Background: A new analytical variation of bioelectrical impedance vector analysis (BIVA), called specific BIVA, has shown to be more accurate in detecting changes in fat mass than classic BIVA. Objective: To compare classic and specific BIVA in order to identify which is more strongly associated with psychofunctional and nutritional indicators in a group of institutionalised elderly patients with dementia. Subjects and methods: Cross-sectional study. Fifty-four patients (34 women, 20 men) with dementia in moderately severe to very severe stages and aged 60-95 years underwent geriatric nutritional assessment, including body mass index calculations, the Mini Nutritional Assessment, the Geriatric Nutritional Risk Index, and whole body composition analysis. Results: With specific BIVA (unlike with classic BIVA), significant differences were found between women with moderately severe and very severe dementia. In the BIVA conducted for body mass index, the confidence ellipses produced with the classic BIVA approach were highly overlapping; but with specific BIVA, significant differences were observed between the women in different nutritional categories (malnutrition, risk of malnutrition, normal weight and obesity). On the other hand, both approaches distinguished malnourished women from those who were at risk of malnutrition, according to the Mini Nutritional Assessment; and men with a moderate-high risk of malnutrition from men with no risk, on the basis of the Geriatric Nutritional Risk Index. Conclusion: Overall, the findings of the present study suggest that specific BIVA is more effective than classic BIVA in identifying bioelectrical changes associated with psycho-functional and nutritional indicators in institutionalised elderly with dementia.