Background: To embrace the need for freely accessible training data sets originating from the real world, in the ELISE project, we integrate source data from a pediatric intensive care unit and provide it to researchers.
Objective: We present our vision, initial results and steps on a trail towards an evolutionary open pediatric intensive care data set.
Methods: Our evolution plan for the data set comprises three steps. The final data set will include raw clinical data and labels on critical outcomes such as organ dysfunction and sepsis, generated automatically by computerized and well-evaluated methods.
Results: First step resulted in an initial version data set available in a central repository.
Conclusions: Our approach has great potential to provide a comprehensive open intensive care data set labeled for critical pediatric outcomes and, thus, contributing to overcome the current lack of real-world pediatric intensive care data usable for training data-driven algorithms.
Keywords: Data Science; Dataset; Intensive Care Units; Pediatric.