In this study, we established an explainable and personalized risk prediction model for in-hospital mortality after continuous renal replacement therapy (CRRT) initiation. This retrospective cohort study was conducted at Changhua Christian Hospital (CCH). A total of 2932 consecutive intensive care unit patients receiving CRRT between 1 January 2010, and 30 April 2021, were identified from the CCH Clinical Research Database and were included in this study. The recursive feature elimination method with 10-fold cross-validation was used and repeated five times to select the optimal subset of features for the development of machine learning (ML) models to predict in-hospital mortality after CRRT initiation. An explainable approach based on ML and the SHapley Additive exPlanation (SHAP) and a local explanation method were used to evaluate the risk of in-hospital mortality and help clinicians understand the results of ML models. The extreme gradient boosting and gradient boosting machine models exhibited a higher discrimination ability (area under curve [AUC] = 0.806, 95% CI = 0.770–0.843 and AUC = 0.823, 95% CI = 0.788–0.858, respectively). The SHAP model revealed that the Acute Physiology and Chronic Health Evaluation II score, albumin level, and the timing of CRRT initiation were the most crucial features, followed by age, potassium and creatinine levels, SPO2, mean arterial pressure, international normalized ratio, and vasopressor support use. ML models combined with SHAP and local interpretation can provide the visual interpretation of individual risk predictions, which can help clinicians understand the effect of critical features and make informed decisions for preventing in-hospital deaths.

Serum potassium (K+) levels between 3.5 and 5.0 mmol/L are considered safe for patients. The optimal serum K+ level for critically ill patients with acute kidney injury undergoing continuous renal replacement therapy (CRRT) remains unclear. This retrospective study investigated the association between ICU mortality and K+ levels and their variability. Patients aged >20 years with a minimum of two serum K+ levels recorded during CRRT who were admitted to the ICU in a tertiary hospital in central Taiwan between January 01, 2010, and April 30, 2021 were eligible for inclusion. Patients were categorized into different groups based on their mean K+ levels: <3.0, 3.0 to <3.5, 3.5 to <4.0, 4.0 to <4.5, 4.5 to <5.0, and ≥5.0 mmol/L; K+ variability was divided by the quartiles of the average real variation. We analyzed the association between the particular groups and in-hospital mortality by using Cox proportional hazard models. We studied 1991 CRRT patients with 9891 serum K+ values recorded within 24 h after the initiation of CRRT. A J-shaped association was observed between serum K+ levels and mortality, and the lowest mortality was observed in the patients with mean K+ levels between 3.0 and 4.0 mmol/L. The risk of in-hospital death was significantly increased in those with the highest variability (HR and 95% CI = 1.61 [1.13–2.29] for 72 h mortality; 1.39 [1.06–1.82] for 28-day mortality; 1.43 [1.11–1.83] for 90-day mortality, and 1.31 [1.03–1.65] for in-hospital mortality, respectively). Patients receiving CRRT may benefit from a lower serum K+ level and its tighter control. During CRRT, progressively increased mortality was noted in the patients with increasing K+ variability. Thus, the careful and timely correction of dyskalemia among these patients is crucial.