We found that individuals from low-income neighborhoods were less likely to present with TIA, had longer times from stroke onset to hospital arrival, and had slightly lower rates of neurologist care compared to those from higher income neighborhoods. However, all other measured aspects of stroke care delivery were similar across income groups. At 1 year after stroke there was a persistent survival advantage for those in the highest compared to the lowest income areas, even after adjustment for age, sex, cardiovascular risk factors, stroke type and severity, in-hospital processes of care, and hospital and physician characteristics. These identifiable risk factors only accounted for a small proportion (13.9%) of the observed disparity in survival between the low- and high-income groups.
An important finding is that mortality at 30 days was similar among socioeconomic groups, and mortality differences were not observed until after the acute care period. This may be due to lower event rates at 30 days with consequent reduced power to observe smaller differences, or due to the fact that stroke severity—rather than processes of care or baseline risk factors—is the principal determinant of early stroke mortality.31
However, this observation also suggests that stroke aftercare may be an important determinant of late mortality after stroke, and is consistent with the structure of our health care system where, in a broad sense, acute care is similar for all.
The finding of lower stroke case fatality rates with higher socioeconomic status has been previously documented, even in Canada and in other countries with universal access to health care.6, 9, 13, 15
Our results suggest that better outcomes among those from high-income areas are not due to marked differences in acute stroke care delivery, and are only partially explained by baseline risk factors and stroke characteristics. This contrasts with findings from a Canadian study of patients with acute myocardial infarction, where age and vascular risk factors appeared to account for the majority of the income-mortality gradient.30
We note that within the CIs, a larger effect of baseline risk factors is possible, and risk factor medication should therefore be a target for interventions designed to improve outcomes in low socioeconomic groups. It is also likely that unmeasured risk factors, including important stroke risk factors such as blood pressure control, physical activity, and waist-to-hip ratios, could account for some of the difference in survival between those from low- and high-income areas, as could differences in adherence to medications and access to follow-up care.32
Of note, in our cohort, higher socioeconomic status was associated with a clustering of risk factors likely to be associated with better outcomes, including low smoking levels and less diabetes, as well as a clustering of sociodemographic factors which could affect access and adherence to care, such as urban residence, higher education, male sex, and fluency in English or French.
Although socioeconomic status has been shown to be an important predictor of the use of certain medical services, such as angiography after myocardial infarction, previous studies have found no consistent association between socioeconomic status and stroke care delivery.2,13,33–36
Given the association between lower socioeconomic status and rural residence, with the potential for limited access to larger hospitals with more stroke-specific resources, one might have anticipated lower rates of specialized interventions such as carotid imaging, neuroimaging, and stroke unit care among patients from low-income neighborhoods. However, Ontario has developed a provincial stroke system aimed at optimizing and coordinating care across the province, and although the system had not been fully implemented during the study period, this might have permitted some patients from rural areas to have been transferred for care and evaluation at regional stroke centers.37
Thus, our results may not be generalizable to jurisdictions without a coordinated stroke care strategy, or to countries without universal access to health care. Of note, the finding of a longer time from stroke onset to hospital arrival in those from lower compared to higher income areas—despite similar rates of ambulance use and similar transport times—suggests that those from lower income neighborhoods may have delays in symptom recognition or in activation of emergency medical services. Knowledge of stroke symptoms and the need for rapid assessment of transient symptoms could also explain the more frequent presentation with TIA in the higher income groups.
Our study has a number of limitations which merit comment. First, we used area-level rather than individual-level measures of socioeconomic status. This may result in nondifferential misclassification of socioeconomic status and lead to underestimates of social gradients in mortality. However, neighborhood-level income measures have the advantage of capturing aspects of neighborhoods, such as the availability of parks, schools, and hospitals, that may affect health, and have been found to provide results that are complementary to those found using individual level data.38,39
Second, we used income as our primary measure of socioeconomic status rather than wealth, occupation, social class, or other composite measures. Although most measures of socioeconomic status tend to be highly correlated and to predict mortality in a similar direction, they may not be consistent in direction and magnitude for all outcomes.14,17,40
In addition, although the socioeconomic status of both an individual and an area may fluctuate over time, the cross-sectional nature of our study meant that we were only able to capture neighborhood income at a single time point. Finally, our data were collected through hospital chart audits rather than prospective data collection or patient or provider interviews. Thus, we were unable to capture information on stroke symptom awareness and potential barriers to stroke care delivery, and we were unable to evaluate risk factor modification and treatment adherence after hospital discharge. In addition, we do not have information on potentially important explanatory variables such as obesity, social isolation, and mental illness. Despite these limitations, our database contains high-quality clinical data on a population-based patient sample, and is strengthened by its linkage to administrative databases to provide long-term follow-up outcomes.
In Ontario's universal health care system, we found higher survival rates after stroke for individuals from the richest compared to the least wealthy neighborhoods. This was not fully explained by differences in stroke type, stroke severity, or processes of stroke care delivery, and may be related in part to a lower burden of unmeasured baseline risk factors for poor health. The superior health outcomes seen for the wealthiest members of society may provide a benchmark for what could be achieved for all through targeted interventions to reduce socioeconomic disparities in health.