Predicting the risks of kidney failure and death in adults with moderate-to-severe chronic kidney disease (KDpredict)

https://storage.unitedwebnetwork.com/files/1099/314b1ad4cdf0637b864751daac5fd852.pdf
Predicting the risks of kidney failure and death in adults with moderate-to-severe chronic kidney disease (KDpredict)
Pietro
Ravani
Simon Sawhney simon.sawhney@abdn.ac.uk University of Aberdeen Aberdeen Centre for Health Data Science Aberdeen
Robert Quinn Rob.Quinn@albertahealthservices.ca University of Calgary Medicine Calgary
Uffe Heide-Jørgensen uhj@clin.au.dk Aarhus University Clinical Epidemiology Aarhus
Simon Jensen skj@clin.au.dk Aarhus University Clinical Medicine Aarhus
Andrew Mclean andrew.mclean1@abdn.ac.uk University of Aberdeen Aberdeen Centre for Health Data Science Aberdeen
Christian Christiansen cfc@clin.au.dk Aarhus University Clinical Medicine Aarhus
Thomas Gerds tag@biostat.ku.dk University of Copenhagen Department of Public Health Copenhagen
Ping Liu ping.liu1@ucalgary.ca University of Calgary Medicine Calgary
 
 
 
 
 
 
 

Chronic kidney disease (CKD) disproportionally affects older individuals, who are more likely to die than develop kidney failure. However, treatment decisions are currently supported by tools that predict risks of kidney failure in isolation only. We designed, trained and tested a super-learner strategy for risk prediction of kidney failure and death in people with incident moderate-to-severe chronic kidney disease, stage G3bG4-CKD (KDpredict).

We used routinely recorded population health data from Canada (training and temporal testing), and Denmark and Scotland (geographical testing) to create and evaluate a super-learner predicting 1-to-5-year risks of kidney failure and all-cause death in adults with incident G3bG4-CKD (newly recorded eGFR 15-44 mL/min/1.73 m2). Super-learner is a meta-algorithm that uses cross-validation to select the best performing learners among many pre-specified regression models and machine-learning algorithms, based on their ability to minimize prediction error. Learners included standard Cox (for mortality) and cause-specific Cox models (for kidney failure) with different settings (i.e., different interactions, stratification factors, spline functions) and random survival forests with different configurations (number of trees, predictors tried at each split, size of terminal leaves, etc). Predictors included age, sex, eGFR, albuminuria, without or with diabetes and cardiovascular disease. We used the index of prediction accuracy (IPA) to compare KDpredict with the benchmark, kidney failure risk equation (KFRE).

The study included 67,942 Canadians, 17,528 Danish and 7,740 Scottish residents with G3bG4-CKD (median age 77-80 years; median eGFR 39 mL/min/1.73 m2). Median follow-up times were 5-6 years in all cohorts. Rates of kidney failure and death were 0.8-1.1 and 10-12 per 100 person-years, respectively. KDpredict outperformed the KFRE in kidney failure risk prediction: 5-year IPA 27.8% (95% CI 25.2 to 30.6%) vs 18.1% (15.7 to 20.4%) in Denmark and 30.5% (27.8 to 33.5%) vs 14.2% (12.0 to 16.5%) in Scotland. KFRE and KDpredict provided substantially different predictions, potentially leading to diverging treatment decisions. An 80-year-old man with an eGFR of 30 mL/min/1.73 m2 and an ACR of 100 mg/g (11 mg/mmol) would receive a 5-year kidney failure risk prediction of 10% from KFRE (above the current 5% threshold for nephrology referral). The same man would receive 5-year risk predictions of 2% for kidney failure and 57% for mortality from KDpredict. A 75-year-old woman with an eGFR of 20 ml/min/1.73 m2 and an ACR of 500 mg/g (56 mg/mmol) would receive a 2-year kidney failure risk prediction of 18% from KFRE (above the current 10% referral threshold for multidisciplinary kidney care). The same woman would receive 2-year risk predictions of 8% for kidney failure and 27% for mortality from KDpredict. Individual risk predictions from KDpredict with 4 or 6 predictors were accurate for both outcomes. The KDpredict models retrained using older data provided accurate predictions when tested in temporally distinct, more recent data.

KDpredict provides accurate risk predictions of kidney failure and death for people with incident G3bG4-CKD. This tool could be incorporated into electronic medical records or accessed online to support more holistic decision-making in this patient population. The KDpredict learning strategy is designed to be adapted to local needs and regularly revised over time to account for changes in the underlying health system and care processes.

E-Poster Format Requirements
  • PDF file
  • Layout: Portrait (vertical orientation)
  • One page only (Dim A4: 210 x 297mm or PPT)
  • E-Poster can be prepared in PowerPoint (one (1) PowerPoint slide) but must be saved and submitted as PDF file.
  • File Size: Maximum file size is 2 Megabytes (2 MB)
  • No hyperlinks, animated images, animations, and slide transitions
  • Language: English
  • Include your abstract number
  • E-posters can include QR codes, tables and photos