Modified creatinine index and risk for long-term infection-related mortality in hemodialysis patients: ten-year outcomes of the Q-Cohort Study

Association of the modified creatinine index with quality of life in haemodialysis patients

Aims/Background The evaluation of health-related quality of life in patients undergoing maintenance haemodialysis has garnered increasing attention. The modified creatinine index, a surrogate marker for muscle mass, has been linked to various clinical outcomes. However, the relationship between modified creatinine index and health-related quality of life in maintenance haemodialysis patients remains unclear. This study aims to elucidate the association between modified creatinine index and health-related quality of life in individuals receiving maintenance haemodialysis. Methods This cross-sectional study included 217 maintenance haemodialysis patients. Health-related quality of life was assessed using the Kidney Disease Quality of Life Instrument. Collected data included general patient information, laboratory results, and haemodialysis-related parameters. The modified creatinine index was calculated based on gender, age, single-pool Kt/V (spKt/V), and pre-dialysis serum creatinine levels. Multiple linear regression models and smooth curve fitting were used to investigate the relationship between modified creatinine index and health-related quality of life. Subgroup analyses and interaction tests were performed to identify potential effect modifiers. Results The 217 maintenance haemodialysis patients had a mean age of 53.66±13.15 years and a median dialysis vintage of 39 (25-84) months; 120 (55.30%) were male. The mean health-related quality of life score was 55.76±10.33, and the mean modified creatinine index was 22.72±2.95 mg/kg/day. After adjusting for confounding factors, an increase in modified creatinine index was associated with an improvement in health-related quality of life (β=0.55, 95% CI: 0.04, 1.06, p = 0.033). No nonlinear relationship was identified between modified creatinine index and health-related quality of life by smooth curve fitting. Subgroup and interaction analyses indicated that the relationship between modified creatinine index and health-related quality of life was stable and not significantly influenced by age, gender, dialysis vintage, diabetes status, or body mass index (p > 0.05). Conclusion Modified creatinine index is positively correlated with health-related quality of life in maintenance haemodialysis patients, suggesting its potential utility in evaluating patient quality of life. Modified creatinine index could be clinically useful to improve the predictability of health-related quality of life in maintenance haemodialysis patients.

Disturbance in the potential cardiovascular-bone-skeletal muscle axis and morbidity and mortality in patients undergoing haemodialysis: the Q-Cohort Study

Background

Disturbances in the cardiovascular system, bone and skeletal muscle are independent risk factors for death among patients receiving haemodialysis (HD). However, the combined impact of disorders of these three organs on morbidity and mortality is unclear in the HD population.

Methods

A total of 3031 Japanese patients on maintenance HD were prospectively followed. The outcomes were all-cause mortality, major adverse cardiovascular events (MACE) and bone fracture. Patients were divided into four groups (G1-G4) according to the baseline number of diseased organs represented as histories of cardiovascular disease and bone fractures and the presence of low skeletal muscle mass as follows: G1, no organ; G2, one organ; G3, two organs; G4, three organs. Multivariable-adjusted survival models were used to analyse associations between the number of diseased organs and outcomes.

Results

During a 4-year follow-up, 499 deaths, 540 MACE and 140 bone fractures occurred. In the Cox proportional hazards model, the risk for all-cause mortality was significantly higher in G2, G3 and G4 than in G1 as the reference {hazard ratio: G2, 2.16 [95% confidence interval (CI) 1.65-2.84], G3, 3.10 [95% CI 2.27-4.23] and G4, 3.11 [95% CI 1.89-5.14]}. Similarly, the risks for developing MACE and bone fractures were significantly elevated as the number of organ disorders increased.

Conclusions

Multiple disorders of the cardiovascular-bone-skeletal muscle axis are strong predictors of morbidity and mortality in patients undergoing HD.

Creatinine-cystatin C ratio and death with a functioning graft in kidney transplant recipients

Death with a functioning graft is important cause of graft loss after kidney transplantation. However, little is known about factors predicting death with a functioning graft among kidney transplant recipients. In this study, we evaluated the association between post-transplant creatinine-cystatin C ratio and death with a functioning graft in 1592 kidney transplant recipients. We divided the patients into tertiles based on sex-specific creatinine-cystatin C ratio. Among the 1592 recipients, 39.5% were female, and 86.1% underwent living-donor kidney transplantation. The cut-off value for the lowest creatinine-cystatin C ratio tertile was 0.86 in males and 0.73 in females. The lowest tertile had a significantly lower 5-year patient survival rate and was independently associated with death with a functioning graft (adjusted hazard ratio 2.574, 95% confidence interval 1.339-4.950, P < 0.001). Infection was the most common cause of death in the lowest tertile group, accounting for 62% of deaths. A low creatinine-cystatin C ratio was significantly associated with an increased risk of death with a functioning graft after kidney transplantation.

