Mini nutritional assessment-short form test: criterion and predictive validity in older adults from a long-term care unity

Introduction: the Mini Nutritional Assessment Short-Form test (MNA-SF) is valid for malnutrition screening and diagnosis of older adults, but few studies evaluated if it predicts hospital length of stay (LOS) and were conducted in long-term care units.Objective: this study aims to evaluate the criterion and predictive validity of MNA-SF.Methods: a prospective observational study was conducted in older adults from a long-term care unity. MNA Long Form test (MNA-LF) and MNA-SF were applied, at admission and at discharge. Percentage of agreement, kappa and intra-class correlation coefficients (ICC) were determined. Sensitivity and specificity of MNA-SF were calculated. The independent association of MNA-SF with LOS (adjustment for Charlson index, sex, age, education) was assessed by Cox regression analysis [results presented as hazard ratio (HR) and 95 % confidence intervals (CI)].Results: this sample is composed of 109 older adults (62.4 % women), aged 66-102 years. According to MNA-SF at admission, 7.3 % of participants presented normal nutrition status, 55.1 % were at risk of malnutrition and 37.6 % were malnourished. Agreement, kappa and ICC were 83.5 %, 0.692 and 0.768 at admission, and 80.9 %, 0.649 and 0.752 at discharge. MNA-SF sensitivities were 96.7 % at admission and 92.9 % at discharge; specificities were 88.9 % and 89.5 %, at admission and at discharge. According to MNA-SF at discharge, being at risk of malnutrition (HR = 0.170, 95 % CI: 0.055-0.528) or malnourished (HR = 0.059, 95 % CI: 0.016-0.223) lowered the odds of being discharged to home or to usual residence.Conclusions: a high agreement was found between MNA-LF and MNA-SF. MNA-SF revealed high sensitivities and specificities. An independent association was found between risk of malnutrition or malnutrition by MNA-SF and LOS. The use of MNA-SF instead of MNA-LF should be considered in long-term care units given its criterion and predictive validity.