Objective: To explore the ADHD diagnostic performance of a screening instrument, and which DSM-5 ADHD number of symptoms (criterion A) was best associated with impairment in a sample of students from 106 primary schools in Nampula, Mozambique. Methods: A random sample of 748 students were assessed using SNAP-IV and 152 youths (76 positive and 76 negative screeners) were invited for psychiatric diagnostic confirmation. Results: The performance of the screening instrument for predicting ADHD diagnosis was poor (all AUCs < 0.53). No other cut-off worked best in predicting impairment than the six symptoms cutoff suggested by DSM-5 for both inattention (AUC = 0.78; 95% CI [0.69, 0.86]) and hyperactivity/impulsivity (AUC = 0.75; 95% CI [0.67, 0.84]). Conclusion: Our findings highlight the adequacy of the DSM-5 ADHD criterion A in an African culture but indicate low diagnostic performance of a screening instruments only based in parent or teacher reports on symptoms to predict ADHD diagnosis.