The authors examined baseline data from the SWHS, an ongoing population-based prospective cohort study. The study was approved by the Institutional Review Boards of all participating institutions, and written informed consent was obtained from all participants. A detailed description of the study methodology has been published elsewhere 44. Briefly, eligible participants for the SWHS were all women 40–70 years of age, residing in seven geographically defined areas of urban Shanghai, China. The study communities were selected because they were similar to urban Shanghai in terms of disease incidence rates and demographic characteristics. Before the study was initiated, rosters of all female residents between the ages of 40 and 70 were obtained from administrative offices of the study communities. All eligible women were approached in their homes by a local community health worker and a trained interviewer to determine interest in participating in the study. After obtaining informed consent from participants, a second home visit was scheduled to collect the self-administered questionnaire provided at the first visit, and to conduct an in-person interview, take anthropometric measurements and collect biological samples.

Of 81,170 eligible participants, 75,221 (92.7%) completed the baseline survey between 1997 and 2000. Reasons for nonparticipation included: refusal (n = 2,407, 3.0%), absence during the recruitment period (n = 2,073, 2.6%), and other miscellaneous reasons (n = 1,469, 1.8%). Completion of the self-administered questionnaire identified 279 women (0.4%) younger than 40 or older than 70 years that were further excluded, leaving 74,942 women in the baseline cohort.

Physical activity data were collected via an in-person interview using a physical activity questionnaire (PAQ) that was validated in a sample of approximately 200 women from the SWHS 45. Reproducibility (2-year test-retest) for adolescent and adult exercise participation (k = 0.85 and k = 0.64, respectively), and adult exercise energy expenditure (ICC = 0.70) were reasonable. Validity evaluated by comparing the results of the first and second administrations of the PAQ with the 7-day physical activity logs and the 7-day physical activity recalls was also good; moderate to strong correlations were found for most physical activity variables (Spearman correlation coefficients between 0.33–0.87) 45.

At the baseline survey, participants were first asked if they had engaged in regular exercise/sports (at least once a week for at least three months a year) during the preceding five years. Exercisers were asked to report details for up to three types of exercise/sports (i.e. type, hours/week, and years of participation in each activity), as well as how often they had sweated during exercise. Likewise, they were asked whether they had engaged in regular exercise during adolescence (13–19 years of age), the average duration (hours/week), and length of participation (years). Participation in athletic teams during adolescence was also evaluated. In terms of non-exercise activities, women were asked about participation in active transportation (i.e. walking and cycling to/from work, daily errands), stair climbing (flights of stairs/day), and housework (hours/day). Finally, women were asked to report the proportion of housework that they did by themselves. The time frame for these non-exercise activities was the preceding year.

Physical activity energy expenditure was estimated using standard metabolic equivalent (MET) values 46. Adult exercise/sports energy expenditure was estimated by the weighted average of energy expended in all activities reported over the five years preceding the interview (MET-h/week/year). For non-exercise activities, the authors estimated the energy expenditure in MET-h/week using appropriate MET values (i.e. walking: 3.3, cycling: 4, stair climbing: 9, and housework: 2) and the corresponding reported duration of the activity (hours/week).

Activities of 3 to 6 MET-h were classified as moderate and those >6 MET-h as vigorous intensity 46. In order to evaluate adherence to CDC/ACSM guidelines, the authors calculated the proportion of women that met current recommendations of 150 minutes per week of moderate activity (e.g. Tai Chi, dancing, walking, cycling) or 60 minutes of vigorous activity (e.g. jogging, martial arts gymnastics, stair climbing) per week. Housework, categorized as light activity (2 MET-h), was not included in these calculations.

Occupational activity in the last job held by participants was categorized as low, medium, or high based on more than 300 different job titles using a coding system previously developed for the Shanghai urban population 47. Clerks and accountants are examples of low activity occupations, while cooks and machine operators are examples of high activity occupations.

Baseline information regarding demographics (age, marital status, family size (number of people living together), education, income, and employment history), lifestyle behaviors (cigarette smoking, alcohol consumption, tea consumption, ginseng intake, TV watching), menopausal status (defined as cessation of menstrual periods for 12 months or longer excluding lapses caused by pregnancy and breastfeeding), and medical history was collected through a self-administered questionnaire. Dietary data were collected by in-person interviews using a validated food frequency questionnaire 48 administered at the same time as the physical activity questionnaire. The food frequency questionnaire used in the SWHS includes 71 food items and food groups that cover about 86% of commonly consumed foods in urban Shanghai. For each food item or food group, subjects were asked how frequently (daily, weekly, monthly, yearly, or never) they consumed the food or food groups over the past year, followed by a question on the amount consumed in lians (50 g/lian) per unit of time. Daily energy intake and other nutrient intake data were derived from reported food intake and nutrient content listed in the Chinese Food Composition Tables 49. Energy intake was included as a continuous variable in the current analysis.

Weight and height were measured by interviewers using a standardized protocol at the time of interview.

The prevalence of various domains of physical activity was calculated as the percentage of women involved in that activity. Median, 25^th, and 75^th percentiles are presented for continuous variables because their distributions were skewed. Body mass index (BMI) was calculated as weight (kg)/squared height (m²). BMI was categorized according to Chinese standards, that is: underweight to normal (<24 kg/m²), overweight (24.0–27.9 kg/m²), and obese (≥ 28 kg/m²) 50.

