•Recurrence strongly associated with liver and renal failure, diabetes and immune-suppressive drugs.•A persistent focus was judged the primary reason for recurrence in 23(74%).•A risk score based on BMI, Immunosuppression, Renal disease, Diabetes, Liver disease predicted recurrence.•Older age and higher neutrophil counts predicted early treatment failure and S. aureus-attributed mortality. Adjunctive rifampicin did not reduce failure/recurrence/death as a composite endpoint in the ARREST trial of Staphylococcus aureus bacteraemia, but did reduce recurrences. We investigated clinically-defined 14-day treatment failure, and recurrence and S. aureus-attributed/unattributed mortality by 12-weeks to further define their predictors. A post-hoc exploratory analysis using competing risks models was conducted to identify sub-groups which might benefit from rifampicin. A points-based recurrence risk score was developed and used to compare rifampicin's benefits. Recurrence was strongly associated with liver and renal failure, diabetes and immune-suppressive drugs (p < 0.005); in contrast, failure and S. aureus-attributed mortality were associated with older age and higher neutrophil counts. Higher SOFA scores predicted mortality; higher Charlson scores and deep-seated initial infection focus predicted failure. Unexpectedly, recurrence risk increased with increasing BMI in placebo (p = 0.04) but not rifampicin (p = 0.60) participants (pheterogeneity = 0.06). A persistent focus was judged the primary reason for recurrence in 23(74%). A 5-factor risk score based on BMI, Immunosuppression, Renal disease, Diabetes, Liver disease (BIRDL) strongly predicted recurrence (p < 0.001). Rifampicin reduces recurrences overall; those with greatest absolute risk reductions were identified using a simple risk score. Source control and adequate duration of antibiotic therapy remain essential to prevent recurrence and improve outcomes.