RESUMEN
Aim: Chronic kidney disease (CKD) is a major complication of diabetes and a significant disease burden on the healthcare system. The aim of this work was to apply a predictive model to identify high-risk patients in the early stages of CKD as a means to provide early intervention to avert or delay kidney function deterioration. Materials and methods: Using the data from the National Diabetes Database in Singapore, we applied a machine-learning algorithm to develop a predictive model for CKD progression in diabetic patients and to deploy the model nationwide. Results: Our model was rigorously validated. It outperformed existing models and clinician predictions. The area under the receiver operating characteristic curve (AUC) of our model is 0.88, with the 95% confidence interval being 0.87 to 0.89. In recognition of its higher and consistent accuracy and clinical usefulness, our CKD model became the first clinical model deployed nationwide in Singapore and has been incorporated into a national program to engage patients in long-term care plans in battling chronic diseases. The risk score generated by the model stratifies patients into three risk levels, which are embedded into the Diabetes Patient Dashboard for clinicians and care managers who can then allocate healthcare resources accordingly. Conclusion: This project provided a successful example of how an artificial intelligence (AI)-based model can be adopted to support clinical decision-making nationwide.