Mortality Risk Assessment in End-Stage Renal Disease

End‐stage renal disease (ESRD) represents the final phase of chronic kidney failure, marked by dependence on renal replacement therapies and a markedly elevated mortality rate. Prognostic assessment combines demographic and clinical variables such as age, comorbidity burden, cardiovascular disease, nutritional and inflammatory status, and dialysis access type. Conventional risk scores, exemplified by the Charlson Comorbidity Index and its renal‐specific adaptations, have long guided clinicians in estimating survival and tailoring treatment pathways. More recently, advances in data science have enabled the integration of large‐scale laboratory, imaging and electronic health record datasets into predictive models. Machine‐learning approaches, including decision trees, random forests and gradient‐boosting algorithms, have demonstrated improved discrimination over traditional regression methods. Simultaneously, clinical tools such as nomograms and recalibrated comorbidity indices permit personalised risk stratification at dialysis initiation. These innovations support informed shared decision‐making, aid in the selection of dialysis modality and inform the timing of transition to palliative care. The global burden of ESRD underscores the necessity of accurate mortality prediction to allocate resources, optimise patient outcomes and standardise care in diverse healthcare settings.

Research from Nature Portfolio

A 2023 study introduced a two‐stage machine‐learning scheme for chronic haemodialysis patients, analysing routine serum laboratory data across 44 indicators. By combining random forest, gradient‐boosting and extreme gradient‐boosting methods, the authors generated stepwise models that outperformed logistic regression in predicting one‐year and three‐year mortality. The approach identified key predictors—such as serum albumin, haemoglobin and inflammatory markers—and achieved superior area‐under‐curve metrics, offering real‐time risk alerts for clinicians.

Building on this, a 2020 investigation of peritoneal dialysis cohorts applied classification algorithms, including survival trees and deep neural networks, to forecast multi‐year mortality. The modified Charlson Comorbidity Index and patient age emerged as dominant risk factors, while neural networks yielded the highest concordance indices. This model underscored the potential of machine learning to refine prognostic scoring beyond conventional regression techniques, particularly by capturing nonlinear interactions among comorbidities and demographic variables.

Mortality Risk Assessment in End-Stage Renal Disease publication trend

The graph below shows the total number of articles in mortality risk assessment in end-stage renal disease across all publications each year (not limited to Nature Index journals).

Technical terms

End-stage renal disease (ESRD): The final phase of chronic kidney failure requiring dialysis or transplantation for survival.

Haemodialysis: A blood‐cleansing process that removes waste and excess fluid via an external dialyser.

Peritoneal dialysis: A home‐based therapy using the peritoneal membrane as a semipermeable filter to clear solutes.

Charlson Comorbidity Index (CCI): A weighted scoring system that quantifies comorbidity burden to predict mortality risk.

Random forest: An ensemble machine‐learning method that builds multiple decision trees for classification or regression tasks.

Nomogram: A graphical representation of a prognostic model that assigns points to clinical variables to estimate outcome probabilities.

Machine learning: Computational techniques that enable predictive modelling by identifying complex patterns in large datasets.

References

1. Data-driven, two-stage machine learning algorithm-based prediction scheme for assessing 1-year and 3-year mortality risk in chronic hemodialysis patients. Scientific Reports (2023).
2. Machine Learning to Identify Dialysis Patients at High Death Risk. Kidney International Reports (2019).
3. Risk factors for mortality in elderly haemodialysis patients: a systematic review and meta-analysis. BMC Nephrology (2020).
4. Prediction of the Mortality Risk in Peritoneal Dialysis Patients using Machine Learning Models: A Nation-wide Prospective Cohort in Korea. Scientific Reports (2020).