Introduction Because of the pronounced and fast population ageing of low and middle-income countries, dementia is now a leading cause of disability. Verbal Fluency Tests (VFT) were used to assess cognitive functions in ELSA (Longitudinal Study of Adults’ Health - Brasil), a cohort that investigates incidence and predictors of chronic diseases among 15 000 civil servants (35–74 years old). The performance in VFT needs to be evaluated by a trained investigator, in order to count the words excluding repetitions and intrusions. As multicentric studies have to concern about the homogeneity of their interpretation, this study investigates the reliability of VFT scoring by supervisors of the six ELSA research centers, who independently judged 120 category (animals) and 120 phonemic (F letter) tests.

Methods The scores were compared to a reference standard score obtained by independent judgement of two experts and ratings reliability obtained by Intraclass Correlation Coefficient (ICC) and Bland-Altman plot.

Results Scores were very similar among centers and a high level of agreement was observed between each center and the reference standard. ICC values for both tests varied from 0.979 (95% CI 0.970 to 0.985) to 0.987 (95% CI 0.981 to 0.991). The Bland-Altman plot showed that the mean difference is small for both tests, and 95% of scores are located between the mean and 2 SDs.

Conclusion These results show the importance of the implemented measures of quality assurance and control and allow ELSA to proceed upon its main objective, to identify social, psychological and biological predictors of cognitive decline in a Brazilian adult population.