Data-driven, two-stage machine learning algorithm-based prediction scheme for assessing 1-year and 3-year mortality risk in chronic hemodialysis patients

Life expectancy is likely to be substantially reduced in patients undergoing chronic hemodialysis (CHD). However, machine learning (ML) may predict the risk factors of mortality in patients with CHD by analyzing the serum laboratory data from regular dialysis routine. This study aimed to establish the mortality prediction model of CHD patients by adopting two-stage ML algorithm-based prediction scheme, combined with importance of risk factors identified by different ML methods. This is a retrospective, observational cohort study. We included 800 patients undergoing CHD between December 2006 and December 2012 in Shin-Kong Wu Ho-Su Memorial Hospital. This study analyzed laboratory data including 44 indicators. We used five ML methods, namely, logistic regression (LGR), decision tree (DT), random forest (RF), gradient boosting (GB), and eXtreme gradient boosting (XGB), to develop a two-stage ML algorithm-based prediction scheme and evaluate the important factors that predict CHD mortality. LGR served as a bench method. Regarding the validation and testing datasets from 1- and 3-year mortality prediction model, the RF had better accuracy and area-under-curve results among the five different ML methods. The stepwise RF model, which incorporates the most important factors of CHD mortality risk based on the average rank from DT, RF, GB, and XGB, exhibited superior predictive performance compared to LGR in predicting mortality among CHD patients over both 1-year and 3-year periods. We had developed a two-stage ML algorithm-based prediction scheme by implementing the stepwise RF that demonstrated satisfactory performance in predicting mortality in patients with CHD over 1- and 3-year periods. The findings of this study can offer valuable information to nephrologists, enhancing patient-centered decision-making and increasing awareness about risky laboratory data, particularly for patients with a high short-term mortality risk.

Creatinine index: a retrospective cohort study in an urban Australian dialysis context

AIM

This study aimed to described the relationship between the CI and mortality in an Australian context.

INTRODUCTION

Maintenance haemodialysis is a catabolic state associated with a significant decrease in lean body mass (LBM) and protein energy wasting. LBM can be derived or estimated from creatinine kinetic modelling, specifically the creatinine index (CI). This has been demonstrated in cohort studies to predict mortality.

METHODS

One hundred seventy-nine patients undergoing haemodialysis in 2015 were included in this cohort. They were followed for 5 years with pertinent clinical data collected to calculate the CI as of December 2015. For analysis, patients were split into a high and low CI group based on the median (18.32 mg/kg/day). The primary outcome of interest was all-cause mortality, and secondary outcomes included myocardial infarction, stroke and transplantation.

RESULTS

During follow-up, 69 (76.7%) patients in the low CI group and 28 (31.5%) patients in the high CI group died (P < 0.001). The relative risk (RR) of mortality within the low compared with the high CI group was 2.43 (95% confidence interval, 1.75-3.38). Fully adjusted Cox proportional hazards modelling demonstrated a hazard ratio (HR) of 0.498 (95% CI, 0.292-0.848) for survival in the high CI group. Lower CI was associated with increased risk of stroke (RR, 5.43 [95% CI, 1.24-23.84]), whereas transplant was more likely in the high CI group (RR, 6.4 [95% confidence interval, 1.96-20.88]).

CONCLUSIONS

In a single-centre Australian haemodialysis cohort, the CI was strongly associated with mortality and stroke risk. The CI is an accurate and simple method to identify patients with low LBM at risk for significant morbidity and mortality.

