4.1. Advantage of the Cross-Sectional Analysis
Based on 3930 subjects from the cohort study, we examined the relationship between social vulnerability and geriatric depression by using an environmentally adjusted model. We observed that the most significant risk was found in the population who had low educational status.
We reviewed previous epidemiological literature and found that our results are more conclusive than the other studies [6
]. Previous studies found that geriatric depression risk is associated with social vulnerability [6
] but none of these studies used environmental variables as controlling factors, resulting in conflicting results in some studies [6
]. For example, Roberts et al. [6
] found significant geriatric depression risk in population who received low education (OR: 1.62) but Forsell [57
] and Østbye et al. [15
] suggested that there are no significant associations between educational status and geriatric depression. Although these studies somewhat used similar epidemiological designs, the results were not comparable due to the difference in community settings and underlying physical environment. It implies that, without considering environmental factors, their results can only partially portray the actual risk of geriatric depression from social vulnerability.
In addition, biases from community influence could not be normalized, adjusted or resolved, based on their simple model design [59
]. This bias was also found in other health studies when they used either environmentally adjusted or unadjusted models to correlate with different health outcomes. For example, there is a slight underestimation of high education to general health from models unadjusted by green space, compared to the adjusted model [60
]. Therefore, our study can be regarded as an innovative design which comprehensively includes the morphological parameters of the built environment (building density, vegetation, building height, and urban form) in our model of geriatric depression risk.
Another strength of our study is that we determined the relationship between the built environment and geriatric depression risk, and developed a socio-environmental vulnerability index that can be used to obtain the spatial variation of geriatric depression risk. Based on the results, we found that in high-density cities, a small neighborhood with lower buildings and greater variation of building height has positive association with geriatric depression risk, instead of green space which contributes to better mental health in mid- or low-density cities. In high-density cities, the congested living environment is often perceived to have detrimental effect on mental well-being. The present study implies that building density can be balanced by proper urban design. For instance, better design of urban spaces can encourage the use of such spaces and hence provide opportunities for recovery from psychological stresses [19
Results of the present study also provides a more advanced and reliable understanding than previous studies which most of the vulnerability indices were either conceptually developed without validation by health outcome data [35
] or only locally calibrated without environmental adjustment [27
]. Therefore, previous indices either require further proof from public health analysis [43
] or may only be able to represent a local scenario that cannot be universally applied. The present study provides a methodology of developing a socio-environmental vulnerability index that can be applied in high-density cities since both social and environmental factors are considered. The conceptual framework of this study can also be used to develop similar indices for other types of urban environment.
4.2. Relationship between Geriatric Depression and Local Environment
The results of the cross-sectional analysis were then used to construct the socio-environmental vulnerability index. The importance of geographical scale in spatial mapping was widely discussed in previous studies [61
]. In this study, individual responses are capable of representing the socio-economic characteristics of the neighborhood due to the large sample size. TPUs, the mapping units used in this study, are designated for small-scale planning purpose so the built environment is relatively homogeneous in the same TPU. As such, the 400-m buffer used in acquiring the information about the built environment is representative of the TPUs that the subjects reside in. Moreover, the subjects mainly live in urban areas so our mapping results can represent the geriatric depression risk at neighborhood scale in Hong Kong.
The spatial pattern of the socio-environmental vulnerability index was found to be associated with the variations in demographic and socio-economic characteristics. In addition, previous studies suggested socio-economic conditions and access to community facilities were associated with different health impacts [18
]. We found that areas with relatively higher vulnerability are the districts with lower socio-economic status, for example, Sham Shui Po and Kwun Tong [63
]. On the other hand, richer people commonly reside in urban areas with lower vulnerability such as the Peak and Kowloon Tong. Nonetheless, there were independent contributions from the built environment after adjusting for socio-economic factors. Our environmentally adjusted model can therefore take into account the effect of the built environment for better estimation of geriatric depression risk in Hong Kong.
The present study indicates that there is a significant relationship between the built environment and geriatric depression risk and the density of residential buildings, average and variation of building height were identified as significantly influential factors. It highlights the characteristics of the compact living environment associated with low resilience or social vulnerability. In this study, the 25th percentile of average building height within a 400-m radius is approximate to 27.9 m which corresponds to buildings with less than ten floors. In Hong Kong, a particular type of the historical buildings called “Tong Lau” is typically with less than ten stories and located in socially deprived districts such as Sham Shui Po, San Po Kong, and Kwun Tong (Figure 2
). Unlike the low-rise buildings in North America or Europe, “Tong Lau” represents a compact built environment with extremely poor living conditions, in term of building services, social environment, and living quality [64
]. Some flats in these buildings could be split into only 3 to 5 m2
with extremely inadequate facilities. As such, it is expected that older adults living in these tiny flats would become socially isolated and suffer from depression.
In addition, single high-rise buildings are replacing some of the “Tong Lau” as these old urban areas experience redevelopment in recent years (Figure 3
), resulting in higher variation in building height and irregular urban form. Our results indicate that high variation in building height was associated with an increased risk of geriatric depression. This can be attributed to the gentrification process that the middle class starts to occupy these areas. It accentuates social isolation due to the difference in lifestyles and changes in local communities, resulting in higher geriatric depression risk.
The variations in building height imply the current situation of redevelopment in Hong Kong. Single-building redevelopment may result in gentrification in the older urban areas and it does not provide sufficient opportunities to improve social services and facilities in the neighborhood. As for strategic planning of redevelopment in these areas, a holistic approach should be adopted to ensure that social services and facilities, even green space, are in place. It can also echo the concept of ageing in place by taking redevelopment of older districts as opportunities to improve social services and facilities in the neighborhood.
