Background: Despite the continual improvements in dialysis treatments, mortality in end-stage kidney disease (ESKD) remains high. Many mortality prediction models are available, but most of them are not precise enough to be used in the clinical practice. We aimed to develop and validate two prediction models for 3-month and 1-year patient mortality after dialysis initiation in our population. Methods: Using population-based data of insurance claims in Taiwan, we included more than 210,000 patients who initiated dialysis between January 1, 2006, and June 30, 2015. We developed two prognostic models, which included 9 and 11 variables, respectively (including age, sex, myocardial infarction, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, peptic ulcer disease, malignancy, moderate to severe liver disease, and first dialysis in intensive care unit). Results: The models showed adequate discrimination (C-statistics were 0.80 and 0.82 for 3-month and 1-year mortality, respectively) and good calibration. In both our models, the first dialysis in the intensive care unit and moderate-to-severe liver disease were the strongest risk factors for mortality. Conclusion: The prediction models developed in our population had good predictive ability for short-term mortality in patients initiating dialysis in Taiwan and could help in decision-making regarding dialysis initiation, at least in our setting, supporting a patient-centered approach to care. Graphical abstract: [Figure not available: see fulltext.]

Original languageEnglish
Pages (from-to)1005-1013
Number of pages9
JournalJournal of Nephrology
Issue number3
Publication statusPublished - Apr 2022


  • Dialysis initiation
  • End-stage kidney disease
  • Mortality
  • Population-based study
  • Prediction

ASJC Scopus subject areas

  • Nephrology


Dive into the research topics of 'Predicting 3-month and 1-year mortality for patients initiating dialysis: a population-based cohort study'. Together they form a unique fingerprint.

Cite this