Abstract
Surgical site infections impact hospital readmission rates, length of stay, and patient and hospital expense. The use of computational intelligence methods can help to predict the risk of SSIs and provide an early warning, enabling hospitals to prepare in advance to respond to these infections. The objective of this paper is to present a machine learning-based predictive model for surgical site infections. This paper also reviews the most recent machine learning-based models developed for the prediction of SSIs. When these predictive models are applied correctly and used effectively, they can be helpful for clinical surveillance teams. However, the implementation of these models and related tools requires quality data to be stored in electronic health records, which may not be available in all health information systems. The limitations of clinical data and the absence of labels adding several challenges with the implementation of the predictive models using machine learning; an imbalanced dataset is also a common issue that can influence the performance of the model, thus requiring an improved strategy to address this concern. Recently, the interpretability of predictive models has become important for hospital physicians to translate the models into real practices.