Associations of environment and lifestyle factors with suboptimal health status: a population-based cross-sectional study in urban China

We conducted a multi-city cross-sectional survey using a four-stage stratified sampling method from September 2017 to September 2018. In the first stage, we selected Guangdong Province (Southeast China), Harbin City, Heilongjiang Province (Northeast China), Sichuan Province (Northwest China), Tianjin City (East China), and Lanzhou City, Gansu Province (Southwest China) as representatives of different Chinese regions based on their geographic distribution, economic characteristics, and populational demographics. The second stage included sampling of 3 ~ 5 representative cities in each province based on demographic, economic, and geographic factors of which two cities were randomly selected in the next stage, respectively. In the final stage, 1 ~ 3 streets were randomly selected from each city and the residents were selected using sampling method, who were administrated questionnaires. To ensure representativeness, the participants on each street were stratified by male and female respondents and age (i.e., brackets of 14–24, 25–34, 35–44, 45–54, and 55+). As such, survey participants were representative of the level of SHS in their respective urban areas.

Oral informed consent was obtained from each participant prior to the data collection. This consent was deemed sufficient as participants volunteered participation and were told they still could withdraw. All data were kept strictly confidential. This experiment has obtained approval of the Ethics Committee of Nanfang Hospital (Approval number: NFEC-2019-196).

This study used an SHS questionnaire to investigate urban Chinese residents. It contained two sections (i.e. both a self-designed and standardized questionnaire) [14]. The self-designed questionnaire asked for general demographic characteristics as well as information on lifestyle and environment. Here, the demographic characteristics included age, gender, and marital status and there were ten lifestyle variables (i.e., smoking, second-hand smoke, alcohol consumption, bad dietary habits, breakfast consumption, sun exposure, physical exercise, early bedtime (before 11 pm), sleep duration, and surfing the internet), and eight environmental variables (i.e., air quality, noise, housing conditions, living conditions, neighborhood harmony, fitness facilities, and supporting facilities). All variables were reported themselves in recent 3 months. The standardized components were based upon the Sub-Health Measurement Scale V1.0 (SHMS V1.0), which were designed by our research group to assess participant health status [20]. Uniform instructions were provided by trained investigators. Each participant was asked to complete the questionnaire in approximately 25 min.

Health-status assessments were performed in accordance with the SHMS V1.0. Testing procedures revealed that it had high reliability and validity (Cronbach’s α and split-half reliability coefficients of 0.917 and 0.831, respectively) [20]. The SHMS V1.0 consists of 39 items in total. Respondents were asked to answer each of these items according to a 5-point scale (1 to 5, from very bad to very good). The SHMS V1.0 was used to assess general SHS (GS) based on three symptom dimensions, including physical SHS (PS), mental SHS (MS), and social SHS (SS). Of the 39 items, Nos. 4–12, 15, 20–25, 28, and 38–39 were reverse scored (six plus the original score). The original subscale score was the sum of all items: higher scores indicated better health status. We calculated and analyzed transformed scores to further understand and compare the data. Transformed scores were determined using the formula: (original score - theoretically lowest score) / (theoretically highest score - theoretically lowest score) × 100. Following our previous study, SHS prevalence was calculated based on transformed scores [21].

Considering that difference provinces may have different rate of sub-health, the generalized linear mixed models (GLMM) were established to analyze the group effect of sampling areas. The intra-class correlation coefficient (ICC) close to 0 and 95% confidence interval (95%CI) indicated no significant group effect and general regression model rather than multilevel model, suggesting that the GLMM model could be used in the association analysis. A general linear model was used to analyze the association of lifestyles (environmental factors) and SHS as adjusted by demographic characteristics and environment factors (lifestyles). Furthermore, a path model of latent variables was constructed based on a hypothesized relationship between items. Structural equation modeling (SEM) was then used to analyze the complexity of associations between lifestyle factors, environment factors, and SHS to estimate model fitness and analyze the direct and indirect effects of the multiple factors used in the hypothesized model [22]. Model fitness was assessed using the five indices commonly applied in SEM analyses (i.e., the relative X² (CMIN/DF), root-mean-square error of approximation (RMSEA), comparative fit index (CFI), goodness-of-fit index (GFI), and adjusted goodness-of-fit index (AGFI)) [23]. The bootstrapping method [22] of repeat sampling (i.e., 2000 times) was applied to verify statistical significance and calculate confidence intervals for the direct, indirect, and total effects (P < 0.05). All statistical analyses were conducted using (SPSS Statistics version 20.0, SPSS Inc., Chicago, IL). Two-tailed p-values < 0.05 were considered statistically significant.

