Predicting Emergency Admissions

fluffy kitten
We expect that contextual and event data will improve our predictions.

We are working with hospitals to provide a better prediction of how many patients arrive at the Emergency Department of a hospital on a given day. Being able to make predictions about the expected number of admissions can greatly improve a hospital’s operations, and ensure that enough resources are available at the time when they are needed.

Previous work in this area has attempted to make this prediction using data from the hospital’s own information systems. Other work has also considered using environmental data such as temperature to make this prediction.

In our work we are working towards developing a predictive model that utilises contextual information from the capture area of the hospital. In addition to historical admission data and weather phenomena, we are also integrating data from public events, local council websites, news headlines, and social media hashtags.

Our intention is to generate a model that predicts in the short term (next 24-72 hours) the number of expected arrivals. Besides offering this prediction as a “black box”, we are aiming to provide explainable predictions which can help hospital staff make informed decisions. For example, the predictions should be able to account for an expected spike in admissions, for example following a major sport event taking place in a nearby area.

We are closely working with hospitals’ data analysis to develop, evaluate, and improve our prediction model.