Lim et al e405 FIGURE 1. Flow diagram of the study participants. SGLTi, sodium-glucose cotransporter 2 inhibitors.

Background Acute kidney injury (AKI) is commonly encountered in clinical practice and is associated with poor patient outcomes and increased health care costs. Despite it posing significant challenges for clinicians, effective measures for AKI prediction and prevention are lacking. Previously published AKI prediction models mostly have a simple design without external validation. Furthermore, little is known about the process of linking model output and clinical decisions due to the black-box nature of neural network models. Objective We aimed to present an externally validated recurrent neural network (RNN)–based continuous prediction model for in-hospital AKI and show applicable model interpretations in relation to clinical decision support. Methods Study populations were all patients aged 18 years or older who were hospitalized for more than 48 hours between 2013 and 2017 in 2 tertiary hospitals in Korea (Seoul National University Bundang Hospital and Seoul National University Hospital). All demographic data, laboratory values, vital signs, and clinical conditions of patients were obtained from electronic health records of each hospital. We developed 2-stage hierarchical prediction models (model 1 and model 2) using RNN algorithms. The outcome variable for model 1 was the occurrence of AKI within 7 days from the present. Model 2 predicted the future trajectory of creatinine values up to 72 hours. The performance of each developed model was evaluated using the internal and external validation data sets. For the explainability of our models, different model-agnostic interpretation methods were used, including Shapley Additive Explanations, partial dependence plots, individual conditional expectation, and accumulated local effects plots. Results We included 69,081 patients in the training, 7675 in the internal validation, and 72,352 in the external validation cohorts for model development after excluding cases with missing data and those with an estimated glomerular filtration rate less than 15 mL/min/1.73 m2 or end-stage kidney disease. Model 1 predicted any AKI development with an area under the receiver operating characteristic curve (AUC) of 0.88 (internal validation) and 0.84 (external validation), and stage 2 or higher AKI development with an AUC of 0.93 (internal validation) and 0.90 (external validation). Model 2 predicted the future creatinine values within 3 days with mean-squared errors of 0.04-0.09 for patients with higher risks of AKI and 0.03-0.08 for those with lower risks. Based on the developed models, we showed AKI probability according to feature values in total patients and each individual with partial dependence, accumulated local effects, and individual conditional expectation plots. We also estimated the effects of feature modifications such as nephrotoxic drug discontinuation on future creatinine levels. Conclusions We developed and externally validated a continuous AKI prediction model using RNN algorithms. Our model could provide real-time assessment of future AKI occurrences and individualized risk factors for AKI in general inpatient cohorts; thus, we suggest approaches to support clinical decisions based on prediction models for in-hospital AKI.

Background: Copeptin is secreted in equimolar amounts as arginine vasopressin, main hormone regulating body fluid homeostasis. A recent study reported a copeptin-based classification of osmoregulatory defects in syndromes of inappropriate antidiuresis that may aid in prediction of therapeutic success. We investigated usefulness of copeptin for differentiating etiologies of hyponatremia and predicting efficacy and safety of hypertonic saline treatment in hyponatremic patients. Methods: We performed a multicenter, prospective cohort study of 100 inpatients with symptomatic hyponatremia (corrected serum sodium [sNa] ≤ 125 mmol/L) treated with hypertonic saline. Copeptin levels were measured at baseline and 24 hours after treatment initiation, and patients were classified as being below or above median of copeptin at baseline or at 24 hours, respectively. Correlations between target, under correction, and overcorrection rates of sNa within 24 hours/24-48 hours and copeptin levels at baseline/24 hours were analyzed. Results: Mean sNa and median copeptin levels were 117.9 and 16.9 pmol/L, respectively. Ratio of copeptin-to-urine sodium allowed for an improved differentiation among some (insufficient effective circulatory volume), but not all hyponatremia etiologic subgroups. Patients with below-median copeptin levels at baseline achieved a higher target correction rate in 6/24 hours (odds ratio [OR], 2.97; p = 0.02/OR, 6.21; p = 0.006). Patients with below-median copeptin levels 24 hours after treatment showed a higher overcorrection rate in next 24 hours (OR, 18.00, p = 0.02). Conclusion:There is a limited diagnostic utility of copeptin for differential diagnosis of hyponatremia. However, copeptin might be useful for predicting responses to hypertonic saline treatment in hyponatremic patients.

Background: Serum uric acid (SUA) is recognized as a risk factor for chronic kidney disease (CKD) and mortality. However, there is controversy as to whether a high or low level of SUA is related to the risk of CKD progression or death, and whether it differs between males and females. Methods: We included 143,762 adults who underwent voluntary health screening between 1995 and 2009 in Korea. For each sex, we divided participants into sex-specific quintiles according to SUA levels and compared end-stage renal disease (ESRD) incidence and mortality between the groups with low and high SUA levels and those with middle SUA levels. Sex-specific Cox proportional hazard analyses were performed for ESRD and all-cause mortality. Results: Among the 143,762 participants, 0.2% (n = 272) developed ESRD. The hazard ratio (HR) of ESRD was higher in the highest (adjusted HR, 2.13; 95% confidence interval [CI], 1.18-3.84) and lowest (adjusted HR, 1.90; 95% CI, 1.02-3.51) SUA quintiles than in the middle SUA quintile in males and the highest SUA quintile in females (adjusted HR, 2.31; 95% CI, 1.10-4.84). Four-point three percent (n = 6,215) of participants died during a mean follow-up period of 157 months. The hazard ratio (HR) of all-cause mortality was higher in the highest SUA quintile than in the middle SUA quintile in males (adjusted HR, 1.15; 95% CI, 1.03-1.28) and females (adjusted HR, 1.17; 95% CI, 1.01-1.35). Conclusion: Elevated levels of SUA are associated with increased risk for ESRD and all-cause mortality in both sexes. Low levels of SUA might be related to ESRD and death only in males, showing U-shaped associations. Our findings suggest sex-specific associations between SUA levels and ESRD development and mortality.

Background: The significance of ambulatory blood pressure (ABP) in Korean patients with chronic kidney disease (CKD) in relation to renal outcome or death remains unclear. We investigated the role of ABP in predicting end-stage renal disease or death in patients with CKD. Methods: We enrolled 387 patients with hypertension and CKD who underwent ABP monitoring and were followed for 1 year. Data on clinical parameters and outcomes from August 2014 to May 2018 were retrospectively collected. The composite endpoint was end-stage renal disease or death. Patients were grouped according to the mean ABP. Results: There were 66 endpoint events, 52 end-stage renal disease cases, and 15 mortalities. Among all patients, one developed end-stage renal disease and died. Mean ABP in the systolic and diastolic phases were risk factors for the development of composite outcome with hazard ratios of 1.03 (95% confidence interval [CI], 1.01-1.04; P < 0.001) and 1.04 (95% CI, 1.02-1.07; P = 0.001) for every 1 mmHg increase in BP, respectively. Patients with mean ABP between 125/75 and 130/80 mmHg had a 2.56-fold higher risk for the development of composite outcome (95% CI, 0.72-9.12; P = 0.147) as compared to those with mean ABP ≤ 125/75 mmHg. Patients with mean ABP ≥ 130/80 mmHg had a 4.79-fold higher risk (95% CI, 1.68-13.70; P = 0.003) compared to those with mean ABP ≤ 125/75 mmHg. Office blood pressure (OBP) was not a risk factor for the composite outcome when adjusted for covariates. Conclusion: In contrast to OBP, ABP was a significant risk factor for end-stage renal disease or death in CKD patients.