Neighborhood’s locality, walkability, and residents’ multimorbidity: Evidence from the China’s middle-aged and older population

Background Neighborhood factors have gained increasing attention, while the role of neighborhood’s walkability and location have not been clarified in multimorbidity. In this study, we estimated the prevalence of 14 non-communicable chronic disease (NCDs) and depicted variations in the number of NCDs as a function of road type, urban-rural settings, neighborhood characteristics, and individual confounders. Methods Data came from China Health and Retirement Longitudinal Study 2011 National Baseline Survey. Negative binomial regression with clustered robust standard errors was employed to analyze variations in the number of NCDs among 13,414 Chinese middle-aged and older population. Results First, over 65% subjects had at least one NCDs, and over 35% had multiple NCDs. Arthritis (33.08%), hypertension (24.54%), and digestive disease (21.98%) were the most prevalent NCDs. There existed no urban-rural differences in multimorbidity after adjusted for neighborhood clustering variations. Lastly, living with paved road was associated with fewer NCDs relative to living with unpaved road. Conclusion Findings suggest that urban-rural disparities in the number of NCDs appeared to result from within-neighborhoods characteristics. Living with walkable road is important for middle-aged and older population.


Abstract
Background Neighborhood factors have gained increasing attention, while the role of neighborhood's walkability and location have not been clarified in multimorbidity.
In this study, we estimated the prevalence of 14 non-communicable chronic disease (NCDs) and depicted variations in the number of NCDs as a function of road type, urban-rural settings, neighborhood characteristics, and individual confounders.

Methods Data came from China Health and Retirement Longitudinal Study 2011
National Baseline Survey. Negative binomial regression with clustered robust standard errors was employed to analyze variations in the number of NCDs among 13,414 Chinese middle-aged and older population.
Results First, over 65% subjects had at least one NCDs, and over 35% had multiple NCDs. Arthritis (33.08%), hypertension (24.54%), and digestive disease (21.98%) were the most prevalent NCDs. There existed no urban-rural differences in multimorbidity after adjusted for neighborhood clustering variations. Lastly, living with paved road was associated with fewer NCDs relative to living with unpaved road.
Conclusion Findings suggest that urban-rural disparities in the number of NCDs appeared to result from within-neighborhoods characteristics. Living with walkable road is important for middle-aged and older population.

