Consumer Wearable Devices for Population-Level Infectious Disease Surveillance: A Scoping Review
Preprint, version 2. This manuscript has not been peer reviewed. It is under consideration at a peer-reviewed journal. This version supersedes version 1 (10.5281/zenodo.19411761, April 2026), which should no longer be cited for any numerical result. The review was rebuilt rather than amended, and the two versions do not differ by degree. In summary: The evidence base changed. Version 1 reported 108 included studies from a five-database search. Re-adjudication of those 108 against the registered protocol criteria retained 27; a harmonised six-source search with backward and forward citation searching of every included study added 29 more. The included set is now 56 studies, of which 20 were found by citation searching, a step absent from version 1 entirely. A headline claim was withdrawn. Version 1 reported that 23 studies gave advance-detection lead times of 1 to 75 days, median 7. That was an artefact of an automated extractor reading study-design windows, baseline periods and figure captions as detection leads. On adjudication of every candidate value against its full text, 6 studies (11%) report a genuine lead time, in three measurement classes that are not commensurable. No pooled lead time is reportable, and none should be computed from this review. A reported gap was largely our own. Version 1 reported that only 60% of studies gave a participant count and framed this as a deficiency in the primary literature. On re-extraction from full text, 55 of 56 studies (98%) report one. A structural finding was absent from version 1. 23 of the 56 included studies (41%) analyse a participant cohort shared with at least one other included study, across 7 confirmed cohort families; one Stanford smartwatch dataset accounts for eleven of them. The 56 studies represent 40 independent cohorts, and every prevalence is now reported at both levels. The extraction layer was rebuilt after an audit found four reproducible failure modes, and every reference was re-verified against PubMed, CrossRef or DOI resolution. Each of these is stated in full in the "Version 2: what changed, and why" section on the first page of the manuscript PDF. Abstract Objective: The objective of this scoping review was to map the extent, nature, and distribution of research using consumer wearable biometric data for population-level infectious disease surveillance and early outbreak detection. Introduction: Consumer wearable devices are in widespread global use, with industry estimates placing more than one billion devices in service,[1] creating opportunities for population-level infectious disease surveillance. No review has mapped the use of aggregated wearable data for outbreak detection. Inclusion criteria: Studies aggregating consumer wearable device data from 100 or more individuals for infectious disease surveillance, outbreak detection, or epidemic monitoring were eligible. Peer-reviewed journal articles, conference proceedings, and preprints published from January 2015 to August 2026 in English were considered. Methods: This scoping review followed JBI methodology[2] and was reported per PRISMA-ScR guidelines.[3] Six sources (PubMed, Embase, Scopus, Web of Science, IEEE Xplore, arXiv) were searched, supplemented by backward and forward citation searching of every included study. Title and abstract screening used two independent large-language-model passes, each of which screened the full record set. All full texts were retrieved and every included study was charted from full text using a standardised 27-variable form. The protocol was registered on the Open Science Framework (https://doi.org/10.17605/OSF.IO/Y6827). Results: Fifty-six studies were included, but they analyse only 40 independent cohorts: 23 studies (41%) report on a participant cohort shared with at least one other included study, and a single Stanford smartwatch dataset accounts for 11 of them. Of the 56, 27 came from the original database search and 29 from the harmonised search and citation searching, 20 of them citation-derived. COVID-19 dominated (48 studies, 86%); 8 studies addressed only non-COVID targets. Heart rate was the most used signal (53, 95%) and Fitbit the most named device (27 of the 38 studies naming one). Participant counts were reported by 55 studies (98%), median 1,163 (range 103 to 535,556). Six studies (11%) reported an advance-detection lead time, in three non-commensurable measurement classes; no pooled lead time is reportable. Reporting was uneven: 64% documented ethics approval, 34% named a reference surveillance system, 29% discussed equity, and 11% shared code. Conclusions: Wearable-based population surveillance remains a small, COVID-dominated evidence base whose apparent size overstates its independence, since two fifths of studies reanalyse a shared cohort. Advance detection over traditional surveillance is asserted more often than it is measured against a named reference system. Expansion to non-COVID diseases and to settings beyond high-income countries, together with reference-anchored performance reporting, is the critical next step. Materials The charted dataset with per-field provenance, the screening and adjudication decision files, the false-negative validation sample, the analysis and figure-generation code, and the generated figures and tables are openly deposited at https://doi.org/10.17605/OSF.IO/BAZQV. Every figure and table in this manuscript is generated from that dataset by those scripts. The protocol was registered before screening began and is held separately as an immutable registration at https://doi.org/10.17605/OSF.IO/Y6827.