Introduction: Rates of cannabis use disorder (CUD) among vulnerable populations have increased in recent years, highlighting a need to equip providers with an efficient screening tool.

Materials and Methods: A short form of the Cannabis Use Disorder Identification Test-Revised (CUDIT-R) was developed by using item response theory and traditional statistical methods, with data from two community samples of cannabis users representing two countries. Four item selection methods (Rasch regression, test characteristic curve, logistic regression, discriminant function analysis) were employed to identify the optimal three-item shortened version. The diagnostic ability of the short form was evaluated by using receiver operating characteristic curves.

Results: Using a cut score of 2, the 3-item CUDIT-Short Form (CUDIT-SF; reliability alpha = 0.66, Sample 1; 0.80, Sample 2) identified 78.26% of participants in Sample 1 and 78.31% of participants in Sample 2 who met DSM-5 criteria for CUD, with 98% agreement in Sample 1 and 93% agreement in Sample 2 with the full CUDIT-R on CUD classifications using a cut score of 13. Specificity was 76.70 and 78.00 in Samples 1 and 2, respectively.

Conclusions: The CUDIT-SF may be useful in busy clinical settings for a stepwise screening. Further validation of this shortened version with larger samples and in different settings is warranted.