Abstract

Consumer health wearables increasingly let a user see how a personal biometric (resting heart rate, heart-rate variability, cardiorespiratory fitness, respiratory rate, blood-oxygen saturation, sleep duration) compares to a population. This disclosure describes a method by which a user opts in to such a comparison and is shown two distributions — one for a general population and one for a self-reported disease cohort (paradigmatically Parkinson's disease), each optionally filtered by sex and age band — without any raw, identifiable health value ever leaving the user's device. The novelty is not in any single component, all of which are established prior art (population-percentile comparison on wearables; disease-cohort percentile comparison; on-device local differential privacy; secure aggregation; randomized-response histograms). The novelty is the deliberate composition: an on-device-privatized biometric contribution, combined across devices by secure/anonymous aggregation requiring no trusted curator, with repeated daily releases governed by a streaming (w-event / w-day) differential-privacy budget that recycles out-of-window budget, and every sex × age-band × disease-cohort cell gated behind a minimum equivalence-class size with small-cell suppression, defaulting to a within-person baseline when a requested cell is too sparse to release privately. The least-attested and most defensible sub-combination — and the spine of this disclosure — is the use of streaming-DP budget recycling together with minimum-cohort small-cell gating to release a self-reported-disease-cohort biometric histogram from consumer wearables. The method is framed as a general-wellness comparison, not a diagnosis, a severity score, or a clinical reference range.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS