QUESTION: Are improvements in walking disability, measured with the walking component of the Oswestry Disability Index, mediated by walking confidence, capacity, fear avoidance, pain severity and/or physical performance? PARTICIPANTS: Community-dwelling adults aged ≥ 65 years. DATA ANALYSIS: A causal mediation analysis was conducted. A directed acyclic graph identified potential pre-treatment confounders. Missing data were imputed, and intervention-mediator and mediator-outcome effects were estimated, along with the natural indirect effects (NIEs) through each mediator, the natural direct effect and the total intervention effect. RESULTS: The analysis included 435 participants. At 12 months, participants receiving the BOOST intervention had greater odds of walking more than one-quarter of a mile (and less than a mile) than those in the control arm (OR 1.38, 95% CI 0.85 to 2.16). The combined NIE via all mediators favoured the intervention (OR 0.71, 95% CI 0.56 to 0.91). The intervention improved walking disability in part by increasing walking confidence at 6 months (NIE OR 0.87, 95% CI 0.76 to 1.00) and reducing pain severity (NIE OR 0.90, 95% CI 0.79 to 1.02), which accounted for 41% and 32% of the combined indirect effect from all mediators, respectively. CONCLUSION: The BOOST intervention improved walking disability primarily through improving the walking confidence and pain severity of older adults with neurogenic claudication. Clinicians should prioritise interventions that build walking confidence (such as walking practice and dynamic balance exercises) and reduce pain severity, to optimise long-term mobility outcomes. TRIAL REGISTRATION: ISRCTN12698674.