In March 2003, the health care system in Toronto, Canada, was confronted with an outbreak of Severe Acute Respiratory Syndrome (SARS), a highly contagious and severe atypical pneumonia. [1–3] The infection control measures employed in response to this crisis were extraordinary and unprecedented. Actions to control SARS in the Greater Toronto Area included the closure of four hospitals, the cancellation of all non-emergency surgical services, an almost complete curtailment of inter-hospital patient transfers, and the postponement of most hospital-based outpatient clinics. [4, 5] Moreover, access to physicians, hospital laboratories, imaging studies, and other high-technology services was limited by infection control procedures. [5] The effect of these measures on access to health care was underscored by media reports of inappropriate delays for cancer therapy and the inability to perform urgent surgeries, sometimes resulting in death. [6–9]

The effects of widespread limitations of access to medical services have been studied from different perspectives. Studies of influenza outbreaks from the early 20^th century provide insight into healthcare restrictions and are applicable to the SARS experience because both are contagious respiratory illnesses with high mortality rates that are largely controlled through isolation procedures. Previous work has documented increased cardiovascular and other non-influenza related deaths during influenza outbreaks. [10–12] However, these increases were likely related to the effects of the infection itself rather than limited access to care. [13–17] Studies of diminished health care access due to withdrawal of medical services during physicians' strikes have documented significant alterations in health service utilization, intermediate outcomes, and various health-related processes of care. [18–20] However, no rigorous study has examined the effects of physician job action on overall population mortality. [21]

While there is little debate regarding the necessity of a concerted response to control the SARS outbreak, the delivery of health care to the general population was clearly reduced. Evidence suggests that in Toronto more essential services were less affected than services for low-acuity conditions. [5] An important question is whether sharply curtailed access to health care had an impact on population mortality due to causes other than SARS. This issue will be highly relevant in the event of a future infectious disease emergency, such as a recurrence of SARS, a highly virulent influenza epidemic, avian flu, the emergence of a new infectious organism, or a bioterrorist attack with an agent such as smallpox. We therefore sought to determine if the infection control measures undertaken during the 2003 SARS outbreak in Toronto were associated with a change in overall population mortality.

We examined overall mortality for the region most affected by the extraordinary infection control measures taken to contain the Toronto SARS outbreak in 2003. Despite a prolonged period of intense clinical service restrictions, we found no significant change in mortality rates compared with corresponding periods in previous years. The widespread limitation of access to medical services did not appear to have any short-term effects on deaths within the population.

Did the infection control measures and policies employed to control the SARS outbreak produce collateral damage to the delivery of health care to the population? The containment strategy did result in dramatic decreases in non-urgent inpatient and outpatient procedures and surgeries, outpatient diagnostic tests, and overall physician visits. [5] Although overall emergency department visits declined substantially during the SARS outbreak, this decline apparently reflected significant reductions in the volume of patients with low-acuity complaints. [5, 26] Hospitalizations in Toronto during this period were 12% lower than expected, representing a relatively modest decrease. [27]

Studies conducted in other jurisdictions suggest that the changes in health care delivery and care-seeking behavior associated with the SARS outbreak may have had negative health effects. In Taiwan, use of ambulatory care fell by 24%, inpatient care decreased by 35%, and childbirths shifted from larger medical centers to less well-equipped district hospitals and clinics. [28, 29] In Hong Kong, the number of individuals diagnosed with active tuberculosis decreased significantly during the outbreak, suggesting that the population's avoidance of medical care led to delays in diagnosis. [30] On the other hand, another study conducted in Hong Kong found that the proportion of positive samples for influenza and respiratory viruses other than SARS decreased significantly during the SARS outbreak, a change that the investigators attributed to the population's heightened attention to respiratory hygiene. [31] Because these acute viral infections can lead to serious complications among individuals with chronic medical conditions, this trend may have had a beneficial effect on the overall mortality rate.

Our data are consistent with the assertion that patients with severe illnesses retained the ability to access life-saving services during the SARS outbreak in Toronto. Fortunately, the 2003 outbreak was relatively short-lived and limited in scope. Future outbreaks of influenza or other emerging infectious diseases may be much more severe and prolonged. The establishment of data capture and reporting systems in advance of such an event would provide decision-makers with timely information on possible increases in morbidity and mortality attributable to impaired health care access, in addition to data on the direct effects of the outbreak itself. This information could be used to determine which services should be prioritized during a period of severely restricted health services.

This work has certain limitations. First, our study was designed to detect increases in short-term mortality that would constitute an immediate health challenge, but not longer-term effects. Although such delayed consequences are not implausible, they would be difficult to identify and attribute to impaired access to care during the SARS outbreak. Second, our study did not consider outcomes such as health status, quality of life, or other measures of morbidity that may have been affected by SARS-related infection control measures. Instead, we focused on mortality, an outcome that is both readily measured and of unquestioned importance. Future studies are needed to assess the effects of extraordinary infection control measures on morbidity. Third, we chose a clinically plausible lag of four weeks before the onset of change in mortality rates for our interrupted time-series analysis. The selection of a different lag period might have resulted in slightly different findings, although our results were corroborated by the Poisson regression analysis. Finally, this study examines a "natural experiment" in which restrictions on health care services were not applied uniformly, instantaneously, or in a randomized manner. Nonetheless, our use of a control group consisting of the population of Ontario outside the Greater Toronto Area, which was not affected by severe restrictions on access to hospital-based care, serves to minimize this potential bias.

A coordinated response to severe infectious disease outbreaks requires an approach that balances an infection control mandate with the need to preserve access to essential health services. In the situation of a widespread disease outbreak, health care decision-makers understandably concentrate upon the immediate threat of the infectious disease. The SARS experience in Toronto suggests that the preservation of the delivery of health care for other urgent conditions is equally important. This concept is particularly relevant because future infectious disease outbreaks, such as avian influenza or a bioterrorist attack, could limit access to health care resources to a much greater extent than did the SARS outbreak. Further research is needed to examine the possible effects of such events on cause-specific mortality rates and on health outcomes other than mortality.