The primary aim of the research was to develop a systematic process for synthesising rehabilitation evidence that incorporates both qualitative and quantitative research and clinical experience. A second aim was to use this evidence-base to develop a coherent conceptual framework for rehabilitation that can inform future interventions. These aims were articulated through the worked example of low back injury (LBI). Evidence-based practice (EBP) has received unequivocal support as a way of ensuring best practice in health and rehabilitation. However, EBP has been interpreted in a way that demands evidence conforms with stringent empirical criteria, as would typically be applied to effectiveness studies (i.e., randomised controlled trials, (RCTs)). Consequently, qualitative research and clinical experience have largely been excluded as valuable sources of evidence. However, there is some suggestion that EBP has failed to translate into improved outcomes following LBI. This thesis has argued that the relevance of EBP will be enhanced if its conceptualisation of knowledge can shift from the stringent criteria that are often applied to evidence to a more inclusive approach. The current thesis employed a program of research comprising three separate studies that contributed to the cumulative development of a coherent evidence-base. Study One systematically explored the empirical support for return-to-work predictors in the quantitative research, using the method of Narrative Systematic Review. Study Two reviewed the clinical utility of these predictors, using a method based on practitioner consultation and consensus. Study Three reviewed the qualitative rehabilitation research using Meta-ethnography. The findings from Study One revealed 38 key predictors that were associated with return-to-work outcome following LBI. Significant empirical support was found for predictors relating to age, the subjective experience of injury and expectations of return-to-work. Trends towards significance were found in relation to workplace factors. Study Two identified nine predictors that were considered to be relevant to rehabilitation practice and, therefore, an important focus for future interventions. Study Two also demonstrated the disparity among experts in their ability to distinguish between work factors, rehabilitation processes and individual characteristics. These diverse attributions may influence practice and, therefore, outcomes. Most importantly, Study Two revealed the fact that practitioners understood predictors of outcome according to the way in which they contributed to three critical processes that underpin rehabilitation. Specifically, practitioners focused on supportive and collaborative processes that attended to issues of time, context and engagement. Study Three identified five key components that provided a framework for understanding rehabilitation as a system. Sixteen themes clustered together to form these five components, which, taken together, defined a rehabilitation system marred by tensions and multiple competing processes. Components such as policy and philosophy interacted and competed with the demands of the workplace and the individual’s personal responses. Partnerships between key stakeholders were lacking, but offered an important strategy for improving outcomes. The current thesis utilised an innovative methodology to efficiently synthesise an extensive literature base and develop clinically meaningful conclusions. This thesis has demonstrated that the knowledge cycle is an evolving process, informed by both empirical research and clinical experience, and located within broader frameworks of understanding. It has also supported the view that evidence and its interpretation must be grounded in the context of the discipline and developmental stage of knowledge-base.