×

Browse

Can a Machine Learning Algorithm Accurately Predict Cardiac Arrest?

OUT-OF-HOSPITAL cardiac arrest (OHCA) is a substantial global health burden, with extremely low rates of survival. Therefore, accurately predicting the daily incidence of OHCA may provide a significant public benefit. Since the risk of OHCA is affected by prevailing weather conditions, the application of high-resolution meteorological and chronological datasets to medicine might provide ways to improve prediction of the daily incidence. 

For this reason, a team of Japanese researchers conducted a population-based study to assess the ability of a machine learning model to predict OHCA incidence using daily weather (e.g., temperature, relative humidity, rainfall, snowfall, cloud cover, wind speed, and atmospheric pressure readings) and timing (e.g., year, season, day of the week, hour of the day, and public holidays) data.  

Of the 1,299,784 OHCA cases occurring between 2005 and 2013, machine learning was initially applied to a training dataset (using either timing or weather, or both) of 525,374 cases. A dataset for 2014–2015, which comprised 135,678 cases, was then used to test the predictive model. 

Overall, the machine learning model with combined meteorological and chronological variables was found to have the highest predictive accuracy in both the training and testing datasets. It forecasted that Sundays, Mondays, public holidays, winter, low temperatures, and sharp temperature drops within and between days were more strongly associated with OHCA than either the weather or timing data alone.  

The academics acknowledged several study limitations. These included a lack of data on pre-existing medical conditions and an absence of detailed information concerning the location of cardiac arrests (with the exception of Kobe city). Despite this, the authors stated: “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.”  

Going forward, this predictive model may prove useful for preventing OHCA and improving the prognosis of patients. This could be achieved through the development of an early warning system for both citizens and emergency services on high-risk days.