ULI calculation and linkage

The ULI was calculated for 1,550,641 residential address proxy sample points across the urban portion of Greater Melbourne (Fig. 1). The final ULI-linked analytical data set comprised of 8691 Victorian Population Health Survey 2014 (VPHS) geocoded respondents with complete data on socio-demographic covariates (Supplementary Fig. 1). Sample size for each analysis after restriction to complete outcome data and exclusions is further reported in Supplementary Table 1.

The spatial distribution of the urban liveability index for Melbourne, Australia, with a target time point of 2011–2012 (data sources dated from 2011 to 2014, as per Table 5); local government area and Australian continent boundaries displayed for reference purposes, sourced from the Australian Bureau of Statistics under CC BY 4.0 licence as per the data availability statement.

Participant characteristics

Participant socio-demographic characteristics are summarised in Table 1 and cardiometabolic outcomes of interest in Table 2. Both tables include the associated liveability (ULI) distribution summary for each item. The sample was largely comprised of an older cohort (3.9% aged 18–30 years and 69.9% aged 50+ years), the majority of whom either owned or were purchasing their own home (86.3%); renting was strongly associated with living in more liveable areas, whether privately or through social housing.

While 46.4% of participants reported undertaking ≥10 min of walking for transport and slightly fewer (39.3%) met physical activity requirements by walking, achieving these respective outcomes was strongly associated with living in a more liveable neighbourhood. Most participants (53.4%) had a body mass index (BMI) >25 kg/m², with 1.5% recording a BMI <18.5 kg/m². Diagnosis of hypertension was reported in 38.0% of participants, while 8.5% reported a diagnosis of type 2 diabetes mellitus (T2DM). Excellent or very good health status was reported by 41.6% of participants, while 14.4% had Kessler 10 results indicative of psychological distress, and fewer still (6.3%) reported not being satisfied with life as a whole.

Multilevel regression analyses

Fully adjusted models examining the association between the health and wellbeing outcomes of interest with both the ULI and its component indicators are reported in Table 3 (physical activity and health outcomes) and Table 4 (wellbeing outcomes). To facilitate comparison of the ULI with previously used measures, Tables 3 and 4 also include results for the pilot ULI and the walkability index.

Results for each exposure were scaled according to its interquartile range (IQR), as observed in the VPHS sample. For each analysis, this represents the difference in living in an area with high exposure value (e.g. highest quartile of ULI), compared with an area with low exposure value (e.g. lowest ULI quartile). Defined in this way, after adjusting for socio-demographic characteristics and geographic clustering, living in a higher liveability area was associated with increased odds of walking at least 10 min per week for transport between 48% and 78%, increased odds of meeting physical activity requirements by walking between 8% and 25% and having a lower BMI between 0.16 and 0.55 kg/m². The point estimate for odds of T2DM diagnosis was estimated to be 10% lower in high- compared with low-liveable areas although the broad credible interval (CI) indicated broad uncertainty of magnitude and direction of this effect in the population, consistent with a 21% reduction, up to a 1% increase in odds of diagnosis. There was no clear evidence of adjusted associations between urban liveability and hypertension or the mental wellbeing outcomes of self-rated health, psychological distress and life satisfaction.

Associations between ULI sub-indicators and cardiometabolic outcomes

All ULI component indicators with the exception of housing affordability and living and working in the same area were associated with walking for transport and to a lesser extent achieving recommended levels physical activity. Notably, those who lived in areas with greater local employment were estimated to have 8% increased odds of achieving recommended levels of physical activity through walking (adjusted OR (AOR) 95% CI 1.01, 1.16). Overall, indicators with larger estimated effect sizes for physical activity (ULI; ‘Community, culture and leisure’ access; supermarket access; convenience access; street connectivity) were also associated with larger reductions in BMI, with point estimates ranging from −0.24 kg/m² (street connectivity) to −0.59 kg/m² (Community, culture and leisure) per IQR change in exposure. In particular, models employing the ‘Community, culture and leisure access’ indicator often resulted in the best fit and largest estimated effect sizes. This was also the only indicator where the posterior CIs for association with hypertension (AOR 0.89; 95% CI 0.82, 0.97) and T2DM (AOR 0.82; 95% CI 0.71, 0.94) diagnosis were not ambiguous with regard to the direction of estimated association.

To aid visualisation of the spatial distribution of distinct regions across the Melbourne metropolitan region with address points within recommended distance of community, culture and leisure destinations, a further spatial analysis was conducted on these points generating the convex hull of those clustered within 400 m of each other, where the resulting geometry collection consisted of ≥10 points (Fig. 2). Labels were annotated through concatenation of the names of 2011 suburb geometries²⁹ with either a >10% intersection with the convex hull polygon or an absolute area of intersection of at least 3 km². In contrast to the ULI, for which the highest scoring addresses (scores of ≥110; 42,666 address points or 2.8%) were clustered in the inner city region (Fig. 2), clusters of address points with access to community, culture and leisure destinations all within recommended distance (54,389 address points or 3.5%) were also identified in some middle and outer suburbs. Correlation of sub-indicators with composite measures is further reported in a supplementary correlation analysis (Supplementary Table 2).

