Our data were collected in the early 1990s and may therefore not reflect current practice patterns. Although the Canadian Study of Health and Aging was not designed to study influenza vaccination, it does provide a large and unique data set of primarily community-dwelling older Canadians, and is therefore potentially useful in the examination of health-related risk factors and demographics that may influence decisions to vaccinate. Given the importance of influenza vaccination in the prevention of significant morbidity and mortality in populations at risk, the vaccination rate of 55.2% in our community-dwelling sample of older adults is concerning. People who were not vaccinated tended to be younger, non-smokers and to have fewer co-morbid illnesses. They were also found to have a lower level of education, not to be married and not to engage in regular exercise. These were the factors that retained statistical significance in the multivariable analysis, suggesting that these associations are unlikely to have arisen due to confounding by any of the variables investigated.
Vaccination has previously been studied in the CSHA, but only in relation to its status as a potentially protective factor with respect to cognitive impairment [21]. The weight of evidence derived from re-examination of large databases is less than that derived from specifically designed trials, however this method still maintains an important role in epidemiological research. For example, evaluation of systematic problems can be used to help develop targeted efforts in improving vaccination rates. Our data do not include information on institutionalized older adults, where it might be expected that at-risk profiles would vary, and where vaccination rates are generally higher [16].
Another important source of potential error is our reliance on self-reported immunization status. However, self-report of influenza vaccination status in elderly outpatients has been found to be highly sensitive and moderately specific when checked against medical record documentation [22].
Other studies have identified predictive factors for vaccination in other countries [8, 12]. An American study found that patients with more health conditions, higher rates of use of health care resources, and history of pneumonia were more likely to be vaccinated, while non-vaccinated individuals were older and more likely to have dementia or stroke [8]. An Iowa study identified a number of factors associated with the receipt of both influenza and pneumococcal vaccines: age over 70, self-owned residence, working, increased number of medical conditions, current prescription medication, and a physician visit within the past year. Geography (rural vs. urban living) was unrelated to vaccine receipt [12].
Our data suggest that there appears to be an important degree of targeting of vaccination resources within the older adult population to people who are less healthy. Even within these higher risk groups, however, immunization is incomplete. Perhaps such people are perceived by their health care providers as being at higher risk from influenza infection, and are thus more likely to be immunized. Similar explanations may account for the higher vaccination rates among smokers and those who consumed more alcohol. However, according to the current Canadian guidelines [9], influenza vaccination is indicated for all persons 65 years of age and over. If it is the case that those at the younger end of this age group are less likely to be vaccinated, as our study suggests, more must be done to ensure that vaccination is reaching its entire target population.
Regular exercise was shown to be a factor predictive of influenza vaccination. Interestingly, this association may have been confounded by generally health-protective behaviour (which might be expected to be associated with both regular exercise, the explanatory variable, and vaccination, the outcome), given that individuals exhibiting healthy lifestyle choices (such as exercise) may have been more likely to have sought preventative health care and to have visited their health care providers more regularly, thus providing more opportunity to be vaccinated. Regular exercise did, however, retain significance in our multivariate analysis.
The regional differences identified in our study may point to geographic differences in access to influenza vaccination, although the general milieu and level of awareness in both the medical community and society at large may also be significant. These regional differences suggest that certain areas may benefit from targeting of vaccination efforts. However, over and above questions of health policy and public health education, the question of access to vaccination is of vital importance if we are to achieve acceptable rates of coverage in this target population. The finding that rural residence is a negative predictor of vaccination is particularly concerning, and points to the larger equity issue of how uptake of influenza vaccination can be improved outside of major urban centers. However, our finding that rural vs. urban residence did not retain statistical significance in the multivariable regression model suggests that this crude association may be due to confounding by other factors.
Our analysis comparing non-responders with those for whom influenza vaccination status was known on the basis of the SARFQ showed that response bias was indeed likely. Non-responders were more likely to drink more alcohol, a factor which was associated with being immunized. However, several factors which were associated with not being immunized, including younger age, good self-rated health, having fewer comorbidities, and residence in a rural area, were more prevalent among non-responders. This over-representation of a number of factors predictive of non-immunization among non-responders suggests that our results may well have been influenced by response bias. For example, we may have over-estimated the prevalence of vaccination in this population.
The analysis comparing regular users of the influenza vaccine with those who reported first-time immunization in the SARFQ demonstrated a number of factors that differed between the two groups. The finding that younger age is associated with first time vaccination may be at least partially explained by those at the younger end of the > = 65 age group having spent less time in this "target group" for vaccination. As such, they would be more likely to be receiving the vaccination for the first time. The effect of age may also play a role in the finding that those with better self-assessed health and fewer comorbidities were more likely to be first-time users (as they may also have been younger). However, being in better health was also a predictive factor for non-vaccination within the previous two years, as was rural residence, suggesting that these factors may influence decisions to initiate as well as to sustain influenza vaccination over subsequent years.