External Validation of the Klinrisk Model in US Commercial, Medicare Advantage, and Medicaid Populations

Background: Chronic kidney disease (CKD) is typically undiagnosed till the majority of kidney function (eGFR) is lost. Accurate risk prediction tools for progressive CKD can enable early intervention for high risk individuals. The Klinrisk machine learning model accurately predicts progressive CKD using routinely collected laboratory data. We aimed to validate this model in US commercial, Medicare Advantage, and Medicaid populations. Methods: The Klinrisk random survival forest model predicts progressive CKD (40% decline in eGFR or kidney failure) using the values of age, sex, and 20 laboratory variables, including results from complete blood cell counts, chemistry panels, comprehensive metabolic panels, and urinalysis. We assessed model performance at 2- and 5- years post-index (first available serum creatinine result) in patients with/without urinalysis results (albumin-to-creatinine ratio, protein-to-creatinine-ratio, and semi-quantitative dipstick) in a large representative US population. Performance was assessed with discrimination (area under the receiver operating characteristic curve), Brier scores, and calibration plots. Results: A total of 4,410,131 patients were evaluated with commercial insurance, 341,666 with Medicare Advantage, and 93,056 patients with Medicaid coverage. Discrimination was excellent across all forms of payor and with or without the results of urinalysis. In all cohorts, for prediction of the progression, AUCs ranged between 0.80 to 0.83 at 2 years, and 0.78-0.83 at 5 years. When urinalysis data were available, AUCs ranged between 0.81 to 0.87 at 2 years, and 0.80 to 0.87 at 5 years (Table). Brier scores were below 0.071 (0.068 to 0.075) for each combination of urinalysis availability and insurer type. Conclusions: A machine model trained on routine laboratory data can predict progression of CKD in a large representative US population of adults with or at risk for kidney disease. Implementation of the Klinrisk model can help identify patients who benefit from early intervention to delay CKD progression and reduce health care costs. Funding: Commercial Support - Boehringer Ingelheim AUC at 2- and 5- years (95% confidence interval) - Insurer All patients Commercial, n = 4,410,131 Medicare, n = 4,410,131 Medicaid, n = 93,056 UACR directly measured Commercial, n = 178,266 Medicare, n = 25,954 Medicaid, n = 9,353 Urine ACR or urine PCR Commercial, n = 193,992 Medicare, n = 28,120 Medicaid, n = 10,108 Urine ACR, urine PCR, or semi-quantitative dipstick result Commercial, n = 1,061,762 Medicare, n = 92,410 Medicare, n = 38,867 Commercial (2 years) Commercial (5 years) 0.83 (0.82 - 0.83)0.81 (0.81 - 0.81) 0.86 (0.85 - 0.87)0.84 (0.83 - 0.85) 0.86 (0.85 - 0.87)0.85 (0.84 - 0.85) 0.87 (0.86 - 0.97)0.85 (0.84 - 0.85) Medicare (2 years) Medicare (5 years) 0.80 (0.79 - 0.80)0.78 (0.78 - 0.79) 0.79 (0.77 - 0.80)0.78 (0.77 - 0.79) 0.79 (0.78 - 0.81)0.78 (0.77 - 0.80) 0.81 (0.80 - 0.82)0.80 (0.79 - 0.80) Medicare (2 years) Medicare (5 years) 0.83 (0.83 - 0.83)0.83 (0.83 - 0.83) 0.84 (0.81 - 0.87)0.87 (0.84 - 0.90) 0.84 (0.81 - 0.87)0.86 (0.83 - 0.90) 0.84 (0.83 - 0.86)0.87 (0.85 - 0.89)

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.