain. The overlap of age, perceived health, and physical pain across both groups indicates that these factors are central to subjective life expectancy regardless of activity level. Meanwhile, education and pension insurance appear only in the active group, suggesting that different psychosocial factors become relevant when older adults are physically active. This demonstrates how machine learning models can reveal subgroup-specific patterns that are not shown in traditional methods. Yang et al.'s findings support the potential use of ML-based risk identification in primary care and public-health screening to flag older adults at elevated perceived life-expectancy risk. Integrating such tools could allow for more efficient targeted interventions in aging populations.