This study examined the associations between real-world walking speed, duration, step regularity, longest walk, running, and step counts derived from a wrist device with the risk of cardiovascular death. We also assessed if these digital gait biomarkers added predictive value beyond established risk factors.
Participants aged 60 to 78 from the UK Biobank who wore a wrist-worn motion sensor for at least five consecutive days were included. Digital gait biomarkers were derived from the sensor data using the Watch Walk Platform. Cardiovascular death, defined as deaths within ten years due to ischemic heart disease, hypertensive disease, cardiac arrest, arrhythmias, stroke, or vascular diseases, were retrieved from the NHS England and the NHS Central Register databases. Established risk factors were collected through questionnaires, including age, sex, BMI, smoking status, blood pressure, and diabetes history. Cox proportional-hazard models were used to evaluate the associations between gait characteristics and cardiovascular death, adjusting for age, sex, and other risk factors. Model performance was measured using Harrell's concordance statistic.
Among 38,766 participants, 485 cardiovascular deaths (1.3%) were recorded. Maximal walking speed had the strongest association with reduced cardiovascular death risk, with each standard deviation (SD, 0.08 m/s) increment reducing the hazard by 25%. Other significant factors included daily running duration (21% reduction per SD), step counts (19% reduction per SD), longest walk duration (18% reduction per SD), proportion of walks ≥ 8 seconds (13% reduction per SD), and step regularity (12% reduction per SD). After adjusting for established risk factors, maximal walking speed, daily running duration, and proportion of walks ≥ 8 seconds remained independently predictive, while step counts did not. The multivariable model’s concordance statistic was 0.76, comparable to established risk scores like SCORE2 and the Framingham Risk Score (both 0.74).
Real-world walking speed and daily running duration were stronger predictors of cardiovascular death than step counts. These gait characteristics, along with completing a greater proportion of longer walks, provide additional predictive value beyond traditional risk factors.
With the rising popularity of smartwatches, gait characteristics can be collected non-invasively and automatically, offering a low-cost tool for early cardiovascular death risk screening. Clinicians might consider assessing real-world gait characteristics beyond step counts when identifying individuals at risk of cardiovascular death.
Wearable sensor
Walking