Background
Multimorbidity is the coexistence of two or more chronic diseases, which has gained increasing attention especially in the context of population aging (Banister 2012).
Relative to younger generation, population aged over 60 has a higher proportion with chronic kidney disease (Zhang 2012), stoke (Teh et al. 2018), and heart disease (Moran et al. 2008), which are all major causes of deaths of the elderly (Couser et al. 2011). In particular, patients with multimorbidity have a higher probability of premature death (Menotti et al. 2001), longer hospital length of stay (Vogeli et al. 2007), functional disability (Fortin et al. 2004), and mental problems (McLean et al. 2014). Care for patients with multimorbidity is challenging, since most of current clinical guidelines are specific for single chronic diseases (Boyd et al. 2005;WHO 2016). In addition, multimorbidity requires patients to take multiple medications and have higher medication adherence, which is more difficult for older patients.
Prior observations have explored several individual-level attributes associated with multimorbidity. First, prior investigators note that multimorbidity is increasing with age (McLean et al. 2014;Violan et al. 2014). Women appear to have higher odds for multimorbidity relative to men . Lifestyle behaviors including smoke history (Menotti et al. 2001) and physical exercise (McLean et al. 2014), physiological attributes such as mental distress (McLean et al. 2014) and body mass index (Menotti et al. 2001), socioeconomic status like income and education status ) are considered to be potential indicators of multimorbidity.
A fair number of observations suggest that urban residents have higher odds for being multimorbidity such as studies from Spain , Myanmar (Aye et al. 2019), and Korea (Yi et al. 2019). In contrast, a number of observations did not find the association between urban-rural disparities and multimorbidity such as a previous study from Portugal (Prazeres and Santiago 2015), and some studies even indicate that rural residents have higher probability of multimorbidity such as the one from Canada (Roberts et al. 2015). There are two relevant studies based on Chinese. One study did not directly introduce urban vs.
rural settings in the model, thereby being unable to investigate urban-rural disparities after adjusting for variations in environment and socioeconomic inequality (Yi et al. 2019). In the other study, results suggest that urban residents have higher odds for multimorbidity, while investigators failed to include several important individuals' characteristics such as physical activity and BMI as well as environment variations in the study (Garin et al. 2015). Whether the urban-rural disparities indeed exist in China after controlling for neighborhood disparities and individuals' attributes is unclear.
Neighborhood factors that may affect health status have gained increasing attention. Hale et al. (2013) suggest that individual's perceived neighborhood quality such as perceptions of crime, litter, and pleasantness in the neighborhood may be associated with health status, and the link could be partially mediated by sleep quality. Steptoe and Feldman (2001) have demonstrated that lower neighborhood scores may be associated with psychological distress. Neighborhood's environmental attributes such as greenness and open space are recognized as space for walking for recreation and social coherence, which may lead to better mental and physical health (Sugiyama et al. 2008;Sugiyama et al. 2009). Other neighborhood attributes including socioeconomic status (Bethea et al. 2016;Bosma et al. 2001;Steptoe and Feldman 2004), access to public transport (Cummins et al. 2005), political climate (Cummins et al. 2005), street noise (Parra et al. 2010), and safety (Johnson et al. 2009;Ou et al. 2018) have been considered to be associated with residents' health status. Despite a significant number of studies focusing on the impact from neighborhood's environment and security on individuals' health (Cummins et al. 2005;Sugiyama et al. 2008;Sugiyama et al. 2009), a limited number of studies have investigated the association between walkability and older population's health status as a reflection of non-communicable chronic disease, care of which may be tremendously affected by neighborhood's walkability and access to health care.
To bridge the gaps, this study analyzed variations in health status for Chinese middle-aged and older population. Here, we offered two hypotheses for discussion.
First, urban subjects may have more NCDs relative to rural subjects since they may live unhealthy life style and have a higher probability of exposure to air population (Steyn et al. 1997). Second, living with walkable road type, such as paved road, may be significantly associated with fewer NCDs since walkability may affect residents' access to health care particularly for population with limited mobility such as the elderly (Satariano et al. 2012).

Data source
We derived data from China Health and Retirement Longitudinal Study (CHARLS) 2011 National Baseline Survey, which recorded data on 17,708 Chinese aged over 45 from 28 provinces in mainland China (Yang et al. 2012). Information regarding sampling, recruitment, response rate, and procedures for data collection could be retrieved from prior study (Yang et al. 2012). Excluding 4295 observations with missing data, we derived data on 13,413 subjects from 432 neighborhoods. In our definition, neighborhoods refer to villages in rural areas and communities in urban areas.

Measures
To measure health status, the overall number of non-communicable chronic diseases was calculated as the dependent variable. Whether a subject had ever been diagnosed with the fourteen non-communicable chronic diseases was recorded in the survey, the 14 non-communicable chronic diseases including hypertension, dyslipidemia, diabetes, cancer or malignant tumor, chronic lung diseases, chronic liver disease, heart problems, stroke, chronic kidney disease, stomach or other digestive disease, mental problems, memory-related disease, arthritis or rheumatism, and asthma. Data were derived from answers to the question "Have you been diagnosed with the following 14 NCDs?". In addition,the data on hypertension, chronic lung disease, and mental problems also included answers to the question "Do you know if you have hypertension, chronic lung disease, and mental problems, respectively?".
Neighborhood's walkability was reflected by road type. In this study, there were four types of road including unpaved road, paved road, sand-stone road and others.
Urban versus rural setting was introduced to measure neighborhoods' locality and urbanization.
The number of primary care institutions in the neighborhood (community health centers, community health care medical posts, township health clinics, and village medical posts) were obtained to measure residents' access to primary care since prior studies have documented that access to health care resources could be associated with population health (Autier et al. 2011;Chen et al. 2010;Cossman et al. 2017). Data were derived from the question "How many community health centers, community health care medical posts, township health clinics or village medical posts in the village or community?". Last, water sanitation (groundwater system) was introduced to reflect neighborhood' living conditions. Individual-level confounders including age, sex, marital status, education status, household income, body mass index (BMI), exercise, and health care insurance were introduced as covariates (Haas et al. 2003;Kuh et al. 2005;Trani et al. 2011).
Health care insurance was classified as uninsured, rural cooperative medical insurance (RCMI), and others including business medical insurance, Urban Residents Medical Insurance, and Urban Employees Medical Insurance due to a limited number of subjects with the last three types of medical insurance in CHARLS.

