Reported frequency of physical activity in a large epidemiological study: relationship to specific activities and repeatability over time

How overall physical activity relates to specific activities and how reported activity changes over time may influence interpretation of observed associations between physical activity and health. We examine the relationships between various physical activities self-reported at different times in a large cohort study of middle-aged UK women. At recruitment, Million Women Study participants completed a baseline questionnaire including questions on frequency of strenuous and of any physical activity. About 3 years later 589,896 women also completed a follow-up questionnaire reporting the hours they spent on a range of specific activities. Time spent on each activity was used to estimate the associated excess metabolic equivalent hours (MET-hours) and this value was compared across categories of physical activity reported at recruitment. Additionally, 18,655 women completed the baseline questionnaire twice, at intervals of up to 4 years; repeatability over time was assessed using the weighted kappa coefficient (κweighted) and absolute percentage agreement. The average number of hours per week women reported doing specific activities was 14.0 for housework, 4.5 for walking, 3.0 for gardening, 0.2 for cycling, and 1.4 for all strenuous activity. Time spent and the estimated excess MET-hours associated with each activity increased with increasing frequency of any or strenuous physical activity reported at baseline (tests for trend, P < 0.003), although the associations for housework were by far the weakest (Spearman correlations, 0.01 and -0.03 respectively for housework, and 0.11-0.37 for all other activities). Repeatability of responses to physical activity questions on the baseline questionnaire declined significantly over time. For strenuous activity, absolute agreement was 64% (κweighted = 0.71) for questionnaires administered less than 6 months apart, and 52% (κweighted = 0.51) for questionnaires more than 2 years apart. Corresponding values for any physical activity were 57% (κweighted = 0.67) and 47% (κweighted = 0.58). In this cohort, responses to simple questions on the frequency of any physical activity and of strenuous activity asked at baseline were associated with hours spent on specific activities and the associated estimated excess MET-hours expended, reported 3 years later. The weakest associations were with housework. Agreement for identical questions asked on two occasions about the frequency of physical activity decreased over time.