Background Palliative care is known to improve patient outcomes and reduce health care utiliza- tion in patients with cancer. But we know little on how to deliver palliative care to the large and growing population of older patients with multiple chronic conditions. Palliative care clinicians are a scarce resource, so care must be targeted to the subset of patients who would benefit most: those at highest risk of near-term death. This is a major challenge outside of specific diseases with known trajectories. Clinicians struggle with prognosis, and current statistical models perform poorly.
Aims We will use novel predictive modeling methods (`machine learning') to identify complex old- er patients at high risk of one-year mortality, drawing on our team's prior work in data analytics and machine learning. We will apply these methods to a diverse population of older patients with multi- ple chronic conditions, in a large academic primary care network. Building on our team's track rec- ord of successful clinical trials, we will conduct a randomized controlled trial of palliative care inte- grated with primary care, targeting older patients at the highest predicted risk of death. We will as- sess impact on a range of measurable patient-reported outcomes and health care utilization. Study design We will develop a model to predict one-year mortality in primary care patients over 65, using a rich set of variables from electronic health records. Our preliminary data indicate that machine learning models are highly accurate for predicting mortality out-of-sample, i.e., in patients the model has never seen. We will identify patients at the highest risk of death?who would benefit most from scarce palliative care resources?and approach them to participate in a randomized trial, comparing usual primary care to primary care integrated with palliative care. The intervention, a series of home-based visits by palliative care clinicians, will build a longitudinal relationship with the patient and primary care team. This strategy is designed specifically to meet the needs of older pa- tients, as well as busy primary care clinicians. We will power the study to detect changes in two primary outcomes: quality of life and care intensity, measured by hospital and emergency visits. Other outcomes include symptom burden, advanced care planning, hospice use, and mortality. Implications This project will generate the first evidence on a new model of palliative care for older adults with multiple chronic illnesses, delivered `upstream' in the disease trajectory. We will build the technical and clinical infrastructure needed to target palliative care interventions for older adults outside of specific disease-based programs. A successful trial would facilitate broader adop- tion of similar interventions for older adults, and fundamentally transform the scale and scope of palliative care efforts in this population.

Public Health Relevance

Despite major advances in palliative care for patients with specific diseases, we know little about how to deliver palliative interventions `upstream'?earlier in the disease trajectory for older adults with multiple chronic conditions. We will apply advanced predictive modeling techniques (`machine learning') to identify older patients in a primary care setting who would benefit most from palliative care: those whose complex interplay of chronic conditions puts them at high risk of near-term death. Building on our team's strong infrastructure for clinical trials in palliative care, we will enroll the highest-risk patients in a randomized controlled trial, comparing usual primary care to primary care integrated with palliative care.

National Institute of Health (NIH)
National Institute on Aging (NIA)
High Priority, Short Term Project Award (R56)
Project #
Application #
Study Section
Health Services Organization and Delivery Study Section (HSOD)
Program Officer
Bhattacharyya, Partha
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Brigham and Women's Hospital
United States
Zip Code
Einav, Liran; Finkelstein, Amy; Mullainathan, Sendhil et al. (2018) Predictive modeling of U.S. health care spending in late life. Science 360:1462-1465