Assessing Palliative Care Needs Using Machine Learning Approaches

Palliative care improves quality of life for patients and their families when facing a life-threatening illness. Studies have shown that physicians tend to over-estimate prognoses, which in combination with treatment inertia results in underutilization of end-of-life care plans. In our study, we apply machine learning methods to patients’ Electronic Health Record (EHR) data to predict patients who are likely to benefit from palliative care. Our work is based on a real-world dataset of 1,462,862 hospitalization records from the State Inpatient Database (SID) of Florida from 2012 to 2015. We examine the performance of five popular machine learning algorithms on all-cause and ten individual disease subgroups. Our experimental results demonstrate that machine learning models can achieve approximately 68% and 60% predictive accuracy for the patients who "need" and "do-not-need" end-of-life care respectively. Given the successful palliative needs identification rate is less than 50% among current hospital practices, our models can be used as an auxiliary tool to assist in palliative care resource allocation.