Evaluation of a short retrospective questionnaire for physical activity in women

Physical activity (PA) is discussed as a preventive factor for many chronic diseases. Thus, in epidemiological studies it often is an important covariate. Due to frequently long latency periods, long-term PA in the past is of greater interest than current PA. However, there is a lack of retrospective questionnaires that are validated for long-term PA, including occupational, household, and leisure activities. We therefore evaluated a short questionnaire for the comprehensive assessment of usual PA in distant age periods, administered with cognitive interviewing techniques. From an ongoing study on postmenopausal breast cancer 110 cases and 101 controls, age 50-74 years, were randomly selected. Our questionnaire was administered in a telephone interview more than two months after the main study interview, which included a detailed questionnaire on PA performed in the age periods 30-49 years and 50+ years. Total PA scores were derived from both interviews as MET-hours per week. Degree of agreement was assessed using Bland-Altman analyses. Further, potential sources of systematic and random error were investigated. The mean difference between both questionnaires was 3 MET hours/week, and 53.6% of absolute differences were below 35 MET hours/week, i.e. showing good agreement. Further 28.9% of differences could be considered acceptable agreement. Measurement errors seem to be non-differential with respect to cancer status. The median interviewing time was 10 min. Overall, this short questionnaire appears to be a useful and valid tool to distinguish between high and low levels of women's physical activity in the distant past.