In our study, we found that there was no significant relationship between green space and geriatric depression, which is contradictory to previous studies [17
] and likely due to the collective consideration of urban and rural areas into a single regression model. In Hong Kong, the hilly topography limits the extent of urban areas and results in very significant separation between urban and rural areas. As rural areas in Hong Kong are generally surrounded by country parks and extensive coverage of green space, combining the data of both urban and rural areas into a single model may induce biases to describe the linear relationship between green space and geriatric depression. Further studies can be conducted by using non-linear models and machine learning techniques for environmental health studies [66
] to determine the relationship between green space and geriatric depression so that the biases of location and selection can be avoided. Nonetheless, we found that the availability of green space was coincidentally lower in more socially vulnerable areas such as Sham Shui Po, Wong Tai Sin and Kwun Tong (1.9, 1.7 and 1.3 m2
per each older adult, respectively), compared to 3.4 m2
of green space per each older adult in Hong Kong [68
]. Such a high percentage of socially deprived population in these old urban areas contributes to psychological distress, which in turn increases the risk of developing depression [69
4.3. Connection with Local Geriatric Studies
Consistent with previous studies in Hong Kong [70
], low education was found to be associated with depressive symptoms in this study. Older adults with lower education level were likely to be depressed, indicating the importance of family relations and social support in protecting older adults from depression [70
]. Therefore, strategies should be developed to facilitate social connectedness. Building social networks for older adults in vulnerable areas could be useful to prevent depression among the senior population.
Early identification and management of older adults with potential depression risk is important to reduce the risk of morbidity, disability and mortality. In Hong Kong, the prevalence of depressive symptoms/depression in the senior population is common [70
]. However, these conditions are generally unrecognized and masked in patients with cognitive impairment or being seen as less influential. According to Lam et al. [71
], only 26% of the individuals with common mental disorders in Hong Kong consulted mental health services in the past years while less than 10% of the individuals consulted general practitioners and family physicians. Mental health services in Hong Kong were especially inadequate [74
], as reflected by the long waiting time (81 weeks) for a first appointment at a psychiatric out-patient clinic in 2013 [75
], which may cause further geriatric depression risk to the senior population [74
]. Our vulnerability map locating the areas with high geriatric depression risk estimates the levels of depression and provides useful information that can be used to guide resource allocation for mental health services.
One limitation of this study is that we do not have sufficient data to analyze shifts of geriatric depression risk through time. While we normalized or adjusted the model to a setting with “same living environment”, understanding the shift will improve the flexibility of applying the socio-environmental vulnerability index. It is particularly useful for the planning of public health policy which targets long-term mitigation. The present study, nonetheless, estimates the spatial variation of the socio-environmental vulnerability under the typical scenario, which is important to preventive healthcare and emergency response [25
]. In order to combine the advantages of both types of studies, future studies should focus on repeatedly re-collecting information of health outcomes from the participants of the cohort study and analyzing the geriatric depression risk through time based on the health outcome data collected at different times.
Another limitation of our study is that our cohort data were mostly obtained from urban areas, which induced very different results in extremely rural areas. Although our results are consistent with previous studies that the highest vulnerability in Hong Kong was found in rural areas with limited access to transportation and community facilities [18
], some rural areas exhibited extremely low values based on the estimation of the socio-environmental vulnerability index. It is partially because our model only included the average and variation of building height and the results were largely determined by social variables due to the lack of building data in the rural areas. As social vulnerability highly contributes to depression risk, the spatial differences in the socio-economic structure between rural areas are relevant to the geriatric depression risk in Hong Kong.
In this study, we followed the methodology of previous studies for the development of vulnerability indices, by first examining the relationship between health risk and individual-level socio-economic status, and further applying the results with area-level data for mapping. This approach provides better ability for cohort data and health surveys to facilitate the biostatistical analyses, while it may also have issue of ecological fallacy because of the unclear relationship between individual-level data and area-level information. In this study, based on the assumption that the characteristics of the built environment in the neighborhood are homogeneous at TPU level, individual data are representative of the neighborhood conditions. Future studies can be conducted to compare the individual-level and area-level data for the representativeness to improve the development of local vulnerability indices. In addition, models for casual relationship may be an option for index improvement if comprehensive local data are available for such studies.
In addition, we did not include certain factors of environmental deprivation such as extreme temperature and air pollution in this study [77
]. However, there are uncertainties in using fine-scale mapping of temperature and air pollution in high-density urban settings [79
], which may introduce random errors to the cross-sectional model. For future studies, more information about the spatial variation of temperature can be obtained using satellite images and numerical modeling of energy fluxes in urban settings [80
], while the spatial variation of air pollution can be obtained through remote sensing techniques and land use regression models [81
]. These approaches provide high-quality spatial datasets that could be used for cross-sectional analyses to adjust for different factors of environmental deprivation in the future.
Finally, this study provided a framework of applying data-driven techniques to map geriatric depression risk across a city. However, variables that we used in this framework may not be applicable in other cities. For example, variation in building height is a specific variable to predict geriatric depression risk in Hong Kong due to the historical urban development of the city. This variable may not be commonly useful across other cities with similar sizes if these cities have been practicing different land use policies for regional development. Therefore, variables should be specifically chosen for the development of local vulnerability indices that can be applied to other cities.