DEVELOPMENT AND VALIDATION OF THE BURTON SCORE: A TOOL FOR NUTRITIONAL ASSESSMENT


Authors

Linda Russell,
Tissue Viability Nurse Specialist,
Queen's Hospital,
Burton Hospitals NHS Trust,
Belvedere Road,
Burton-on-Trent,
DE13 0RB,
Tel: 01283 566333,
Fax: 01283 517614,

Also:
JENNY TAYLOR,
Head of Dietetics,
JUNE BREWITT,
Sister, ITU,
MIKE IRELAND,
Pharmacist,
TIM REYNOLDS,
Consultant Chemical Pathologist,

Contents


Published in World Wide Wounds: October 2000
Version: 1.0

Summary

This paper describes the development and validation of the Burton Score, a nutritional assessment tool based on the Waterlow score, with the rationale that since nurses already collect data for one score, it would only lead to unnecessary duplication of effort if a totally different scoring scheme were to be used for nutritional assessment. Initial cut offs were determined by a pilot study of 26 patients on an elderly care ward and validated by comparing the nutritional status of 263 patients estimated by the Burton score with a dietitian's assessment of nutrition. The validation study showed that although there was significant correlation between the Burton and Waterlow scores the Burton score correlated more closely with the dietitian's assessments. The King's Fund report of 1992 stated that all patients should have assessment of nutritional status on admission to hospital: we believe the Burton score could provide a simple tool to achieve this goal.

Introduction

Malnutrition in hospitals was highlighted in a King's fund report in 1992 [1] and is still thought to be a very big problem. Clinical surveys of hospitalised patients continue to show an unacceptably high incidence of malnutrition in medical, surgical, paediatric and care-of-the-elderly wards. [2] [3] [4] [5] [6] [7] Furthermore, a general sense of weakness and illness impairs appetite and ability to eat. Malnourished patients become apathetic and depressed and this may lead to loss of morale and loss of will for recovery. Inability to concentrate may also affect the patient's ability to remember instructions for self-care after discharge. Weakness may affect the respiratory muscles [8] making it more difficult for the patient to cough and expectorate with a consequent liability to chest infection [9] . Similarly, this impaired ventilatory drive may make it difficult to wean a critically ill patient off a ventilator1 [10] . There are many other effects of malnutrition including impaired wound healing [11] , impaired resistance to infection [12] , impaired gastro-intestinal function [13] , immobility and increased susceptibility to pressure sores [14] .

The deleterious effects of malnutrition described above have shown to be minimised by nutritional support. Thus it has also been shown that improved nutrition leads to improved wound healing [15] , an improved immune response [16] , improved respiratory muscle function [17] , and a reduction in post-operative complications [18] [19] . Some studies have shown that the average length of hospital stay of malnourished patients admitted to a general medical unit can be up to 5 days longer than those not malnourished [20] . Furthermore, it has also been shown that a malnourished patient who develops a major complication can cost four times as much as a well nourished patient with an uncomplicated disease course [21] . Analyses of the cost-effectiveness of nutritional support have emphasised the importance of a reduction in the incidence of complications and thus the length of stay and cost per day in patients undergoing moderate-to-major gastro-intestinal surgery [22] [23] . It has been estimated that up to £450 per patient admission could be saved merely by prescribing oral dietary supplements in the postoperative recovery period [17].

It is obvious therefore, that nutritional status assessment on admission should be carried out for all patients but there are few simple tools available for this task. It was for this reason that we developed the Burton score.

In many hospitals pressure risk is assessed using the Waterlow score. This does not necessarily indicate which patients require nutritional support. The Nutritional Group at Queen's Hospital, Burton, developed a simple nutritional scoring system based on and complementary to the Waterlow score which was named the Burton score. This article describes its development and validation.

