Spine (Phila Pa 1976). 2008 (Feb 15); 33 (4 Suppl): S170–175
Eleanor Boyle, PhD, Pierre Côté, DC, PhD, Alexander R. Grier, DC, MBA, J. David Cassidy, DC, PhD, DrMedSc
Centre of Research Expertise for Improved Disability Outcomes, University Health Network Rehabilitation Solutions, Toronto Western Hospital, Toronto, Ontario, Canada. email@example.com
STUDY DESIGN: Ecological study.
OBJECTIVES: To determine the annual incidence of hospitalized vertebrobasilar artery (VBA) stroke and chiropractic utilization in Saskatchewan and Ontario between 1993 and 2004. To determine whether at an ecological level, the incidence of VBA stroke parallels the incidence of chiropractic utilization.
SUMMARY OF BACKGROUND DATA: Little is known about the incidence and time trends of VBA stroke diagnoses in the population. Chiropractic manipulation to the neck is believed to be a risk factor for VBA stroke. No study has yet found an association between chiropractic utilization and VBA diagnoses at the population level.
METHODS: All hospitalizations with discharge diagnoses of VBA stroke were extracted from administrative databases for Saskatchewan and Ontario. We included incident cases that were diagnosed between January 1993 and December 2004 for Saskatchewan and from April 1993 to March 2002 for Ontario. VBA cases that had previously been hospitalized for any stroke or transient ischemic attack (TIA) were excluded. Chiropractic utilization was measured using billing data from Saskatchewan Health and Ontario Health Insurance Plan. Denominators were derived from Statistics Canada's annual population estimates.
RESULTS: The incidence rate of VBA stroke was 0.855 per 100,000 person-years for Saskatchewan and 0.750 per 100,000 person-years for Ontario. The annual incidence rate spiked dramatically with a 360% increase for Saskatchewan in 2000. There was a 38% increase for the 2000 incidence rate in Ontario. The rate of chiropractic utilization did not increase significantly during the study period.
CONCLUSIONS: In Saskatchewan, we observed a dramatic increase in the incidence rate in 2000 and there was a corresponding relatively small increase in chiropractic utilization. In Ontario, there was a small increase in the incidence rate; however, chiropractic utilization decreased. At the ecological level, the increase in VBA stroke does not seem to be associated with an increase in the rate of chiropractic utilization.
From the Full-Text Article:
We measured the incidence rate of VBA stroke in 2 Canadian provinces. Overall, the incidence was similar in Ontario and Saskatchewan ranging from 7.5 to 8.6 per 1,000,000 person-years, respectively. In both provinces, there was a spike in the annual incidence rate of VBA in 2000. The annual incidence rates ranged from 0.48 to 1.05 per 1,000,000 person-years in Ontario and the rate ranged from 0.39 to 1.79 per 1,000,000 person-years in Saskatchewan. Chiropractic utilization differed between the 2 provinces. In Ontario, there was a steady decrease in the utilization rate overtime whereas, Saskatchewan experienced fairly stable rates.
The incidence rates reported for Saskatchewan in our study are artificially increased because of the spike in the annual incidence from 1999 to 2002. Despite this, the incidence rates reported in our study are slightly lower than the rates reported by Lee et al, in Olmstead County, Minnesota, and Rochester, Minnesota for the period 1995 to 2003.  The hospitals in the state of Minnesota may have had greater availability and accessibility to diagnostic imaging than the provinces in Canada which may have resulted in increased reporting of VBA. In Canada, VBA may have been misclassified as another type of stroke because of the problems associated with accurately diagnosing this type of stroke. 
A large increase in the annual incidence rate of VBA stroke was found in Saskatchewan and a smaller increase in Ontario after 1999. This increase subsequently dropped in 2002 for Ontario and in 2003 for Saskatchewan. Such a large population increase in the incidence rate of noninfectious and rare health problems is puzzling. Four possible hypotheses were considered to explain this large increase in incidence: (1) changes in diagnostic methods; (2) changes in diagnostic coding; (3) a dramatic increase in the prevalence of a risk factor; and (4) change in diagnostic/reporting behaviors.
