Towards a General Intelligence and Interface for Wearable Health Data

TL;DR

This paper introduces a large-scale foundation model trained on extensive wearable sensor data to improve personalized health insights, enabling efficient few-shot learning and adaptable health prediction tasks.

Contribution

The work presents a novel foundation model trained on over a trillion minutes of unlabeled data, enhancing health prediction across diverse tasks and enabling autonomous downstream model optimization.

Findings

Model trained on over one trillion minutes of data shows systematic performance improvements.

Population-scale representations enable label-efficient few-shot learning and generative capabilities.

Integration into a Personal Health Agent improves relevance, context-awareness, and safety of health predictions.

Abstract

While ubiquitous wearable sensors capture a wealth of behavioral and physiological information, effectively transforming these signals into personalized health insights is challenging. Specifically, converting low-level sensor data into representations capable of characterizing higher-level states is difficult due to high phenotypic diversity and variation in individual baseline health, physiology, and lifestyle factors. Moreover, collecting wearable data paired with health outcome annotations is laborious and expensive, and retrospective annotation remains practically unfeasible, contributing to a scarcity of data with high-quality labels. To overcome these limitations, we propose a foundation model for wearable health that is pretrained on more than one trillion minutes of unlabeled sensor signals drawn from a large cohort of five million participants. We demonstrate that the joint scaling of model capacity and pretraining data volume leads to systematic improvements in performance, as evaluated on a diverse set of 35 health prediction tasks, spanning cardiovascular, metabolic, sleep, and mental health, as well as lifestyle choices and demographic factors. We find that this population scale representation unlocks label-efficient few-shot learning and generative capabilities for robust daily metric estimation. To further leverage this learned representation, we deploy a classroom of LLM agents to autonomously search the space of downstream predictive heads built on the model embeddings, showing broad performance improvements that increase with LLM model capacity. Finally, we show how integrating these downstream predictors into a Personal Health Agent can support model responses that are more relevant, contextually aware, and safe, and we validate this via 1,860 ratings from a cohort of clinicians.