Acute kidney injury (AKI) is a serious complication of major, non-cardiac surgery. Risk prediction models for AKI may be useful for informed consent processes and to identify high-risk patients to target with perioperative prevention and early intervention strategies. This thesis includes a systematic review of published prediction models for AKI following, major, non-cardiac surgery. Although seven models were identified, most were derived in small, single center cohorts, and none were externally validated. The second part of the thesis reports derivation and internal validation of five risk prediction models and a risk index based on readily available preoperative variables for predicting severe AKI requiring dialysis after major, non-cardiac surgery. The final risk index showed excellent discrimination (c-statistic 0.89) and was well calibrated. Further research to externally validate this risk index and evaluate the clinical impact of its use is needed to establish its role in perioperative care.