FEATURE ARTICLEModel for Differential Nursing Diagnosis of Alterations in Urinary Elimination Based on Fuzzy LogicDE MORAES LOPES, MARIA HELENA BAENA PhD; SIQUEIRA ORTEGA, NELI REGINA PhD; MASSAD, EDUARDO PhD; MARIN, HEIMAR DE FÁTIMA PhDAuthor Information Author Affiliations: Nursing Department, Faculty of Medical Sciences, University of Campinas (Dr Lopes); School of Medicine, University of São Paulo (Drs Ortega and Massad); Nursing Department, Federal University of São Paulo, Brazil (Dr Marin). This work was supported in part by a grant from Informatics for Global Health (1#1 D43 TW7015-01), Fogarty International Center, NLM, NIBIB, NIH (principal investigator, Dr Lucila Ohno Machado). It is also supported by the National Council of Scientific and Technological Development (Conselho Nacional de Desenvolvimento Científico e Tecnológico-CNPq). Corresponding author: Maria Helena Baena de Moraes Lopes, PhD, Rua Conceição, 552-apto. 25-Cambuí, 13010-050 Campinas, SP, Brazil (firstname.lastname@example.org; email@example.com). CIN: Computers, Informatics, Nursing: September-October 2009 - Volume 27 - Issue 5 - p 324-329 doi: 10.1097/NCN.0b013e3181b21e6d Buy Metrics Abstract Nursing diagnoses associated with alterations of urinary elimination require different interventions. Nurses, who are not specialists, require support to diagnose and manage patients with disturbances of urine elimination. The aim of this study was to present a model based on fuzzy logic for differential diagnosis of alterations in urinary elimination, considering nursing diagnosis approved by the North American Nursing Diagnosis Association, 2001-2002. Fuzzy relations and the maximum-minimum composition approach were used to develop the system. The model performance was evaluated with 195 cases from the database of a previous study, resulting in 79.0% of total concordance and 19.5% of partial concordance, when compared with the panel of experts. Total discordance was observed in only three cases (1.5%). The agreement between model and experts was excellent (κ = 0.98, P < .0001) or substantial (κ = 0.69, P < .0001) when considering the overestimative accordance (accordance was considered when at least one diagnosis was equal) and the underestimative discordance (discordance was considered when at least one diagnosis was different), respectively. The model herein presented showed good performance and a simple theoretical structure, therefore demanding few computational resources. © 2009 Lippincott Williams & Wilkins, Inc.