JMIR Publications

ABSTRACT

Background:

Cardiovascular disease (CVD) is highly prevalent in the U.S. and adults ages 40-60 with specific health conditions are at particularly elevated risk. Physical activity (PA) is a key cardioprotective behavior and many interventions exist to promote PA in this group. Importantly, effective promotion requires accurate assessment of PA behavior to correctly evaluate intervention effects. As PA is variable and often estimated by averaging across multiple days, a potential threat to accurate assessment is measurement reactivity, or an atypical increase in PA behavior at the start of measurement periods that may bias conclusions. Evidence for PA measurement reactivity is equivocal, though concern has resulted in recommendations to add or drop PA measurement days from inclusion, which may introduce undue burden on participants. At present, the extent of PA measurement reactivity and the behaviors most likely to be affected (e.g., steps vs. minutes of exercise) are unclear. In addition, participant characteristics such as gender and the study expectations (e.g., intervention vs. observation only) may contribute to differences in these patterns, though the extent of these differences is unknown.

Objective:

An improved understanding of the extent of PA measurement reactivity and potential moderators among adults in midlife with CVD risk factors may inform decisions about whether and how to account for PA measurement reactivity, and in what contexts.

Methods:

To achieve this goal, we will conduct a coordinated multilevel analysis across 6 studies among adults ages 40-60 who have >1 risk factors for CVD (total N =1844; 13,434 total days of PA measurement with 6-7 days per person). The aims of this project are (1) to estimate the extent of PA measurement reactivity in each study and its influence on estimates (activity units, steps per day, minutes of activity in different intensities), (2) to examine individual difference characteristics as moderators of these patterns (e.g., gender, BMI, PA motivation), and (3) to examine differences in these patterns by study and measurement characteristics (e.g., PA monitoring device, location of wear, observation only vs. intervention baseline). Changes in PA behavior across the measurement period will be examined at the day level, using 2-level multilevel models (days nested within persons) and cross-level interactions (for moderation effects).

Results:

Data to be included are from publicly available datasets (observation only: NHANES and MIDUS) and baseline weeks of observation from behavioral weight loss clinical trials (K = 4), all collected in the U.S. Dataset acquisition is complete and data cleaning in preparation for analysis is underway.

Conclusions:

Results from this coordinated analysis project will provide the first large-scale estimation of the extent of PA measurement reactivity in an at-risk group. Findings will inform best practices for mitigating potential measurement reactivity in multi-day assessments of PA behavior.