Aim Acute kidney injury (AKI) is diagnosed by a 50% increase in creatinine. For patients without a baseline creatinine measurement, guidelines suggest estimating baseline creatinine by back‐calculation. The aim of this study was to evaluate different glomerular filtration rate (GFR) equations and different GFR assumptions for back‐calculating baseline creatinine as well as the effect on the diagnosis of AKI. Methods The Modification of Diet in Renal Disease, the Chronic Kidney Disease Epidemiology (CKD‐EPI) and the Mayo quadratic (MQ) equation were evaluated to estimate baseline creatinine, each under the assumption of either a fixed GFR of 75 mL min−1 1.73 m−2 or an age‐adjusted GFR. Estimated baseline creatinine, diagnoses and severity stages of AKI based on estimated baseline creatinine were compared to measured baseline creatinine and corresponding diagnoses and severity stages of AKI. Results The data of 34 690 surgical patients were analysed. Estimating baseline creatinine overestimated baseline creatinine. Diagnosing AKI based on estimated baseline creatinine had only substantial agreement with AKI diagnoses based on measured baseline creatinine [Cohen's κ ranging from 0.66 (95% CI 0.65–0.68) to 0.77 (95% CI 0.76–0.79)] and overestimated AKI prevalence with fair sensitivity [ranging from 74.3% (95% CI 72.3–76.2) to 90.1% (95% CI 88.6–92.1)]. Staging AKI severity based on estimated baseline creatinine had moderate agreement with AKI severity based on measured baseline creatinine [Cohen's κ ranging from 0.43 (95% CI 0.42–0.44) to 0.53 (95% CI 0.51–0.55)]. Conclusion Diagnosing AKI and staging AKI severity on the basis of estimated baseline creatinine in surgical patients is not feasible. Patients at risk for post‐operative AKI should have a pre‐operative creatinine measurement to adequately assess post‐operative AKI.