In this chapter, we have built a predictive model for predicting outcomes in the emergency department. While there are many machine learning problems in healthcare, this exercise has demonstrated the issues typically faced as one preprocesses healthcare data, trains and scores models, and makes predictions with unlabeled data. This chapter marks the end of the coding portion of this book.
Now that we have seen the construction of a predictive model firsthand, the next logical question to ask is how predictive models have fared when compared with traditional statistical risk scores in predicting clinical outcomes. We explore that question in the next chapter.