Background/Purpose
A large classification discrepancy has been reported in commonly used physical activity (PA) measures (e.g., according to Troiano et al., 2008, % of U.S. adult females meeting PA recommendations was 46.4 based on BRFSS self-report, was 42.5 according to NHANES self-report, but was only 3.2 according to NHANES Actigraph accelerometer findings!), which made the construct validity of these measures questionable. The purpose of this study was, through a meta-analysis, to examine the concurrent validity of a set of commonly used PA measures by examining their correlations with the doubly labeled water (DLW) method, the gold standard for measuring energy expenditure in a free-living condition.
Method
A comprehensive literature search was conduced via several literature databases (e.g., MEDLINE and Web of Science database) with the following key words: Doubly labeled water, accelerometer, heart rate, pedometer, questionnaires, diary, activity record, and armband. Seventy-six publications met all the inclusion/exclusion criteria, studies were 35 using accelerometers, four heart rate monitors, four pedometers, 24 questionnaires, seven activity records/diaries and two armbands. The collected information was analyzed using the meta-analysis procedure by Hunter and Schmidt (2004).
Analysis/Results
The correlations between the PA measures with DLW were ranked as follows (PA measure, N, raw-correlation mean, estimate of the population correlation, variance of sample correlation, average sample size, sampling error variance, & estimate of population-correlation variance): Armband (51, 0.91±0.08, 0.8729, 0.0012, 25.5, 0.0024, -0.0012),Pedometer (68, 0.50±0.30, 0.6358, 0.0409, 15, 0.0092, 0.0317),Heart rate (42, 0.52±0.31, 0.6107, 0.1096, 11.71, 0.0367, 0.0729), Accelerometer (1584, 0.62±0.22, 0.5389, 0.0455, 34.4, 0.0150, 0.0305), Questionnaire (1261, 0.29±0.26, 0.2590, 0.0517, 31.95, 0.0281, 0.0236), Activity record/diary (773, 0.34±0.33, 0.1332, 0.0362, 96.63, 0.0101, 0.0261).
Conclusions
Among the PA measures examined, only armband's validity was confirmed. Pedometers, Heart rate monitors and Accelerometers showed moderate validity evidence while low validity evidence was found for questionnaires and activity records/diaries. The observed discrepancy is likely caused by poor construct validity of some of these measures.