The bacterium Staphylococcus aureus is a major human pathogen, where the emergence of antibiotic resistance is a global public-health concern. Host factors such as age and the presence of co-morbidities have been implicated in a worse outcome for patient. However, this is complicated by the highly complex and multi-faceted nature of bacterial virulence, which has so far prevented a robust mapping between genotype, phenotype and infection outcome. To investigate the role of bacterial and host factors in contributing to S. aureus bacteraemia-associated mortality we sequenced a collection of clinical isolates (of the MLST clonal complex CC22) from patients with bloodstream infections and quantified specific virulence phenotypes. A genome-wide association scan identified several novel virulence-affecting loci, which we validated using a functional genomics approach. Analysing the data comprising bacterial genotype and phenotype as well as clinical meta-data within a machine-learning framework revealed that mortality associated with CC22 bacteraemia is not only influenced by the interactions between host and bacterial factors but can also be predicted at the individual patient-level to a high degree of accuracy. This study clearly demonstrates the potential of using a combined genomics and data analytic approach to enhance our understanding of bacterial pathogenesis. Considering both host and microbial factors, including whole genome sequence and phenotype data, within a predictive framework could thus pave the way towards personalised medicine and infectious disease management.