The studies in Table 2 and others (89–91) demonstrate that the substance use disorders criteria represent a dimensional condition with no natural threshold. However, a binary (yes/no) diagnostic decision is often needed. To avoid a marked perturbation in prevalence without justification, the work group sought a threshold for DSM-5 substance use disorders that would yield the best agreement with the prevalence of DSM-IV substance abuse and dependence disorders combined. To determine this threshold, data from general population and clinical samples were used to compute prevalences and agreement (kappa) between DSM-5 substance use disorders and DSM-IV dependence or abuse, examining thresholds of two or more to four or more DSM-5 criteria (Table 3). As shown, prevalence was very similar, and agreement (ranging from very good to excellent) appeared maximized with the threshold of two or more criteria, so it was selected. Another recent large independently conducted study further supported this threshold (92).
