SES & diabetes prevalence
These findings demonstrate that neighborhoods with low income have a higher prevalence of diabetes than do wealthy neighborhoods. This socio-economic gradient in diabetes prevalence has been shown previously across studies and across cultures [3, 4, 6]. The link between income and diabetes risk is complex. It has been speculated that the increased diabetes risk seen in low income groups is related to the increased prevalence of obesity within this group. It has already been clearly shown that low SES is associated with a much higher prevalence of obesity, especially among women .
Obesity remains a potent risk factor for the development of diabetes; however, low income has been shown to be an independent risk factor for the development of diabetes among women – even after controlling for body mass index and physical activity level . Alternatively, low SES could be a result of diabetes in so far as disability related to diabetes complications may limit work and educational opportunities.
Neighbourhood and community level factors also contribute to the increased diabetes risk seen in low income populations. The "built" environment has been shown to be a clear barrier to physical activity in poorer neighbourhoods. Low income communities have been shown to have less biomass and park-space compared to wealthier communities . There may also be a perception that it is less safe to walk in a poorer neighbourhood – this not only deters physical activity but erodes the sense of community among residents [3, 13, 14]. This sense of community, along with established social networks, has been shown to be protective against certain negative health outcomes .
SES and diabetes care utilization
Previous studies examining the association of income on access and/or utilization of health care services have suggested that even within single payer systems such as Canada's, access may not be universal. Dunlop and colleagues showed that poor individuals are more likely to visit their family physician, but that the wealthy are nearly twice as likely to be referred on to specialty care . The wealthy are also likely to have a shorter wait time for procedures such as coronary angioplasty .
Consistently reaching therapeutic targets in diabetes usually requires the support of a multidisciplinary team (diabetes educators, registered dieticians, and social worker and diabetes medical specialists) and the use of several medications [15–17]. Diabetes education centres allow patients to access the relevant health care professionals and education services within a single centre . Previously, little was known about how individuals access such diabetes care services in this centralized model of care, particularly in relation to that individual's socio-economic standing.
The present study shows that people in the lowest income strata were more likely to be referred for structured diabetes education and care. Our study shows that low income patients are approximately 30% more likely to be referred to this DEC and that this seems to appropriately reflect burden of disease.
SES and utilization of diabetes care controlling for prevalence of diabetes
Our unique ability to study DEC access while also knowing prevalence of diabetes indicates that referral of patients with diabetes is quite consistent across income quintiles. Therefore, the utilization gradient seen truly reflects disease burden and implies that there is no access bias based on income. This is a positive finding but somewhat surprising in light of a history of studies suggesting that less affluent individuals have impaired access to care. We speculate that increasing patient awareness of the "diabetes epidemic" may be leading to more patients requesting referral. It is also possible that the DEC may be viewed as an extension of primary care. Low income populations have a higher burden of health problems in general, and the primary care physicians who serve these communities may view the DEC as the most efficient way to provide complex patients (ie those with diabetes) with the care they require. Given the finding of Dunlop and colleagues  of good access to primary care for lower income individuals, it is perhaps not that surprising that less affluent patients who are visiting their family physician frequently are also accessing the DEC.
Primary care physicians' threshold for referral warrants further examination. It was notable that among those with diabetes, the level of general education was significantly associated with referral. It is possible that better educated patients are better advocates for their health and as a result perhaps more likely to be referred earlier in the natural history of their condition. While we did not see a socioeconomic gradient in overall access to the DEC, we cannot exclude the possibility that wealthy individuals may have been referred earlier and with less co-morbid disease.
This study is limited by its cross-sectional nature. We also examined the association of SES and referral in the context of a centralized model of diabetes care. Our findings, while applicable to the health region and period under study, may not be indicative of how services are utilized elsewhere. It should be noted that the city under study is relatively wealthy. Only 6% of this study population lived at or below the national poverty line. However, post-hoc analyses demonstrated that while there was a higher prevalence of diabetes among those who live in poverty, access to diabetes care was not significantly different to those with higher income. This study also examined data aggregated to the level of dissemination area, and therefore was ecologic in nature. We must always be cautious when making inferences about individuals when a study examines a grouping of individuals (e.g., individuals living in a geographic area) rather than the individuals themselves (i.e., the ecologic fallacy). While finding that low income was associated with diabetes and referral to diabetes care, we cannot say that it was indeed the lowest income individuals in these neighborhoods that were the most likely to have diabetes or to be referred. It is plausible that within low income groups, it was those with relatively higher earnings that accessed care. Finally, the validity of using neighborhood income as a surrogate for individual level income has been called into question by previous studies [18–20]. There is emerging evidence, however, that neighborhood-level income is in and of itself an independent SES construct that is a valid predictor of health outcomes, over and above any effect relating to individual income .
In spite of these limitations, the present study provides encouraging data that diabetes care services are being accessed and utilized by those who require it. This study involved a unique combination of several data sources, and with this richness of data was able to show that while diabetes may be more prevalent in lower income regions, individuals living in such regions appear to be as successful at accessing diabetes care services as their wealthier counterparts.