OR WAIT null SECS
Rachael Zimlich is a freelance writer in Cleveland, Ohio. She writes regularly for Contemporary Pediatrics, Managed Healthcare Executive, and Medical Economics.
The benefits and drawbacks of predictive modeling in end-of-life care decision making.
There’s no easy way to discuss end-of-life care. And there’s no easy way for patients and families to make decisions related to it. But there are advances in technology that could help providers frame the discussion to help patients and families better understand their situation and gain confidence in the choice they are making.
“I do believe that we’re really at a point that we’re starting to see data analytics and predictive modeling for individuals, especially as we start to look at population health,” says Michael L. Munger, MD, a family physician in Overland Park, Kansas, and president of the American Academy of Family Physicians. “These tools help you not only with your end-of-life discussion, but they are also going to lead to more and better utilization of palliative care services.”
Providers sometimes come to a point where they see that there aren’t many options left for a patient to experience clinical improvement or any meaningful quality of life, but with the high level of medical intervention that is possible, some patients’ families find it difficult to accept. Other times, providers may want to continue with interventions at the request of families, without a clear picture of how those actions might actually help patients.
In these cases, predictive analytics may prove useful. It can offer providers risk stratification scores based on a patient’s conditions, medications, hospitalizations, and age. That information can then be shared with families and patients. “To me that’s almost like the next frontier,” Munger says. “It allows everyone to really focus on what they want and what is reasonable. It addresses the question of how do you want to live the rest of your life in the best manner and what’s important to you. Hopefully we can shift the conversation from having to do everything possible to one of having great quality of life.”
Predictive modeling can be used even when end-of-life isn’t imminent. For example, in primary care offices providers can discuss end-of-life care plans with patients using data about their age, comorbidities, functional level, and more. “It’s then a lot easier when you’re sitting down with mom and adult children to say, ‘This is what we see based on all of the previous information and studies together,” says Munger.
These discussions, while difficult, can result in patients living their final years with the best quality of life with fewer hospitalizations, he says. “There is peer-reviewed research that shows that if you have that true advanced care planning earlier, it leads to better care and much higher patient satisfaction. Families no longer feel like they are making the decision to withdraw care,” Munger says. “You have to have these discussions sooner. You can’t wait until it’s time; we have to have it ahead of time.”
Some think predictive analytics tools are too new and untested to take the chance of using at such a sensitive time. “I think we are some distance from the use of data analytics in end-of-life discussions,” says Linda Harrington, RN-BC, PhD, an independent consultant on health informatics and digital strategy; professor at Baylor College of Medicine; past chair of the American Association of Critical-Care Nurses (AACN) Certification Corporation national board of directors; and technology department editor for AACN Advanced Critical Care. “Leading healthcare organizations, grappling with the use of data analytics to solve issues, are confronting challenges largely surrounding data quality, analytic tools, and talent to do the work.”
Today, data analytics alone are insufficient to counsel patients and families about chances of recovery or survival, she says. “End-of-life decisions are very individual and complex, requiring data that may not be currently available or held in one database, such as an electronic health record. In addition to the patient’s medical and psychosocial data, an analysis of data in related research, family history, genetics, resource availability, and more can impact survivability. Data analytics that create a holistic view may one day enable better support for patients and families.”
More than 40,000 studies have been published over the last decade on end-of-life care, and 10,000 on data analytics, according to Teresa Rincon, RN, enterprise critical care champion for the EHR design team at UMassMemorial Health Care. Still, she says, fewer than 100 of these studies have specifically investigated the use of analytics in end-of-life care. While EHRs contain an enormous amount of data, she cautions that this data varies in its completeness and detail. She adds that EHRs are used primarily for clinical and financial functions, and may not contain the data elements and formatting necessary to use in data analytics for the purposes of end-of-life care.
Rincon says data analytics also lack the ability to take into account the emotional and physical aspects of death and dying. Harrington agrees, adding that it is an interesting time to discuss technology in end-of-life care because she is seeing a growth in narrative medicine in response to the shortcomings of the digital world.
“The focus is on the experience of illness, in this case end of life, the meaning of which can be lost in hard data stores,” Harrington says. “The lesson here is to balance technology and data analytics with the larger picture of the patient and family experience.”
How one processes end-of-life discussions is also influenced by emotional, psycho-social, and spiritual aspects, Rincon says. Although survival rates and statistics may help some in the decision-making process, the same information may create false hope for others. “There are also those that seem to defy statistics, living much longer than predicted. These outliers may cause caregivers, patients, and families to lose trust in predictive models derived from data analytics,” she says.
Data analytics may provide some use in providing individualized care and interventions in end-of-life care, Harrington says, but research is also conducted in controlled environments and doesn’t often take into account the human element. To truly apply data analytics to sensitive end-of-life care discussions and planning, there has to be a full understanding of the source of the data and how it was compiled, and the unique needs of the patient and their family, she says.
These concerns underscore Munger’s advice that end-of-life discussions happen sooner, and with a provider who has a relationship with the patient and family. “This is where having a relationship with a patient and family really pays off because when I take care of a family, I’ve had other challenging conversations that are delicate,” he says. “If you have someone that’s trusted and that you’ve shared things with before, now we can sit down with good data and statistics and have that relationship because I’ve become a trusted voice.”
Rachael Zimlich, RN, is a writer in Columbia Station, Ohio.