Diagnostic accuracy of the PG-SGA short form and NRS 2002 in internal medicine ward

Rationale: The PG-SGA is a validated instrument to assess malnutrition and its risk factors. Its patient component, i.e. the PG-SGA Short Form (SF), can be used as screening instrument. In this multicenter study, we aimed to assess diagnostic accuracy of the PG-SGA SF and NRS 2002, in patients at the Internal Medicine ward.Methods: In 192 patients (76.0±13.5 years; 53% female) in 9 Portuguese internal medicine wards, malnutrition risk was assessed by PG-SGA SF and NRS 2002. PG-SGA SF ≤8 was defined as low/medium malnutrition risk and NRS 2002 ≤2 as low risk. PG-SGA SF ≥9 and NRS 2002 ≥3 were defined as high malnutrition risk. Nutritional status was assessed by the full PG-SGA (reference method). Malnutrition was defined as PG-SGA Stage B (moderate/suspected malnutrition) or Stage C (severely malnourished). Diagnostic accuracy was tested by sensitivity, specificity, positive and negative predictive value, and receiver operating curve. Agreement between PG-SGA and NRS-2002 was tested by McNemar’s test and Cohen’s kappa (κ).Results: Forty-six % and 53% were categorized as at risk of malnutrition by PG-SGA SF and NRS 2002, respectively. In total, 55% were malnourished. Sensitivity, specificity, positive and negative predictive value of PG-SGA SF and NRS 2002 were 0.84, 1.00, 1.00, 0.83 and 0.74, 0.74, 0.77 and 0.70, respectively. Area under curve of PG-SGA SF and NRS 2002 was 0.987 and 0.778 respectively. McNemar’s test showed no significant disagreement (p=0.86) between PG-SGA SF and NRS 2002. Cohen’s kappa showed weak agreement (κ=0.492; p<0.001) (Table 1).Conclusion: Our findings indicate that in patients at the internal medicine ward, PG-SGA SF shows better diagnostic accuracy than NRS 2002, i.e. better sensitivity and specificity.