The first 3 hypotheses unlikely account for the increase and the subsequent decrease in the VBA incidence rate. Firstly, between 1998 and 2000, there were no significant technological improvements in the diagnostic methods or procedures used to diagnose VBA strokes that could account for observed increase in incidence. Secondly, there were no changes in the diagnostic codes used to report VBA strokes during the study period. The introduction of the ICD-10-CA codes for the provinces occurred slightly later. It was first introduced for some hospitals in Saskatchewan on April 1, 2001 and all Saskatchewan hospitals were using it by April 1, 2002. Ontario hospitals were using ICD-10-CA in 2002. Thirdly, an increase in the prevalence of risk factors for stroke could not subsequently explain the drop in the VBA incidence rate after the 2000 spike. The fourth hypothesis may be the most plausible. The 2 highly publicized deaths after chiropractic neck manipulations may have influenced the diagnostic and reporting behaviors of physicians resulting in a reporting bias.  A reporting bias occurs when patients are examined differently depending on the exposures they had before their stroke. In this case, it is possible that physicians may have focused their diagnostic investigation on VBA stroke in patients who had recently reported seeing a chiropractor before their hospital admission. This could result in more cases being diagnosed than what would have been diagnosed if the physicians were following their usual standard of care. Another possibility is that ischemic strokes that were considered “probable,” or “unlikely” VBA stokes before 1998 to 2000 may have been coded as VBA strokes during that period.
Under the assumption that chiropractic manipulation to the cervical spine is a risk factor for VBA stroke, it is possible that the increase in incidence may have been related to an increase in the utilization of chiropractic services. However, 2 observations do not support this hypothesis. First, we found that the incidence rate of VBA strokes was similar in Saskatchewan and Ontario even though chiropractic utilization was 10 times higher in Saskatchewan than in Ontario. Second, the sharp increase in the rate of VBA stroke occurred despite a decrease in chiropractic utilization in Ontario. In Saskatchewan, the sharp increase in the incidence of VBA strokes occurred whereas the chiropractic utilization remained fairly stable. These findings are in agreement with the results of Cassidy et al who found that there was no significant added risk associated with antecedent exposure to a chiropractor before VBA compared against the risk associated with exposure to a primary care physician in the ambulatory setting. 
Provincial payments of chiropractic services differed between the 2 provinces. Residents of Saskatchewan may have been required to pay additional fees above the government share for their chiropractic care unless covered by supplementary health benefits, and there are no limits to the number of encounters they may have in a given fiscal year. Ontario residents also had to make a copayment, but the provincial government placed a cap on the number of visits an individual may make, after which patients were responsible for the entire fee and the visit may not be captured in the administrative database. The number of visits was capped at 22 per year, until the fiscal year of 1999/2000 when it was capped at $150 per year (i.e., around 15 visits). This cap in coverage would have resulted in the under-reporting of the number of chiropractic services in Ontario. This may partly explain the drop in utilization rate in 1999.
This study is ecological in nature and is therefore liable to ecological bias. Therefore, we cannot draw inferences at the individual level on the association between chiropractic utilization and the risk of VBA stroke. We are unable to determine the exact cause of the increase in the annual incidence rate during the latter part of the 90s. We can only hypothesize that it may have resulted from the increased awareness of the “potential” risk of stroke associated with chiropractic care.
Another limitation of our study is that many strokes are likely misclassified in the administrative data.  This may explain the differences in the incidence rates with the American study.1 The strength of our study is that the administrative database used in the analysis captures all discharge diagnoses for all residents of the 2 provinces. It also involved using over 10 million person-years of data from Saskatchewan and slightly over 109 million person-years of data from Ontario to calculate the cumulative incidence rate.
The annual incidence rate of VBA strokes was fairly stable in Ontario and Saskatchewan between 1993 and 2004, except for 2000 when a sharp increase was observed. The increase in the incidence rate of VBA stroke could not be explained by a proportional increase in exposure to chiropractors at the ecological level of analysis. It may have been a reporting bias, influenced by media attention resulting from a coroner's inquest into a death from VBA stroke after chiropractic care.