Spatial distribution of community, culture and leisure access, with annotated convex hull clusters of address points with full access to community, culture and leisure destination types within recommended distances. The Australian continental digital boundary for 2011 was sourced from the Australian Bureau of Statistics under CC BY 4.0 licence as per the data availability statement.

Comparison of the ULI with the pilot-ULI

Compared with its pilot version, the ULI was found to have a broader IQR and in turn stronger magnitudes of associations with cardiometabolic outcomes. Unlike the ULI, the pilot ULI was not found to be associated with BMI after adjustment. Nevertheless, the walkability index was associated with lower BMI (AOR −0.29; 95% CI −0.44, −0.12); however, the magnitude per change in IQR was lower than that of the ULI.

Main findings

Residing in higher liveability neighbourhoods was found to be positively associated with undertaking at least 10 min of weekly walking for transport, meeting physical activity recommendations by walking, negatively associated with BMI, and although not statistically significant, lower levels of T2DM diagnosis. Living in more walkable neighbourhoods and those with greater access to local destinations was associated with higher odds of walking and achieving recommended levels of physical activity but no other cardiometabolic risk factor. Living in areas with access to diverse community, culture and leisure destinations was associated with lower odds of diagnosis for both hypertension and T2DM. The associations observed for the ULI were of greater magnitude than for the pilot liveability index and walkability index.

The combined role of community, culture and leisure

Notably, more so than street connectivity, dwelling density or other types of destination, we found that living in a neighbourhood with access to a mix of community, culture and leisure destinations (i.e. a community centre, a cinema or theatre, a library and a museum or art gallery) was uniquely predictive of reduced odds of both hypertension and T2DM, following adjustment for socio-demographic characteristics. The geographic particularity of a residential location with proximal access to each of these distinct amenities may serve as a marker for a vibrant, well-provisioned activity centre offering a broad range of services and amenities supporting local, walkable living; and this requires further investigation.

Liveability and the social determinants of health

Meaningful associations with the ULI or the other spatial indicator exposures considered were not observed for self-rated health, psychological distress or life satisfaction outcomes. However, the observed pattern in associations between the ULI and its constituent indicators is consistent with the theoretical social determinants of health pathways and outcomes. Considered through this framework, it is expected that associations with built environment exposures will be strongest upstream for behavioural and lifestyle factors such as meeting physical activity recommendations, rather than more distal outcomes such as subjective wellbeing, which develop over longer periods with complex pathways; yet, it is broadly accepted that interventions which target behaviours as underlying causes contributing to downstream conditions can improve overall population health and reduce health system burdens^30,31.

Generalisability

The findings from this study were based on a spatially representative cohort of adults in Melbourne. The general pattern of the spatial distribution of amenities observed in this study has been observed in other Australian cities⁴; however, extension of the ULI methods to other contexts should consider applicability to local environmental and social contexts: cultures and customs, climate and topography, histories of built environment development and equity and regional economic integration are important factors to consider when translating liveability³². The goal of absolute or relative ‘liveability’, and the equity of its spatial distribution for a city’s population, is arguably an inherent human response to the challenges of urbanism, which is a global phenomenon; formulas for liveability are not necessarily new or culturally specific. The usage of the term, in a pamphlet comparing London unfavourably with Paris, by John Storer in 1870 presents a definition of liveability resonant with that used in this study 150 years later, with appeal not only to the health impacts of urban form but also the importance of cultural and social life for population wellbeing: ‘When we speak of a city being liveable, we mean that there is in it room and space for all; that the blessings of fresh air and sunlight are accessible to the poor and rich alike; that life is not robbed of all its grace and beauty by everlasting confinement in close alleys, or in noisy damp dark streets; that the atmosphere is unpolluted by smoke; and beyond this we mean a city which wise rulers have adorned, and beautified with noble intellectual works of art; so that the people lifting up their eyes, and beholding what is beautiful and refining, may themselves become beautiful and refined in mind and body, and worthy of the high place man fills upon this planet. A high ideal is indicated by this word liveable, if we use it rightly’³³.

The Economist Intelligence Unit measures ‘liveability’ globally as a basis for remuneration of expatriates³⁴. However, the ULI measured in this study reflects the aspirations for liveability for all of a city’s residents, shared by communities, stakeholders and researchers from diverse contexts^32,35,36,37. It is noted that if approached uncritically, and without an equity perspective, the application of liveability in policy may at times be fraught^18,19. Our ULI method provides a flexible framework for calculating the spatial distribution of liveability for distinct residential address points across a study region, allowing for identification of local inequities with high precision, or, as in this study, for linkage with geocoded health survey data to investigate implications of variation in local liveability for cardiometabolic health and wellbeing. Critical application of this method in diverse contexts with adaptations aligns with comparative urbanism, which respects difference and variation³⁸. A universally applicable definition of liveability allowing for generalised statements on sustainable health outcomes across diverse contexts is perhaps apocryphal; however, adaptation of the ULI method to local contexts can support both within-city and between-city comparisons, as long as the variations and contingencies in these contexts are acknowledged.

