How much undernutrition is there in hospitals?

Many studies over the last 30 years have emphasised the importance of disease-related undernutrition in hospitalised patients (Bistrian et al. 1974; Hill et al. 1977; McWhirter & Pennington, 1994; Corish et al. 2000; Kelly et al. 2000). This is largely because undernutrition adversely affects clinical outcome, and well-being, and is responsible for a disproportionately large consumption of health service resources (Consumer's Association, 1999). Although the deleterious consequences of undernutrition, which include increased morbidity, delayed recovery from illness and increased length of hospital stay are generally accepted (Lennard-Jones, 1992; Elia, 1993; Green, 1999), there is still controversy about the exact incidence of undernutrition in the hospital setting. One of the major reasons for this controversy is that there is no universal agreement about the de®nition of undernutrition. Since various workers have used different criteria to screen for the presence of undernutrition, it is not surprising that the reported magnitude of the problem has also been highly variable both in hospitals (10±60 %) (McWhirter & Pennington, 1994; Naber et al. 1997; Strain et al. 1999; Vlaming et al. 1999; Watson, 1999; Weekes, 1999; Corish et al. 2000; Kelly et al. 2000) and in nursing homes (10±80 %) (Kerstetter et al. 1992). However, it is also feasible that there are major variations in the incidence of undernutrition in different hospitals in Republic of Ireland and the UK. Corish et al. (2000) have recently screened for the presence of undernutrition in patients admitted to two teaching hospitals in Dublin, Republic of Ireland. They used the same criteria as those employed by another group of investigators, who published their results from Dundee, Scotland, UK 6 years earlier (McWhirter & Pennington, 1994). This was to ensure more reliable comparisons of the incidence of malnutrition between the two geographic locations. In both studies, patients were classi®ed as `undernourished' if they had a BMI , 20 kg/m and a triceps skinfold thickness or mid-arm muscle circumference , 15th percentile (Bishop et al. 1981). The incidence of undernutrition in newly admitted patients in Dublin was reported to be more than three-fold lower (11 %) (Corish et al. 2000) than in Dundee (40 %) (McWhirter & Pennington, 1994). What could be responsible for this striking difference? Although a simple answer cannot be provided, several possibilities are worth exploring. One possibility is that the population in Dublin is leaner than in Dundee. However, it is unlikely that this was the main reason for the differences between studies, because the data available from national surveys of Republic of Ireland (Lee & Cunningham, 1990), Scotland (The Scottish Of®ce, 1995) and England (Of®ce of Population Census and Survey, 1994), suggest that the proportion of these populations with a BMI , 20 kg/m is similar. Furthermore, a recent study in Glasgow (Kelly et al. 2000), which is less than 100 miles (160 km) from Dundee, reported that 18 % of patients admitted to hospital had a BMI , 20 kg/m (compared to 37×4 % in the Dundee study (McWhirter & Pennington, 1994)). Another possibility relates to the characteristics of patients and the criteria for admission to hospital. Amongst the patient characteristics that could have differed between studies are age (the average age of the group of patients admitted to the Dundee hospital was not documented, but in Dublin 40 % of patients were over 65 years) and the diagnostic case-mix (which was generally similar between the two studies). The criteria for admission to hospital might vary depending on the local resources, including the number of hospital beds relative to the population served, the extent to which hospitals act as primary, secondary and tertiary referral centres, and the degree to which private healthcare schemes and community services operate in different regions. Furthermore, there could also be more than one acute hospital in each region so that malnourished patients with more severe or more prolonged disease are directed into one hospital in preference to the other. If that was the case, then comparisons between single hospitals in different regions could be misleading. Speci®cally, Corish et al. (2000) recruited every third patient admitted to a large hospital and every tenth patient admitted to a smaller hospital in Dublin to ensure proportional representation, whereas in the study in Dundee, only patients admitted to the main hospital of the region were recruited. Yet another possibility concerns methodological differences between the two studies. For example, which arm was used for anthropometric measurements (not reported in either study)? Was the presence or absence of oedema taken into account, which could confound measurements of weight, anthropometry and calculations of BMI (Corish et al. 2000) excluded patients with gross ̄uid retention but it is unclear whether such patients were excluded in the Dundee study)? Were surrogate measures used, and if so, was their use validated in the patient group? In the study by Corish et al. (2000) 17 % of patients did not have their height measured (n 95 patients) and 9 % (n 50) were not weighed and so surrogate measures were used. However, in the study in Dundee the number of patients who did not have their weight or height measured, and the proportion in whom surrogate measures were used instead, was not recorded. Furthermore, it is unclear if the measurement of knee height, which was used to estimate height in both centres, was undertaken using the equipment and procedure employed by Chumlea et al. (1985). Neither study appeared to have validated their measurement of knee height against the British Journal of Nutrition (2000), 84, 257±259 257