Statistical analysis
First, we stratified study subjects into urban and rural groups to exam the distribution of the baseline characteristics (independent variables). Statistic tests including t test, Mann-Whitney U test, and Chi-squared test were employed according to data characteristics. The analysis was used for in-sample interpretations; thereby CHARLS survey weights were not used (table 1). For multimorbidity and prevalence of each NCD, CHARLS sampling weights were used for interpreting results as China's population representative parameters (table 2).
Next, negative binomial regression was employed instead of Poisson regression, since the dependent variable's variance was larger than its mean value. Univariate analysis was performed to examine disparities of NCDs as a function of each independent variable. Multivariate negative binomial regression analysis was employed with all covariates (Model1 in table 3). Clustered robust standard errors were generated in the model 2 to take individuals nested within neighborhoods into account. Variance inflation factors were calculated to exam collinearity among independent variables, which suggested slight collinearity. Models' significance was examined by Pearson chi-square.
In addition, we undertook sensitive analysis by performing a multinomial logistic regression with 5 responses (Y = 0, 1, 2, 3, and > or = 4) with robust standard errors. Results were qualitatively similar with those from negative binomial regression.
Statistical analyses were performed with Stata/SE 15.0 (StataCrop, TX, USA). A twotailed P-value of less than 0.05 was considered statistically significant. Table 1 presents the baseline descriptive characteristics for participants. We had 5639 urban subjects and 7774 rural subjects. Of the 13,413 participants, 3104 subjects (23.14%) lived with unpaved roads with 1001 subjects from urban areas (17.75%) and 2103 from rural areas (27.05%). There were 7.87% urban subjects and 13.07% rural subjects lived with sand-stone roads, and 74.00% urban subjects and 59.09% rural subjects relied on paved roads. Results from statistical test suggest that urban-rural disparities existed in road type (table 1). In addition, 24.05% subjects lived in neighborhoods without primary care institution, while 14.17% subjects lived with over three primary care institutions (table 1). Furthermore, 29.21% subjects lived with groundwater system, while over 70% subjects lived without groundwater system (table 1). A higher proportion of urban population (41.66%) lived with groundwater system relative to rural population (20.18%), which was statistically significant (P < 0.001). Last, individual-level attributes including household income, BMI, and exercise varied across urban and rural subjects (table   1). <Table 1 about here> According to weighted analysis in table 2, over 65% study subjects had at least one NCDs, followed by 19.11% with two NCDs, 9.81% with three NCDs, and around 10% with at least 4 NCDs. The top 3 NCDs were arthritis or rheumatism (33.08%), hypertension (24.54%), and digestive disease (21.98%). In contrast, cancer, memory-related disease, and mental problems were the least prevalent NCDs. There were 0.90%, 1.45%, and 1.78% subjects self-reporting to have cancer, memoryrelated disease, and mental problems, respectively. Furthermore, there existed urban-rural disparities in the prevalence of hypertension, heart problems, dyslipidemia, diabetes, and asthma in unweighted statistical tests. Urban subjects were more likely to have hypertension, heart problems, dyslipidemia, and diabetes, whereas rural subjects had a greater probability of asthma (P < 0.05). After controlling for CHARLS sampling weights, there existed no urban-rural disparities in the prevalence of each NCD since the 95% CI of the proportion in urban residents overlapped with that in rural residents across all NCDs. <Table 2 about here>

