Two or three mornings of wrist accelerometry predicted five-year mortality

Design

secondary analysis of a national cohort with mixed effects location scale modelling and two-stage regression

Population

584 community-dwelling older adults wearing wrist accelerometers for 72 hours; 104 deaths at 5 years

Primary outcome

five-year mortality and frailty phenotype score among survivors

Effect

mean morning activity OR 0.64 per SD for mortality (P = 0.04); morning activity variance OR 0.67 per SD for frailty (P = 0.01)

Accelerometry tracks frailty well but the usual five to seven day wear protocols lose adherence in exactly the older people they are meant to assess. This analysis asked whether a much shorter window carries the same information, using 584 community-dwelling older adults from a national American cohort who wore wrist accelerometers for 72 hours in 2010-11, with frailty measured then and in 2015-16. A hundred and four died within five years.

Two morning measures were derived from the 7am to noon window: mean hourly activity, and how much that activity varied from day to day within the same person. After adjustment for baseline frailty, demographics, comorbidities, body mass index and cognition, higher mean morning activity predicted lower five-year mortality (odds ratio 0.64 per standard deviation, P = 0.04). Higher day-to-day morning variability predicted better frailty scores among survivors (0.67 per standard deviation, P = 0.01).

The variability finding runs against intuition — irregularity is usually read as instability — and the authors' interpretation is that day-to-day variation reflects a person still doing different things on different days, while a flat, uniform pattern reflects a life narrowed by frailty. That is plausible and unproven.

The practical significance is feasibility rather than discovery. If 72 hours carries the signal, a device can be issued at a clinic visit and returned at the next one, which is a workable proposition in a way that a week of wear is not. This remains a research finding in one cohort, with the mortality association at P = 0.04, and no threshold exists that anyone could act on for an individual patient.

• Note that a 72-hour wear period may be sufficient for frailty-related accelerometry, which changes what is feasible in clinic
• Do not use accelerometry output to assess an individual patient's risk; no clinically usable threshold exists
• Ask about morning activity in the history — what a patient actually does between waking and midday
• Read a flat, unvarying daily pattern as a possible marker of narrowing function rather than of stability
• Continue to use validated frailty assessment; this does not replace it

Why it matters

The reason accelerometry has not entered practice is the wear period, and this suggests the wear period can be much shorter.

Don't overread it

One cohort, a borderline P value for mortality, and no validated threshold — this cannot be applied to an individual patient.

The statistics, in plain English

An odds ratio of 0.64 per standard deviation with P = 0.04 sits just inside conventional significance, so this is a finding that could shift with a slightly different model or cohort. Adjusting for baseline frailty is important because it means the accelerometry adds information beyond what a frailty assessment already gives — but the covariates are those recorded, and reverse causation remains live: people who are already declining move less. With 104 deaths among 584 participants there is reasonable power for the mortality analysis and less for subgroup questions.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.