Background: Physical fitness is important for health and independence in older adults, but existing systems do not always provide a continuous multidomain score. Objective: To develop and preliminarily evaluate a Physical Fitness Index (PFI) for older adults in a Chinese retirement community. Methods: A total of 561 participants aged 65–89 years completed assessments of body composition, strength, endurance, balance, mobility, and flexibility. Scores were normalized within sex and 5-year age groups and combined using a fixed radar-axis order. Results: The PFI averaged 39.24 ± 13.43 points (range, 4.16–84.47). In a reduced surrogate regression model, R² was 0.825 and ten-fold cross-validation yielded an RMSE of 5.79 points and an MAE of 4.33 points. Conclusion: The PFI provides a continuous overall score with a domain profile, but its calibration and measurement properties require external and longitudinal validation.

Background: Multidimensional evidence on exercise participation and frailty in older adults is limited. Methods: This 12-week preference-based, non-randomized study included 145 participants (mean age 82.25 years): walking (n = 50), resistance training (n = 20), Traditional Chinese exercise (n = 23), and control (n = 52). Frailty was the primary outcome; physical-performance and molecular analyses were exploratory. Paired trajectories, baseline-stratified binary transitions, and adjusted associations were examined. Results: Among paired participants, pre-frailty/frailty prevalence changed from 70.0% to 60.0% with walking, 90.0% to 70.0% with resistance training, 50.0% to 63.6% with Traditional Chinese exercise (n = 22), and 50.0% to 69.2% in controls. Baseline status constrained opportunities for binary improvement or deterioration. All adjusted frailty risk-ratio confidence intervals included 1. Walking was associated with faster Timed Up and Go and Five-Yard Walk Test performance after correction across eight outcomes. Molecular changes also occurred in controls; intervention timing and potential laboratory-batch confounding limited between-group interpretation. No microbiome taxon survived multiplicity correction; small paired samples limited precision. Conclusions: Walking and resistance-training groups showed descriptively more favorable prevalence trajectories, but adjusted frailty comparisons remained inconclusive. These findings do not establish frailty prevention, modality superiority, or exercise-specific molecular mechanisms.