The Impact of Built and Social Environmental Characteristics on Diagnosed and Estimated Future Risk of Dementia

INTRODUCTION

Dementia has a significant impact not only on individuals but also on their carers, communities, and health and social care systems, with the global economic cost of dementia care estimated at US$1 trillion per year [1]. There is no disease modifying treatment for dementia; therefore, focus is on dementia prevention, intervention, and care [1]. Although genetic factors are important for Alzheimer’s disease [2–4], there is evidence that 40% of dementia cases potentially related to modifiable risk factors [1]. In addition to lifestyle and cardiovascular risk factors, there is now evidence that air pollution is a major risk factor for dementia [1] and emerging evidence of other potential social and environmental factors although several candidate risk factors have not yet been fully evaluated such as public open space, and level of social fragmentation [5–8].

There is evidence that people living in a neighborhood with more public open space had slower cognitive decline [9]. Further, green spaces—a specific type of open space—have a restorative effect, improving mental alertness and enhancing recovery from stressful experiences [10]. They can also provide attractive settings to build capacities through physical and social recreation [11]. A growing body of epidemiologic evidence has identified urban green space as a potential modifiable factor that can improve human health, especially mental health. However, we know little about whether urban green space exposure is associated with reduced risk of dementia in primary care settings [12, 13].

Air pollution is another aspect of the built environment that has been linked with dementia [1]. Although air pollution has both natural and anthropogenic sources, exposures to traffic-related air pollutants, for example, are generally highest in dense urban areas. Air pollution has been a focus of several studies on cognitive decline and dementia risk [14]. A Canadian study reported that living near busy roads and traffic could increase the risk of dementia [15]. Further, an Australian systematic review study highlighted that increased exposure to airborne pollutants is related to increased dementia risk [16]. There is a need for further research into effect of neighborhood level air pollution on dementia risk across communities [1].

Since physical inactivity is a risk factor associated with dementia [1], it is plausible that environments that support physical activity may also reduce dementia risk. Walkability is a commonly used measure of the built environment that has demonstrated relationships with increased walking and physical activity in adults [17, 18]. A study by Cerin et al. found that objectively determined indices of neighborhood walkability were associated, cross-sectionally or longitudinally with brain imaging markers of better cognitive health [19]. However, this important study was not based on a representative sample of the population and was relatively small. There is now a need to extend this work to examine environmental associations with dementia risk on larger population-based samples and to consider a wider range of potential risk factors.

Further, characteristics of the social environment such as social isolation and neighborhood level of social fragmentation may have an effect into dementia risk. There is some evidence that social isolation is associated with poor cognitive outcomes [20–22], but this evidence is primarily based on observational studies of older adults, with insufficient evidence from interventions to support recommendations about social engagement in the WHO Guidelines on Risk reduction of cognitive decline and dementia. Importantly, we do not yet know whether neighborhood level social fragmentation is related to areas of high risk of dementia across communities.

The ANU-ADRI is an evidence-based, validated tool aimed at assessing individual exposure to risk factors known to be associated with an increased risk of developing Alzheimer’s disease in late-life, that is, over the age of 60 years [23]. The ANU-ADRI is intended to provide a systematic individualized assessment and estimates on Alzheimer’s disease risk factor exposure. The estimated dementia risk score may be useful for individuals who wish to know their risk profile and areas where they can reduce their risk. It may also be useful to clinicians who would like their patients to record their current risk profile for discussion at their next medical appointment. The ANU-ADRI is also used in research projects that aim to evaluate methods of reducing risk of Alzheimer’s disease.

This study aims, for the first time, to investigate: 1) the associations between built and social environments’ characteristics (i.e., public open spaces, air pollution, density of busy roads, walkability, and social fragmentation) with diagnosed cases of dementia and the estimated dementia risk in a primary health care setting using general practice clinical data; 2) spatial variation in dementia risk and diagnosed cases of dementia is investigated across neighborhoods (SA1s) in the study area.

