Sir,

The article by Maberley et al1 adds to our understanding of ophthalmic and concomitant general medical conditions experienced by inner city residents in Vancouver's downtown eastside. Although this report is useful in its own rights, it is also important to highlight some of the methodological limitations of this study that will have a bearing on the interpretation of the findings. Firstly, the way study participants were recruited raises questions as to how representative this group is to the general community. It is problematic when community prevalence of a disease/condition has to be generated from attendees of health care, as these are unlikely to represent fairly the general population from which they come from. So from the study, it can be said that the prevalence estimates obtained would represent better the population of patients who attend this health facility for eye care. Even when we do that, the participants in this study were not randomly recruited, as they were consecutive patients on special days when recruitment occurred. How representative these special days and times (2 h) are raise further questions on the representativeness of the ‘sample’.

The authors also report that ocular examinations were conducted by a single ophthalmologist and this can be a source of systematic error as compared to when more than one person makes a diagnosis on the same patient. The comparison of the prevalence obtained in this study to an earlier study aimed to study general community prevalence2 needs to be made with the differences in the study designs in mind. In the 2005 study, all patients attending a health facility over a 5-year period were eligible. Diagnoses were verified through use of more than a single ophthalmologist. Owing to the large sample size (N=962), these results would be more representative of the study population than in the current (inner city) study.

Finally, the authors need to be commended for presenting not just ‘point prevalence values’ but also confidence intervals (CI) for this parameter. This obviously helps the reader to see that virtually all of the values have wide CI raising questions on the precision of these estimates. For example, nonstandardized prevalence of visual disability was 500 per 10 000 (95% CI, 242–900 per 10 000), and nonstandardized prevalence of low vision and blindness 400 per 100 000 (95% CI, 174–770).