Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0688
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-08T06:26:56.381202Z
status: published
lens: trace
sector: health
headline: Machine learning-assisted prediction of 5-year mortality in chronic kidney disease: the KoreaN cohort study for Outcome in patients With Chronic Kidney Disease (KNOW-CKD)
dek: Mortality prediction models for patients with non-dialysis chronic kidney disease (CKD) remain limited despite their clinical importance. While machine learning (ML) offers the potential to improve prediction accuracy, its "black-box" nature has hindered clinical adoption. This study aimed to develop and validate an interpretable ML model for predicting 5-year all-cause mortality in patients with non-dialysis CKD and to deploy it as a user-friendly web-based risk stratification tool. We analyzed 1,858 patients (…
gain_title: A CatBoost model trained on 1,858 non-dialysis CKD patients predicted 5-year all-cause mortality with AUC 0.813, outperforming logistic regression, and a simplified 5-feature version maintained AUC 0.795 in external validation.
problem_title: Machine learning's black-box nature has hindered clinical adoption of mortality prediction models for non-dialysis CKD despite their clinical importance.
trace_subject: machine learning prediction of 5-year all-cause mortality in non-dialysis chronic kidney disease patients
gain_reading: A CatBoost model trained on 1,858 non-dialysis CKD patients predicted 5-year all-cause mortality with AUC 0.813, outperforming logistic regression, and a simplified 5-feature version maintained AUC 0.795 in external validation.
gain_evidence: The CatBoost model demonstrated superior performance (area under the curve [AUC], 0.813; 95% confidence interval [CI], 0.737-0.888; p = 0.04) compared to logistic regression (AUC, 0.747; 95% CI, 0.655-0.840). | The CatBoost-based ML model accurately predicts 5-year mortality in non-dialysis CKD patients using five readily available clinical parameters.
problem_reading: Machine learning's black-box nature has hindered clinical adoption of mortality prediction models for non-dialysis CKD despite their clinical importance.
problem_evidence: its "black-box" nature has hindered clinical adoption.
quick_read: Researchers developed and validated an interpretable machine learning model to predict 5-year all-cause mortality in non-dialysis chronic kidney disease using data from 1,858 patients in the KNOW-CKD prospective cohort, with 94 deaths observed. The CatBoost model achieved AUC 0.813 versus 0.747 for logistic regression, and a simplified version using age, eGFR, albumin, urine protein-to-creatinine ratio, and total calcium retained AUC 0.795 in an external cohort of 348 patients.

The work matters because accurate mortality risk stratification could guide earlier personalized management in a population where prediction tools remain limited, and the use of SHAP to surface five routine labs plus a U-shaped calcium risk makes the tool more clinically usable. Uncertainty remains about generalizability beyond Korean cohorts, performance with only 94 events, and whether web-based deployment will translate into changed care or outcomes.
limitation: 
tag: Automated dual reading
key_points: Study analyzed 1,858 patients (94 deaths) from prospective KNOW-CKD cohort of non-dialysis CKD patients. | Simplified model used top-5 SHAP-ranked features: age, estimated glomerular filtration rate, serum albumin, spot urine protein-to-creatinine ratio, and serum total calcium. | External validation performed in independent cohort of 348 CKD patients with AUC 0.795. | SHAP visualization revealed U-shaped relationship for serum calcium, identifying hypocalcemia as under-recognized mortality risk factor.
rundown: Researchers trained several ML algorithms including CatBoost and compared them to conventional logistic regression on 1,858 KNOW-CKD participants, using SHapley Additive exPlanations to rank features and improve interpretability, then built a simplified model from the top five features.

The simplified model was externally validated in 348 patients and deployed as a user-friendly web-based risk stratification tool intended to enable early risk stratification and personalized management in non-dialysis CKD care.
sources:
- peer_reviewed | Kidney Research and Clinical Practice | https://doi.org/10.23876/j.krcp.25.471 | 2026-08-07
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

4591b50000d05befb6434cd8786c31e27c793777e1c951fa3e645b95cfa30489

previous

0000000000000000000000000000000000000000000000000000000000000000