Association between leisure time physical activity preference and behavior: evidence from the China Health & Nutrition Survey, 2004–2011

Findings from a number of studies have suggested that food preference is a good indicator of actual food intake and is also found to be associated with cardiovascular disease [1‐4]. Exploring the association between physical activity preference and physical activity behavior thus appears to be promising and necessary. Studies on sedentary activity preference indicate that it is a significant predictor of lower physical activity level [5, 6], and preferences for physical activity and psychological variables such as anxiety, depression, and avoidance explained significant physical activity changes among children [7]. To the best of our knowledge, however, no studies have examined the direct relationship between preference for LTPA and actual LTPA behavior in adults. To further understand the determinants and predictors of physical activity behavior, it becomes necessary to investigate the association between LTPA preference and behavior, especially by using longitudinal data to examine causal effects.

Such a step is important in consideration of the complexity and difficulty in reliable measurements of LTPA. Comprehensive and relevant measurement of LTPA is fundamental to health promotion [8]. It can be challenging, however, to obtain accurate and reliable measurements of LTPA. One major issue relates to whether LTPA is subjectively or objectively assessed. There is evidence that correlations between self-reporting measures of LTPA and direct measures were low to moderate, suggesting that measurement methods may have a substantial impact on the observed level of LTPA [9]. This finding points to the importance of identifying alternative measures of LTPA, including LTPA preference.

In China, LTPA level is low and with recent, continuous evidence of decrease. Data from the China Health & Nutrition Survey shows that LTPA in Chinese adults has dropped from 382MET-h/week in 1991 to 264MET-h/week in 2011 [10]. Examining the association of LTPA preferences and association to actual behavior would provide valuable information for the design of future interventions that focus on preference education as one means to reverse this alarming trend. In addition, a growing number of studies have been focused on the disparities between urban and rural residents in LTPA in China. Specifically, urban adults were more physically active than their rural counterparts during leisure time [11]. Urban adults exercised more regularly, whereas a considerable increase was seen in rural residents who owned televisions [12, 13]. However, there is no evidence showing that LTPA has increased significantly in either urban or rural residents during the last several decades. In consideration of the substantial urban-rural disparities in China in terms of occupational structure, culture, access to recreational and exercise facilities, and the like, this study seeks to assess the association between LTPA preference and behavior in urban and rural residents by conducting separate stratified analyses.

We hypothesized that LTPA preference is predictive of actual physical activity behavior, after adjusting for likely confounding variables. A secondary hypothesis is that, due to the strong degree of labor-intensive farming by rural Chinese, the association between LTPA preference and actual LTPA behavior would be less apparent in rural China than it would in urban China.

Table 1 shows descriptive statistics for all outcome variables and covariates in the whole sample, for urban residents, and for rural residents. Overall, the prevalence of leisure-time physical activity in 2011 was low in our sample (14.1%), and it was much more prevalent among urban residents (20.2%) than their rural counterparts (4.0%). The LTPA preferences of the participants were low in 2004 (mean preference score was 6.16 in the whole sample, and again, the possible preference score is between 3 and 15). Urban residents were more likely than rural citizens to report higher LTPA preference (6.58 vs. 5.45). In ethnic composition of the study, the Han dominated our sample (almost 90% of respondents were of Han ethnicity across different community groups), which is consistent with the national ethnic distribution in China. Urban residents were more likely to be widowed than rural participants (2.7% vs. 1.6%). Education level was low in our sample: specifically, less than one third (30.3%) of the sample received a high school education or higher. The disparity between urban and rural residents in education was substantial. Thirty nine point 2 % (39.2%) of urban residents reported high school or above education, compared to only 15.6% among rural residents. Over one-third (37.7%) of the whole sample was from a rural area (village), while 62.3% were from urban areas (including urban, suburban, and town). Urban residents were less likely to be smokers (31.7% vs. 34.1%), and more likely to drink alcohol (36.0% vs. 32.7%) than their rural counterparts. Urban residents also tended to have higher Body Mass Index (BMI): they specifically were less likely to be underweight (3.1% vs. 5.1%) or have normal weight (49.0% vs. 58.2%), and more likely to be overweight (37.7% vs. 29.3%) and obese (10.9% vs. 7.4%). Urban residents were also more likely to be covered by health insurance than rural people (43.6% vs. 22.3%).

The Spearman correlation coefficients between LTPA behavior (in 2004 and 2011) and LTPA preference in 2004 shows that the time spent on LTPA in 2004 was positively and significantly related to their preference in the same year among the entire sample (coefficient = 0.202, p < 0.001), and among urban residents (coefficient = 0.201, p < 0.001) and rural residents (coefficient = 0.111, p < 0.001). However, the association was stronger in urban residents. Furthermore, the amount of time spent in physical activity in 2011 was positively and significantly associated with physical activity preference in 2004 among the whole sample (coefficient = 0.126, p < 0.001) and urban residents (coefficient = 0.111, p < 0.001), but not among rural residents.

