Fig. 1.

Study Design

Note. A - Data collection process. B - Machine Learning approach. After randomly undersampling, machine learning algorithms were generated with data from participants with severe (n = 142) and non-severe (n = 142) withdrawal syndrome at the end of treatment, based on the cut-off of 21 points in the CSSA. Nested cross-validation was performed with six rounds in the outer loop and three rounds in the inner loop. ASI-6 - Addiction Severity Index, 6 edition; CSSA - Cocaine Selective Severity Assessment; CTQ - Childhood Trauma Questionnaire.