Objective: To determine the contribution of reassessment in the attribution process of neuropsychiatric (NP) events to SLE or other aetiologies in a large, prospective and multidisciplinary... Show moreObjective: To determine the contribution of reassessment in the attribution process of neuropsychiatric (NP) events to SLE or other aetiologies in a large, prospective and multidisciplinary assessed NPSLE cohort and to compare these results with other available attribution models for NP events occurring in SLE. Methods: Three hundred and four consecutive SLE patients presenting NP events were evaluated. All subjects underwent standardized multidisciplinary medical, neuropsychological, laboratory and radiological examination on the inclusion and reassessment dates. Diagnosis was always established by multidisciplinary consensus. The final diagnosis after reassessment also took into account disease course and response to treatment. These data were compared with currently available attribution models for NP events in SLE. Results: A total of 463 NP events were established. After reassessment, attribution to SLE was discordant in 64 (13.8%) NP events when compared with the first visit. We show that 14.5% of NP events previously attributed to SLE reclassified as non-NPSLE. In 86.4% of these patients immunosuppressive therapy was started after the first visit. When reassessment and available attribution models were compared, NPSLE cases overlapped considerably. Although specificity was high for all comparisons (0.81–0.95), an important variation in sensitivity (0.39–0.83) and agreement estimates (κ = 0.29–0.68) was observed. The Italian algorithm showed the highest sensitivity and specificity (>0.80) and moderate agreement (0.59–0.64). Conclusión: In clinical practice NP events presenting in SLE are too often attributed to an immune-mediated origin. Multidisciplinary reassessment avoids misclassification in NPSLE. Multidisciplinary reassessment is the reference standard in NP events presenting in SLE and cannot be replaced by available attribution models Show less