Predicting System for Nosocomial Infections in Peruvian Public Hospitals using Neural Networks

One of the most important challenges for public health systems in Peru is the control of Healthcare-Associated Infections (HAI), which has had a high economic and social impact in recent years. Due to the limited resources available in the health sector, it is critical to find solutions to strengthen medical surveillance and have valuable information that allows for timely decision-making. In this context, artificial intelligence (AI) systems, consisting of logical algorithms trained to make decisions, offer alternatives to improve medical management by predictive models implemented in different scenarios. This applied research proposes the implementation of an HAI prediction system using artificial neural networks. The proposed system was developed in a five-layer architecture: capture, logic, training, storage, and presentation. The validation process for the proposal was carried out in three phases: selection, registration, and training. Two experiments were conducted, for which data from two and three weeks were taken from eight hospitals belonging to the Social Health Insurance of Peru. Additionally, with the support of experts, the results were evaluated, this demonstrated that the system allows for the prediction of HAI's based on the fourteen selected variables with values above 95% in each metric, reaching sensitivity values of 99.1%, specificity of 99.6%, and AUROC of 99.2%.