I haven't been admitted to hospital many times in my life, but every time the only thing I really cared about was: when am I going to get out? It's also a question that weighs heavily on hospital managers: by knowing ahead of time how long each patient's stay is likely to be, they can better manage facilities and staff, and know whether the hospital is likely to reach maximum capacity in the near future.
To help hospital administrators better predict how long patients are likely to stay, Microsoft has published the Predicting Length of Stay in Hospitals solution on the Cortana Intelligence Gallery. Clicking on “deploy” creates an instance of the Data Science Virtual Machine with simulated patient data in SQL Server, and a model implemented with R Services to predict the length of stay. The predictions are then presented as a Power BI dashboard to a Care Line Manager or a Chief Medical Information Officer as shown below. (Click the Try It Now button on this page to interact with the dashboard.)
The underlying model is a Random Forest implemented with
rxDForest in the RevoScaleR package. (The data science process behind training the model is described here, and the R code is available in this Github repository). The model is trained using simulated patient lab and diagnosis data, and observed lengths of stay. For hospitals that want to build a similar model using real data, the Predicting Hospital Length of Stay solution page provides detailed background on configuring a similar service in a cloud-based architecture or installed in a private data center.
GitHub (Microsoft): Predicting Hospital Length of Stay