A Random Forest Algorithm for Assessing Risk Factors Associated With Chronic Kidney Disease: Observational Study.

Area of Science:

• Medical research
• Public health
• Machine learning applications in healthcare

Background:

• Chronic kidney disease (CKD) prevalence and mortality are rising globally.
• CKD poses a significant and increasing economic burden.
• Early detection and intervention are vital for mitigating CKD progression and patient costs.

Purpose of the Study:

• To investigate the utility of the random forest (RF) algorithm for assessing CKD risk factors.
• To compare RF algorithm performance against traditional logistic regression models.

Main Methods:

• Analysis of 40,686 individuals' screening records (2015-2020) in Shanghai, China.
• Classification of participants based on glomerular filtration rate and albuminuria.
• Application of logistic regression and RF algorithms to identify and rank CKD risk factors.

Main Results:

• Logistic regression identified gender, age, obesity, abnormal eGFR, retirement, and insurance status as significant CKD risk factors.
• RF algorithm highlighted age, albuminuria, working status, and urine albumin-creatinine ratio as top predictors.
• The RF model achieved a high predictive accuracy with an AUC of 93.15%.

Conclusions:

• The RF algorithm demonstrates significant predictive value for CKD risk factor assessment.
• RF enables effective screening of individuals at risk for CKD.
• These findings support the use of RF for early CKD intervention and prevention strategies.