Readmission prediction

Predicting the likelihood of all-cause patient readmissions is outside the scope of a typical clinician's knowledge base, since it is not tied to a specific organ system or disease. However, it is becoming a problem of increasing importance in the healthcare world, since preventable hospital readmissions are a major cause of elevated healthcare expenditures in the United States and other countries. We discussed the incentive and rationale for predicting hospital readmissions and the US government's Hospital Readmission Reduction Program (HRRP) in Chapter 6, Measuring Healthcare Quality. Let's now review how machine learning algorithms can be used to augment simpler readmission risk scores.

LACE and HOSPITAL scores

The most well-known readmission risk score is the LACE score, which was developed in 2010 by Canadian researchers (van Walraven et al., 2010). "LACE" stands for the four predictors used to calculate the score, and the full score calculation ranges from 0-19...