This study validated a walkability index for Sydney that was comparable to the PLACE index frequently used for walkability research . The PLACE index combines four built environment attributes associated with walking for utilitarian purposes: residential dwelling density, intersection density, land use mix, and retail floor area ratio. A limitation of this and similar four-attribute indexes is that floor space data are frequently unavailable to calculate retail floor area ratio . This was the case in the current study for which floor space data were only available for a part of the study region. We therefore tested a three-attribute abridged index and found it to have similar measurement properties to a full index. This has international implications because retail floor area data are often difficult to source [1, 12] or unavailable  for index construction and applications of abridged indexes that exclude retail floor area ratio may allow for greater use of walkability indexes in research [22–25].
The innovative observation in this study was that the abridged walkability index retained 87% of the variability in the full index, assigned all analysis units to within one walkability quartile of the full index, and found associations of similar magnitude to the full index between walkability and prevalence of walking to work after adjusting for demographic and socioeconomic confounders. Thus, in the absence of retail floor space data, an abridged index comprising residential dwelling density, intersection density and land use mix only may be used to characterise walkability. This would be advantageous in the many global locations where retail floor space data are not available . We recommend researchers with data on the four walkability components in only a subset of spatial units also compare three and four-attribute indexes to further validate this finding.
Principal component analysis of the abridged Sydney Walkability Index attributes extracted a single component with high loadings for all attributes; similar component structures and loadings were also observed for City of Sydney full and abridged indexes. This appears to be the first time that a latent variable structure of a PLACE/NQLS index has been described, and supports the validity of the Sydney Walkability Index as a cohesive measure of walkability. Internal consistency of the abridged Sydney Walkability Index is also acceptable for research purposes , especially given the small number of items included in the index .
These results demonstrate the feasibility of a Sydney Walkability Index, the utility of a three-attribute derived index, and a consistent relationship between walkability and walking to work that is only partially moderated by socioeconomic status. Walking to work increased monotonically with increasing abridged walkability index score decile, and was higher for high walkability areas compared to low walkability areas in both lower and higher income areas. These findings concur with NQLS index validation outcomes that found increasing walk trips with increasing decile of walkability, and more walking in high versus low walkability areas for both high and low income strata , providing additional support for the validity of the abridged Sydney Walkability Index.
Although the prevalence of walking to work in the Sydney Metropolitan Region increased with increasing walkability decile, this association was more pronounced at the upper deciles of walkability. Excluding low population density Census Collection Districts as suggested by Leslie et al.  did not alter this trend, and may indicate homogeneity in the distribution of urban sprawl outside the inner city area. This is consistent with the adjusted odds for walking to work, which were significantly higher for high and very high walkability areas compared to low walkability areas, but similar for medium compared to low walkability areas. Further study into possible walkability threshold effects may provide useful information for planning and policy interventions to improve built environments to support walking.
Visualisation of choropleth maps indicated consistent patterns of clustering across the study area for Sydney Walkability Index scores and its component environmental variables. This was supported by correlation analyses that indicated all variables were strongly associated with one another. High residential density, street connectivity and land use mix were concentrated in the central, eastern and north Sydney areas, and decreased along an east–west gradient to a ring of low walkability areas on the outer fringes of the Sydney Metropolitan Region. This patterning is consistent with the spatial distribution of population density and socioeconomic disadvantage in the study area , and highlights the planning potential of the Sydney Walkability Index to target walkability infrastructure upgrade and development initiatives in the Sydney Metropolitan Region.
Understanding the features of the built environment that facilitate or constrain walking is important for research, planning and policy aimed at increasing the proportion of adults who attain recommended levels of physical activity . Linking the Sydney Walkability Index to land use and transport planning strategies such as the Sydney Metropolitan strategy  has the potential to create more walkable communities, and have a greater population impact on reducing physical inactivity than individual-level interventions [5, 37].
Spatially referenced objective walkability measures such as the one constructed here may also be linked to existing administrative or epidemiological data collections with location information to add both research and policy value. For example, the Sydney Walkability Index is being used in the 45 and Up study to profile the independent health effects of environmental factors such as walkability, to compare with self-report (PANES) items, and to assess changes in activity behaviours when mid to older aged adults change residence [26, 38]. From the Sydney urban planning perspective, objective indexes of the built environment could also be used to monitor, inform and evaluate policy through desktop simulations of proposed developments for walkability based on their urban design features, identify “best buy” areas for infrastructure upgrades and residential development to maximise active transport use, and monitor changes in the walkability of geographical areas over time and following environmental interventions [1, 39]. In this regard, the Sydney Walkability Index provides an “out-of-the-box” resource for researchers, planners and policy makers that is evidenced-based and derived using the best-available spatial data.
The main limitation of this study is that comparability analyses between full and abridged walkability indexes were confined to the City of Sydney local government area as it was the only area for which retail floor space data were available. It is feasible that the similarity in performance of three and four-attribute indexes is unique to this area and may not be as comparable in other areas. However, the generalizability of our results beyond the City of Sydney area is supported by our corresponding analysis for the entire Sydney Metropolitan Region, which produced similar associations between walkability and prevalence of walking to work using the three-attribute index, and identified similar factor structures and explained variance for the Sydney Metropolitan Region abridged index. It would be advantageous for researchers to confirm this finding in other cities where data are available for all four walkability components.
Another limitation of this study is that GIS derived estimates of walkability were not compared to the physical reality on the ground via site visits, so the level and nature of any measurement error is unknown. Previous studies using similar indexes have included field verification as the indexes were used to generate sampling frames for interventions [1, 12]. Field validation in these cases comprised “informal windshield observations”  and systematic observations . While both studies observed some discrepancies in walkability classifications, Leslie et al. concluded that the PLACE index had good face validity and that field observations were concordant with index classifications for the majority of their study units .