METHODS

Outcome measures

Diagnosed dementia

Diagnosed cases of dementia recorded by general practitioners after undertaken complete investigation using standard and validated screen tools, pathology and brain imaging and using the International Classification of Primary Care (ICPC) which is the most widely used international classification for systematically capturing and ordering clinical information in primary care [26]. In primary care dataset, P70 code of ICPC codes were used to classify patient with diagnosed dementia. The diagnosed dementia cases in the current study include those patients who diagnosed with dementia within the study period and those who diagnosed with dementia prior to study time and had visited GPs within study period.

Dementia risk

The individual life-long dementia risk score was estimated using the Australian National University-Alzheimer’s Disease Risk Index (ANU-ADRI) tool, assessing both risk and protective factors [23]. Seven risk factors were considered, including age, sex, body mass index (BMI), blood cholesterol, smoking status, type 2 diabetes, and depression and three protective factors including physical activity, social engagement (marital status was used as proxy for social engagement), and alcohol intake [5]. The value of dementia risk score ranged from –3 to 49 with SD of 10.6 at individual level.

Figure 2a and 2b are representing the high-high clusters (hot spots) and low-low clusters (cold spots) of the estimated dementia future risk and the diagnosed cases of dementia using the local Moran’s I method at the study area.

OPEN IN VIEWER

Fig. 2

Spatial clusters of the estimated future risk of dementia and diagnosed cases of dementia.

RESULTS

The sociodemographic characteristics of the patients are described in Table 1. In the GP practice sample of patients 65 years and over, 99.6% of gender data were available and indicated 56.10% of the population were women. Of the 37.5% who completed data on marital status, 51.5% were married. Of the 65.2% who reported smoking status, 8.2% were smokers. BMI data were available for 39% patients in which 75.5% were overweight or obese and of the 69.73% of data completed for blood pressure, 44% were recorded as hypertension. Diagnosed dementia and estimated dementia risk rose as age increased and for women. Married participants had slightly more diagnosed dementia and risk compared to singles. Dementia was prevalent in smoker, and underweight groups. Patients who were diagnosed as hypertensive had a higher dementia risk. The level of missingness impacts the estimated risk score and our estimated risk score slightly is under-estimated when the number of missing variables increases (Supplementary Table 2).

We observed weak to moderate correlations between the SA1-level built and social environment measures (Fig. 1), most notably between social fragmentation and socio-economic status (SES). SA1-level NO₂, social fragmentation. The correlations between all exposures were statistically significant except for association between social fragmentation (mobility component), public open spaces and busy road length.

OPEN IN VIEWER

Fig. 1

Correlations matrix between SA1-level characteristics. All relationships are significant at < 0.05. SF_fam, social fragmentation family structure; SF_mob, social fragmentation mobility structure; NO2, nitrogen dioxide; SES, socio-economic status; POS, public open spaces; LBR, length busy road.

Summary statistics on the exposure and outcome variables are provided in Table 2. The estimated dementia risk ranged from –3 to 49 (mean = 7.5. SD = 10.2). The observed mean for dementia risk and public open spaces was higher than the median (median = 3) representing a positively skewed distribution.

Table 3 shows the odds ratios (ORs) for any versus zero diagnosed dementia cases and estimated dementia risk. Focusing on the full model, a 1-SD increase in NO₂ was associated with any (versus no) diagnosed dementia, with increased odds of 10% (95% CI: 1%, 21%). Contrary to our hypothesis, a 1-SD increase in walkability score was associated with increased odds of 10% (95% CI: 1%, 22%). Associations for the social fragmentation index (family and mobility components), dwelling density, busy roads, public open spaces, and socio-economic status were weaker and non-significant. Weak associations were detected between decreased social fragmentation (mobility component), decreased air pollution (NO₂), increased SES, and reduced dementia risk were observed. Air pollution (NO₂) was the only exposure variable, with evidence of an association with both diagnosed dementia and estimated dementia risk score.

The analysis of clusters and outliers using Anselin’s Local Moran’s I showed a spatial heterogeneity pattern in the estimated risk of dementia and for diagnosed cases of dementia in primary care setting in the study areas. There were 51 SA1s with clusters of high–high and 62 with low–low clusters, nine were low–high outliers and 14 were high–low outliers in the pattern of the estimated future risk of dementia (see Fig. 2a). Further, this analysis identified 11 high-high, 15 low-low, 9 high-low, and 16 low-high clusters in the pattern of diagnosed dementia recorded in GP practice data (see Fig. 2b).