Methods

Design of the Burton Score

To ensure a simple system that would not excessively increase nurses' workload, the Burton score was designed by examining the Waterlow score card to determine which elements could be used in a nutrition score. The elements to be included in the Burton score were decided by committee discussion of those factors which members thought, from their clinical experience, to have a significant influence on nutrition. The scores for each element were similarly decided with values being chosen in similar proportions to those used in the Waterlow score; the object being to design a balanced estimate of the contribution of each element. The Waterlow score card design [24] [25] was then modified to distinguish Waterlow-specific data from general data which was felt to be relevant to nutritional status: eg the demographics, patient's build, visual assessment of skin areas and appetite. The build question from the Waterlow card was modified to include body mass index (BMI) because this improved the definition of who should be considered average, above/below average or obese. The simplicity of the BMI was though to be important as the aim was to design a tool for rapid routine use rather than a complex research tool that would be useless in a clinical situation. It was recognised that BMI is not a perfect indicator of nutrition because there are changes in body composition with age, eg muscle mass tends to fall in the elderly but this is to some extent compensated for by extra points added for elderly patients as defined in the demographics section. In common with the Waterlow score, each question has an associated value. The nutrition specific questions which were added were: Unintentional weight loss was specified to exclude patients who had been dieting etc. Weights were deliberately given in imperial measures since these are understood by patients; substitution for metric weights should be possible either using an accurate 6.4 kg per stone conversion factor or even using round weights of 5kg.

The final design thus has three columns: Waterlow-specific information, general information and nutrition specific information (figure 1). To evaluate a patient's risk, the total score for each of these columns must be determined. Then, to derive the Waterlow score, the Waterlow-specific and general columns are summed and to derive the Burton score the nutrition-specific and general columns are summed.

Pilot Study

To determine preliminary cut-offs for the Burton score, all patients in one elderly care ward (n=26) were assessed for their Burton score by a ward nurse and by a dietitian to see if the patients were identifiable as nutritionally deficient. From this information cut-offs were chosen that adequately described the patients' nutritional status and could be used to identify patients who would benefit from nutritional supplementation or required a detailed nutritional assessment from a dietitian (table 1). These cut-offs took into account the availability of dietetic support at Burton and therefore could be modified by other users of the score to suit local conditions.

Table 1. Interpretation of the Burton Score

A second pilot study was carried out to assess whether the chosen cut offs were appropriate. During a hospital pressure sore point prevalence survey, the Burton score was calculated in addition to the Waterlow score for all 350 patients surveyed. As would be expected, there was a correlation between patients' age and the Waterlow score and with the presence of pressure sores. The Waterlow and Burton scores also correlated with each other but most importantly, the cut offs chosen in the first pilot study were found to be acceptable. Therefore, since it was believed that nutritional screening could produce significant benefits for patients and that the Burton Score may be a useful tool for a screening program, further validation was carried out.

Validation Study

Validation of the Burton score was carried out in conjunction with an audit of nutritional supplementation. The aims of the study presented here were to allow estimation of sensitivity and specificity assessments for the Burton score versus a gold standard (formal dietetic assessment) for different cut offs.

The study evaluated 263 patients. Assessments were initially made using a nurse-led interview in which Waterlow and Burton scores were collected in addition to other data (table 2), and a recall food diary in which the patient was asked to provide a detailed list of their food intake prior to their hospital admission. It was found that the food diary approach was not practicable because, due to their illnesses, many patients were unable/unwilling to provide the necessary detail. therefore a dietitian-led interview in which a coded estimate of dietary intake for 3 key areas was made was introduced. Intake was estimated in 4 bands: adequate, probably adequate, probably inadequate and inadequate. Banding was based on the patient's description of their food intake interpreted according to the dietitian's clinical experience against the recommended daily intakes for each nutrient class. A single dietitian was used throughout the study to prevent inter-observer biases. Nutrients were assessed in three groups: kilocalorie intake, protein intake and vitamin/mineral intake. Later, analysis was carried out to determine whether this refinement was necessary.

Results of the Validation Study

1. Correlation between the Waterlow score and the Burton score.

Since the two scores have a central core of general data, some degree of correlation is to be expected. The correlation coefficient (r) is 0.683 (n = 207) with t = 13.02 (P<0.01). This indicates significant correlation but since r2 = 0.46, a significant proportion of each score (54%) is independent.

2. Correlation between the Waterlow score/Burton score and serum albumin.

Serum albumin concentration is a recognised marker of poor nutrition [26] [27] . Although it is not a particularly good marker, it has the advantage of being measured on most patients on admission to hospital [28] . Therefore, relationships between the scores and albumin concentration were tested. As the Waterlow or Burton score increased, so the albumin concentration decreased. The correlation coefficients between albumin concentration and the Waterlow score was -0.372 and between albumin concentration and the Burton score was -0.384 (n=176 in both cases). The significance of these was tested using Student's t-test giving significance's of t = 5.28 (p<0.05) and t = 5.48 (p<0.05) respectively.

