Predictive Analytics in Health Care: Methods and Approaches to Identify the Risk of Readmission

The increasing focus on evidence-based healthcare services as well as rising health expenditures for inpatient treatment forces hospitals to introduce new approaches to allow for a more efficient delivery of said services. As a new measure of healthcare quality, readmission rates are increasingly used to determine the quality of care, benchmark hospital performance and determine funding rates or even issue penalties. It is therefore key to determine patients at high risk of readmission. This can be done by using predictive risk models that are able to predict the risk of readmission to the hospital for individual patients using various data mining techniques and algorithms. Based on these models and with the increasing amount of data collected in hospitals, clinicians and hospital management can be supported in their daily decision-making to reduce readmission rates. Ultimately, the implementation of such prediction models can help avoid unnecessary costs as well as improve the quality of healthcare services. This work aims at identifying and analysing state-of-the-art risk prediction models in healthcare with regard to their specific application areas, applied algorithms and resulting accuracy to determine the suitability of different methods in different healthcare contexts.