Selection bias is of critical concern in the study of influenza vaccine effectiveness when using an observational study design. This bias is attributable to the inherently different characteristics between vaccinees and non-vaccinees. The differences, which are related both to vaccination and signs of clinical disease as an outcome, may lead to erroneous estimation of the effectiveness. In this report, we describe how selection bias among elderly nursing home residents may lead to a spurious interpretation of the protective effect of influenza vaccine. Our results should be a lesson in the importance of regarding selection bias when assessing influenza vaccine effectiveness.