Adapting the Intensity Gradient for Use Across Commonly Derived Accelerometer Epoch Summary Measures: A Learning Network for Advanced Behavioural Data Analysis (LABDA) Project

The intensity gradient (IG) quantifies the distribution of time spent across accelerometer-assessed physical activity intensity and is positively associated with health. It was developed using the Euclidean norm minus one (ENMO) acceleration measure. This study aimed to enable generation of comparable IGs across other measures (mean amplitude deviation [MAD], monitor independent movement summary [MIMS], and counts) by addressing a key step in the IG algorithm of dividing physical activity intensity into incremental intensity bins. Two methods of creating analogous bins for MAD, MIMS, and counts were explored: (a) linear scaling (“naïve”) and (b) nonlinear modeling (“modeled”). Generated IGs were compared with the original IG (IG_ENMO) using limits of agreement and intraclass correlation. Forty-three adults (age: median [interquartile range]: 23 [21, 26], 61% female) were included. Relative to IG_ENMO, the naïve approach led to higher IGs (+0.27, +0.39, +0.54 for MAD, MIMS, and counts, respectively). In contrast, the modeled approach led to lower IGs (bias: −0.43, −1.23, and −0.91, respectively). For MAD and counts, limits of agreement were slightly wider for naïve bins (95% limits of agreement: ±0.26 and ±0.34) versus modeled bins (±0.21 and ±0.28) but for MIMS were slightly wider for modeled bins (modeled: ±0.35, naïve: ±0.31). For the naïve approach, IG_MAD was most consistent (intraclass correlation 95% CI: 0.49–0.82) and IG_Counts least consistent (0.09–0.61). Intraclass correlations were higher for the modeled approach with IG_MAD most consistent (0.72–0.91) and IG_MIMS least consistent (0.59–0.86). Results indicate that consistency of the IG between measures is improved with appropriate scaling to create analogous intensity bins, but agreement is limited.