Use of the Patient-Generated Subjective Global Assessment to Identify Pre-Frailty and Frailty in Hospitalized Older Adults

Objectives: The Scored Patient-Generated Subjective Global Assessment (PG-SGA) and Edmonton Frail Scale (EFS) are widely used in acute care settings to assess nutritional and frailty status, respectively. We aimed to determine whether the scored PG-SGA can identify pre-frailty and frailty status, to simultaneously evaluate malnutrition and frailty in clinical practice.

Design: Cross-sectional study.

Settings And Participants: A convenience sample of 329 consecutive patients admitted to an acute medical unit in South Australia.

Measurements: Nutritional and frailty status were ascertained with scored PG-SGA and EFS, respectively. Optimal cut-off scores to identify pre-frailty and frailty were determined by calculating the Scored PG-SGA's sensitivity, specificity, positive and negative predictive values, Youden Index (YI), Liu index, Receiver Operator Curves (ROC) and Area Under Curve (AUC). Nutritional status and patient characteristics were analysed according to frailty categories.

Results: The optimal cut-off PG-SGA score as determined by the highest YI, to identify both pre-frailty and frailty was >3, with a sensitivity of 0.711 and specificity of 0.746. The AUC was 0.782 (95% CI 0.731-0.833). In this cohort, 64% of the patients were well-nourished, 26% were moderately malnourished and 10% were severely malnourished. Forty-three percent, 24% and 33% of the patients were classified as robust, pre-frail and frail, respectively. Bivariate analysis showed that those robust were significantly younger than those who were pre-frail (-2.8, 95% CI -5.5 to -0.1, p=0.036) or frail (-3.4, 95% CI -5.9 to -1.0, p=0.002). Robust patients had significantly lower Scored PG-SGA than those who were pre-frail (-2.5, 95%CI -3.8 to -1.1, p<0.001) or frail (-4.9, 95% CI -6.1 to -3.7, p<0.001).

Conclusion: The Scored PG-SGA is moderately sensitive in identifying pre-frailty/frailty in older hospitalized adults and can be useful in identifying both conditions concurrently.