DISCUSSION

Our study investigated the associations of built and social environments on dementia diagnosis and estimated future risk of developing dementia using clinical data from primary care setting. We observed that NO₂ and walkability were associated with increased diagnosed dementia cases. We found direct associations between social fragmentation (mobility component) NO₂ and a reverse relationship between public open spaces and socio-economic status, with estimated future risk of dementia.

To our knowledge, this is the first study to investigate the associations between diagnosed dementia and estimated dementia risk with built and social environments using a large scale primary health care clinical data [37]. The estimated future risk of dementia will help to identify high risk individual and screening for early diagnosis of dementia in primary care setting. Further, this provides an opportunity to general practitioners to monitor risk factor for their patient and reduce or delay the risk of developing dementia.

The relationship of NO₂ with diagnosed dementia and estimated dementia risk is in line with previous studies’ findings [15, 37]. A study by Carey et al. (2018) [38] investigated various measures of air pollution (NO₂, PM_2.5, and O₃) in an urban primary care data set and highlighted a positive relationship between NO₂, and PM_2.5 with dementia. NO₂ is a marker of traffic- and other combustion-related pollution (including road traffic, industry, and non-road emissions from shipping and aviation) and it is important to investigate its impact on dementia in urban environment. Although the mechanisms through which air pollution exposure might affect cognitive decline are unknown, oxidative stress and systemic inflammation may explain such effect [15]. Given the potential relationship of air pollution exposure such NO₂ on diagnosed dementia and estimated dementia risk, understanding the mechanisms of air pollution effects on brain health merit further investigation.

Contrary to our hypothesis and previous evidence [19], we found that higher levels of walkability were associated with greater odds of a diagnosed dementia. It may be that while more walkable areas potentially support greater levels of physical activity [39], they may also have greater levels of pollution [40, 41]. However, this is unlikely to be the only explanation since NO₂ and walkability were only weakly correlated in our dataset. This unexpected finding might also be explained by the specific study setting where older people, at greater risk of dementia, also live in older suburbs with traditional gridded street networks and higher levels of walkability. Further research is needed to clarify relationships between walkability and dementia. Exploring the separate components of walkability would be an important first step.

We found that an increased proportion of public open spaces within a SA1 was associated with a decrease in estimated risk of dementia. Since open spaces support increased levels of green vegetation, our finding of reduced dementia risk [42] is consistent with other studies that have demonstrated relations between increased greenness and tree canopy and Alzheimer’s disease and dementia [13, 43]. It is plausible that exposure to public open spaces may increase physical activity, leading to a protective effect for dementia risk. Understanding the potential effects of open public spaces on brain health and cognitive ageing is important and opens new research avenue to study direct and indirect effects of public open space is warranted. Having said that we found a non-significant relationship between already diagnosed cases of dementia and public open space. This may relate to the use of cross-sectional data rather than looking for longitudinal and life course effects of public open spaces on diagnosed dementia. Another explanation for this relationship might relate to people with diagnosed dementia living in residential care, which may be located in areas with a high proportion of public open spaces.

There is a growing body of literature that highlighted the benefits of green and blue spaces on wellbeing and health [44, 45]. Further, given the clusters of low diagnosed dementia cases and the estimated dementia risk in the western suburbs of Adelaide city bordering the coast, further research on the impact of access to blue space is warranted.

A growing body of literature has supported a relationship between greenspace and health. However, various greenspace metrics exist; some are based on subjective measures while others are based on an objective assessment of the landscape such as land use, vegetation indices [46, 47], and tree canopy [43, 48]. The different metrics capture diverse elements of greenery and may therefore have different relationships with dementia. For example, presence of tree canopy and vegetation in the residential neighborhood (measured using a satellite based vegetation index) has been shown to have mental health benefits [49] and is also associated with less cognitive decline [9], and lower odds of Alzheimer’s disease [50], and dementia mortality [51], and dementia risk [43]. Our study did not measure vegetation, but instead assessed access to public open space and public open spaces was not constructed as a measure of greenness [52] but was specifically built to measure publicly accessible open spaces [47], they do provide important physical activity settings [53], and physical activity is in turn associated with lower dementia risk [54]. Despite this, we did not find evidence of a relationship between public open space and diagnosed dementia.

