Predicting life expectancy using machine learning

 

Predicting life expectancy using machine learning

Predicting life expectancy using machine learning, But measuring mortality for people with disabilities can prove fatal.

Predicting life future

Everyone wants to know what's to come—right? But our obsession with predictions reveals much more about us.

In Neolithic China, seers practiced pyro-ostomy, or the reading of bones; The ancient Greeks predicted the future by the flight of birds; The people of Mesopotamia also tried to read fortunes in the indigestible entrails of dead animals. We've observed the motion of the stars and planets, we've observed weather patterns, and we've even observed physical predictions such as the "baby born with a cowl" superstition to ensure future good fortune and long life. By the 1700s, the art of prediction had become a bit more scientific, 

with mathematician and probability expert Abraham de Moivre attempting to calculate his death by equation, but truly accurate predictions remained out of reach.

Then, in June 2023, De Moivre's fondest wish appeared to come

true: Scientists discovered the first reliable measurement to determine the length of your life. Using a dataset of 5,000 protein measurements from nearly 23,000 Icelanders, researchers working for DECODE Genetics in Reykjavik, Iceland developed a predictor for the time of death—or, as stated in their press release, "

How much is left in one person's life." It's an unusual
claim, and it comes with particular questions about methodology, ethics, and
what we mean by life.

 

For most people, most of the time, death remains a nebulous thought, haunting the shadowy corners of our minds. But knowing when our life ends, having a sense of the days and hours remaining, removes the comforting shield of abstraction. It also prompts us to look at risk differently; For example, we are more likely to try unproven treatments in an attempt to overcome the odds. If the prediction comes early enough, most of us may even try to prevent the likely event or avoid the outcome. 

Science fiction often confronts us with that possibility; Movies like Minority Report, Thrill Seekers, and the Terminator franchise use advanced knowledge of the future to alter the past, averting death and mayhem (or not) before it happens. Indeed, when healthy and able-bodied people think about predicting death, they think of these sci-fi possibilities – futures where death and disease are eradicated before they even begin. 

But for people with disabilities like me, the technology of predicting death serves as a reminder that we are already considered better off dead. Predicting the length of life is a science to judge its value: longer life equals a better or more meaningful life. It is not difficult to see the ramifications of a technocratic authority on the most vulnerable.

 

This summer's discovery was the work of researchers Kari Stefansson and Tjodbjörg Erikdottir, who found that different proteins in our DNA are related to overall mortality—and that different causes of death still have similar "protein profiles." Eiriksdottir claims they can measure these profiles in a single draw of blood, a kind of clockwork for the time remaining in the plasma. Scientists call these mortality tracking indicators biomarkers, and there are up to 106 of them that help predicts all-cause (instead of specific to disease) mortality.

 But for Stefansson, Eriksdottir, and their research team, success measures up. The process they developed is called the summer-based multiplex proteomic assay, and it means the group can measure thousands and thousands of proteins at once.

Not all of these measurements result in an accurate date and time. Instead, it provides medical professionals with the ability to accurately predict the top percent of patients most likely to die (at the highest risk, about 5 percent of the total) and also the top percent least likely to die (at the lowest risk). ), just a prick of a needle and a small vial of blood. It may not look like a crystal ball, but it is clear that it is only a jumping-off point.

The deCODE researchers plan to improve the process to make it more "useful," and the effort joins other earlier projects in death-predicting technology, including an artificial intelligence algorithm for palliative care. Is. The creators of this algorithm hope to use the "cold calculus of AI" to nudge physicians' decisions and force loved ones to have the dreaded conversation—because "I'm dying"

In their press release, deCODE researchers predict biomarkers for a large portion of the population

 

According to Stephenson's Clinical Trials, "You can easily compare large groups in a standardized way." But a standardized treatment is not something that applies well to the deeply diverse needs of individual patients. What happens when such technology—complemented by AI algorithms—leaves the research lab and is put into use in real-world situations?

This is the first time death-predicting data has been employed on such a large scale – and it has revealed the deeply troubling limits of “cold calculus”.

