Exploring New Methods of Identifying the Most Complex Patients for Targeted Interventions
About the Presenters
Somayeh Sadat, Senior Methodologist Ali Vahit Esensoy, Senior Methodologist Informatics Centre of Excellence, Cancer Care Ontario
Abstract
It is well known in the healthcare field that a small proportion of patients account for the majority of healthcare costs. Earlier identification of these patients could enable clinicians as well as policymakers to target clinical or policy interventions aimed at reducing costly health system use. In this talk, we present two use cases of using machine learning to build Bayesian Network predictive models of complex patients. In one case, clinical and past health system use data readily available on the first day of dialysis is used to predict whether or not a dialysis patient would be among the top 10% of dialysis patients in terms of inpatient hospital use during the first year of dialysis. Given that these top 10% of patients account for more than 55% of inpatient hospital costs in their first year of dialysis, proactive identification and targeted clinical interventions for these patients as well as using policy levers to impact drivers of high inpatient hospital use can drastically reduce health system cost. In another use case, we present ongoing work in building a predictive model of the top 10% of patients who wait the longest for long-term care homes while occupying acute hospital beds. Not only occupying acute hospital beds without a medical need for these beds imposes unnecessary costs to the system, but also the beds occupied for such long periods of time practically become blocked, potentially creating overcrowding in emergency departments with patient awaiting admissions to inpatient beds as well as cancellations of elective surgeries due to insufficient availability of inpatient beds. Understanding the predictors of such long waits for long-term care beds is a major step in designing interventions that can reduce health system costs and optimize patient flow throughout the health system.