Predicting mortality in patients with diabetes starting dialysis

Area of Science:

• Nephrology
• Diabetology
• Epidemiology

Background:

• Limited prediction models exist for mortality in diabetic dialysis patients.
• Diabetic end-stage renal disease (ESRD) patients on dialysis face high mortality risks.
• Accurate mortality prediction is crucial for managing this population.

Purpose of the Study:

• To identify predictors of 1-year all-cause mortality in diabetic dialysis patients.
• To develop and validate a predictive model for 1-year mortality in this cohort.
• To provide a tool for improved patient counseling and clinical decision-making.

Main Methods:

• Utilized data from the Netherlands Cooperative Study on the Adequacy of Dialysis (NECOSAD).
• Employed multivariate logistic regression and backward selection to build the prediction model.
• Assessed model performance using calibration, discrimination (c-statistic=0.810), and internal validation via bootstrapping.

Main Results:

• The final model included seven predictors: age, smoking, macrovascular complications, diabetes duration, Karnofsky scale, serum albumin, and hemoglobin.
• 82 out of 394 (21%) diabetic dialysis patients died within one year.
• The model demonstrated good predictive performance with a c-statistic of 0.810.

Conclusions:

• A validated 7-predictor model effectively forecasts 1-year mortality in incident diabetic dialysis patients.
• The model shows good predictive accuracy, aiding in prognosis assessment.
• External validation is recommended prior to clinical implementation for patient counseling.