Discussion
By employing a national cohort of China's middle-aged and elder population, this study depicts the prevalence of multimorbidity and 14 NCDs among Chinese middleaged and older population. Although prior studies have focused on the association between residents' walkability and mental health as well as quality of life (Sugiyama et al. 2008;Sugiyama et al. 2009), this study contributes to literature around the association between walkability and the middle-aged and the elderly's health status reflected by the number of NCDs, care of which may require pedestrian-friendly road to reach health services as well as social coherence. In contrast, we found that there existed no urban-rural disparities in residents' multimorbidity after controlling for within neighborhoods' variations.
Our results call for attention directed on non-communicable diseases in China.
Relative to other society, the situation of non-communicable diseases in China is more urgent for its demographic shift and large population. According to our analysis, over 65% middle-age and the elderly had at least one NCDs and about 35% subjects had comorbidity, whereas prior observation in Canada concluded that around only 20% aged from 45 to 64 had comorbidity (Pefoyo et al. 2015). Results show that arthritis or rheumatism was the most prevalent NCDs, followed by hypertension and digestive disease. However, our study only includes observed NCDs, while neglecting those unobserved. One prior study, which analyzed data from medical questionnaire in CHARLS and included unobserved hypertension, suggested that there were around 40.9% subjects with hypertension (Feng XL 2014). Therefore, it is unwarranted to compare the prevalence of the 14 NCDs due to the exclusion of unobserved NCDs. However, results that Chinese middle-aged and elder population have a significant proportion with arthritis or rheumatism are consistent with a prior study (Zhang et al. 2001). Even though arthritis may not contribute to disease burden as considerably as hypertension and diabetes do (Yang et al. 2008), suffering from arthritis significantly impact elderly's mobility and quality of life (Badley and Wang 1998;Husted et al. 2001).
Our results offer some clues for urban-rural disparities in multimorbidity. Although urban population have a greater probability of encountering more NCDs relative to rural population (univariate analysis and model 1 in table 3), the differences appeared to be fully explained by within-neighborhood confounding. One possibility is that those with higher risk of multimorbidity, such as white collars that undertake greater life pressure, tend to be nested in several neighborhoods in urban cities.
They may prefer to live closely to offices in order to save time to work, thereby forsaking their living conditions. It's unwarranted to compare our results (model 2 in  With reference to the second hypothesis, results confirm that living with walkable road was positively associated with individual's health status (table 4). Results here support findings from prior studies that living with pedestrian-friendly road type may be associated with the elderly's social coherence, mental health, and physical activity (Stafford et al. 2007;Sugiyama et al. 2008). While the other possibility is that walkability may link with access to primary care for the elderly. With a given number of primary care institutions in the neighborhood, population living with walkable road type had a lower probability of more NCDs (table 4), which is particularly important for patients with limited morbidity. For arthritis patients, the middle-aged, and the elderly, their morbidity may be limited and decrease over time (Satariano et al. 2012). Therefore, neighborhoods' walkability may affect their access to primary care, which may indirectly affect their probability of multimorbidity.
Three implications for policy makers or future analysis could be drawn from this study. First, although the prevalence of NCDs among Chinese older population has been extensively discussed, most of these studies focused on hypertension and diabetes (Yang et al. 2008). Therefore, we suggest that local government should pay more attention to the distribution of arthritis or rheumatism-related health resources and health education program. In addition, results in terms of urban-rural disparities were not consistent across models, which suggest that urban-rural disparities may link with within-neighborhood characteristics. Future studies may consider to take within-neighborhood variations into account. Last, our study suggests that government should consider local population's demographic characteristics and health demand, for example, the proportion of the elderly and their mobility. For population in transportation-disadvantaged regions such as rural areas in China's Tibet, Xinjiang, and Sichuan province, the distribution of health care resources should consider not only population density but also geographic attributes including land area, road type, and access to public transportation.
This study employed a good representative sample of Chinese middle-aged and older population with a large sample size. We depicted the prevalence of the 14 NCDs among Chinese middle-aged and older population. We analyzed variations in NCDs as a function of neighborhood attributes as well individual characteristics. We generated robust standard errors in multivariate models to capture withinneighborhood variations.
Despites the strength, this study has several limitations. First, the cross-sectional study design limited the potential to offer strong evidence for causality. Second, this study employed self-reported disease status, which may yield reporting bias and recalling bias. Moreover, CHARLS only recorded information of the 14 chronic diseases; thereby this study may underestimate the prevalence of multimorbidity. In addition, due to unavailable data, this study did not control for other living environment attributes, such as safety and green space, which may affect dwellers' health as well (Johnson et al. 2009;Sugiyama et al. 2008