Ratio of extracellular water to intracellular water and simplified creatinine index as predictors of all-cause mortality for patients receiving hemodialysis

The bioelectrical impedance analysis-measured ratio of extracellular water (ECW) to intracellular water (ICW) reflects fluid volume and malnutrition. It may be an indicator of protein-energy wasting and muscle wasting in hemodialysis patients. We examined the association between the ECW/ICW ratio and simplified creatinine index, which is a new surrogate marker of protein-energy wasting and muscle wasting, and whether their combination can accurately predict mortality. A total of 224 patients undergoing hemodialysis for more than 6 months and having undergone bioelectrical impedance analysis for the assessment of body composition were included. Patients were divided into two groups based on the cut-off values of the ECW/ICW ratio (0.57) and simplified creatinine index (20.4 mg/kg/day) for maximumly predicting mortality. Thereafter, they were cross-classified into four groups with each cut-off point. The ECW/ICW ratio was independently associated with the simplified creatinine index (β = -0.164; P = 0.042). During a follow-up of 3.5 years (2.0-6.0 years), 77 patients died. A higher ECW/ICW ratio (adjusted hazard ratio, 3.66, 95% confidence interval 1.99-6.72, P <0.0001) and lower simplified creatinine index (adjusted hazard ratio, 2.25, 95% confidence interval 1.34-3.79, P = 0.0021) were independently associated with an increased risk of all-cause mortality. The adjusted hazard ratio for the higher ECW/ICW ratio and lower simplified creatinine index group vs. the lower ECW/ICW ratio and higher simplified creatinine index group was 12.22 (95% confidence interval 3.68-40.57, p <0.0001). Furthermore, the addition of the ECW/ICW ratio and simplified creatinine index to the baseline risk model significantly improved the C-index from 0.831 to 0.864 (p = 0.045). In conclusion, the ECW/ICW ratio may be a surrogate marker of muscle wasting. Moreover, combining the ECW/ICW ratio and simplified creatinine index may improve the accuracy of predicting all-cause mortality and help stratify the mortality risk of hemodialysis patients.

Association of the modified creatinine index with muscle strength and mortality in patients undergoing hemodialysis

Background

In the updated consensus, low muscle strength overtook the role of low muscle mass, and probable sarcopenia was diagnosed once low muscle strength was detected. Whether the modified creatinine index (mCI) could identify persons with probable sarcopenia who may be at risk of adverse outcomes remains unknown. We aimed to evaluate the association of the mCI with probable sarcopenia and mortality in patients undergoing hemodialysis.

Methods

In the cross-sectional study (n = 346), univariate and multivariable logistic regression analyses were performed to study the association of mCI with probable sarcopenia. Modified Quantitative Subjective Global Assessment (MQSGA) was used to evaluate the nutritional status. The performance of the mCI value for identifying probable sarcopenia was analyzed using receiver operating characteristic (ROC) curve analysis. The appropriate cutoff points were determined using Youden’s method. In the longitudinal cohort study composed of an independent hemodialysis cohort (n = 218), cox proportional regression models were used to evaluate crude and adjusted hazard ratios and 95% confidence intervals (CIs) of death by mCI and MQSGA.

Results

Cross-sectional results showed that after adjusting for confounders, the association of mCI with low muscle strength remained significant. The area under the curve (AUC) of the mCI to predict probable sarcopenia was 0.804 (95% CI, 0.744–0.863; p < 0.001) for men and 0.787 (95% CI, 0.711–0.864; p < 0.001) for women. The optimal mCI cutoff values were 21.07 mg/kg/d for men and 19.57 mg/kg/d for women, respectively. Longitudinal results showed that compared with those in the high mCI group, subjects in the low mCI group had a higher risk of death for all causes (adjusted HR, 2.51; 95% CI, 1.16–5.41; p = 0.019). Adding the mCI significantly improved the predictive accuracy for death with an increase in C-index from 0.785 to 0.805 (p = 0.026) and improved the net reclassification index (38.6%, p = 0.021), while adding MQSGA did not.

Conclusion

The mCI is a predictor of muscle strength and survival in hemodialysis patients, and is preferable to the MQSGA for predicting death. Assessment of mCI could provide additional predictive and prognostic information to sarcopenia.

Changes in spike protein antibody titer over 90 days after the second dose of SARS-CoV-2 vaccine in Japanese dialysis patients

OBJECTIVES

There is no report on antibody titers after vaccination against SARS-CoV-2 in Japanese dialysis patients. As dialysis is different between Japan and other countries, changes in antibody titers were examined.

METHODS

Baseline characteristics and anti-spike protein antibody titers (Roche) over 90 days after administration of the BNT162b2 messenger RNA vaccine were investigated in dialysis patients.

