LAUSR

Critical review and evaluation of the article “Mental health and substance use in an urban First Nations population in Hamilton, Ontario” by Firestone et al. brought an important issue to the forefront, namely the use of respondent-driven sampling for participant recruitment. The use of respondent-driven sampling during the study’s recruitment period raises questions about the presence of any underlying biases that were not controlled for, nor addressed in any way throughout the article. This is problematic as unaddressed biases can hinder the external validity of the study. This study in particular is at risk for developing sampling bias as a consequence of respondent-driven sampling. Respondent-driven sampling is used when hard-to-reach populations are being studied. Although an accurate representation of the sample population is produced with respondent-driven sampling, the same cannot be said about its corresponding target population. Findings from respondent-driven sampling are often questionable with regard to their generalizability and reliability. Thus, cautionmust be considered when interpreting results, as respondent-driven sampling does not control for, or reduce, bias once it has occurred. Respondent-driven sampling was useful in reaching the hidden urban Aboriginal population in Hamilton, Ontario. It was also useful in collecting representative data on the urban Aboriginal sample population. However, problems with accuracy and representativeness arise when applying the study’s results to other urban Aboriginal populations elsewhere in Canada. Without abandoning respondent-driven sampling altogether, alternative considerations within its methodology could diminish the effects of sampling bias. The study used staff-selected seeds to participate and recruit additional participants. This method is common in most respondent-driven sampling studies, but can also influence the further generalizability of studies. The authors of the article could have selected an alternative method to recruit seeds and additional participants. That is, conducting respondent-driven sampling using self-presenting seeds and participant recruits. Self-presenting seeds and participant recruits approach the study independently (e.g., word-of-mouth) without explicit selection from researchers or staff (4). Unlike staff-selected seeds and participant recruits, there is evidence to support that selfselecting seeds reach a broader range and variation of the hard-to-reach population (1). While the self-presenting seeds and participants increase participant variation, they are able to enhance the accuracy and representativeness of respondent-driven sampling (4). Thus, the use of self-presenting seeds in the urban aboriginal population of Hamilton, Ontario could have expanded the range of participants, making the results more generalizable to other aboriginal populations.