Processing electronic medical records to improve predictive analytics outcomes for hospital readmissions

Hospital readmissions are costly but largely preventable. In recent years, many researchers have used predictive analytics to build models that can minimize the adverse economic and social consequences of readmissions in chronic diseases. Most of these studies, however, have focused on improving the results either through the development of better models or through employing richer data sets. A very small number of them have focused on a comprehensive data preprocessing to improve the efficacy of analytics methods for better predictions. In this study, we propose a new data processing approach that extracts individual- and database-level historical information from the medical records to improve the performance of readmission analytics. We test and validate this method using two rather large data sets that belong to chronic diseases with the highest rates of hospital readmissions. We conclude that proper processing of large clinical data sets with analytics and big data technologies can provide competitive advantages to health care organizations.