RESULTS

The maximum anti-spike protein antibody titer after the second dose was 738 (327 to 1143) U/mL and was reached at 19 (17 to 24) days after the second dose. Antibody titers decreased over time, with titers of 770 (316 to 1089) U/mL at 15 days, 385 (203 to 690) U/mL at 30 days, 254 (138 to 423) U/mL at 60 days, and 208 (107 to 375) U/mL at 90 days after the second dose. When an antibody titer of 137 U/mL was assumed to be a measure related to breakthrough infection, the proportion of subjects with antibody titers exceeding this level was 90.1% at 15 days, 85.3% at 30 days, 75.0% at 60 days, and 65.4% at 90 days after the second dose. When a decrease in antibody titers below the assumed breakthrough level was defined as an event, subjects with a pre-dialysis albumin ≥ 3.5 g/dL were significantly less likely to experience an event than subjects with a pre-dialysis albumin < 3.5 g/dL.

CONCLUSIONS

The presence of anti-spike protein levels ≥ 313 U/mL at 30 days after the second vaccine dose might be a factor in maintaining enough antibody titers at 90 days after. Whether an additional vaccine dose is needed should be determined based on indicators serving as factors in maintaining antibody titers as well as the status of the spread of infection.

Machine-learning based prediction and analysis of prognostic risk factors in patients with candidemia and bacteraemia: a 5-year analysis

Bacteraemia has attracted great attention owing to its serious outcomes, including deterioration of the primary disease, infection, severe sepsis, overwhelming septic shock or even death. Candidemia, secondary to bacteraemia, is frequently seen in hospitalised patients, especially in those with weak immune systems, and may lead to lethal outcomes and a poor prognosis. Moreover, higher morbidity and mortality associated with candidemia. Owing to the complexity of patient conditions, the occurrence of candidemia is increasing. Candidemia-related studies are relatively challenging. Because candidemia is associated with increasing mortality related to invasive infection of organs, its pathogenesis warrants further investigation. We collected the relevant clinical data of 367 patients with concomitant candidemia and bacteraemia in the first hospital of China Medical University from January 2013 to January 2018. We analysed the available information and attempted to obtain the undisclosed information. Subsequently, we used machine learning to screen for regulators such as prognostic factors related to death. Of the 367 patients, 231 (62.9%) were men, and the median age of all patients was 61 years old (range, 52-71 years), with 133 (36.2%) patients aged >65 years. In addition, 249 patients had hypoproteinaemia, and 169 patients were admitted to the intensive care unit (ICU) during hospitalisation. The most common fungi and bacteria associated with tumour development and Candida infection were Candida parapsilosis and Acinetobacter baumannii, respectively. We used machine learning to screen for death-related prognostic factors in patients with candidemia and bacteraemia mainly based on integrated information. The results showed that serum creatinine level, endotoxic shock, length of stay in ICU, age, leukocyte count, total parenteral nutrition, total bilirubin level, length of stay in the hospital, PCT level and lymphocyte count were identified as the main prognostic factors. These findings will greatly help clinicians treat patients with candidemia and bacteraemia.

Synergistic Effects of the Geriatric Nutritional Risk Index and the Modified Creatinine Index for Predicting Mortality in Patients on Hemodialysis

This study aimed to investigate whether a combined estimation of the geriatric nutritional risk index (GNRI) and the modified creatinine index (mCI) provides synergistic information for mortality in patients treated by chronic hemodialysis. We analyzed 499 patients on hemodialysis for five years. We set each cut-off value as the high (≥92) and low (<92) GNRI groups and the high (≥21 mg/kg/day) and low (<21 mg/kg/day) mCI groups, and divided them into four subgroups: G1, high GNRI + high mCI; G2, high GNRI + low mCI; G3, low GNRI + high mCI; and G4, low GNRI + low mCI. The survival rate was evaluated and time-to-event analysis was performed. All-cause death occurred in 142 (28%) patients. Kaplan−Meier curves showed that G2 and G4 had a significantly worse outcome (p < 0.05) than G1 but not G3. Using the multivariable-adjusted model, only G4 was significantly associated with all-cause mortality compared with G1. Our study suggests that the synergistic effects of the GNRI and the mCI are helpful in predicting all-cause mortality. The combination of these indices may be superior to a single method to distinguish patients who are well or moderately ill from potentially severely ill.

