The study found that multiple dimensions of health along with environmental, socioeconomic, demographic, and health service spatiality predicted GP access and utilization in Sarnia. The importance of differentiating access and utilization, and furthermore taking a closer look at how they interact is exemplified in a high exposure environment. Within the context of compromised environmental quality, our findings demonstrate that neighbourhoods, individual and environmental characteristics can interact to predict access and utilization of health care. Besides the apparent impact of access to regular care on community satisfaction, our analysis revealed that the a priori, SES, health care needs and other health related variables were significantly associated with GP access. These results confirm findings from previous studies in Canada and abroad . For instance, the findings here support the OMA's report  on the increased cost of health care due to air pollution in the Sarnia region. The results demonstrate that there are significant and specific impacts of different pollutants and their spatial distributions on primary care access and utilization.
A previous study in Sarnia found that high annoyance scores were significantly related to both NO2 and SO2 levels . Other studies have found associations between NO2, cardiac autonomic dysfunction and increased blood pressure, which are also symptoms of stress and can lead to cardiovascular disease [49, 50]. NO2 can also increase the risk of respiratory tract infections through interaction with the immune system and SO2 contributes to respiratory problems for both healthy subjects and those with pulmonary disease . Increased likelihood of GP use from smoking cigarettes was attenuated by NO2 and BTEX, and the effect became insignificant with SO2 in the model. Therefore, we suggest that NO2, BTEX and SO2 in Chemical valley increase health care utilization because of their effects on respiratory and cardiovascular diseases prominently featured in the area . Thompson et al.  reported that daily fluctuations in benzene concentrations predicted acute asthma emergency admissions of children in Belfast, Northern Ireland, but our analysis showed no direct significant relationship between BTEX and our outcomes. This may be because the predicted distribution of these compounds when taken together does not reveal their individual impacts since their modes of dispersion and the way people respond to them differ. Most health effects associated with VOCs are observed over longer periods of time.
Our results suggest that low SES and stress enhanced by odour annoyance may be compounding health care demands, possibly due to increased susceptibility to adverse pollution impacts (e.g., compromised immune system) . This 'double burden of deprivation' has been identified in studies conducted in Worcester, U.S.A., and Montreal, Canada [15, 54]. We observed that the low income only predicted access significantly within the medium and high exposure zones and not the low exposure area. This finding signifies a worrisome interaction between burdens of social and environmental stress, and access to primary health care services. We also found that high exposure ranked respondents were more likely than low ranked respondents to lack GP access if renting, which provides further support for low SES being a barrier to health care.
The role of odour annoyance as a stressor in Sarnia is consistent with work by Shusterman et al.  who found that odour mediated mechanisms and annoyance contribute to how people judge and cope with air quality, and furthermore provide important diagnostic information in appraising the potential threats to health and well-being. Research on beliefs regarding toxicity of environmental pollution suggests that "if environments smell bad, they're probably damaging to health"  or at the very least, they may reinforce annoyance. Consistent with earlier studies , we found females were more likely to report high odour annoyance than their male counter parts. This could be because of gender differences in cognitive and affective processes .
Negative perceptions about the environmental health of one's neighbourhood can influence health outcomes along with predisposing determinants of health [58, 59]. Consequently, perceived personal susceptibility and severity of health threats are modified by psychosocial factors, which thereafter can influence compliance with medical recommendations and perceived benefits of preventive action . Benefits of access to regular care are not only preventive, but also associated with the responsibility placed on health care providers to inform and educate their patients about hazards of air pollution . It is therefore particularly concerning to find that residents in Sarnia who live in high exposure areas spend more time travelling and waiting for GP consultations. This may be due to primary care providers in Sarnia locating their practices in less polluted areas, thus leaving those in highly polluted areas to travel long distances to seek care. Further research is required to determine the apparent challenges of delivering primary health care services to these areas and their populations.
Although levels of estimated exposure for respondents in the current study were within provincial guidelines, our monitoring methods were not able to capture the impact of individual release events of airborne toxins or "bad air days"; they are not uncommon and are occasionally accompanied by warning sirens and emergency response guides that demand individual coping mechanisms . We can assume that residents at risk from chronic exposure are the same residents at most risk from these events. Residential areas within the high rank zone of Figure 1 provides target areas for improving health care delivery as population characteristics in this zone predicted inequitable access and coincide with high NO2 and SO2 concentrations that increased the likelihood of utilizing services.
There are a few limitations to this analysis that are worth noting. We note that our conceptualization of access related to a source of regular care at the time of the study and did not account for barriers to acquiring a family doctor or the fact that some people do not attempt to find regular care. Our measure of access did not include alternative sources of care such as walk-in clinics, Telehealth Ontario (free telephone consultation with a registered nurse) and nurse practitioners, and these providers arguably represent important points of access for certain deprived populations identified in this study. A study that looked at access to family physicians in Southwestern Ontario found that of the 9.1% of the population that did not have regular care, 55% used walk-in clinics and 13% used emergency rooms as their source of care . The study was also limited by lacking longitudinal measures of health and SES as the study used a cross-sectional design. O'Neill et al.  suggest studies on air pollution that include SES measures consider how they change through the course of life. They also propose that exposure assessment include the effect of daily movement, which our study design did not permit. Furthermore, we found that our model predicting health care utilization was relatively weak, but there were nonetheless significant associations with air pollution that have potential implications for policy.