The clinical implementation of urinary and plasma renal injury biomarkers has been hampered by the variability associated with nonstandardized commercially available biomarker assays, uncertainty and variations in patient selection criteria, and the absence of context-specific cutoffs for biomarker concentrations. These limitations are increased by comparison with serum creatinine to define acute kidney injury. The critical problem affecting biomarker performance is patient heterogeneity involving the cause, context (including comorbidity and baseline renal function), and timing of the injury. We suggest strategies for stratifying subjects to provide appropriate context, and illustrate a creatinine-independent method for defining thresholds for biomarker concentrations in these contexts which utilizes the same sensitivity for the clinical outcomes of dialysis or death. Large multicenter cohort studies are needed to validate the proposed cutoffs.