Claim Check: "Wearables Detect Illness Before Symptoms"

The claim

"Your wearable can tell you are getting sick before you feel it." The claim sounds like a smoke alarm for infection. This page checks what was actually tested.

What the intervention actually is

Nothing is detected directly. A wearable records heart rate, skin temperature, activity and sometimes breathing rate. Software compares tonight's numbers with your own recent baseline and flags a change. A review of studies of this idea reported that increased heart rate was most frequently linked with SARS-CoV-2 infection, along with increased skin temperature and respiratory rate. [1] So the device measures body signals, and an algorithm estimates "something may be off." It does not identify a virus.

What the studies found

• A 2022 systematic review of 12 published articles (plus 12 unfinished study protocols) found accuracy varied greatly, with area under the curve from 0.52 to 0.92, and that the ability to catch presymptomatic infection ranged from 20% to 88% of cases. [1] The authors described the evidence as early stage.
• A study of symptomatic people (30,529 enrolled; 54 symptomatic people tested positive and 279 negative) found that combining symptom and sensor data reached an area under the curve of 0.80. [2] Note that this separated positive from negative among people who already had symptoms.
• A prospective smartwatch study of 3,318 people, 84 of whom were infected, produced alerts for 67 (80%) of the infected people, with presymptomatic signals at a median of 3 days before symptoms. [3] The same study found that other respiratory infections and non-infection events such as stress, alcohol and travel could also trigger alerts. [3]
• A randomized trial of prompts to test at home (DETECT-AHEAD) enrolled 450 people with a smartwatch. Of the 300 people in the two groups offered home tests, 118 (39%) were prompted to test, and 61 (52%) completed a self-test. [4] The authors showed early feasibility and also identified barriers to participation and performance to solve before large-scale use. [4]
• A hospital-based algorithm for lung infection, tested in 87 patients with lung infection and 408 healthy people, reported 81.4% sensitivity and 90.4% specificity against clinical diagnosis. [5] It combined several signals, including cough sound, and some co-authors were affiliated with the watch maker. [5]

Plausible mechanisms (separate from the evidence)

Researchers tried these signals because subtle changes in heart rate, breathing rate and skin temperature could act as early digital biomarkers of infection. [1] That makes the idea plausible. It does not prove that a particular consumer device reliably catches it.

What the marketing leaves out

• "Detects illness days early" usually reflects a subset of cases in one study, at a cost of false alarms.
• A wellness feature that sends notifications is not the same as a diagnostic test. In the United States, the FDA says it generally does not examine low-risk general wellness products to decide whether they are devices. [7]