Construct and Longitudinal Validity of Real-World Walking Distance in Older Adults Recovering from Hip Fracture

Background: Although the ability to walk longer distances is critical to regaining independence after hip fracture surgery, capacity for walking distance is typically measured in clinical settings and it remains unclear what distances people cover in their daily lives. Wearable devices can measure daily-life walking behaviour, but methods to quantify real-world walking distance remain limited and clinical validation is required. This study assessed the construct and longitudinal validity of real-world walking distance metrics derived from both single walking bouts (WBs) and a novel method for clustering WBs, using a single wearable device during recovery after hip fracture surgery. Methods: This multicentre prospective cohort study recruited patients from inpatient and outpatient lists, with real-world mobility and clinical- and patient-reported outcomes collected at first visit, 6- and 12 months follow-up. A total of 8 digital mobility outcomes (DMOs) descriptive of the walking distance were computed for both Single WBs > 30s and Clustered WBs (Distance DMOs: Mean, SD, Median and 95th percentile/P95) and were examined for construct (convergent, divergent, known-groups) validity. In addition, the two P95 DMOs were examined for longitudinal validity (ability to detect change and predicted trajectories). Analyses were performed comparing Spearman correlations and effect sizes with a priori expectations, and using linear mixed-effects models. Results: The total sample (N = 505) had a median age of 79 years (P25-P75: 72–84), with 66% women, and a 6-minute walk test (6MinWT) distance of 284 ± 126 m. Seven out of eight Distance DMOs demonstrated evidence for construct validity, showing expected weak-to-strong correlations with related constructs, and distinguishing between four hip fracture recovery phase groups. The two P95 Distance DMOs showed longitudinal validity, including ability to detect change, comparable to the 6MinWT distance. Conclusion: Real-world Walking Period Distance DMOs, derived from single WBs > 30s and a novel WB clustering approach, demonstrated evidence of construct and longitudinal validity in patients after hip fracture. The upper-range Distance DMOs showed correlations, effect sizes, and longitudinal trajectories comparable to 6MinWT distance, with the Cluster Distance consistent with patient-reported unchanged or improved walking ability. These findings support the clinical relevance of upper-range Distance DMOs for characterising real-world walking performance in hip fracture rehabilitation.