Serum albumin concentration and clinical assessments of nutritional status in hospitalized older people: Different sides of different coins?

OBJECTIVES: Malnutrition is common in hospitalized older people, and some have advocated routine nutritional screening. Serum albumin and clinically based measures such as the Subjective Global Assessment (SGA) are two potential methods of assessing nutritional status in hospitalized older people. Although both measures are strongly associated with prognosis, it is not clear whether they measure similar or different clinical constructs. Our goal was to assess the degree of clinical concordance between these measures. DESIGN: Cross-sectional study. SETTING: The inpatient medical service of a university teaching hospital. PARTICIPANTS: Three hundred eleven older (aged :70) patients. MEASUREMENTS: We independently measured serum albumin and performed the SGA on 311 older medical patients (aged 2:70) shortly after hospital admission. The SGA classified patients as well nourished, moderately malnourished (generally 5% weight loss with mild examination findings), or severely malnourished (generally >10% weight loss with marked findings) based on findings from a directed history and examination. We compared the distribution of clinical rating in patients with differing albumin levels and examined diagnostic test characteristics of albumin as a predictor of malnutrition as diagnosed on clinical examination. RESULTS: The mean age of subjects was 79.9; 64% were women, 42% were African American. Discordance between albumin and the SGA was common. For example, 38% of patients with albumin levels of 4.0 g/dL or higher were at least moderately malnourished on the SGA, whereas 28% of patients with albumin levels lower than 3.0 g/dL were rated as well nourished. No choice of albumin level was associated with simultaneously acceptable sensitivity and specificity as a predictor of SGA ratings. The area under the receiver operating characteristic curve for albumin level as a predictor of SGA rating was 0.58, suggesting that the ability of either measure to predict the other measure is only marginally better than chance. CONCLUSIONS: Albumin levels and clinical assessments, two possible measures of nutritional status in hospitalized older people, are often discordant. To some extent, this reflects limitations in both measures as markers of nutritional status. However, it also demonstrates that, in this population, albumin and clinical assessments of nutritional status reflect fundamentally different clinical processes.