Table 2 shows the multivariate logistic regressions assessing the association between LTPA behavior in 2011 and preference in 2004, after adjusting for age, sex, ethnicity, marital status, community types, region of residence, education, employment status, annual household income, smoking status, alcohol consumption, current health status (self-reported), health insurance coverage, and BMI. LTPA preference in 2004 was a significant predictor of LTPA behavior in the whole sample (OR = 1.05, 95% CI = 1.01–1.09) and urban sample (OR = 1.06, 95% CI = 1.01–1.10), but not among rural residents (OR = 1.04, 95% CI = 0.92–1.17). Age was also significantly associated with higher probability of physical activity in the whole sample (OR = 1.01, 95% CI = 1.00–1.03) and urban sample (OR = 1.02, 95% CI = 1.00–1.03). Married rural residents significantly related to lower probability of physical activity (OR = 0.12, 95% CI = 0.03–0.53) compared to their counterparts who never married. In the whole sample, village (OR = 0.30, 95% CI = 0.20–0.47) and town residents (OR = 0.69, 95% CI = 0.49–0.99) had a lower probability of physical activity. Participants in the whole sample and urban sample with higher education levels tended to be more involved in physical activity. Specifically, an elevated probability of physical activity, compared to participants who never attended school or could not read and write, was associated with those who had attended primary school (OR = 1.92, 95% CI = 1.14–3.25 in the whole sample; OR = 1.81, 95% CI = 1.02–3.22 in urban residents), middle school (OR = 2.53, 95% CI = 1.51–4.23 in the whole sample; OR = 2.46, 95% CI = 1.41–4.30 in urban residents), and high school or above (OR = 3.72, 95% CI = 2.22–6.22 in the whole sample; OR = 3.25, 95% CI = 1.86–5.68 in urban residents) as their highest education level. The associations were diminished in rural residents except for those who had high school or above education, who had a higher probability of physical activity (OR = 9.33, 95% CI = 2.20–39.55). Of note, participants covered by health insurance were more likely to report physical activity (OR = 1.77, 95% CI = 1.35–2.33 in the whole sample; OR = 1.53, 95% CI = 1.14–2.06 among urban residents; OR = 4.07, 95% CI = 1.92–8.60 for rural participants).

The odds ratios and 95% CI of physical activity with and without confounders have been compared. After removing BMI and current health status (self-reported) from multivariate models, odds ratios and 95% CI of physical activity among the whole sample, urban residents, and rural residents did not dramatically change. The result is available upon request.

This study examined the association between LTPA preference and LTPA behavior among a longitudinal sample of adults who participated in the China Health & Nutrition Survey in both 2004 and 2011. The longitudinal sample allowed us to examine the effects of participants’ LTPA preference in 2004 on their LTPA behavior in 2011 by setting the temporal order. Overall, LTPA was not commonly practiced in our sample, which is consistent with previous studies [7, 31]. In addition, the results reveal that LTPA preference was a significant predictor of LTPA behavior, which supports our original hypothesis, but only among the whole sample and urban residents, and not among rural participants.

Preferring LTPA was strongly and positively correlated with performing LTPA. This is not surprising in light of existing research on the causal relationship between food preference and food intake, and between sedentary activity preference and sedentary activity behavior [1‐6]. Those studies specifically show the predicting value of food preference on actual food intake, and causal relationships between sedentary activity preference and lower physical activity level. This study confirms that measuring participants’ preference on LTPA provides evidence of their actual activity level. The results are consistent with the original hypothesis, although future multiple genetic analyses and longitudinal analyses might provide more information on the pathway from preference to behavior.

The urban-rural difference was observed in our study. LTPA was substantially higher among urban residents than their rural counterparts, and urban participants were more likely to prefer LTPA. This could be explained by previous research [11‐13], which shows that urban adults were more physically active during leisure time than their rural counterparts, while rural residents were more involved in TV watching. In addition, the significant association between LTPA preference and activity was found in urban residents, but not in rural participants. A possible explanation could be that the urban residents had more access to facilities and more spaces for physical activity during leisure time than those in the rural areas [32, 33]. Therefore, it would not be surprising to see that rural residents who prefer LTPA might find it impractical to do exercise, thus leading to lower activity level. Alternatively, the unadjusted confounding effects (such as occupational physical activity) might contribute to the association difference. For instance, rural residents may be busy with physical farm-work, and this made leave them with little leisure time and less perceived need for physical activities. Both of these factors could result in an observation of diminished association in rural residents.

The confounding effects of BMI and current health status (self-report) had been assessed by sensitivity analysis, given that a number of existing studies suggest the two variables have a significant impact on physical activity [34‐37]. The difference between before and after removing them was however not substantial, suggesting that the causal relationship between LTPA preference and behavior observed in the study was not confounded by the two variables.

This study has several limitations. First, there was no qualitative data on physical activity related to beliefs about the purpose of performing physical activity, the perceived usefulness of physical activity, the impact physical activity has on other health behaviors, and so forth. Thus, though physical activity preference predicts physical activity behavior, one cannot fully disentangle why and how this is so. Furthermore, the use of self-reported data can potentially introduce recall bias, especially for variables based on participants’ long-term memory; for example, duration of physical activity [38]. In addition, the questions on LTPA preference and behavior were based on different types of activity, which may further obscure their association, especially among rural residents for whom a significant correlation was not found. For example, individual sports activities such as soccer and tennis were surveyed in the behavior section, but sport in general was queried in the preference section. Furthermore, regardless of the substantial statistical significance in the Spearman correlation coefficients, the associations in the analyses were not very strong. This may also be due to the discrepancy of types of activity surveyed regarding LTPA preference and actual behavior. Finally, this study was based on a sample in China, where cultural norms and patterns of LTPA may be unique to the Chinese context, and therefore may limit the ability to generalize our findings to other countries.

Despite these limitations, this population-based study is unique in establishing the causal relationship between LTPA preference and behavior. To the best of our knowledge, this is the first scientific attempt to examine the predictive value of preference for LTPA. In addition, our study used longitudinal data, which allows for inferring causality.

The study has proved the predictive value of LTPA preference on actual LTPA behavior. This may provide researchers an alternative to the traditional physical activity assessment. Changing people’s LTPA preference to increase LTPA may be helpful in chronic disease prevention and control in China, and may be able to be applied to other countries and contexts. In addition, health interventions in rural areas may increase residents’ access to LTPA facilities.