Pallialytics, a machine learning model to determine palliative-care eligible patients
Palliative care is specialized medical care that is focused on providing seriously ill patients with relief from symptoms, pain and stress. Palliative care can be provided at any time during an illness with curative or supportive therapies, and it is often the last care the patient receives. In existing practice, the means of identifying patients with palliative care needs are highly inefficient. First, physicians may not refer patients who are likely to benefit from palliative care for multiple reasons such as over-optimism, time pressures, or treatment inertia. This may lead to patients failing to have their wishes carried out at end of life and overuse of aggressive care. Second, a shortage of palliative care professionals makes proactive identification of candidate patients via manual chart review an expensive and time-consuming process. Therefore, there is a need for a technology that would play a crucial role by efficiently identifying patients who may benefit most from palliative care but might otherwise be overlooked under current care models.
Duke Connected Care (DCC) partnered with researchers at Duke Institute for Health Innovation to design a novel machine-learning predictive model that combines medical claims and clinical data to effectively target DCC’s existing palliative care interventions among the Medicare Shared Savings Plan (MSSP) population. Their Pallialytics model uses the CMS Medicare data such as Durable Medical Equipment charges, diagnoses, medications, and information from past claims to predict a patient's mortality risk. In parallel, Duke researchers have developed a regression model to support and validate the novel model’s predictions. Using the dual-model approach, they are able to predict up to four outcomes per patient over a 12-month time horizon: mortality, hospitalization, high Medicare costs, increasing rate of costs.
- More than doubling of success rate for accurately identifying candidate patient compared to conventional methods (improves from 20% to 45%).
- Will correctly rank two random patients to benefit from palliative care 81% of the time.
- Saves time and resources in healthcare system