There is growing evidence of the importance of social environments for brain health [55, 56]. However, no research has reported on the impact of social fragmentation on dementia risk. Although the observed relationship between social fragmentation and diagnosed dementia was not significant, we found a significant relationship between social fragmentation (mobility component) and estimated dementia risk. This suggest that social inclusion and social cohesion measured as ‘neighborhood’s social environment structure’ may be associated with reduced risk of dementia. This is the first study to investigate the relationship between social fragmentation and dementia incident and estimated dementia risk. However, studies investigating subjective loneliness and dementia risk [57] suggest that loneliness may put one at risk for dementia through higher inflammation [57]: poorer health behavior (such as heavy drinking or being sedentary), and lack of social interaction (stimulating cognitive health). Future research could shed light on additional pathways through which social fragmentation increases risk.

Spatial cluster analysis showed that there is a great deal of spatial variation in the estimated future risk of dementia and diagnosed cases of dementia patterns. This suggesting a possible link between contextual factors such as built and social environmental characteristics and the clusters (hot spots) of dementia risk. Further, the spatial clusters analyses revealed a spatial convergence between the estimated future risk of dementia and already diagnosed cases of dementia in some neighborhoods (i.e., in the central locations of the study area). Investigating spatial clusters and exploring drivers of dementia clusters generate an evidence on spatial variability of dementia risk and diagnosed case of dementia, and potential link with the natural and built environments characteristics of areas [58]. Additionally, dementia cluster maps highlight local variations in dementia pattern and understanding of place effect on health outcomes [59] and generate new knowledge for local policy planning and geographically targeting of interventions [60]”. These findings are particularly important in light of current global policy efforts, in which the topic of geographical epidemiology of dementia and environmental risk factors feature prominently [1, 6].

The present study had several strengths. It included a large, relatively diverse sample, clinically measured risk factors and a validated measure of dementia risk. These strengths need to be put in the context of some limitations of this research. First, while we used a well-validated dementia risk prediction (ANU-ADRI) tool to estimate the probability of developing dementia for individuals, this measure may over-estimate the dementia risk score since it did not record important protective factors such as fish consumption and social engagement although these items are rarely included in dementia risk scores [61].

We used cross-sectional data and this data cannot capture the longitudinal effect of social and built environment exposure on developing dementia risk. Longitudinal and life course data are needed to fully quantify the effects of such exposures on dementia risk. Further, the missing data for risk factors was likely to cause significant changes in the individual risk profile, however, it was less likely to cause significant impact at the community level.

The number of diagnosed dementia cases from 2012 to 2014 includes incidence (new case of dementia) plus those cases diagnosed prior to this period if they had three visits to GPs within two years. However, the pattern of age distribution of dementia cases was in line with the expected national levels of dementia.

The multilevel models used in this study aggregate some exposure measures (i.e., busy road length) to the small area level and this may have an impact on the ability of the model to capture the relationship between exposure and outcome variables. An alternative approach could be measuring the distance to an exposure within a specified distance threshold.

The Local Moran’s I analyses showed a potential spatial structure in the estimated dementia risk and diagnosed dementia outcome measures. However, the multilevel models used do not take this into account other than what can be explained by the individual small area level exposure being analyzed. Spatial autoregressive models could take this structure into account, and they could consider as a possible alternative for future research.

Conclusion

A positive association was observed between built environment characteristics (NO₂ and walkability score) and diagnosed dementia. We observed a reverse relationship between built environments exposures (access to public open spaces and socio-economic status) and the estimated dementia risk outcome. Further, the study found these associations were largely unchanged after controlling for participants’ socio-demography and lifestyles. The study adds strength to the growing international evidence that there is a potential opportunity to delay or reduce dementia risk, through structural changes to our built and social environments.