The ways in which technological advancements can help to improve end-of-life care are numerous. One such breakthrough comes from Stanford University, where researchers recently developed a deep neural network algorithm to help doctors decide when to implement palliative care.
The algorithm uses a patient’s medical records to predict their chance of dying within the next three to 12 months. The algorithm also explains its predictions to doctors. This is important because doctors are much more willing to accept an automated system’s recommendations if they know the reasons behind its conclusions.
In a report detailing their findings, the Stanford researchers discuss the fact that access to palliative care has been on the rise in the United States during the past 10 years. According to the researchers, 53 percent of all hospitals with 50 or more beds had palliative care teams in 2008, and that number rose to 67 percent in 2015. The National Palliative Care Registry estimates, however, that less than half of the 7 to 8 percent of hospitalized patients who require palliative care actually receive it.
The researchers cite a palliative care workforce shortage that contributes to this disparity, making delays in much-needed palliative care almost inevitable. But they also believe the new algorithm can help streamline the process of identifying which patients would benefit most from palliative care. This would let doctors and their teams focus on patients who need care the most.
“Our predictions enable the Palliative Care team to take a proactive approach in reaching out to such patients,” the researchers write, “rather than relying on referrals from treating physicians, or conduct time-consuming chart reviews of all patients.” They hope the algorithm will allow palliative care teams to more easily hone in on which patients will benefit most from care.
How It Works
Not surprisingly, a deep neural network algorithm is not the simplest thing to describe. So the report itself can be difficult to understand at times. The algorithmic system analyzes an individual’s personal health records. It then compares it to data from roughly 2 million other patients from the last 20 years, and then predicts when the individual will die.
The researchers mention that it’s important for the doctors implementing this system to feel comfortable taking action based upon its findings. So the system also provides a report describing how it came to its conclusion. This includes details on how certain factors, such as the number of days spent in a hospital, medications prescribed and severity of the diagnosis were part of the prediction.
The researchers make it clear that the algorithm is still in its early stages. However the initial results have been positive, and they believe that the system can be beneficial for both doctors and patients alike. The algorithm is currently being used in a pilot program at a university hospital.