New Risk-Prediction Tool Enhances Mortality Estimates for Advanced CKD Patients

Researchers led by Dr. Connie Rhee at UCLA and the Veterans Affairs Greater Los Angeles Healthcare System developed and validated a mortality risk-prediction tool for patients with advanced chronic kidney disease (CKD). The model estimates a patient's 1-year mortality risk comparing outcomes of conservative care versus dialysis treatment. The study analyzed data from over 61,000 veterans in the VA database with an eGFR below 25, comparing those who started dialysis within two years to those who received conservative care. The tool was externally validated using data from over 76,000 patients in the OptumLabs Data Warehouse, including commercially insured and Medicare Advantage patients. The model demonstrated consistent predictive accuracy across these cohorts, with C-statistics around 0.69 to 0.70, indicating good discrimination between higher and lower mortality risk patients. Key factors associated with increased mortality included older age, advanced and rapidly declining kidney function, elevated albuminuria, frailty, low BMI and serum albumin, recent hospitalizations, heart disease, sepsis, and whether the patient transitioned to dialysis. This multifactorial risk assessment supports personalized treatment considerations for advanced CKD patients. The prediction tool aims to improve shared decision-making among patients, healthcare providers, and care partners by providing a clearer risk profile for conservative management versus dialysis. Such tools can assist in aligning treatment plans with patient-specific clinical factors and preferences. This development arrives amid ongoing efforts to optimize CKD management and reduce unnecessary interventions. The validated model’s incorporation of real-world data from both veterans and insured populations enhances its applicability to diverse clinical settings. It represents a step toward more data-driven, patient-centered care strategies in nephrology.