The effects of measurement error on the relation between physical activity and cardiovascular disease mortality in the Canada Fitness Survey cohort

grantor: University of TorontoThe purposes of the present study were to examine the relation between physical activity and mortality in women and men in the Canada Fitness Survey cohort, to assess the reliability of the physical activity components of the Canada Fitness Survey questionnaire (CFSQ) and to adjust estimates of the hypothetical true associations between physical activity and mortality, using estimates of reliability for the CFSQ. Results indicate a significant inverse association between physical activity and mortality in women due mainly to the contribution of non-leisure (household chores) energy expenditure. The association between energy expenditure and mortality in men was not as strong as in women. The CFSQ is moderately reliable for most measures of physical activity. Estimates of reliability vary considerably among the various activities and components of these activities and between males and females. The adjustment procedure resulted in estimates which showed a stronger association with mortality than did the crude estimates. The change in the adjusted estimates and the width of their confidence intervals (CI) were related to the size of the measurement error. Results of simulation studies showed that when the error is too great, the adjustment procedure resulted in unpredictable estimators. Nevertheless, in the presence of moderate measurement error and large enough sample sizes, results from the current study indicate that, adjustment for measurement error is a feasible and desirable procedure.Ph.D