Predicting survival time for kidney dialysis patients: a data mining approach

Area of Science:

• Nephrology
• Data Science
• Health Informatics

Background:

• End-stage kidney disease (ESKD) treatment via hemodialysis is costly.
• Improving patient survival and reducing healthcare expenses are critical goals.
• Dialysis involves complex care with numerous monitored parameters influencing outcomes.

Purpose of the Study:

• To investigate the intricate relationships between demographic, clinical, and treatment parameters and patient survival in hemodialysis.
• To develop a data-driven approach for predicting individual patient survival.
• To identify key factors impacting survival for personalized intervention strategies.

Main Methods:

• Employed data preprocessing and transformation techniques.
• Utilized a data mining approach with two distinct algorithms to extract decision rules.
• Developed a decision-making algorithm for survival prediction in new patients.

Main Results:

• Successfully elicited knowledge regarding the interaction between measured parameters and patient survival.
• Identified and interpreted important parameters influencing patient survival based on data mining insights.
• Validated the approach using data from four distinct dialysis centers.

Conclusions:

• Data mining can uncover complex relationships between patient parameters and survival in hemodialysis.
• The developed decision-making algorithm shows promise for predicting patient survival.
• Findings support the potential for individualized treatment plans to improve outcomes in ESKD patients.