Background: The lack of consensus criteria of acute on chronic kidney injury (ACKI) affects the judgment for its clinical prognosis. Methods: In this study, we analyzed the data from 711,615 hospitalized adults who had at least 2 serum creatinine (SCr) tests within 30 days. We estimated the reference change value (RCV) of SCr given initial SCr level in adults without known risks of acute kidney injury other than chronic kidney disease (CKD). We proposed a criterion for ACKI based on the RCV of SCr (cROCK), which defined ACKI as a ≥25% increase in SCr in 7 days. We validated cROCK by its association with the risks of in-hospital mortality, death after discharge, and CKD progression in a large cohort of patients with CKD stage 3. Results: In 21,661 patients with CKD stage 3, a total of 3,145 (14.5%), 1,512 (7.0%), and 221 (1.0%) ACKI events were detected by both cROCK and Kidney Disease Improving Global Outcomes (KDIGO), cROCK only, and KDIGO only, respectively. cROCK detected 40% more ACKI events than KDIGO. Compared with patients without ACKI by both definitions, those with cROCK- but not KDIGO-defined ACKI had a significantly increased risk of in-hospital mortality (hazard ratio [HR] 5.53; 95% CI 3.75–8.16), death after discharge (HR 1.51; 95% CI 1.21–1.83), and CKD progression (OR 5.65; 95% CI 3.05–10.48). Conclusions: RCV-based criterion (cROCK) for ACKI is clinically valid in that it has a substantially improved sensitivity in identifying patients with high risk of adverse outcomes.