RESUMO
Introduction: Studies have indicated a global high prevalence of hospital malnutrition on admission and during hospitalization. Clinical Nutritional Risk Screen [CNRS] is a way to identify malnutrition and manage nutritional interventions. Several traditional and non-computer based tools have been suggested for screening nutritional risk levels. The present study was an attempt to employ a computer based fuzzy model decision support system as a nutrition-screening tool for inpatients
Method: This is an applied modeling study. The system architecture was designed based on the fuzzy logic model including input data, inference engine, and output. A clinical nutritionist entered nineteen input variables using a windows-based graphical user interface. The inference engine was involved with knowledge obtained from literature and the construction of [IF-THEN] rules. The output of the system was stratification of patients into four risk levels from [No] to [High] where a number was also allocated to them as a nutritional risk grade. All patients [121 people] admitted during implementing the system participated in testing the model. The classification tests were used to measure the CNRS fuzzy model performance. IBM SPSS version 21 was utilized as a tool for data analysis with alpha = 0.05 as a significance level
Results: Results showed that sensitivity, specificity, accuracy, and precision of the fuzzy model performance were 91.67% [ +/- 4.92], 76% [ +/- 7.6], 88.43% [ +/- 5.7], and 93.62% [ +/- 4.32], respectively. Instant performance on admission and very low probability of mistake in predicting malnutrition risk level may justify using the model in hospitals
Conclusion: To conclude, the fuzzy model-screening tool is based on multiple nutritional risk factors, having the capability of classifying inpatients into several nutritional risk levels and identifying the level of required nutritional intervention