The ability to predict rapid kidney function decline (RKFD) in patients with early stages of type 2 diabetic kidney disease (T2DKD) can improve long-term health outcomes through earlier intervention. Our objective was to develop a robust risk score for patients with early stages of T2DKD. We obtained EHR and plasma from patients with T2DKD (Kidney Disease Improving Global Outcomes [KDIGO] G1-G2, A2-A3 or G3a-G3b, A1-A3) from the Mount Sinai BioMe Biobank and the Penn Medicine Biobank and measured soluble tumor necrosis factor receptor (sTNFR1, sTNFR2) and kidney injury molecule-1 (KIM-1) via a validated quantitative electrochemiluminescent assay. A machine learning model was trained using the EHR/biomarker data to predict RKFD, defined as eGFR decline of ≥5 ml/min/year or ≥40% sustained decline in eGFR or kidney failure within 5 years, and performance was compared to a clinical model and KDIGO risk-matrix. A total of 1146 subjects with a mean age of 63, baseline eGFR 54 ml/min, and median UACR 61 mg/g were randomly divided into 60% training (n=686) and 40% validation (n=460) sets. The median follow-up was 4.3 years and 241 (21%) experienced RKFD. The AUC of KidneyIntelX for RFKD was 0.85 (95% CI 0.84-0.86) in training and 0.77 (95% CI 0.76-0.79) in the validation set vs. AUC of 0.67 (95% CI 0.64-0.70) for a clinical model. Using pre-specified cutoffs from the training set, KidneyIntelX stratified 47%, 37% and 16% of the validation set into low-, intermediate- and high-risk strata, with a PPV of 62% (vs. 41% for KDIGO) in the high-risk group (p <0.001), a net reclassification improvement for events of 41% (p <0.05), and a NPV of 91% in the low-risk group (vs. 85% for KDIGO). KidneyIntelX, a machine learning model combining biomarkers and EHR data, improved prediction of RKFD over standard clinical metrics in patients with T2DKD. Future studies will prospectively evaluate KidneyIntelX in clinical decision-making and impact.


L. Chan: None. G.N. Nadkarni: Advisory Panel; Self; Renalytix AI. Consultant; Self; AstraZeneca, Reata, Renalytix AI. Stock/Shareholder; Self; Renalytix AI. Other Relationship; Self; Renalytix AI. F. Fleming: Board Member; Self; Renalytix AI plc. F.E. Salem: None. B. Murphy: Board Member; Self; RenalytixAI. M.J. Donovan: Consultant; Self; Renalytix AI plc. S. Coca: Advisory Panel; Self; Renalytix AI plc. Consultant; Self; Bayer Healthcare Pharmaceuticals Inc., CHF Solutions, Relypsa, Inc., Takeda Pharmaceutical Company Limited. Stock/Shareholder; Self; Renalytix AI plc. S.M. Damrauer: Consultant; Spouse/Partner; Rhythm Pharmaceuticals. Research Support; Self; CytoVas LLC, Renalytix AI plc.


Renalytix AI plc

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at