Background: In a previous study we tested the predictive value of the Mini Nutritional Assessment (MNA) in an Italian population of frail elderly in long-term hospital care. The results of our study confirmed the MNA's excellent overall predictive value and sensitivity. Unfortunately we had a large number of false positive judgments, hence our study's low specificity, which we think was caused by two factors: 1. in most cases it was impossible to conduct a reliable subjective assessment of the patients' nutritional and health status. 2. most patients failed to respond to some of the MNA questions, which as a consequence received a "0" score. The result was an artificially low global MNA score even in well-nourished patients.
Objective and design: We tried to neutralize the effects of the defective answers by modifying the total score and the cut-off points of the test. Thus, we: 1. replaced the subjective assessment of health and nutritional status with an objective evaluation; 2. replaced the total score of MNA with the ratio of this value with the maximum of points that each subject can obtain without including the items for which we could not have a response. Similarly, the cut-off points (17 and 24) were replaced with the ratio of these values with the maximum of points obtainable by a complete MNA (30). Patients are classified as "malnourished" below 0.56, "at risk of malnutrition" between 0.56 and 0.79, and "well-nourished" from 0.8 up.
Results: This way, the overall predictive value of MNA is increased from 80.3 to 85.4% and the specificity from 12.8 to 25%, whereas the sensitivity increase is modest (from 98 to 98.1%).