Author:
Toyama Tadashi,Shimizu Miho,Yamaguchi Taihei,Kurita Hidekazu,Morita Tetsurou,Oshima Megumi,Kitajima Shinji,Hara Akinori,Sakai Norihiko,Hashiba Atsushi,Takayama Takuzo,Tajima Atsushi,Furuichi Kengo,Wada Takashi,Iwata Yasunori
Abstract
AbstractAssociation rule is a transparent machine learning method expected to share information about risks for chronic kidney disease (CKD) among diabetic patients, but its findings in clinical data are limited. We used the association rule to evaluate the risk for kidney disease in General and Worker diabetic cohorts. The absence of risk factors was examined for association with stable kidney function and worsening kidney function. A confidence value was used as an index of association, and a lift of > 1 was considered significant. Analyses were applied for individuals stratified by KDIGO’s (Kidney Disease: Improving Global Outcomes) CKD risk categories. A General cohort of 4935 with a mean age of 66.7 years and a Worker cohort of 2153 with a mean age of 47.8 years were included in the analysis. Good glycemic control was significantly related to stable kidney function in low-risk categories among the General cohort, and in very-high risk categories among the Worker cohort; confidences were 0.82 and 0.77, respectively. Similar results were found with poor glycemic control and worsening kidney function; confidences of HbA1c were 0.41 and 0.27, respectively. Similarly, anemia, obesity, and hypertension showed significant relationships in the low-risk General and very-high risk Worker cohorts. Stratified risk assessment using association rules revealed the importance of the presence or absence of risk factors.
Funder
Japan Society for the Promotion of Science
Publisher
Springer Science and Business Media LLC
Cited by
2 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献