Apple Watch can detect stress

Apple Watch devices are already capable of predicting stressful situations in people. This is because the Series 4, 5, Series 6, Series 7, Series 8 and Ultra models can generate an electrocardiogram (ECG).Apple's ECG sensor is similar to a single-channel ECG. Via software, users can record a 30-second ECG, the data of which can also be exported.In addition to the ECG, Apple also offers other health sensors such as pulse measurement and blood oxygen measurement, but so far no software function that allows conclusions to be drawn about the wearer's stress. Competitors like Garmin or Fitbit already offer the possibility to evaluate the user's stress level on some of their devices. The function could therefore be added to the Apple Watch via a software update. This is the conclusion reached by researchers at Canada's University of Waterloo following a pilot study with 36 participants. They used machine learning (ML) to evaluate stress-typical signs based on ECG data from the Apple Watch, using a self-developed app based on the HealthKit developer interface (API) that study participants used to record an ECG every 3 hours for 2 weeks. The data was then used to train a machine learning model.

Of interest to health authorities

The pilot study sees the benefit of such a stress function using machine learning (ML) not only in the user of the watch. In particular, health authorities could gain anonymized, important insights into how the population is faring in terms of stress. In this way, appropriate initiatives and measures can be taken, as high stress levels are directly linked to health problems such as cardiovascular disease, depression or obesity. In Canada, one in five citizens complains of high stress levels. Digital watches are widespread, socially accepted and therefore already have the advantage of being perfectly suited to collect such data in real time.

Models with high specificity

The current Apple Watch models have "high specificity," he said, and researchers can use various characteristics, such as heart acceleration or heartbeat delays, to determine when a stressful situation is occurring. Here, the detection of conditions in which no stress is present proves to be more accurate than when a high stress level is actually present. Stress detection ranged from 52 to 64 percent. The stress-free situations were correctly detected by more than 60%.

A longer duration of ECG recordings could further improve the measurement results. Combining other data, such as sleep tracking, could also further improve detection, the researchers say. A wearable device capable of continuous stress monitoring in real time allows users to react early to changes in their mental health.