Prognostic Value of the Controlling Nutritional Status (CONUT) Score in Patients at Dialysis Initiation

Protein-energy wasting (PEW) is common in patients with chronic kidney disease (CKD), and affects their prognosis. The Controlling Nutritional Status (CONUT) score is a nutritional screening tool calculated using only blood test data. This study aimed to investigate the prognostic value of CONUT score in patients just initiating dialysis. A total of 311 CKD patients who stably initiated dialysis were enrolled. Only 27 (8.7%) patients were classified as having normal nutritional status. The CONUT score was also independently correlated with elevated C-reactive protein levels (β = 0.485, p < 0.0001). During the median follow-up of 37 months, 100 patients (32.2%) died. The CONUT score was an independent predictor of all-cause mortality (adjusted hazard ratio 1.13, 95% confidence interval 1.04−1.22, p < 0.0024). As model discrimination, the addition of the CONUT score to a prediction model based on established risk factors significantly improved net reclassification improvement (0.285, p = 0.028) and integrated discrimination improvement (0.025, p = 0.023). The CONUT score might be a simplified surrogate marker of the PEW with clinical utility and could predict all-cause mortality, in addition to improving the predictability in CKD patients just initiating dialysis. The CONUT score also could predict infectious-disease mortality.

Emerging cross-talks between chronic kidney disease-mineral and bone disorder (CKD-MBD) and malnutrition-inflammation complex syndrome (MICS) in patients receiving dialysis

Chronic kidney disease–mineral and bone disorder (CKD–MBD) is a systemic disorder that affects multiple organs and systems and increases the risk of morbidity and mortality in patients with CKD, especially those receiving dialysis therapy. CKD–MBD is highly prevalent in CKD patients, and its treatment is gaining attention from healthcare providers who manage these patients. Additional important pathologies often observed in CKD patients are chronic inflammation and malnutrition/protein-energy wasting (PEW). These two pathologies coexist to form a vicious cycle that accelerates the progression of various other pathologies in CKD patients. This concept is integrated into the term “malnutrition–inflammation–atherosclerosis syndrome” or “malnutrition–inflammation complex syndrome (MICS)”. Recent basic and clinical studies have shown that CKD–MBD directly induces inflammation as well as malnutrition/PEW. Indeed, higher circulating levels of inorganic phosphate, fibroblast growth factor 23, parathyroid hormone, and calciprotein particles, as markers for critical components and effectors of CKD–MBD, were shown to directly induce inflammatory responses, thereby leading to malnutrition/PEW, cardiovascular diseases, and clinically relevant complications. In this short review, we discuss the close interplay between CKD–MBD and MICS and emphasize the significance of simultaneous control of these two seemingly distinct pathologies in patients with CKD, especially those receiving dialysis therapy, for better management of the CKD/hemodialysis population.

Combined Evaluation of Geriatric Nutritional Risk Index and Modified Creatinine Index for Predicting Mortality in Patients on Hemodialysis

The geriatric nutritional risk index (GNRI) and modified creatinine index (mCI) are surrogate markers of protein-energy wasting in patients receiving hemodialysis. We aimed to examine whether a combined evaluation of these indices improved mortality prediction in this population. We retrospectively investigated 263 hemodialysis patients divided into two groups, using 91.2 and 20.16 mg/kg/day as cut-off values of GNRI and mCI, respectively. The resultant four groups were reshuffled into four subgroups defined using combinations of cut-off values of both indices and were followed up. During the follow-up period (median: 3.1 years), 103 patients died (46/103, cardiovascular causes). Lower GNRI and lower mCI were independently associated with all-cause mortality (adjusted hazard ratio (aHR) 4.96, 95% confidence intervals (CI) 3.10–7.94, and aHR 1.92, 95% CI 1.22–3.02, respectively). The aHR value for the lower GNRI and lower mCI group vs. the higher GNRI and higher mCI group was 7.95 (95% CI 4.38–14.43). Further, the addition of GNRI and mCI to the baseline risk assessment model significantly improved the C-index of all-cause mortality (0.801 to 0.835, p = 0.025). The simultaneous evaluation of GNRI and mCI could be clinically useful to stratify the risk of mortality and to improve the predictability of mortality in patients on hemodialysis.

Current status of the assessment of sarcopenia, frailty, physical performance and functional status in chronic kidney disease patients

PURPOSE OF REVIEW

Low physical function, frailty, and sarcopenia are common complications of chronic kidney disease (CKD). In this article, we review the epidemiology and pathogenesis of low physical function, as well as its associations with adverse outcomes in CKD patients. Additionally, we present various traditional and novel methods for assessment of physical function in CKD patients.

