The probability of diabetes and hypertension by levels of neighborhood walkability and traffic-related air pollution across 15 municipalities in Southern Ontario, Canada: A dataset derived from 2,496,458 community dwelling-adults

Individuals’ risk for cardiovascular disease is shaped by lifestyle factors such as participation in physical activity. Some studies have suggested that rates of physical activity may be higher in walkable neighborhoods that are more supportive of engaging in physical activity in daily life. However, walkable neighborhoods may also contain increased levels of traffic-related air pollution (TRAP). Traffic-related air pollution, often measured through a surrogate marker (e.g. NO2), has been associated cardiovascular disease risk and risk factors [1], [2], [3], [4]. The higher levels of TRAP in walkable neighborhoods may in turn increase the likelihood of developing conditions like hypertension and diabetes. Our recent work assessed how walkability and TRAP jointly affect the odds of diabetes and hypertension in a sample of community-dwelling adults from Southern Ontario, Canada [5]. This article contains additional data on the probability and odds of hypertension and diabetes according to their walkability and TRAP exposures. Data on cardiovascular risk factors were collected using health administrative databases and environmental exposures were assessed using national land use regression models predicting ground level concentrations of NO2 and validated walkability indices. The included data were generated using logistic regression accounting for exposures, covariates, and neighborhood clustering. These data may be used as primary data in future health risk assessments and systematic reviews, or to aid in the design of studies examining interactions between built environment and TRAP exposures (e.g. sample size calculations).

contain increased levels of traffic-related air pollution (TRAP). Traffic-related air pollution, often measured through a surrogate marker (e.g. NO 2 ), has been associated cardiovascular disease risk and risk factors [1e4]. The higher levels of TRAP in walkable neighborhoods may in turn increase the likelihood of developing conditions like hypertension and diabetes. Our recent work assessed how walkability and TRAP jointly affect the odds of diabetes and hypertension in a sample of community-dwelling adults from Southern Ontario, Canada [5]. This article contains additional data on the probability and odds of hypertension and diabetes according to their walkability and TRAP exposures. Data on cardiovascular risk factors were collected using health administrative databases and environmental exposures were assessed using national land use regression models predicting ground level concentrations of NO 2 and validated walkability indices. The included data were generated using logistic regression accounting for exposures, covariates, and neighborhood clustering. These data may be used as primary data in future health risk assessments and systematic reviews, or to aid in the design of studies examining interactions between built environment and TRAP exposures (e.g. sample size calculations

Data
The raw data used here are held by ICES [6]. These data were derived from a cross-sectional study of 2,496,458 adults aged 40e74 years who were living in one of 16 urban municipalities in Southern Ontario on January 1, 2008. All individuals were eligible for provincial health insurance for at least two years at inclusion, had not resided in a long-term care facility within the previous 5 years, and were free from cardiovascular disease at baseline (i.e. history of prior myocardial infarction, stroke, congestive heart failure or cardiovascular revascularization procedure). Associations between neighborhood walkability, traffic-related air pollution, hypertension and diabetes are displayed in Table 1. All estimates were adjusted for baseline sociodemographic variables, chronic obstructive pulmonary disease, the total number of individual comorbidities, and city/region. Adjusted probabilities from models including an interaction between walkability and traffic-related air pollution were estimated across levels of NO 2 (0 ppb, 5 ppb, 10 ppb, 20 ppb, 30 ppb, 40 ppb; range of NO 2 in sample: 3.94 ppbe51.47 ppb) and walkability (quintiles e Q1 lowest 20%, Q5 highest 20%; range in underlying walkability scores (unitless): 6.26, 27.78) ( Tables 2 and 3).
Associations observed among members of this same cohort who had a longer duration of exposure (resided in their residential neighborhood for at least 5 years, N ¼ 1,609,247) are shown in Table 4. Estimated probabilities from this sample by level of walkability and traffic-related air pollution are shown in Tables 5 and 6.
Finally, we estimated models among a larger sample of individuals from the general population of adults aged 40e74 living in the same region at baseline, including those with and without a prior history of cardiovascular disease (N ¼ 2,592,646) ( Table 7). The estimated probabilities of hypertension and diabetes across levels of walkability and traffic-related air pollution, adjusted for sociodemographic variables, are found in Tables 8 and 9. Table 10 includes the parameter estimates from logistic regression models assessing the interaction between traffic-related air pollution and neighborhood walkability on the odds of hypertension and diabetes. These models were fit among individuals free from cardiovascular disease at baseline.

