Respondents from LAC and SDC have similar characteristics compared to the entire CHIS sample in California, except that the LAC/SDC sample is more racially/ethnically diverse (). Compared to a U.S nationwide survey, the combined study area and California as a whole are more racially/ethnically diverse, younger, have a lower average income, and have more immigrants, more college graduates, and more residents who did not graduate from high school. LAC has fewer non-Hispanic whites, more immigrants, and lower income and education levels than SDC. These comparisons are not relevant to the implementation of the spatial scan approach, but they may influence interpretation of the findings of such analyses.
Demographics of subject counties (based on respondents only), California and the U.S., % unless otherwise indicated
Overall, the percentage of walking/biking and the average walking/biking time over all respondents in LAC is higher than it is in SDC (42.0% vs 36.1% and 84 vs 80 minutes per week, respectively). Without age-adjustment, there are more significant clusters detected in LAC than in SDC, perhaps because of greater spatial variation of active transportation or larger sample size in LAC (). Five clusters with more walkers/bikers and two clusters with fewer walkers/bikers are detected in LAC using 3% of the number of total respondents as the maximum search window (). Cluster 5 is significant when looking for clusters of only more walkers/bikers; it is not quite significant (p=0.066) when testing for both more and fewer walkers/bikers. All other clusters are very significant (p<0.01). The relative risks of being a walker/biker were 1.7, 1.5, 1.6, 1.5, 1.5, 0.4, and 0.6 for Clusters 1 to 7, respectively. Three significant clusters of more walkers/bikers (p-values<0.05) and a marginally significant one with fewer walkers/bikers (p-value=0.06) were detected in SDC () when searching for clusters of either more or fewer walkers/bikers. The relative risks of being a walker/biker in SDC were 2.2, 2.8, 2.0, and 0.09 for Clusters 1 to 4, respectively. Among walkers/bikers, average walking/biking times per week were 188 minutes (n=3573, SD=319) in LAC and 173 minutes (n=680, SD=232) in SDC. One area with lower walking/biking duration was detected in LAC (), but no significant clusters were detected in SDC, suggesting that the duration of activities among those walkers/bikers is spatially homogeneous or that there is a lack of power because of modest sample sizes, particularly in SDC.
To illustrate the capacity of the spatial scan approach to explore contributions of specific variables to clustering, the clustering pattern after adjusting for age was also evaluated. Because younger people tend to have more active transportation than older people, spatial heterogeneity in age distributions alone could result in spatial clustering of active transportation. In LAC, age-adjustment accounts for some but not all of the significant clusters of active transportation ( vs 1B). The clusters with high prevalence in SDC were similar with and without age-adjustment (maps after age-adjustment not shown). However, the marginal cluster of low prevalence (No. 4 in ) is no longer found after age-adjustment.
Characteristics of Clusters
The patterns of the other covariates in the analysis without age-adjustment ( and ) were similar to those with age-adjustment (see Appendixes A and B
, available online at www.ajpm-online.net
). This discussion focuses on cluster properties for clusters identified without age-adjustment, allowing for description of age characteristics inside/outside clusters. Neighborhood characteristics were consistently different between high- and low-walking clusters in both counties, whereas comparisons of individual characteristics varied geographically ( and ). Population and employment densities were higher in high-walking clusters and lower in low-walking clusters. Similarly, higher street, block, and intersection densities; shorter block lengths; and the presence of a bus route were associated with more walking and biking for transportation, compared to the rest of each county. Most of these street connectivity differences were highly significant.
Individual and contextual features associated with clusters of high and low walking prevalence in Los Angeles County, % unless otherwise indicated
Individual and contextual features associated with clusters of high and low walking prevalence in San Diego County, % unless otherwise indicated
People were slightly younger in the high-walking areas and older in the low-walking areas, but gender and BMI did not seem to be related to the pattern of active transportation. The associations between walking/biking and other individual characteristics generally varied among the clusters. In LAC, high-walking Clusters 1, 2, 4, and 5 had lower income and education levels, whereas Cluster 3 residents had higher education and income levels. Clusters 1, 2, and 5 had predominantly Hispanic residents, whereas Clusters 4 and 5 had more black residents; Cluster 3 is predominantly non-Hispanic white compared to areas outside the clusters (). In SDC, high-walking Clusters 1 and 3 had lower income and education levels; the small Cluster 2 had more highly educated residents but not a significantly different income distribution compared with the rest of the county. Cluster 1 had more Hispanic and black residents; Cluster 3 had more Hispanic residents; and Cluster 2 had more non-Hispanic whites than areas outside the clusters. In contrast, the low-walking clusters in both counties have more non-Hispanic white residents and higher income levels.
Self-reported health status showed mixed associations with walking/biking prevalence. For example, LAC high-walking Clusters 1, 2, 3, and 4 had fewer residents who responded that they were limited a lot in climbing stairs compared those to outside the clusters, but more residents had such limitations in Cluster 5. In Cluster 4, respondents reported the highest health status among all cluster areas and outside cluster areas (with 76.5% without limitation of climbing stairs). Street connectivity is very high in Cluster 4, so despite significantly fewer bus stops, freeways, and bus routes, it still appeared to be a cluster of more walkers/bikers, and their average duration walked/biked is the longest (264 minutes per week).
No significant clusters of longer or shorter walking/biking times were found in either county, independent of the prevalence of active transportation, but one area in LAC had a shorter walking/biking duration that was of borderline significance (). This area had residents with more education (86% vs 59% college education), higher income (75% vs 44% with income more than 300% of poverty level), fewer immigrants (23% vs 29%), and more non-Hispanic whites (75% vs 39%).
Although overall the length of time spent in active transportation did not cluster significantly in the study counties, the average walking/biking time did vary among the high- and low-prevalence clusters. The average time spent walking/biking among walkers/bikers in LAC Cluster 3 is lower than the average time spent outside the clusters (181 vs 192 minutes), whereas the time spent is longer inside than outside the other clusters (239, 193, 264, and 269 minutes for Clusters 1, 2, 4, and 5, respectively). Even though more people were found to walk/bike in SDC Clusters 2 and 3, their time spent is not long compared with those people who do walk/bike outside the clusters (118 and 158 vs 170 minutes per week). However, SDC high-walking Cluster 1 had a longer time spent walking/biking than the area outside clusters (263 minutes in Cluster 1 vs 170 minutes outside SDC clusters).