The present findings should be interpreted in light of common limitations in most large-scale surveys. First, information on substance use and SUD was based on self-report and not confirmed by objective methods. Second, diagnoses may be subject to recall bias (the longer the time interval between the event and assessment, the higher the probability of incorrect recalls) and to cognitive impairment associated with the use of drugs (Grant et al., 2003c; Hasin and Liu, 2003). Third, in some cases, substance use may have been intermittent, rather than continuous, which would produce inexact estimates of the time elapsed from use onset to dependence. Fourth, the NESARC excludes institutionalized populations, who have an increased probability of having a diagnosis of substance dependence (McCutcheon et al., 2009). Fifth, factors that may help explain the racial-ethnic differences, such as income could not be included in the models as time-dependent covariates since the NESARC, as most large scale surveys, does not include information on changes in income overtime.
