Stanford University's "death calculator" may now be able to answer the question, "When will I die?" with accuracy.

A new AI deep learning system could provide groundbreaking development in Palliative Care.  Researchers at Standford University have tested a new artificial intelligence algorithm, which can help hospitals improve palliative health care delivery for cancer patients and patients with terminal illnesses.

How The 'Death Calculator' Works

The algorithm based on a Deep Neural Network learning machine can analyze important medical records or Electronic Health Records of admitted terminally ill patients to know if they are likely to benefit from end-of-life care or Palliative Care.

The algorithm can predict the mortality of patients from three to 12 months and use that prediction to refer patients for palliative care.

"Our predictions can enable the palliative care team (of hospitals) to take a proactive approach in reaching out to such patients, rather than relying on referrals from treating physicians, or conduct time-consuming chart reviews of all patients," the study indicates.

Previous studies showed that approximately 80 percent of Americans would like to spend their final days at home but only 20 percent are able to do so and more than half of patient deaths occur in acute care hospitals.

In fact, terminally ill patients often receive aggressive medical care in their dying days instead of having their end-of-life wishes carried out.

Improving Palliative Care

It is that the capability of hospitals to provide palliative care have improved in recent years. The study says, however, that only 7 to 8 percent of patients actually receive it.

Factors like lack of palliative care professionals analyzing each of all the patient's data, the common overoptimism of doctors on a patient's prognosis, and life expectancy contribute to the said problem. This is where the Deep Learning AI algorithm comes in.

"We could build a predictive model using routinely collected operational data in the healthcare setting, as opposed to a carefully designed experimental study," says Anand Avati, of the Stanford University Computer Science Department. "The scale of data available allowed us to build an all-cause mortality prediction model, instead of being disease or demographic specific."

Deep Learning Technique

The learning technique used referred to as deep learning algorithm utilizes neural networks to filter and analyze a large volume of data.

The research approached the prediction of mortality by being agnostic to the disease type, patient age, the severity of admission among others.  It used the EHR data of the patients in the prior year from first the contact to determine their mortality within 12 months.

For the particular study, the researchers analyzed two million records of adult and child patients who were admitted to the Standford Hospital and Lucile Packard Children's Hospital. They identified 200,000 patients for the study. The EHR of the said patients were analyzed by the system to predict their mortality.

For the pilot study, the algorithm was tasked to predict the mortality of each of the 160,000 patients within 12 months from a given date. The study was able to train the system to predict patient mortality within the next three to 12 months.

Afterward, the algorithm assessed the data of the remaining 40,000 patients and was able to accurately predict mortality within the 3 to 12 months timespan in nine out of 10 cases.

Majority of patients assessed with a low probability of dying within the time span lived beyond a year.

"We want to make sure the sickest patients and their families get a chance to talk about what they want to happen before they become critically ill and they end up in the ICU," says Ken Jung, medical research scientist and co-author of the study.

Findings of the study "Improving Palliative Care with Deep Learning" is published in the preprint server Arxiv.

ⓒ 2021 All rights reserved. Do not reproduce without permission.