Fitbit Charge 3 signals precede 4 of 5 COVID-19 exacerbations

AI system

System: Fitbit Charge 3 estimated oxygen variation (EOV) and resting heart rate, with investigator-defined signal thresholds
Model: Not assessed
Release: Exact release not established.
Application developer: Fitbit (device and EOV algorithm); Kanazawa University investigators (signal definitions)
Role: Not assessed
Configuration: The participant used a Fitbit Charge 3.
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A Fitbit Charge 3 worn for 30 days by patients with PCR-confirmed COVID-19, giving EOV and resting heart rate. An investigator-defined high-EOV signal (a significant dip and recovery of the oxygen level within the index period) reviewed by the study team. Weekly phone calls to discharged patients and access to portable polysomnography.
Human oversight: Not assessed
Relation to AI: The device produced the EOV values. The investigators' own rule flagged high signals, and clinicians decided to order polysomnography for the sleep-apnea patient. The paper does not say whether Fitbit's EOV estimation uses machine learning and notes the functions were not approved as a medical device. The COVID-19 prediction figures were computed after the outcomes were known, not from alerts acted on in real time.
AI contribution: The device produced the EOV values. The investigators' own rule flagged high signals, and clinicians decided to order polysomnography for the sleep-apnea patient. The paper does not say whether Fitbit's EOV estimation uses machine learning and notes the functions were not approved as a medical device. The COVID-19 prediction figures were computed after the outcomes were known, not from alerts acted on in real time.

Benefits

Outcome identifiers: Not assessed
Benefit: EOV signals preceded four of five analyzed cases (sensitivity 80%, specificity 90%)
Category: Improved Physical Health
Status: Not assessed
Beneficiary groups: People. Organizations.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: No
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: EOV signals preceded four of five analyzed cases (sensitivity 80%, specificity 90%)
Comparison: high resting-heart-rate (RHR) signals from the same Fitbit data
Results: Reference: the sensitivity and specificity of high RHR signals were 50% and 80%, respectively. Result: 4 out of 5 cases (sensitivity=80%, specificity=90%).
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Fitbit (device and EOV algorithm); Kanazawa University investigators (signal definitions). Operator: Kanazawa University and JCHO Kanazawa Hospital (DEXTERITY pilot, UMIN000041421).
Jurisdictions: Benefit: Kanazawa, Japan. Developer: United States. Operator: Japan.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Patients can pair the device with a smartphone app (elderly patients unable to use a smartphone were not enrolled). Enough baseline data to separate abnormal signals from normal variation, which the authors list as lacking in the 30-day window.
Replication attempts: Not assessed
Later findings: No later DEXTERITY publication and no follow-up on patient JCHO-023 was found in the cited paper.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0294-source-0001–0004
Citations: Numbered references below.
Language: Not assessed
Source relationships: Developer
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People. Organizations.
Depth: People: 1. Organizations: Not assessed.
Reasons: People: One man received a diagnosis of severe obstructive sleep apnea (3% ODI 33.6/h) from an incidental wearable signal. The paper reports no treatment or follow-up for him. Organizations: The 80% sensitivity for predicting exacerbation is a retrospective pilot result on five events, which the authors say is too small to generalize. No live alerting or changed care is documented.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The makers' claim

Entry details

Kind: Not assessed
Related entries: Not assessed
Field: Healthcare (infectious disease; sleep medicine). Consumer wearables.
Purpose: Fitbit's EOV metric summarizes overnight oxygen variation for consumers. The investigators repurposed it to flag COVID-19 deterioration, and the sleep-apnea finding was incidental to that study aim.
Dates: Date basis unassessed: 2021 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: 1c723e06aa36…

References

Other identifiers: Not assessed
Category mappings: Not assessed
Brief: Not assessed
Revisions: 5 revisions. Latest: 21 October 2026.

Sources

Peer-reviewed study Developer

Yamagami K, Nomura A, Kometani M, Shimojima M, Sakata K, Usui S, Furukawa K, Takamura M, Okajima M, Watanabe K, Yoneda T (2021). Early Detection of Symptom Exacerbation in Patients With SARS-CoV-2 Infection Using the Fitbit Charge 3 (DEXTERITY): Pilot Evaluation. JMIR Formative Research 5(9):e30819 (PMC8448084).