Can smartphones predict your risk of dying?

Elizabeth Barber discusses new research of how smartphones can predict our risk of mortality through sensor data and how the science behind this works

It’s well established that increased daily physical activity reduces mortality, and previous studies have been able to affirm this link through using smartwatches and fitness trackers. While smartphones have similar sensors, people are less likely to always carry their phones, meaning that there are fewer data points available. 

In a journal article that was published in October of this year, researchers collected data from 100,000 participants aged 45-79 who wore motion sensors for a week, with a sample that was demographically representative of the UK population. The researchers ran motion sensor and death data on about one-tenth of participants through a machine learning model. They were able to develop an algorithm with a machine learning model that could estimate five-year mortality risk using acceleration during a six minute walk. 

They were able to develop an algorigthm with a machine learning model that could estimate five-year mortality risk using acceleration during a six minute walk

It was highlighted that cheaper phones are also effective for this measure, increasing the inclusivity of the study, with the researchers positing that similarly large trials with diverse populations and affordable phones are both technically feasible and socially desirable. They also suggest that higher predictive accuracy might be achieved for older patients, especially those with cardiopulmonary diseases. 

They also suggest that higher predictive accuracy might be achieved for older patients, especially those with cardiopulmonary diseases

Lead author Bruce Schatz highlighted that from this research, a highly accurate and scalable screening method could be created that could be used nationally to screen for health risks. He anticipated that this research will be most useful in responding to cardiovascular disease, as it is often difficult to detect without visiting a doctor, and may help identify problems in those who were unaware that they had the condition. Going forward, researchers plan to conduct large population trials using data from cardiopulmonary patients’ smartphones, to investigate whether this will produce more accurate predictive models.