Introduction: The diagnosis of dementia requires documentation of cognitive loss with respect to the patient's previous level. It would therefore be very useful to have models available which would predict the result expected in the tests normally used in the diagnosis of dementia.
Objective: To validate a model for prediction of the results of a test of semantic verbal fluency in persons with no dementia.
Patients and methods: A model for the prediction of semantic verbal fluency deduced from a sample of 138 persons was applied to two other independent samples: the first of 86 persons from the same environment as the original sample and a second multicentric sample of 92 persons. The validity of the model was evaluated by residual analysis.
Results: No sex differences were seen between the samples, but there were differences regarding the other variables, including the observed and predicted verbal fluency. The residuals of the samples did not differ from each other nor vary from zero, but were normally distributed.
Conclusion: The model proposed based on sociodemographic and clinical variables is valid and satisfactorily predicts the verbal fluency to be expected in each case.