BackgroundThe severity of substance use disorder (SUD) is currently defined by the sum of DSM-5 criteria. However, little is known about the validity of this framework or the role of additional severity indicators in relapse prediction. This study aimed to investigate the relationship between DSM-5 criteria, neurocognitive functioning, substance use variables and cocaine relapse among inpatients with cocaine use disorder (CUD). Methods128 adults aged between 18 and 45 years were evaluated; 68 (59 males, 9 females) had CUD and 60 (52 males, 8 females) were healthy controls. For the group with CUD, the use of other substances was not an exclusion criterion. Participants were tested using a battery of neurocognitive tests. Cocaine relapse was evaluated 3 months after discharge. ResultsScores for attention span and working memory were worse in patients compared to controls. Earlier onset and duration of cocaine use were related to poorer inhibitory control and global executive functioning, respectively; recent use was related to worse performance in inhibitory control, attention span and working memory. More DSM-5 criteria at baseline were significantly associated with relapse. ConclusionsRecent cocaine use was the most predictive variable for neurocognitive impairments, while DSM-5 criteria predicted cocaine relapse at three months post treatment. The integration of neurocognitive measures, DSM-5 criteria and cocaine use variables in CUD diagnosis could improve severity differentiation. Longitudinal studies using additional biomarkers are needed to disentangle the different roles of severity indicators in relapse prediction and to achieve more individualized and effective treatment strategies for these patients.