Patient classification and outcome prediction in IgA nephropathy

Area of Science:

• Nephrology
• Medical Informatics
• Data Science

Background:

• IgA Nephropathy (IgAN) is a prevalent kidney disease often leading to End-Stage Kidney Disease (ESKD).
• Predicting the long-term prognosis of IgAN patients at diagnosis is challenging due to complex clinical and laboratory interrelationships.
• Accurate prediction models are crucial for timely intervention and patient management.

Purpose of the Study:

• To identify the optimal data mining tool for predicting ESKD in IgAN patients.
• To develop a model that predicts both the likelihood of reaching ESKD and the timeframe (within or after 5 years).
• To compare the performance of Artificial Neural Networks (ANNs), Neuro Fuzzy Systems (NFSs), Support Vector Machines (SVMs), and Decision Trees (DTs).

Main Methods:

• Utilized the world's largest IgAN cohort dataset (1174 patients from Italy, Norway, Japan) spanning 30 years.
• Compared four data mining models: ANNs, NFSs, SVMs, and DTs.
• Employed 10-fold cross-validation for unbiased model performance evaluation.

Main Results:

• Artificial Neural Networks (ANNs) demonstrated superior performance over other models.
• The ANN model achieved over 90% accuracy, precision, recall, and f-measure for time-to-ESKD prediction.
• The ANN model for ESKD prediction also exceeded 90% accuracy, with slightly lower, yet significant, performance metrics for the ESKD class.

Conclusions:

• Data mining models, particularly ANNs, are effective tools for predicting IgAN patient outcomes.
• The developed predictive model can support clinical diagnosis and prognosis.
• Implementation in a Web-based decision support system (DSS) enhances clinical utility.