Everyone is faced with the prospect of death, though not knowing when the time will come can be unsettling for many. Now, researchers have uncovered a disturbing benefit of artificial intelligence – the ability to more accurately predict a person's demise.
The unsettling research, published in PLOS One, is the result of researchers training algorithms to look at a decade's worth of health data from just over 500,000 people in the U.K. between the ages of 40 and 69 and assess their chances for dying prematurely. During the 2006-2010 time period (which was followed up until 2016), nearly 14,500 people died, largely from cancer and other diseases.
"We mapped the resulting predictions to mortality data from the cohort, using Office of National Statistics death records, the UK cancer registry and 'hospital episodes' statistics," the study's lead author, University of Nottingham assistant professor of Epidemiology and Data Science Dr. Stephen Weng, said in a statement. "We found machine-learned algorithms were significantly more accurate in predicting death than the standard prediction models developed by a human expert."
To aid in the effort to understand premature mortality, Weng and his team of researchers looked at two different types of AI. The first is known as 'deep-learning,' in which information-centric networks help a computer learn from past instances. The second is 'random forest,' which Live Science reports is able to "combines multiple, tree-like models to consider possible outcomes." They took the conclusions from the two different AI models, compared them to the more commonly used 'Cox regression' (which is based on age and gender) and found the returns of the AI models were far superior.
The deep learning algorithm correctly predicted 76 percent of the subjects who died, while the random forest algorithm clocked in at 64 percent. The Cox model only correctly identified 44 percent of those who had died.
The different algorithms used several varying factors, such as body fat, blood pressure and food consumption (random forest), as well as job-related hazards, air pollution and alcohol intake (machine learning). The Cox model, which largely relied on inputs such as ethnicity and physical activity, ultimately proved to be inferior.
"We have taken a major step forward in this field by developing a unique and holistic approach to predicting a person's risk of premature death by machine-learning," Dr. Weng added in the statement. "This uses computers to build new risk prediction models that take into account a wide range of demographic, biometric, clinical and lifestyle factors for each individual assessed, even their dietary consumption of fruit, vegetables and meat per day.
While it may be a bit unsettling to think that a computer knows when your time is up, preventative healthcare is only likely to grow in popularity, as more people want to know what they change about their lifestyle in an effort to live longer and better lives.
"There is currently intense interest in the potential to use 'AI' or 'machine-learning' to better predict health outcomes," said University of Nottingham professor Joe Kai, who also worked on the study, in a statement. "In some situations we may find it helps, in others it may not. In this particular case, we have shown that with careful tuning, these algorithms can usefully improve prediction."
Weng echoed Kai's sentiments, adding that the work has been going on for years to aid humanity.
"Preventative healthcare is a growing priority in the fight against serious diseases so we have been working for a number of years to improve the accuracy of computerized health risk assessment in the general population," Weng added. "Most applications focus on a single disease area but predicting death due to several different disease outcomes is highly complex, especially given environmental and individual factors that may affect them."
In addition to longer life spans, there are also economic benefits to preventative healthcare. The Surgeon General has released a white paper that states a 1 percent reduction in "weight, blood pressure, glucose, and cholesterol risk factors would save $83 to $103 annually in medical costs per person," while a 5 percent reduction in hypertension would save $25 billion over 5 years.
"Research from the Milken Institute suggests that a modest reduction in avoidable risk factors could lead to a gain of more than $1 trillion annually in labor supply and efficiency by 2023," the white paper added.