Using wearables to enable cardiovascular-kidney-metabolic (CKM) risk profiling for personalized coaching

Judgment

Google Research’s CKM team must decide whether to take this investigational coachable-health-age framework into prospective clinical validation, trading faster consumer rollout for evidence that 12-week counterfactual plans improve hard outcomes rather than only coach preference.

Basis

Zeinab Esmaeilpour and Ahmed A. Metwally, Research Scientists at Google Research, introduce an investigational ML framework. The coachable health age is grounded in published epidemiological links and integrates resting heart rate, daily step count, sleep duration, and weekly Active Zone Minutes, with optional VO2max, sleep regularity, and insulin resistance; overlap is adjusted so each habit contributes its own share to health-age delta. Sensor World Model is an action-conditioned world model pre-trained on 2.28 million person-weeks from roughly 100,000 consenting Fitbit users; given two weeks of baseline data, it forecasts 20 physiological markers over 12 weeks and is paired with a feasibility-constrained optimizer and an LLM coaching agent. Validation includes 18,423 disease-free All of Us participants over 8 years and blinded reviews by five certified lifestyle coaches and five board-certified primary care/lifestyle medicine physicians across 49 profiles (490 paired reviews), plus automated evaluations in WEAR-ME (n = 2,780); counterfactual-grounded plans were preferred nearly 3-to-1 and met 99.2% clinical safety standards.