Enhancing UTI severity classification in elderly inpatients with AI: Toward smarter clinical decision-making

Abstract

This study presents a machine learning-based system to assess mortality risk in elderly UTI patients using data from Chonburi Hospital, Thailand (2019–2023), involving 17 predictive variables. Following data preprocessing, which included addressing missing values, removing outliers, and standardizing data, key features were identified using the ReliefF algorithm. In this study, various classification models, including SVM, K-NN, and neural networks, were evaluated. The K-NN model demonstrated superior performance, a sensitivity of 96.3%, specificity of 81.6%, F1-score of 89.6%, and an overall accuracy of 92.2%. These results highlight the potential of the K-NN model for early risk assessment and its utility in enhancing clinical decision-making processes.

Ähnliche Arbeiten