Identifying The Risk Factors For Survival In Hemodialysis Patients Using Random Forests Algorithm
The statistics of chronic kidney diseases (CKD) and end-stage renal disease (ESRD) show the trends and movement in the Taiwan elder population. Hemodialysis is the major treatment for the patients with end-stage renal disease. In the present study, the random forest classification algorithm was used to identify the association between all-cause mortality and clinical risk factor in hemodialysis patients. The random forest algorithm is widely used in operation studies, especially in the identification analysis. The method of random forest can help to determine the strategy that is most likely to reach a goal. Therefore, the clinical risk factors can be analyzed to estimate the survival and mortality for attention. A total of 1,018 hemodialysis patients were included and tracked from January 2009 to December 2013 from Kaohsiung Chang Gung Memorial Hospital. The regression test was used to evaluate the correlation between the survival by forward and backward selection. After selecting the significant factor, the random forest was then used to train a best model which was validated by the testing data. The final best model showed accuracy of 82.31% in the validation result, indicating that the risk factors including elder age, higher cardiothoracic ratio, higher alkaline-phosphate level, and lower albumin level were significant increase in the mortality risk. In the future work, more factors might be identified to determine the risk of mortality cautiously.
Keywords: Hemodialysis, Regression, Random Forest Algorithm