Correlations among energy expenditure estimates from various domains of physical activity were evaluated using the partial Spearman rank-correlation coefficient (r). For practical reasons the authors dichotomized energy expenditure (MET-h/week) from non-exercise activities at the median to create binary variables to be used in logistic regression. The associations of physical activity domains with demographics, behavioral factors, health status, and dietary energy intake were evaluated by odds ratios (OR) and 99% confidence intervals (CI) derived from logistic regression. Given the large sample size available for the analyses, the level of statistical significance was set at 0.01 and all tests were two-tailed. Statistical analyses were performed using SAS version 9.1 for Windows (SAS Institute, Carry, NC).

Physical activity patterns reported during adolescence and adulthood are presented in table 1. More than three fourths of the women participated in exercise/sports during adolescence, with a median duration of two hours per week. This finding may reflect participation in physical education classes in middle and high school. The prevalence of reported regular exercise during the preceding five years was 35.5%, and 50% for women who reported exercising more than two hours per week. Those who exercised during adolescence were more likely to be adult exercisers (OR: 1.5, 99% CI: 1.4, 1.5). Most of the regular exercisers (76.4%) reported participation in only one activity, 19% in two activities and 5% in three activities (data not shown). The most common types of exercise reported were fitness-oriented traditional Chinese exercises of moderate intensity. The prevalence of vigorous activity reported was low (8% of exercisers). In contrast, active commuting to work (referred to as transportation), especially walking, was reported by more than 90% of women who were employed at the time of interview. Among women reporting these activities, the median duration of walking was about five hours per week and cycling to and from work about three hours per week. Almost all women reported walking, and only 8% reported biking for purposes other than transportation to work (referred to as daily activity). A large amount of time was spent in housework (median: 14 hours/week). Only 7% of participants reported recent jobs that required high levels of activity.

When all moderate and vigorous non-occupational activities were taken into account, 63.8% of participants met current CDC/ACSM recommendations 51.

Overall, more than 90% of participants' reported physical activity energy expenditure was from non-exercise activities. Total energy expenditure was, on average, higher in regular exercisers than in non-exercisers (105.5 vs. 96.7 MET-h/week), but the amount of energy expended in non-exercise activities, most of moderate intensity, was higher in non-exercisers (96.7 vs. 92.1 MET-h/week) (table 2). Regular exercisers reported expending less energy in transportation and more energy walking and cycling for habitual daily activities than non-exercisers.

As presented in table 3, energy expenditure from exercise/sports was correlated inversely and weakly with that from transportation (r = -0.11); inverse correlations between transportation and daily activities (r = -0.24) and housework (r = -0.23), respectively, were also noted. A positive correlation was observed between energy expenditures in daily activities and housework (table 3).

Older age and being unemployed (including retired and never employed) were positively associated with participation in exercise/sports and higher levels of housework (tables 4 and 5). Interestingly, some of the associations of physical activity domains with socio-demographic and lifestyle variables were in opposite directions for exercise/sports and non-exercise activities. For example, women with more education, more income, who were professionals, or had a smaller family size were more likely to report exercise participation, but they reported fewer non-exercise activities. Being a widow was not associated with participation in sports/exercise, but widows were more likely to walk or bike for transportation to and from work (OR: 1.26) and were less likely to do housework (OR: 0.87) than currently married women. Women who had never been married were less likely to exercise regularly or to expend energy in daily activities and housework. Chronic disease history was positively associated with exercise/sports (OR: 1.20) and somewhat inversely associated with non-exercise activities (ORs between 0.90 and 0.96). The most common types of exercise among women with a history of chronic disease were Tai Chi and other slow movement exercises (84.8%).

Among behavioral factors, cigarette smoking was negatively associated with regular exercise (OR: 0.75, 99% CI: 0.68, 0.83), while alcohol consumption (OR: 1.26, 99% CI: 1.13, 1.40), tea consumption (OR: 1.15, 99% CI: 1.11, 1.19), and ginseng intake (1.50, 99% CI: 1.45, 1.55) were all positively associated with regular exercise. Watching TV more than four hours/day was negatively associated with participation in sports/exercise, but was positively associated with housework and walking or biking for daily activities. Both exercise and non-exercise activities were significantly associated with increased total dietary energy intake (ORs between 1.21 and 1.38). Food groups had a very weak correlation with energy expenditure; therefore, these data are not presented in the tables.

This study is one of the largest and most comprehensive population-based surveys investigating patterns and correlates of physical activity, and the first such study conducted among urban Chinese women. The population-based design and a participation rate of more than 90% allowed a direct estimation of the prevalence of physical activity levels in Chinese women from Shanghai and also minimized selection bias. However, these results cannot be generalized to non-urban or younger women who were not included in the SWHS 44.

The physical activity questionnaire was interviewer-administered and provided a comprehensive picture of participants' physical activity patterns by collecting extensive and detailed information on the type, duration, and length of various domains of physical activity. A limitation of this questionnaire is that it did not distinguish between differences or changes in exercise habits that might have occurred during the 5-year period covered by the questionnaire and current exercise patterns. However, in the validation study of the PAQ, we found that the 5-year average of exercise habits correlated well with exercise during the preceding year as measured by multiple 7-day PAQ and 7-day logs 45, suggesting that exercise patterns in our study population are relatively stable and that the averaged exercise for the preceding 5 years reflects the participants' current exercise patterns. Because of the cross-sectional nature of data, the temporality of the observed associations cannot be determined and no causal inferences can be made. These limitations should be considered when interpreting the results.