Main content

Development of iChoose Kidney Risk Estimates


The iChoose Kidney tool uses data from more than 700,000 incident, adult ESRD patients in the national cohort of United States Renal Data System (USRDS) surveillance data from January 1, 2005 through December 31, 2015. Logistic regression models were used to predict 1 and 3-year risk of death for dialysis patients vs. transplant patients. Predictive accuracy of the model was assessed using a validation data set by the use of the c-statistic of the associated receiver operating characteristic (ROC) curve, which estimates the probability of concordance between the observed number of deaths and the predicted number of deaths based on the model. Model calibration was assessed by comparing the observed and expected number of deaths for each model.

Translation of Risk Estimates into a Shared Decision Making Tool

We used the following equation to translate model coefficients for 1-year and 3-year mortality estimates into an individualized, estimated risk of mortality represents the sum of the individual's risk factors (risk score) and is the estimated Y-intercept (or baseline risk).

For example, 3-year mortality risk for dialysis patients is derived from the formula below,

-2.5998 (Baseline risk) + -0.0285 (Sex) + 0.0406 (Age) + -0.3882 (Black race) + -0.5724 (Other Race) + -0.5742(Hispanic ethnicity) + 0.4434(Cardiovascular disease)+ -0.3884 (Hypertension) + 0.0353(Diabetes) + 0.3435(Low albumin) + 0.3872(Home hemodialysis) + -0.2477(Peritoneal dialysis)

and 3-year mortality risk for transplant patients who have been on dialysis is derived from the formula below,

-5.1731(Baseline risk) + -0.1281(Sex) + 0.045(Age) + -0.0904(Black race) + -0.3463(Other race) + -0.4295(Hispanic ethnicity) + 0.2872(Cardiovascular disease)+ -0.147(Hypertension) + 0.3604(Diabetes) + 0.3331(Low albumin) + 0.0711(Home hemodialysis) + -0.1893(Peritoneal dialysis) + -0.1753(6-12 months on dialysis) + 0.2045(1-2 years on dialysis) + 0.3119(2-3 years on dialysis) + 0.5461(3-5 years on dialysis) + 0.7652(5-7 years on dialysis) + 1.1532(7-10 years on dialysis) + 1.4644(10-14 years on dialysis) + 1.3654(14+ years on dialysis)

and 3-year mortality risk for transplant patients who have never been on dialysis is derived from the formula below,

-5.0757 (Baseline risk) + -0.2183 (Sex) + 0.0416 (Age) + 0.1147 (Black race) + -0.4672 (Other Race) + -0.176 (Hispanic ethnicity) + 0.283 (Cardiovascular disease)+ -0.6676 (Hypertension) + 0.1437 (Diabetes) + 0.6646 (Low albumin)

where Baseline risk=1; Sex=1 for Male, Sex=2 for Female; 1=yes and 0=no for Black race, Other race, Hispanic ethnicity, Cardiovascular disease, Hypertension, Diabetes, Low albumin, Home hemodialysis, Peritoneal Dialysis, 6-12 months on dialysis, 1-2 years on dialysis, 2-3 years on dialysis, 3-5 years on dialysis, 5-7 years on dialysis, 7-10 years on dialysis, 10-14 years on dialysis, and 14+years on dialysis.

In this case, a 41 year old Non-Hispanic, African American male with a history of diabetes and hypertension who has been on dialysis for 8 months has a risk of dying over the next year of 6% on dialysis and 1% with a kidney transplant (relative risk of dying on dialysis vs. transplant is 6). This patient’s three year risk of dying on dialysis is 16%, and 3% with a transplant (Relative Risk=5). Risk of dying within the next three years with a deceased donor transplant (4%) is 2 times higher than risk of dying with a living donor transplant (2%).

Prediction Model Discrimination and Performance

We performed external validation of the risk prediction models for dialysis and transplantation at 3 years using the validation cohorts from USRDS. The discriminatory ability of the model for 3-year mortality was moderate for dialysis (c-statistic = 0.7055 [95% CI: 0.7038-0.7071] and for transplant (c-statistic = 0.7051 [95% CI: 0.6970-0.7131]) among patients who had been on dialysis and for patients who had never been on dialysis (c-statistic = 0.7259 [95% CI: 0.7028-0.7491]).  The c-statistic was 0.6861 (95% CI: 0.6770-0.6953) for deceased donor transplant and 0.7128 (95% CI: 0.6934-0.7322) for living donor transplant for patients who had been on dialysis. 

The c-statistic was 0.6762 (95% CI: 0.6430-0.7064) for deceased donor transplant and 0.7037 (95% CI: 0.6607-0.7467) for living donor transplant for patients who had been on dialysis. The c-statistics were similar among the derivation and validation datasets, indicating that the predictive models were generalizable among the U.S. ESRD patients and kidney transplant recipients.

Additional information about the development of the original and updated risk prediction models are described in the following publications: 

  • Patzer RE, Basu M, Larsen C, Pastan SO, Mohan S, Patzer MC, Konomos M, McClellan WM, Lea J, Howard D, Gander J, Arriola KJ. iChoose Kidney: a Clinical Decision Aid for Kidney Transplantation vs. Dialysis. Transplantation. 2016; 100(3): 630-9.
  • Gander J, Basu M, McPherson L, Garber M, Pastan SO, Manatunga A, Arriola KJ,  Patzer RE (2018). iChoose Kidney for Treatment Options: Updated Models for Shared Decision Aid. Transplantation.