Objectives: Continuous renal replacement therapy (CRRT) has become the standard modality of renal replacement therapy (RRT) in critically ill patients with acute kidney injury (AKI). However, there is no consensus criterion for discontinuing CRRT. We aimed to evaluate the usefulness of the prediction model from our previous study in a large multicenter retrospective cohort. Methods: Validation was performed using data from 1517 critically ill patients with AKI who underwent CRRT for more than 2 days and survived for 7 or more days after CRRT discontinuation from 2018 to 2020 in five medical centers (one temporal cohort and four external cohorts). Variables included in the model were urine output (≥ 300 mL/day, score 4) on the day before discontinuation and blood urea nitrogen (BUN < 35 mg/dL, score 2), serum potassium (< 4.1 mmol/L, score 1), and mean arterial blood pressure (50-78 mmHg, score 1) on the discontinuation day. Successful discontinuation of CRRT was defined as no RRT requirement for 7 days after CRRT discontinuation. Results: Overall area under the curve of the receiver-operating characteristic (AUC-ROC) curve was 0.74 (95% CI 0.71-0.76). Overall differences between observed and predicted incidence rates were 3.0% (17.7% observed and 16.9% predicted probability), 3.6% (35.2% and 34.8%), and 2.0% (69.3% and 70.3%) in the low- (0-2 points), intermediate- (3-5 points), and high-score (6-8 points) groups, respectively. In a subgroup analysis, 4 centers showed similar discriminatory power (AUC-ROC 0.770, 0.731, 0.735, and 0.725, respectively), while 1 center showed poor discriminatory power (AUC-ROC 0.556). Conclusions: Good performance of our prediction model for successful discontinuation of CRRT in critically ill patients was validated in one temporal and three external cohorts, with the exception of one external cohort. Our results supports the need of an adequate protocol regarding CRRT discontinuation.