Validation of the Global Leadership Initiative on Malnutrition criteria for diagnosis of malnutrition and mortality prediction for people living with HIV or AIDS.

OBJECTIVES

To validate the Global Leadership Initiative on Malnutrition (GLIM) criteria to diagnose malnutrition in hospitalized people living with HIV or AIDS (HA) considering different combinations, using the Subjective Global Assessment (SGA) as the semi-gold standard, and to assess the predictive effects of malnutrition according to the GLIM criteria on hospital length of stay and mortality.

METHODS

Retrospective observational study including hospitalized people living with HA aged >18 y. Forty GLIM combinations were obtained by combining the different phenotypic and etiologic criteria. The concurrent validity was assessed according to the sensitivity and specificity values, and the agreement with the SGA was tested using κ values. Multivariate logistic and Cox regression models were used to test the independent predictors for longer length of stay (LOS) and mortality, respectively.

RESULTS

The sample comprised 320 patients (mean age, 44.6 ± 12.1 y; 69.1% were men, and 68.4% were malnourished, according to the SGA). The prevalence of malnutrition, according to GLIM, varied from 10.3% to 69.1%. The combination of any phenotypic criteria with the etiologic criteria of low food intake and the combination of any phenotypic criteria with the etiologic criteria of disease severity were independent predictors for mortality (Hazard Ratio: 2.09 [95% CI, 1.15-3.77] and 2.09 [95% CI, 1.25-3.51], respectively). The combination of low body mass index and reduced absorption was independently associated with LOS higher than the median value (Oodds Ratio; 2.57; 95% CI, 1.21-5.45).

CONCLUSIONS

Nine GLIM combinations had satisfactory sensitivity and specificity values to determine concurrent validity, all of them including weight loss and low weight; two combinations were independent predictors of mortality (any phenotypic criteria and low food intake or opportunistic infections), and one combination predicted longer LOS. Combining any phenotypic criteria with low food intake resulted in adequate concurrent and predictive validity.