Chronic kidney disease (CKD) and heart failure (HF) are the first and most frequent, and life-threatening comorbidities in type 2 diabetes (T2D) patients. Early intervention is important to prevent CKD and HF onset and to improve prognosis of these patients; however, the risk assessment of CKD and HF remains to be established. We aimed to build a machine learning model to predict the risk of CKD or HF onset in the early-stage T2D patients without a history of CKD or cardiovascular diseases (CVD) . We developed prediction models using light gradient boosting machine (LightGBM) , neural network, logistic regression, and Cox proportional hazards model. The model was derived and validated in a sample of 217,054 T2D patients aged ≥18 years without a history of CKD or CVD extracted from a Japanese hospital-based administrative claims data. The outcomes used for the prediction model were diagnosis of incident CKD or HF and hospitalization for CKD or HF in 1, 2, 3, and 5 years. Based on the importance, 60 characteristics and laboratory data were selected and used for models. LightGBM outperformed other models in all outcomes, with AUROCs 0.777 for diagnosis of incident CKD or HF, and 0.785 for hospitalization for CKD or HF in 5years. (Figure) The present study demonstrated successful development of prediction algorithms to support identifying T2D patients with high-risk of CKD or HF onset, which will contribute to improvement of patient prognosis.


H. Tsubota: Employee; AstraZeneca. A. Suzuki: Research Support; Chugai Pharmaceutical Co., Ltd., Ono Pharmaceutical Co., Ltd., Taisho Pharmaceutical Holdings Co., Ltd., Takeda Pharmaceutical Company Limited. Speaker's Bureau; Amgen K.K., Eli Lilly and Company, Novo Nordisk A/S. M. Makino: None. E. Kanda: None. T. Yajima: Employee; AstraZeneca. Y. Kidani: Employee; AstraZeneca. N. Morita: Employee; AstraZeneca.

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