A Real-World Cohort Study of the Risks of Kidney Failure and Death in Diabetes

Key Points Rates and risks of kidney failure versus death in diabetes have not been previously quantified across a full range of kidney function. In a large population with diabetes, risk of death exceeded kidney failure overall, but kidney failure outpaced death at eGFR <30 ml/min per 1.73 m 2 . Current risk estimates are needed to support public health and clinical strategies to monitor and improve kidney health in diabetes. Background Risks of kidney failure versus death in diabetes have not been previously quantified across a full range of kidney function. The aim of the study was to assess competing risks of kidney failure and death in a real-world cohort with diabetes. Methods Cumulative incidence (CMI) functions for kidney failure and death, stratified by baseline eGFR, were estimated for the diabetes population from electronic health record data at Providence and the University of California Los Angeles health systems. Cox proportional hazards models assessed predictors of kidney failure and death. For kidney failure, cause-specific hazards were modeled with competing risk of death. Results Among 618,739 persons with diabetes followed for a median (interquartile range) of 3.79 (1.80–6.00) years during 2013–2022, 4% ( n =24,097) developed kidney failure and 10% ( n =63,128) died. Five-year CMI of kidney failure increased from 2% (95% confidence interval, 1.4% to 1.5%) for eGFR ≥90 ml/min per 1.73 m 2 to 62% (95% confidence interval, 61.0% to 63.5%) for eGFR 15–29 ml/min per 1.73 m 2 . CMI of all-cause death was higher than kidney failure with eGFR ≥45 ml/min per 1.73 m 2 , whereas kidney failure became more common at eGFR <30 ml/min per 1.73 m 2 . Hazards of kidney failure were higher in men (reference: women), 40–59 and 60–79-year groups (reference: 18–39 years), non-White race or Hispanic or Latino/a ethnicity groups (reference: non-Hispanic White), and by noncommercial (reference: commercial) health insurance or hospitalization (yes/no) within 1 year before follow-up. Hazards of death were similar except that age ≥80 years imparted the highest risk, and only the American Indian or Alaska Native and other race groups had higher risk. Conclusions Death was more likely than kidney failure at eGFR ≥45 ml/min per 1.73 m 2 , but this trend reversed at eGFR <30 ml/min per 1.73 m 2 . These contemporary risk estimates are important for public health and clinical strategies for monitoring and interventions to improve kidney health in diabetes.