The Stanford Machine Learning Algorithm that can tell you when you'll Die with 90% accuracy
The AI that can tell you when you'll DIE: Stanford
reveals 'startlingly accurate' system to predict the end of life for hospital
patients
Stanford researchers trained a deep neural network on 2
million hospital records
With this, the algorithm learned to accurately predict
patients' mortality
Researchers say this can be used to help pre-screen
patients for end-of-life care
By Cheyenne Macdonald PUBLISHED: 15:18 EST, 18 January
2018 | UPDATED: 16:09 EST, 18 January 2018
Stanford researchers have developed an AI that can
predict when a patient will die with up to 90 percent accuracy.
While the idea might sound unnerving, the team behind the
work says it could vastly improve end-of-life care for patients and their
families.
By more accurately pinpointing when a terminal or
seriously ill patient may pass, caregivers can prioritize their wishes and
ensure important conversations are held before it’s too late.
The team trained a deep neural network on Electronic
Health Record data from the Stanford Hospital and Lucile Packard Children’s
hospital, encompassing roughly 2 million adult and child patients, according
toIEEE Spectrum.
Instead, it could be used in conjunction with assessments
by the human doctor to make proactive decisions in pre-screening patients for
end-of-life planning.
While the process may be helpful, there is one challenge
– based on the ‘black box’ nature of the algorithm, the researchers don’t know
exactly what its predictions are based on.
In this type of application, however, they say knowing
why it made the predictions isn’t necessarily important.
As the care is not tied to the reason a person is sick,
'in this setting, it doesn’t matter as much as long as we get it right,' research
scientist Kenneth Jung told IEEE.
In the new study published pre-print to arXiv, the
Stanford team explains that there is often a huge discrepancy between the way a
patient wants to live out the rest of their life, and how it actually happens.
According to the researchers, roughly 80 percent of
Americans wish to spend their final days at home – but, as many as 60 percent
end up dying in the hospital.
In effort to close the gap, the team at Stanford
University trained a deep neural network on Electronic Health Record data from
the Stanford Hospital and Lucile Packard Children’s hospital, encompassing
roughly 2 million adult and child patients, according to IEEE Spectrum.
‘We could build a predictive model using routinely
collected operational data in the healthcare setting, as opposed to a carefully
designed experimental study,’ Anand Avati, a PhD candidate in computer science
at the AI Lab of Stanford University, told IEEE.
‘The scale of data available allowed us to build an
all-cause mortality prediction model, instead of being disease or demographic
specific.’
The tool isn’t designed to work by itself to guide the
care process.
Instead, it could be used in conjunction with assessments
by the human doctor to make proactive decisions in pre-screening patients for
end-of-life planning.
As the researchers explain, it isn’t always easy to
understand who needs palliative care and when.
‘The criteria for deciding which patients benefit from
palliative care can be hard to state explicitly,’ the authors explain in the
paper.
‘Our approach uses deep learning to screen patients
admitted to the hospital to identify those who are most likely to have
palliative care needs.
‘The algorithm addresses a proxy problem – to predict the
mortality of a given patient within the next 12 months – and use that
prediction for making recommendations for palliative care referral.’
While the process may be helpful, there is one challenge
– based on the ‘black box’ nature of the algorithm, the researchers don’t know
exactly what its predictions are based on.
In this type of application, however, they say knowing
why it made the predictions isn’t necessarily important.
‘The palliative care intervention is not tied to why
somebody is getting sick,’ research scientist Kenneth Jung told IEEE.
‘If it was a different hypothetical case of “somebody is
going to die and we need to pick treatment options,” in that case we do want to
understand the causes because of the treatment.
‘But in this setting, it doesn’t matter as much as long
as we get it right.’
Comments
Post a Comment