Physicists rewrite a quantum law that collides with our universe


In October 2021, a study at the University of Copenhagen demonstrated that a particular protein on the cell surface is likely to predict who is at risk of severe infection caused by the novel coronavirus. Once this protein biomarker was employed,

it determined who would become seriously ill with a 78.7 percent accuracy rate. It looked like good news. We need to know which patients will most need care – and triage, or sorting, has traditionally been used more effectively as a means of saving more lives. Everyone will be taken care of; fewer life-threatening cases may wait longer to see a doctor. But as Covid-

 

19 overwhelmed ICU wards and hospitals ran out of supplies and beds, triage was instead employed to decide who got care and who was turned away.

During the height of the pandemic, in New York guidelines aimed to save the most lives, "as defined by a patient's short-term probability of surviving an acute medical episode." Trying to pin down exactly what it means can be difficult; It can refer to "saving as many lives as possible" or saving "the greatest possible number of years of life", or even more problematically, saving "the greatest amount of quality-adjusted life years". Is. In as many models as possible, this could mean privileging those without the protein which predicts a longer Covid hospital stays.

In models about years of life, especially when subjective measures about the quality are included, people with disabilities or chronic conditions, or even mental health problems, may be excluded. Some US states had emergency protocols stating that "individuals with brain injuries, cognitive disorders or other intellectual disabilities may be poor candidates for ventilator support," while a physician in Oregon cited this as a reason for refusing a ventilator. Cited "quality of life". The research now available for the worst outbreaks has shown how deep the implicit bias towards disabled lives really runs.

As the pandemic continues to escalate, people with disabilities continue to fear being denied the care they need because their amount, quality, or value of life has been measured by someone else. If the standardized predictions envisioned by deCODE are first made with a view to providing care to the able-bodied, then measuring mortality is much more than predicting death; For people with disabilities, it can actually accelerate it.

There are better ways to measure a life than to count it to the end. Disability advocates, many of them persons with disabilities, have long documented systemic biases in our healthcare systems, but the COVID crisis has helped bring some of these issues to the fore. As Matthew Cortlandt, an attorney and senior fellow at Data For Progress, explains,

the automated algorithms proposed by AI or deCODE “could be used to determine who to deny care,” as “they’re going to die anyway.”, We should save money." Similarly, Alyssa Burgart, a physician, bioethicist, and clinical director at Stanford,

describes crisis thinking as a way of considering short lives of little value, such as disabled, chronically ill, or elderly people as less human. Or were worth saving less. The assumptions being made will be with us long after Covid has come and (hopefully) gone; Our thinking needs to change in crisis or people with disabilities will always be a secondary consideration.

The problem is the concept of "long-term survival", focusing on the length of life as a means of assessing value. "Technology that predicts death doesn't have to be bad," explains Bergart, "it all depends on human decisions." The technology isn't as objective or accurate as it seems, she says, but when policymakers get death predictions right, she says, "to give more resources to people who are already doing fine." Taking the risk of making unwise decisions: how can we be sure death predictions are accurate." the most needed resources go to those who can benefit most from them?" We should instead protect the most vulnerable
Should do

Cortlandt suggests that similar data could be used to "increase resources" for people who are at "increased risk relative to short-term mortality." For example, when assessing patients for ventilators,

 

Death itself should not be the focus, nor should it be a solution in itself. questions, they explain,
Should be "What keeps people alive?" It's not just ICU beds and ventilators, it's also resource allocation outside hospitals: a safe place to live, enough to eat, and affordable medicine. Predictive algorithms cannot analyze social inequality; Public health and policymakers cannot allow them to inadvertently enforce social determinants of health through denial of care.
A disabled person, a disadvantaged person, an ethnic minority, an elderly person, a woman, a child, or a refugee all lives matter.

 

Prediction tools will continue to be used and can be used for good, but we at least take responsibility to preserve them. When crises come—and they will come, whether through new forms, entirely new diseases, or through the consequences of climate change—we can build new hospitals, temporary wards, and treatment tents; 

We can bring doctors out of retirement or provide them with temporary emergency treatment licenses (as has happened in Canada). We can exhaust the resources to ensure that all life is treated with equality. Furthermore, policy should put in the foreground those who will be most at risk from death prediction technology and put advocates in charge of policymaking to control and incorporate it. Braggart says that the future is always affected by our decisions and priorities in the present.

Death prediction can be useful for the early detection of disease, but in the end, it will never be able to measure the value of life.