help-seeking, such as local availability of SUD treatment services (Cummings et al., 2014), was not available. Fourth, we based our analyses in DSM-IV categories. Recent work proposes cross-walks between DSM-IV and DSM-5 disorders and suggests different potential thresholds (Compton et al., 2013). Our analyses, indicating that more severe SUD categories are associated with greater likelihood of treatment suggest that higher thresholds in those cross-walks tend to be associated with greater rates of treatment. Future studies, collecting information on DSM-5 categories should test this prediction. Fifth, to maximize statistical power and increase the stability of the estimates, we examined jointly bipolar I and II disorders, although SUDs are more strongly associated with bipolar I than bipolar II disorder and they may differ in their clinical characteristics, patterns of comorbidity, course, and prognostic implications (Chamorro et al., 2012; Moreno et al., 2012; Sala et al., 2014, 2012). Similarly, we examined all drugs together, although patterns of treatment-seeking may vary by drug.
