Third, since the survey was interviewer-administered, responses could have been biased towards social desirability, which could underestimate statistical associations. Fourth, given the small sample size, there are limitations pertaining to adequate statistical power for multivariable logistic regression models, including decreased precision of point estimates and higher probability of making statistical errors.