RECENT FINDINGS

In nondialysis dependent (NDD) and dialysis-dependent CKD patients, the prevalence of low physical function, frailty, and sarcopenia are substantially higher than in the general population. The potential mechanisms of low physical function, frailty, and sarcopenia in CKD patients are due to various factors including underlying kidney disease, co-existing comorbidities, and certain therapeutic interventions utilized in CKD. Increasing evidence has also uncovered the ill effects of impaired physical function on clinical outcomes in CKD patients.

SUMMARY

Routine assessment of physical function is an under-utilized yet important component in the management of CKD patients. Future studies are needed to determine how prescription of exercise and increased daily physical activity can be tailored to optimize the health and well-being of NDD and dialysis-dependent CKD patients in pursuit of successful aging.

Association of the nutritional risk index for Japanese hemodialysis patients with long-term mortality: The Q-Cohort Study

BACKGROUND

Protein-energy wasting (PEW) is a risk factor for mortality in patients undergoing hemodialysis. Recently, a nutritional risk index for Japanese hemodialysis patients (NRI-JH) has been proposed as a surrogate index of PEW. However, no study has determined the association of the NRI-JH with long-term mortality in patients undergoing hemodialysis. Furthermore, the validity of the NRI-JH has not been confirmed.

METHODS

In total, 3046 patients undergoing hemodialysis and registered in the Q-Cohort Study were followed up for 10 years. The NRI-JH was calculated on the basis of body mass index and serum levels of albumin, total cholesterol, and creatinine. The patients were divided into four groups according to the NRI-JH scores: 0-3 (G1, n = 1343), 4-7 (G2, n = 1136), 8-10 (G3, n = 321), and 11-13 (G4, n = 246). We examined the association between the NRI-JH and the 4-year and 10-year risks of all-cause, cardiovascular, and infection-related deaths using the Cox proportional hazards model.

RESULTS

During the follow-up period, 647 patients died during the first 4 years, and 1503 patients died within 10 years. The 4-year prognosis was analyzed and compared with the lowest NRI-JH score group. Multivariable-adjusted hazard ratios (95% confidence intervals) for all-cause death were 1.93 (1.57-2.38), 2.68 (2.05-3.50), and 3.16 (2.40-4.16) in the G2, G3, and G4 groups, respectively. Similarly, a higher NRI-JH score was associated with an increased risk of cardiovascular and infection-related deaths.

CONCLUSION

A higher NRI-JH score was associated with an increased risk of long-term mortality in patients undergoing maintenance hemodialysis.

TRIAL REGISTRATION

The study protocol was registered in the University Hospital Medical Information Network (UMIN) clinical trial registry (UMIN ID: 000000556).

Prognostic Machine Learning Models for First-Year Mortality in Incident Hemodialysis Patients: Development and Validation Study

BACKGROUND

The first-year survival rate among patients undergoing hemodialysis remains poor. Current mortality risk scores for patients undergoing hemodialysis employ regression techniques and have limited applicability and robustness.

OBJECTIVE

We aimed to develop a machine learning model utilizing clinical factors to predict first-year mortality in patients undergoing hemodialysis that could assist physicians in classifying high-risk patients.

METHODS

Training and testing cohorts consisted of 5351 patients from a single center and 5828 patients from 97 renal centers undergoing hemodialysis (incident only). The outcome was all-cause mortality during the first year of dialysis. Extreme gradient boosting was used for algorithm training and validation. Two models were established based on the data obtained at dialysis initiation (model 1) and data 0-3 months after dialysis initiation (model 2), and 10-fold cross-validation was applied to each model. The area under the curve (AUC), sensitivity (recall), specificity, precision, balanced accuracy, and F1 score were used to assess the predictive ability of the models.

RESULTS

In the training and testing cohorts, 585 (10.93%) and 764 (13.11%) patients, respectively, died during the first-year follow-up. Of 42 candidate features, the 15 most important features were selected. The performance of model 1 (AUC 0.83, 95% CI 0.78-0.84) was similar to that of model 2 (AUC 0.85, 95% CI 0.81-0.86).

CONCLUSIONS

We developed and validated 2 machine learning models to predict first-year mortality in patients undergoing hemodialysis. Both models could be used to stratify high-risk patients at the early stages of dialysis.