Sample and data sources
Participants were drawn from the Cardiovascular Health in Ambulatory Care Research Team (CANHEART) cohortda cohort of Canadian adults from Ontario, Canada assembled using administrative databases held at ICES in Toronto, Canada. The protocol for creation, individual databases used, and variables available have been described previously [9]. Selection criteria for sample used were described also elsewhere [5]. Briefly, individuals residing within one of 15 municipalities (Toronto & Greater Toronto Area, Hamilton, London, Ottawa) who were between 40 and 74 years of age were Value of the data Previous work examining relationships between the built environment, traffic-related air pollution, and cardiovascular risk factors has generally treated these variables in isolation. These results demonstrate how antagonistic interactions between walkability or traffic-related air pollution and cardiovascular risk factors may occur Researchers investigating healthy community design, public health practitioners, and individuals engaged in urban policy may benefit from these data The results reported here may be used to develop health risk assessments which take into account interactions between environmental variables, in systematic reviews of environmental correlates of cardiovascular disease risk factors, and in planning future studies examining interactions between built environment and air pollution variables Previous analyses (e.g. Refs. [7,8]) have used literature-derived estimates of associations between physical activity, air pollution, and cardiovascular health to assess whether the protective value of physical activity declines in polluted environments. These estimates, however, often do not consider interactions between these pollution and walkable environments. These data may provide more accurate assessments of the value of walkable environments in the context of air pollution. They may also help in the design of policies directed at mitigating air pollution in urban environments.
eligible for inclusion. Individuals who resided within a long-term care facility within the past 5 years were excluded. In total, data from 2,496,458 individuals was included. Information on traffic-related air pollution was drawn from a national land use regression model designed to provide estimates of annual average outdoor NO 2 concentration for locations across Canada [10]. Model predictors included satellite-derived NO 2 , land area for industrial uses within 2 km, total road length within 10 km, and summer rainfall. Estimates were generated for postal codes where individuals within our analytic sample resided. To additionally account for small-scale variation in NO 2 , deterministic gradients were used to adjust NO 2  Hypertension and diabetes were assessed using validated algorithms using diagnostic code information from individual's hospital admissions and outpatient physician billings, or fee codes for diabetes related programs. For hypertension, individuals who had one hypertension-related diagnostic code on a hospital admission or a hypertension-related diagnostic code on two outpatient physician service billings within a 2-year period before January 1 2008 were considered to have hypertension.  Individuals were considered to have diabetes if they had records during a 2-year period before January 1 2008 indicating a diagnosis of diabetes during a hospital admission, two physician service billings, or a billing for a diabetes related program (i.e. insulin therapy support or diabetes management assessment). These algorithms were validated against clinical information, including laboratory testing information, clinical charts, measured blood pressures, and medication information. Validation of the  Notes Covariates included in model: age, sex, ethnicity, immigration history, neighborhood median income). Association estimates for traffic-related air pollution are per 10-unit increase in NO 2 . Independent models include either walkability or trafficrelated air pollution. Joint models include walkability and traffic-related air pollution simultaneously. OR: odds ratio, CI: confidence interval, Ref: reference category.

Table 5
Predicted probability of hypertension at varying levels of walkability and NO 2 among individuals remaining in their neighborhood for 5 or more years adjusted for baseline sociodemographic factors. hypertension algorithm found a sensitivity of 0.72 and specificity of 0.95 [14]. Validation of the diabetes detection algorithm found a sensitivity of 0.89 and 0.98 [15]. Table 6 Predicted probability of diabetes mellitus at varying levels of walkability and NO 2 among individuals remaining in their neighborhood for 5 or more years adjusted for baseline sociodemographic factors.
Walkability Quintiles (Q)  Notes Covariates included in model: age, sex, ethnicity, immigration history, neighborhood median income). Association estimates for traffic-related air pollution are per 10-unit increase in NO 2 . Independent models include either walkability or trafficrelated air pollution. Joint models include walkability and traffic-related air pollution simultaneously. OR: odds ratio, CI: confidence interval, Ref: reference category.

Analysis
Adjusted odds ratios were estimated using logistic regression models accounting for clustering at the neighborhood level (dissemination areas) using generalized estimating equations. Predicted probabilities were generated using logistic regression models including the main effects of walkability and traffic-related air pollution along with their multiplicative interaction term, in addition to covariates. The additional covariates included in the model are listed in the table notes and are further described by Howell and colleagues [5].