Comparative validity of physical activity measures in older adults

Abstract

Purpose: To compare the validity of various physical activity measures with doubly labeled water (DLW)-measured physical activity energy expenditure (PAEE) in free-living older adults.

Methods: Fifty-six adults aged ≥65 yr wore three activity monitors (New Lifestyles pedometer, ActiGraph accelerometer, and a SenseWear (SW) armband) during a 10-d free-living period and completed three different surveys (Yale Physical Activity Survey (YPAS), Community Health Activities Model Program for Seniors (CHAMPS), and a modified Physical Activity Scale for the Elderly (modPASE)). Total energy expenditure was measured using DLW, resting metabolic rate was measured with indirect calorimetry, the thermic effect of food was estimated, and from these, estimates of PAEE were calculated. The degree of linear association between the various measures and PAEE was assessed, as were differences in group PAEE, when estimable by a given measure.

Results: All three monitors were significantly correlated with PAEE (r=0.48-0.60, P<0.001). Of the questionnaires, only CHAMPS was significantly correlated with PAEE (r=0.28, P=0.04). Statistical comparison of the correlations suggested that the monitors were superior to YPAS and modPASE. Mean squared errors for all correlations were high, and the median PAEE from the different tools was significantly different from DLW for all but the YPAS and regression-estimated PAEE from the ActiGraph.

Conclusions: Objective devices more appropriately rank PAEE than self-reported instruments in older adults, but absolute estimates of PAEE are not accurate. Given the cost differential and ease of use, pedometers seem most useful in this population when ranking by physical activity level is adequate.

FIGURE 1

Bland–Altman plots of repeat administrations, 10 d apart, of the physical activity questionnaires. Mean differences in PAEE (first administration − second administration) are plotted against the mean PAEE (kcal·d⁻¹) derived from CHAMPS (A), modPASE (B), and YPAS (C).

FIGURE 2

Comparison of median PAEE (kcal·d⁻¹) values from DLW versus SW armband, Freedson equation from ActiGraph, Crouter equation from ActiGraph, and CHAMPS, modPASE, and YPAS questionnaires. The y axis is presented in the log scale. The solid line represents median value, box edges are the 25th and 75th percentiles, and the whiskers indicate the range. The dashed horizontal line is at the median value (680 kcal·d⁻¹) for DLW. *Significantly different from PAEE based on Wilcoxon rank sum test, P < 0.008 (Bonferroni adjustment: 0.05/6 comparisons = 0.008).

FIGURE 3

Bland–Altman plots of PAEE (kcal·d⁻¹) estimated by different physical activity measures assessed for agreement with DLW-derived PAEE. Differences (PAEE_predicted – PAEE_DLW) are plotted against the mean PAEE (mean of PAEE_predicted and PAEE_DLW) for the SW armband (A), Crouter equation from the ActiGraph (B), Freedson equation from the ActiGraph (C), CHAMPS questionnaire (D), modPASE questionnaire (E), and YPAS questionnaire (F).