Semantic-enhanced models to support timely admission prediction at emergency departments

With the rapid outstripping of limited health care resources by the demands on hospital care, it is of critical importance to find more effective and efficient methods of managing care. Our research addresses the problem of emergency department (ED) crowding by building classification models using various types of pre-admission information to help predict the hospital admission of individual patients. We have developed a framework of hospital admission prediction and proposed two novel approaches that capture semantic information in chief complaints to enhance prediction. Our experiments on an ED data set demonstrate that our proposed models outperformed several benchmark methods for admission prediction. These models can potentially be used as decision support tools at hospitals to improve ED throughput rate and enhance patient care.