Bedside Score Predicts Successful Dialysis Discontinuation in ICU Patients

The UNDERSCORE tool uses six routine clinical variables to help physicians determine if critically ill patients can safely stay off dialysis.

For Doctors in a Hurry

Clinicians lack reliable tools to predict whether patients with severe acute kidney injury will successfully remain off renal replacement therapy after discontinuation.
Researchers analyzed 180 trial patients to develop a prediction model and validated it externally in 415 intensive care unit patients.
The model yielded an area under the curve of 0.86 (95% confidence interval 0.80 to 0.91) internally, and 0.73 during external validation.
The authors concluded that this scoring system effectively estimates the probability of sustained renal replacement therapy independence across diverse patient populations.
Physicians can utilize this validated bedside tool to objectively estimate weaning success and guide post-discontinuation management decisions.

The Uncertainty of Halting Dialysis in the ICU

Severe acute kidney injury frequently necessitates renal replacement therapy (RRT) in the intensive care unit, but managing the trajectory of this life-saving intervention remains a complex clinical challenge [1]. While extensive research has focused on the optimal timing for initiating RRT and the comparative efficacy of different dialysis modalities, clear guidelines on when to safely stop therapy are lacking [2, 3, 4]. Premature discontinuation can lead to dangerous metabolic derangements and fluid overload, whereas unnecessarily prolonging RRT exposes critically ill patients to ongoing risks of bleeding, infection, and hemodynamic instability [5, 4]. To address this gap in post-discontinuation management, researchers evaluated routine clinical predictors to help physicians estimate the probability of sustained RRT liberation after an initial trial of stopping therapy.

Defining Successful Weaning Parameters

Renal replacement therapy (RRT) is a critical intervention for severe acute kidney injury (AKI), yet managing its cessation presents a distinct challenge. After a clinician-initiated discontinuation of RRT, the likelihood of sustained liberation remains uncertain, leaving physicians without clear guidance on whether a critically ill patient will tolerate remaining off the machine. To address this clinical blind spot, researchers aimed to identify predictors of successful RRT weaning and develop a pragmatic bedside tool called the UNDERSCORE to support post-discontinuation management. To build this predictive model, the investigators conducted a post-hoc analysis of two multicenter randomized trials, AKIKI and AKIKI2. The analysis specifically included intensive care unit patients diagnosed with KDIGO stage 3 AKI (the most severe category of acute kidney injury under the Kidney Disease: Improving Global Outcomes criteria, characterized by a massive drop in kidney function or the need for dialysis) who were managed with a conservative RRT initiation approach. By focusing on this specific cohort, the researchers ensured the data reflected a standardized strategy for starting dialysis, providing a uniform baseline for evaluating how patients respond when the therapy is withdrawn. Patients were eligible for the analysis if they underwent an RRT weaning attempt, which the investigators strictly defined as the discontinuation of therapy for 3 or more consecutive days. The primary outcome of the study was successful weaning, defined as no RRT resumption within seven days of the initial stop. To determine which clinical factors most accurately forecasted this outcome, independent predictors of successful weaning were identified using multivariable logistic regression (a statistical method that evaluates multiple patient variables simultaneously to isolate the strongest independent associations).

Six Clinical Variables Drive the UNDERSCORE Tool

To develop the predictive model, the researchers analyzed data from the derivation cohort to understand how frequently patients tolerate the withdrawal of dialysis. Out of 554 patients who received RRT, 180 underwent a weaning attempt. Among these individuals, 101 patients (56 percent) were successfully weaned, meaning they did not require therapy to be restarted within seven days. The investigators isolated the most robust independent variables associated with sustained liberation from dialysis, ultimately retaining six predictors to construct the UNDERSCORE tool. The first three predictors are RRT duration before the attempt, septic shock on admission, and baseline serum creatinine. These factors help establish the severity of the initial illness and the baseline renal reserve before the trial of discontinuation. The remaining three predictors are clinical variables assessed after the weaning attempt: use of vasopressors, invasive mechanical ventilation, and urine output. By combining these routine bedside metrics, the model provides a quantifiable probability of weaning success. The UNDERSCORE showed strong discrimination in the derivation cohort, with an area under the curve (AUC) of 0.86 (95 percent CI 0.80-0.91). An AUC of 0.86 (a metric of diagnostic accuracy where 1.0 is perfect and 0.5 is equivalent to a coin toss) indicates that the tool is highly effective at distinguishing between critically ill patients who will successfully remain off dialysis and those who will require therapy to be resumed. For practicing intensivists and nephrologists, this means having a data-driven framework to decide whether a patient is truly ready to transition away from renal support.

External Validation in a Diverse ICU Population

To ensure the tool's reliability beyond the initial study population, the UNDERSCORE model was externally validated in an independent Swiss intensive care unit cohort. This external validation group included 415 patients. In contrast to the derivation group, this population experienced a higher rate of liberation from dialysis. Specifically, 338 patients (81 percent) were successfully weaned from renal replacement therapy. When applied to this new group, the UNDERSCORE demonstrated fair performance across a broader case mix, achieving an AUC of 0.73 (95 percent CI 0.66-0.80). This performance indicates that the model maintains a clinically useful ability to distinguish between successful and unsuccessful weaning attempts, even when applied to patients outside the strict parameters of the original trials. Ultimately, the UNDERSCORE was derived from a homogeneous cohort with conservative RRT initiation and validated in a diverse ICU population. By bridging these two distinct groups, the tool provides a practical bedside metric to estimate the probability of sustained RRT discontinuation after an initial clinician-initiated stop. For the practicing physician, this scoring system offers an objective, evidence-based method to guide complex post-discontinuation management, potentially reducing the risks associated with both premature cessation and unnecessarily prolonged dialysis.

Study Info

Predictive factors of successful renal replacement therapy weaning : Deciding On patients Orientations after Renal replacement therapy Stopping (The DOORS study).

Khalil Chaïbi, Yanis Akrour, Guillaume Louis, Laurent Martin Lefevre, et al.

Journal Intensive care medicine

Published March 17, 2026