Background: Chronic kidney disease is a health burden for the general population. We designed a cohort study to construct prediction models for chronic kidney disease in the Chinese population.
Methods: A total of 5168 participants were followed up during a median of 2.2 (interquartile range, 1.5-2.9) years, and 190 individuals (3.7%) developed chronic kidney disease, defined by a glomerular filtration rate of less than 60 mL/min/1.73 m(2).
Results: We developed a point system to estimate chronic kidney disease risk at 4 years using the following variables: age (8 points), body mass index (2 points), diastolic blood pressure (2 points), and history of type 2 diabetes (1 point) and stroke (4 points) for the clinical model, with the addition of uric acid (2 points), postprandial glucose (1 point), hemoglobin A1c (1 point), and proteinuria 100 mg/dL or greater (6 points) for the biochemical model. Similar discrimination measures were found between the clinical model (area under the receiver operating characteristic curve, 0.768; 95% confidence interval (CI), 0.738-0.798) and the biochemical model (area under the receiver operating characteristic curve, 0.765; 95% CI, 0.734-0.796). The area under the receiver operating characteristic curve of the clinical model was 0.667 (95% CI, 0.631-0.703) for the external validation data from community-based cohort participants. The optimal cutoff value for the clinical model was set as 7, with a sensitivity of 0.76 and a specificity of 0.66.
Conclusion: We constructed a clinical point-based model to predict the 4-year incidence of chronic kidney disease. This prediction tool may help to target Chinese subjects at risk of developing chronic kidney disease.