Accuracy exists at several levels
- Sensor validity: does the device measure the underlying signal accurately?
- Derived-metric validity: does an algorithm estimate sleep stage, HRV, VO₂ max, or another derived measure reliably?
- Biomarker validity: is that measure associated with aging, function, disease, or mortality in the relevant population?
- Composite-model validity: does the final age score predict or track the outcome it claims to represent?
Consumer wearable performance is heterogeneous
A 2024 living umbrella review synthesized 24 systematic reviews covering 249 non-duplicate validation studies and more than 430,000 participants. Performance varied substantially across devices and outcomes, and only a minority of commercially available devices had validation evidence for at least one biometric outcome.
Sleep staging illustrates the problem. Consumer devices can be useful for longitudinal sleep monitoring, but stage-level performance depends on sensors and algorithms and should not be treated as equivalent to polysomnography.
Longitudinal consistency can still be useful
A measurement does not need to replace a clinical gold standard to provide useful longitudinal context. Repeated measurements under similar conditions may reveal within-person patterns even when absolute accuracy is imperfect.
The critical requirement is to preserve measurement limitations rather than letting a polished composite score create false precision.
What BioMIR exposes
BioMIR uses selected Apple Health observations and keeps contributor-level values visible. Data Confidence summarizes daily input support, while CMA Freshness keeps the recency of persisted cardiometabolic observations explicit. These features support interpretation; they do not substitute for external model validation.