This study surveyed 6750 urban Chinese residents aged 14 year old or more who had lived in an area for the preceding over 6 months. A total of 807 participants who had become ill during 1 month of the study period were excluded. As such, 5943 respondents were either healthy or SHS for at least a period of 1 month prior to the study. However, 62 of these surveys had missing values for lifestyle, environment, and/or SHS items, and were thus excluded. Therefore, 5881 urban residents were finally included in the current study (a valid response rate of 87.13%).

Table 1 presents overall baseline, lifestyle, and environmental characteristics. The participants included 2817 males and 3064 females with a mean age of 40.27 ± 15.69 years. Most participants were married (64.50%). Furthermore, 66.7% were in GS, 67.0% were in PS, 65.5% were in MS, and 70.0% were in SS.

The mean standard deviation (SD) transformed scores for GS, PS, MS, and SS were 67.23 (12.03), 71.08 (12.70), 67.04 (14.63), 61.47 (15.65), respectively. The GLMM model analysis found that the group effect of provinces investigated was insignificant in the analysis of overall sub-health (ICC = 0.019, 95%CI: − 0.018 - 0.057), physical sub-health (ICC = 0.028, 95%CI: − 0.024 - 0.08), psychological sub-health (ICC = 0.011, 95%CI: − 0.011 - 0.033) and social sub-health (ICC = 0.016, 95%CI: − 0.017 - 0.048). So, the general linear model could be an appropriate measure of the association between lifestyle and environment factors with SHS.

Association between each lifestyle factor and SHS was adjusted by other lifestyle behaviors, demographic characteristics, and environment factors (Table 2). Participants who never smoked, had good dietary habits, consumed breakfast daily, did daily physical exercise, and slept 7–9 h per night had the highest GS, PS, MS, and SS transformed scores. Participants who were not exposed to second-hand smoke and never consumed alcohol had the highest GS, PS, and MS transformed scores. Participants with sufficicent sleeping (bedtimes before 11 p.m.) had the highest GS and PS transformed scores. Sun exposure was only associated with PS; the highest PS scores were found in the people with 14 h or more of sun exposure each week. Surfing the internet was only associated with MS; those who surfed less than 3 hours a day had the highest scores.

Table 3 shows the effects of environmental factors after adjusting other environment factors, demographic characteristics, and lifestyle behaviors. Pleasant housing, harmonious neighborhoods, and convenient living conditions were positively associated with GS, PS, MS, and SS. Urban green space was positively associated with GS and SS. We also observed positive associations between fresh air and GS, PS, and MS. The presence of many fitness facilities was positively associated with GS and MS. There were no significant associations between noise and SHS. However, people with spacious homes had lower GS and MS scores.

The total associations of lifestyle behaviors and environment with SHS were analyzed through SEM (Fig. 1). Although noiseless areas were not independently associated with SHS, the model was not deemed fit without an environmental noise component. The SEM thus included a “noiseless” variable. Except for the CMIN/DF, CFI, and AGFI of Model 1 and the CMIN/DF of Model 2, Table 4 presents information about the fitness measurements for all four models. The associations of lifestyle and environment factors with SHS are listed in Table 5. The data demonstrated that unhealthy lifestyle had significantly negative effects on GS, PS, MS, and SS, while good environment factors had a positive impact (P < 0.001). Lifestyle had the largest effect on PS (β = − 0.418) and the least effect on SS (β = − 0.274). On the other hand, environment factors had the largest effect on MS (β = 0.286) and the least effect on PS (β = 0.225). As the influencing effects were standardized, GS had a larger association with lifestyle (β = − 0.371), but less with environment (β = − 0.282); physical health was more associated with lifestyle (β = − 0.418), but less associated with environment (β = − 0.225). The associations of lifestyle behaviors and environment were similar for MS and identical for SS.

This population-based study examined a sample of urban residents (5881 respondents), thus facilitating overall generalizability to the entire urban population in suboptimal health prevention in China. Furthermore, we illustrated the relative strengths of lifestyle behaviors and environment factors on the associations with SHS. We firstly illustrated the important association between environment factors and MS and SS, which had almost the same association with lifestyle behaviors. Furthermore, associated factors were examined comprehensively, including ten lifestyle behaviors and eight environment factors, which can be intervention targets and would be helpful for preventing SHS and NCDs.

First, because of the cross-sectional design it was not possible to confirm causal relationships of SHS with lifestyle behaviors and environment factors. Second, lifestyle factors and environmental variables were self-reported in this study, which may have potential bias and affect the accuracy of the measurement. Third, environment factors included in this study were all life-related, and other environment factors hadn’t been included. What’s more, although we have considered as many factors as possible, bias would inevitably occur because of certain factors not being included.