Development and implementation of nationwide predictive model for admission prevention: System architecture & machine learning

Hospitals in developed countries are experiencing increasing inpatient load due to aging. Many of these readmissions could be prevented through early interventions. Using routinely available data in Electronic Health Records, we developed and implemented a machine learning predictive model to identify patients who are at risk of multiple unplanned readmission within the next 12 months from their index hospital admission. This is the first nationwide Predictive Model for Admission Prevention in Singapore that is deployed in all public acute general hospitals to identify high risk patients for enrollment into a community-centric intervention programme after discharge. In this paper, we describe the approach we have taken to augment the prediction model into a routine patient care process.