Study design

The impact of local neighbourhood urban liveability on downstream health outcomes such as hypertension and T2DM may not be immediate but instead may involve sustained exposure to more or less ‘liveable’ built environments across an individual’s life course^22,39. The present cross-sectional study did not account for participants’ residential histories; however, a meta-analysis of longitudinal studies by Chandrabose et al., mostly conducted in urban settings in developed countries, concluded that there is strong evidence that more walkable environments are associated with improved hypertension—and T2DM—related outcomes⁹. Walkable neighbourhoods underpin a liveable city, providing opportunities for healthy sustainable lifestyles⁴⁰. However, liveable communities offer more than simply being walkable. The ULI incorporates access to a wide range of destinations and was more strongly associated with walking behaviour than the walkability index alone. Assuming there is a unidimensional property, i.e. a single factor of liveability, then ULI and walkability can both be seen as imperfect measures of this quantity. They are strongly correlated with one another and both predict walking behaviour in this cross-sectional study. If ULI’s improved performance can indeed be attributed to being a more precise measure of liveability, then it shall also perform better in longitudinal studies. Such studies have higher power to detect effects on downstream cardiometabolic outcomes (hypertension, T2DM) as the preventative effect of healthy lifestyles improves with sustained activity over multiple years. Due to their correlation, walkability and ULI are both susceptible to the same sources of confounding, but the magnitude of such biases need not be identical in this study. Natural experiments are particularly suitable to overcome bias and changes in at least one ULI component are, by definition, at least as likely as changes in walkability. This provides more opportunities for research. The key contribution of our study has been the development of an enhanced address-level built environment exposure variable associated with healthy, sustainable lifestyles.

Temporality

Both the ULI and its pilot predecessor were constructed using spatial data sources targeting 2012. However, the VPHS survey was conducted between 2012 and 2014. While the temporal sequence of ‘exposure precedes outcome’ may be assumed to be broadly correct, given that ‘real-world’ features will have pre-dated their representation in data, some degree of information bias arising from mismatch of data sources to some participants’ responses is acknowledged. It was not possible to acquire all data sources at perfectly coincident time points. Despite best efforts towards achieving data contemporaneity, it is understood that the time when a data representation of a location could be considered ‘true’ may have expired by the time the data were published or retrieved by us. It is also possible that some participants moved into an area after spatial data recorded for that area had become obsolete, and thus their lived experience of a neighbourhood may differ from that which we have estimated. While we note this limitation, it is assumed that such instances would be either random, meaning that associations observed are likely to have been somewhat attenuated⁴¹, or specifically impacting urban fringe areas with more dynamic development. Overall, given the large sample size and variety of data sources considered, the impact on any location or individual from potential mismatch is considered limited.

Liveability and individual activity space

The ULI has been constructed using the walkable area surrounding individual residences. However, there is an emerging body of research that considers the health impacts on individuals of their broader activity space exposures, with consideration beyond the residential neighbourhood and home life, to also encompass work and recreation locations that may be more important at different life stages⁴². Future research needs to consider the expansion of the spatial ULI to encompass individuals’ broader activity spaces⁴³.

Quality of amenities

A number of the indicators included in the ULI measure access within a walkable distance to an amenity; however, the quality of that amenity—beyond proximity, or size in the case of public open space—was not considered due to data limitations. The consideration of quality is an added complexity, which in principle could be further accommodated in the ULI by calculating indicators using data sources with additional covariates. For example, our public transport indicators measure proximity to public transport stops, which is one dimension of transport service known to influence walking behaviour⁴⁴. However, other dimensions of public transport service not considered in the present study include: frequency of service; accessibility with regard to individual abilities or overcrowding; and overall utility given cost, comfort and travel time for getting to a specific destination of interest. Not accounting for these dimensions may impact the estimation of effect sizes and the attribution of effects to different components of the built environment. The advent of broadly available General Transit Feed Specification data (in 2015, for Melbourne) means that some of these aspects of public transport may be more readily measurable for major urban centres at the time of writing than was the case in 2014⁴⁵. Future studies should aim to make use of quality measures to more deeply understand associations between liveability and health and wellbeing.

Future directions for urban liveability research to inform urban planning interventions

Through linkage of a spatial ULI with a population health survey, we found that more liveable areas were associated with increased walking for transport behaviour and physical activity levels achieved through walking, and lower BMI. Creating more liveable cities therefore has the potential to foster both health and sustainable lifestyles. The ULI displayed stronger health-beneficial associations than a previous pilot version, and the walkability index is commonly used in built environment and active transport studies. Examination of the index sub-components found that access to a mix of community, culture and leisure destinations (treated as an individual indicator within the ULI) was most strongly associated with cardiometabolic outcomes, including reduced diagnoses of hypertension and T2DM. This may be because these types of destination may be a marker for established activity centres. This requires further investigation in future studies. Given global interest in urban liveability in policy and public discourse, these results highlight the utility of the ULI as a tool to inform localised urban planning interventions that would create healthier more sustainable cities.