Comparison of screening tools in patients undergoing haemodialysis

Abstract

Background: Subjective Global Assessment (SGA) had been recommended by the Renal Association (Wright & Jones 2010) as an appropriate nutritional screening tool for the detection of malnutrition in haemodialysis (HD) patients. However, it has been criticised for being time consuming (Kondrup et al. 2003) and lacking sensitivity (Kalantar-Zadeh et al.1999). The Patient Generated Subjective Global Assessment (PG-SGA) requires the patient to complete part of the screening tool; thus reducing the amount of time the dietitian spends completing it. A recent pilot project showed that, compared to SGA, PG-SGA was able to distinguish significant differences between several independent measurements of nutritional status (demonstrating construct validity) (Hyam et al., 2010). In contrast the Malnutrition Universal Screening Tool (MUST) did not identify risk of malnutrition in HD patients. The aim of the current project was to assess a greater number and range of patients (with respect to nutritional status) using the three tools, in order to confirm whether PG-SGA was the most appropriate nutritional screening tool for use in HD patients. Methods: The nutritional status of 65 HD patients (n = 43 male; n = 22 female) attending two renal units; St Georges and Canterbury, was assessed by one investigator at each unit (KF and RP respectively). There was no exclusion criteria, however investigators aimed to assess a greater proportion of patients with suspected malnutrition (identified by SN and SR) as this group was possibly under-represented in the previous study. Data included SGA, MUST, PG-SGA, grip strength, anthropometric and biochemical parameters. Construct validity was examined using ANOVA to compare differences between the nutritional values of those patients classified by the nutritional screening tools as being at no risk, moderate risk or high risk of malnutrition. Criterion validity was assessed using the kappa statistic to measure the level of agreement between SGA (as the gold standard), PG-SGA, and MUST. Results: SGA classifications of nutritional status were able to distinguish significant differences in: Albumin (P = 0.03); CRP (P = 0.023), hand grip strength (P = 0.001), average daily energy intake (P = 0.005), protein intake (P = 0.01) and haemoglobin (P = 0.0001). PG-SGA only identified significant differences in hand grip strength (P = 0.008) and mid arm circumference (P = 0.024). MUST identified differences in current BMI (P = 0.04), mid arm circumference (P = 0.005) and CRP (P = 0.03). A moderate kappa statistic (0.51) was recorded when comparing the level of agreement between PGSGA and SGA classifications of patients. Virtually no agreement was found between SGA and MUST (( = 0.07). The sensitivity of PG-SGA compared to SGA was 89% and the specificity was 50%. When MUST was compared to SGA the sensitivity was 31% and the specificity was 95%. The majority (63%) of patients were unable to complete the PG-SGA; which were then completed with the assistance of the investigators. Discussion: MUST poorly identify malnourished HD patients (as found previously by Hyam et al., 2010). However in contrast to the pilot study, this current study which included a larger more nutritionally diverse group of patients, indicated that SGA was more sensitive to changes within independent nutritional parameters than PG-SGA. Over half the patients were not able to complete the PGSGA so, contrary to expectations, there were no time savings. Conclusion: SGA appears to be more sensitive to the nutritional status of HD patients than PG-SGA and additionally a large proportion of HD patients find PG-SGA too difficult to complete. MUST is not a sensitive method for detecting malnutrition in this group of patients.