Recent advancements in AI are taking health predictions to a whole new level, and it all starts with the Apple Watch. A new study reveals an AI model trained on data from the wearable device that can predict health conditions more accurately than traditional sensor-based methods. 
The research, titled ‘Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions,’ introduces a groundbreaking way to analyze behavior and detect health issues before they become serious.
Earlier AI models focused on real-time sensor data like heart rate or blood oxygen levels. This new model, however, takes a more comprehensive approach by analyzing how you move, sleep, and exercise over time. At the core of the study is the Wearable Behavior Model (WBM), which analyzes key behavioral metrics like step count, sleep duration, heart rate variability, and mobility-all calculated by the Apple Watch using on-device algorithms.
According to the researchers, this new approach allows the AI model to predict health conditions more accurately than relying solely on biometric data. The WBM is particularly good at identifying ‘static health states’ (e.g., someone taking beta blockers) or more transient conditions, such as sleep quality or respiratory infections.
One of the most impressive outcomes of the model was its ability to detect pregnancy with 92% accuracy, when combined with traditional biometric data in a hybrid approach. The model was trained on over 2.5 billion hours of data collected from more than 160,000 participants in the Heart and Movement study, and tested on 57 different health prediction tasks. This vast dataset enabled the AI to identify behavioral changes over time, such as those that occur over days or weeks, rather than sudden health issues.
The study suggests that smartwatches, such as the Apple Watch, have evolved significantly and could be key players in the future of AI-powered health analysis. While it’s unclear if this model will be integrated into future Apple Watches, it shows that current Apple Watch hardware has a lot of untapped potential for health monitoring and prediction.