Predictive analytics systems are gaining ground in the hospice and palliative care fields as a means to identify patients in need of service earlier in the course of their illness. Providers and payers are increasingly using these systems to build their census and ensure that patients and families receive the right care at the right time.
Predictive analytics systems analyze statistical patient data, typically gleaned from electronic health records, that can predict changes in a patient’s condition or a particular outcome.
“This is generally done through the application of machine learning techniques, which create inferences from data,” said Charles Saunders, CEO of health care technology firm Integra Connect. “They take advantage of empirical correlations in the data that are observed. We use it in oncology to predict adverse events or other outcomes so that they can be aggressively managed, or to try and reduce the impact or occurrence of those events, and in particular which patients are likely to die within a particular period.”
Organizations using these systems are seeing results. Integra Connect in 2016 began working with the multi-specialty physician group Michigan Health Professionals (MHP) to apply the company’s predictive analytics system to MHP’s oncology program. MHP, which has been working to build a palliative care and hospice program, saw a 700% increase in hospice utilization among their patients after implementing the Integra’s predictive analytics system.
MHP also saw a 70% drop in the number of hospital admissions as well as a 16% reduction in outpatient and emergency department costs, particularly during the last 30 days of life, according to Saunders.
“The last 30 days of life is very high-cost period in one’s life journey. Patients can die in the hospital or stay in the [intensive care unit] and receive chemotherapy up until the end, or they can go down a different path with palliative and hospice care that consumes far less resources and may in fact be a better experience for the patient,” Sauders told Hospice News. “Patients in this in this stage in their life require a lot of thought and sometimes a lot of care management and support. There aren’t enough resources to provide the same level of service to every single patient, so you need to identify proactively which patients are most likely to benefit from treatments like chemotherapy or be candidates for palliative and hospice care. That means having some level of predictive ability.”
Use of these systems is growing throughout the health care system. Researchers at Penn Medicine, a component of the University of Pennsylvania, have developed a predictive analytics system that increased the organization’s number of palliative care consultations by 74%.
The system, called Palliative Connect, extracts clinical data from an organization’s electronic medical records and uses machine learning to analyze 30 factors of a patient’s likely prognosis during a six month period, which is the time frame physicians use to determine whether a patient is ready for a palliative care consultation. Based on these data, the system assigns the patient a score that indicates their suitability for palliative care, according to a recent study in the Journal of General Internal Medicine.
Fort Worth, Texas-based Community Hospice of Texas and Providence Hospice is collaborating with technology firm Turn-Key Health, a subsidiary of Enclara Healthcare, to form a new community-based palliative care program to serve members of Medicare Advantage plans.
The partnership will use predictive analytics to help identify health plan members who are eligible for palliative care early in their disease trajectory, as well as providing support from nurses and social workers who can assist in guiding care plans, health plan member engagement and care management.
“By using real world evidence and clinical data as well as claims and financial indicators, one can, in the future, predict the likelihood of success and best therapeutic options for a patient in a given circumstance based on their unique characteristics,” Saunders told Hospice News. “I think predictive analytics has a role to play in making those determinations, based on the myriad of complex of variables that a patient might present with. In developing our own algorithms, we look at about 800 different variables and then select a short list of those that have the greatest predictive power. I think it’s going to be something we’ll see a lot of in the future.”