ANN ARBOR, Mich. — Cardiac arrest is a sudden stoppage of the heart, rather than a heart attack or heart failure that can play out over several hours or even days. Approximately 356,000 people experience cardiac arrest in the U.S. each year, of which 90% are fatal. Scientists are now working on a smartphone app that could actually predict cardiac arrest rates — by using weather forecasts and calendar dates.
They have developed an artificial intelligence system, or neural network, that combines both, with more than 90% accuracy for anticipating the number of cases at a district level. According to this system, cardiac arrests are most likely to happen during sharp drops in temperature. Vulnerable individuals should also be wary of Sundays, Mondays, and public holidays.
“We will provide information on high-risk days for vulnerable individuals by smartphone in the future. It could prevent thousands of deaths,” explains first author Dr. Takahiro Nakashima of the University of Michigan, per South West News Service.
This app opens the door to “cardiovascular emergency warnings” on devices linked to the internet. According to Dr. Nakashima, it could also shed light on how changing behavior – such as keeping warm when it’s cold or cooling off when it’s hot – can prevent cardiac arrests. Older people would be reminded to limit time spent outdoors, dress appropriately, not over-exert themselves, take medications, and stay away from booze. Ambulance and hospital staff would also be better prepared – just like motorists are alerted to hazardous driving conditions.
“Our study is the first to predict incidence based on both meteorological and chronological variables using machine learning,” the study explains. The “machine learning” model is based on high-resolution meteorological daily forecasts. “It could use advanced analytics to build a warning system for individuals potentially at risk for cardiac arrest through internet of things (IoT) devices.”
App shows stunning accuracy in forecasting cardiac arrest risk
Tests showed the revolutionary technique was remarkably accurate. Specifically, increases were linked to larger differences and ranges in the average temperature from the previous day – and within a day – respectively.
“We speculated a sudden change in ambient temperature on days with extreme cold or heat plays a key role,” the authors write. Studies have shown that decreasing temperatures reduce blood flow by causing the vessels to constrict, thereby triggering clots. The number of heart attacks are known to rise in winter because of this phenomenon.
Sundays and holidays were also more strongly associated with incidence – as well as Mondays. This may be due to the urge for people to binge drink when they are off – dubbed “holiday heart syndrome.” Also, returning to work after a weekend off can be stressful.
The computer “brain” was built using daily weather data in Japan between 2005 and 2013. It included temperature, relative humidity, rainfall, snowfall, cloud cover, wind speed and atmospheric pressure readings. Information on the year, season, day of the week, hour of the day and public holidays were also added. It was then applied to more than 525,000 cardiac arrest cases occurring over the period – using either weather or timing or both. It was also compared with over 135,000 cases in 2014 and 2015 to test its ability in other years.
The researchers also carried out a “heat map analysis” at the local level using another dataset drawn from out-of-hospital heart attacks in Kobe city. They used a technique called MAPE (mean absolute percentage error) – used to measure the ability of a forecasting method.
“Our data shows the accuracy is more than 90%,” says Dr. Nakashima. In one experiment, it predicted 24 out of 27 heart attacks that occurred during a study week. It identified the exact number in 4 of 7 districts – and was off by only 1 in each of the three others.
“Our predictive model for daily incidence is widely generalizable for the general population in developed countries. This study had a large sample size and used comprehensive meteorological data. The methods could be applied to other clinical outcomes of interest related to life-threatening acute cardiovascular disease,” the authors note.
“A machine learning predictive model using comprehensive daily meteorological and chronological data allows for highly precise estimates of incidence. It may be useful for prevention and improving the prognosis of patients via a warning system for citizens and emergency medical services on high-risk days in the future,” the study concludes.
“Knowing what the weather will most likely be in the coming week can generate ‘cardiovascular emergency warnings’ for people at risk. These predictions can be used for resource deployment, scheduling and planning so emergency medical services systems, resuscitation resources and heart laboratory staff are aware of – and prepared for – the number of expected cases during the coming days,” writes Dr. David Foster Gaieski, of Sidney Kimmel Medical College at Thomas Jefferson University, in an editorial article about the technology.
Findings are published in the British journal Heart.
SWNS writer Mark Waghorn contributed to this report.