Step Towards Predicting Patient Length of Stay in Intensive Care Units

Increasingly, hospitals are collecting huge amounts of data through new storage methods. These data can be use to extract hidden knowledge, which can be crucial to estimate the length of stay of admitted patients in order to improve the management of hospital resources. Hence, this article portrays the performance analysis of different data mining techniques through the application of learning algorithms in order to predict patients’ length of stay when admitted to an Intensive Care Unit. The data used in this study contains about 60,000 records and 28 features with personal and medical information. A full analysis of the results obtained with different Machine Learning algorithms showed that the model trained with the Gradient Boosted Trees algorithm and using only the features that were strongly correlated to the patient’s length of stay, achieved the best performance with 99,19% of accuracy. In this sense, an accurate understanding of the factors associated with the length of stay in intensive care units was achieved.