A New AI System Could Predict When People Are Likely to Die

They come suddenly, without warning, and the question pulsing in your mind "why now?" melts into the grim realization that this could mean the end of your life. Until now.

A new machine-learning system of artificial intelligence (AI) can successfully predict the risk of cardiac arrest — heart attacks — using timing and weather data, according to a new study published in the journal Heart.

However, it's crucially important to note that the new machine learning application doesn't predict when people will die of cardiac arrest. It merely predicts when the risk of a heart attack may rise.

Machine learning AI was trained and tested locally

Machine learning consists of the study of computer algorithms, and grounds itself in the idea that systems can learn and self-improve from data by identifying patterns and adapting with little-to-no human intervention. The study found that the out-of-hospital risk of cardiac arrest was highest on Sundays, Mondays, during sharp drops in temperature during or between days, and on public holidays.

The new findings could serve as an early warning system, lowering the risk of fatal episodes and raising survival odds, in addition to improving emergency medical services' ability to prepare for serious situations, according to the researchers. This is significant because out-of-hospital cardiac arrests are common globally, and are generally linked to low survival rates. Risk is further complicated with weather conditions. Meteorological data is highly complex, but machine learning could eventually find associations conventional or one-dimensional statistical approaches can't, said the Japanese researchers.

To deepen the research, the scientists evaluated the capacity of machine learning to anticipate daily out-of-hospital cardiac arrests via timing, as in the year, season, day of the week, hour of day, or public holidays, and daily weather, like relative humidity, snowfall, rainfall, temperature, wind speed, cloud cover, and atmospheric pressure readings. Between 2005 and 2013, 1,299,784 cases happened, and machine learning was implemented for 525,374 using timing data, weather, or both for a training dataset. These results were then contrasted with 135,678 cases that happened between 2014 and 2015, to examine the model's capacity for accuracy for anticipating the statistical number of daily cardiac arrests in other years.

New AI application combines weather and timing data to predict a high risk of out-of-hospital cardiac arrests

To reveal the accuracy at the local level of this approach, the researchers executed what's called a "heatmap analysis," which used a separate dataset drawn from out-of-hospital cardiac arrests in Kobe city between Jan. 2016 and Dec. 2018. Combining timing and weather data showed a high accuracy of cardiac arrest predictions for out-of-hospital cases, for both testing and training datasets. Specifically, combining weather and timing data yielded the "hotspots" of cardiac arrests of Sundays, Mondays, low temperatures, sharp drops in temperatures, winter, and public holidays.

However, the researchers don't claim to have extensive information about the location of heart attacks outside of Kobe city, and lack data for those with pre-existing medical conditions. Both exceptional cases may have modified their results. "Our predictive model for daily incidence of [out of hospital cardiac arrest] is widely generalizable for the general population in developed countries, because this study had a large sample size and used comprehensive meteorological data," said the researchers in an embargoed release shared with Interesting Engineering.

"The methods developed in this study serve as an example of a new model for predictive analytics that could be applied to other clinical outcomes of interest related to life-threatening acute cardiovascular disease," added the researchers. "This predictive model may be useful for preventing [out of hospital cardiac arrest] and improving the prognosis of patients [...] via a warning system for citizens and [emergency medical services] on high-risk days in the future."

It's critically important to note that this study does not promise a way to predict when people will really die from cardiac arrest. As a leading cause of death, the new machine learning AI merely predicts when the risk of having a heart attack is high. With this in mind, the potential medical, logistical, and personal implications for improving the human condition are vast.