Published activity-count equations for estimating METs and PA intensity

Returns the regression coefficients and count-based cut-points for estimating energy expenditure (METs) and classifying physical activity (PA) intensity from ActTrust(R) (Condor Instruments) or ActiGraph(R) GT3X+ activity counts, hip or wrist placement.

Usage

pa_equations(equation_set = "batista2026")

Arguments

equation_set: character(1). Currently only "batista2026" is available.

Value

A tibble with one row per device x placement combination and columns:

device: "ACTT" or "GT3X+".
placement: "hip" or "wrist".
b0, b1: Intercept and slope of sqrt(MET) = b0 + b1 * sqrt(activity_counts).
cut3, cut6, cut9: Published count/min cut-points (95% CI midpoints) at the 3, 6, and 9 MET thresholds – i.e. the count value at which the fitted equation crosses that MET level. Provided for reference; classify_pa_intensity() classifies on estimated METs directly rather than re-deriving these.

Important caveats

These equations come from a single controlled-laboratory validation study (Batista et al. 2026; N = 56 healthy adults aged 18-35; treadmill walking/running at 3-9 km/h only). Treat them as one available equation set, not a universal standard:

The paper's own Discussion section compares its GT3X+ (hip) cut-points against two other published GT3X+ studies (Sasaki et al. 2011; Santos-Lozano et al. 2013) and finds differences of 2-65% depending on the MET threshold – and those two reference studies differ from each other by 16-39%. The paper attributes that spread mainly to sample-level characteristics rather than device or methodological artefacts, but that reading applies to the comparison between Sasaki and Santos-Lozano – it doesn't fully carry over to comparisons against this paper's own equations, since all three studies used different modelling approaches (Sasaki et al.: ActiGraph's two-regression model; Santos-Lozano et al.: an artificial neural network; this paper: a simple sqrt-transformed linear model with a device x placement interaction). Cross-study cut-point differences therefore reflect model choice as well as sample, not sample alone – and that's checkable to different degrees: this paper's linear coefficients (b0/b1 below) are transparent enough to compare term-by-term against Sasaki et al.'s two-regression model, so a discrepancy there is at least diagnosable. Santos-Lozano et al.'s cut-points come from an ANN, which has no inspectable coefficients – there is no way to say why it disagrees with the equations here, only that it does.
The ACTT (hip)/ACTT (wrist) equations are the first published cut-points for ActTrust(R) at all, so there is nothing yet to cross-check them against.
Validated only for laboratory treadmill walking/running in healthy young adults. Applying these equations to free-living data, other age groups (children, older adults), clinical populations, or other activity types is an extrapolation the source study explicitly flags as untested.

equation_set is exposed as an explicit argument (rather than hard-coding a single table) so a future validation study covering a different population or device can be added as an alternative set without changing the estimate_ee() / classify_pa_intensity() API.

Examples

pa_equations()
#> # A tibble: 4 × 7
#>   device placement    b0     b1  cut3  cut6  cut9
#>   <chr>  <chr>     <dbl>  <dbl> <dbl> <dbl> <dbl>
#> 1 GT3X+  hip        1.06 0.0199  1132  4853  9468
#> 2 ACTT   hip        1.11 0.0088  5057 23339 46410
#> 3 ACTT   wrist      1.23 0.0081  3761 22368 47203
#> 4 GT3X+  wrist      1.21 0.0127  1698  9503 19787