3. Correlation between the Waterlow score/Burton score and body mass index.

The correlation coefficient between body weight and the Waterlow score was -0.080 and between body weight and the Burton score was -0.195 (n = 197 and 198 respectively) ie as score values increased, so body weight decreased. The significance of these was tested using Student's t-test giving significance's of t = 1.12 and t = 2.78 respectively (p = not significant for either value).

4. Evaluation of the Burton score against the three dietary nutrient groups.

The Burton score only provides a single number to estimate nutritional status. To evaluate whether it was necessary to consider each nutrient group separately, the Burton score group and dietitian's assessments were plotted graphically (figure 2). It was immediately apparent that the pattern of results was very similar for all three nutritional categories. A comparison between protein and calorie intake demonstrated that 95% of all patients who were calorie deficient were also protein deficient. For vitamin/mineral intake, the relationship with calorie intake was less clear cut. On balance it was therefore decided to use the estimate of calorie intake as the dietetic gold standard against which further comparisons were made.

5. Comparison of the effectiveness of the Waterlow and Burton scores with the dietitian's assessment of patient's nutritional status.

The Waterlow Score has previously been demonstrated to be effective at identifying those at risk of developing pressure sores [24] [25]. It is necessary to prove that the Burton score is superior at identifying nutritional problems, otherwise there is little point in the extra work involved in calculating the score. A simple way of determining whether a method is an effective screening tool and to determine cut offs, is to draw a receiver-operator curve (ROC) plot [29] . This graphs the detection rate against the false positive rate: any screening test which lies on the identity line where detection equals false positives is no more efficient than tossing a coin, any lying below the line is less effective! Figure 3 shows the ROC plot for the Burton and Waterlow scores. It was derived by assuming that the dietitian's assessment was correct and since only two groups can be compared, all patients identified by the dietitian as having adequate or probably adequate intake were classified as normally nourished (true negative) and all others as inadequately nourished (true Positive). Then the number of patients identified above or below the defined cut offs for the Burton score (6, 11 and 16) or the Waterlow score (10, 15 and 20) were counted, enabling false positive and false negative rates to be estimated for each threshold level. From these values detection and false positive rates can be calculated to derive the graph. Similarly, the graph can be used in reverse to estimate a cut off that will give a desired false positive or detection rate. It is evident that the Burton score is clearly more effective than the Waterlow score in identifying nutritional inadequacy.

6. Anthropometric corroboration of the Burton score.

In addition to the dietitian's assessment of the patient's nutritional status, mid upper arm circumference (MUAC) was assessed. The expected values for this vary between men and women and show a non-linear trend. However, after conversion to Log centiles of normality, the MUAC can be correlated against the Burton score. In men there was a statistically significant correlation such that as the Burton score increases the MUAC tended to decrease (r = -0.628, P<0.05) but in women the correlation did not reach statistical significance.

Discussion

There are some potential weaknesses in our suggested scoring system; for example, confused patients may be unable to provide details of weight loss but an observant nurse may notice that the patient's clothing is extremely loose which could indicate significant weight loss in the absence of other corroboratory information. The remainder of the score does not require detailed information from the patients, so should be less problematical. Another potential weakness of the study was that the dietitian's assessment was considered to be a gold standard: to minimise inter-observer variation all assessments were made by the same person but there remains the possibility that the assessments were biased, even tough the study dietitian's routine clinical recommendations were not significantly at variance to the other dietitians in the Trust. Thirdly, the Burton score has only been validated for adults because it was felt that height/weight centile charts already provided an effective measure for children. Despite these limitations, we feel the Burton score is a valuable first attempt to create a routine nutrition scoring tool. Further research is necessary, particularly on whether additional data need to be included, whether other assessment methods are superior, and whether use of the Burton score targets nutritional support to those at highest risk. Some of this research is being carried out at Burton, but we would be happy to collaborate with any other researchers who wish to use the Burton score. Finally, we should consider the question of whether using the score has improved patient care: anecdotally, since its introduction we have noticed an increase in appropriate referrals to the dietetics department but this has yet to be formally assessed.

In conclusion therefore, we have presented a simple tool for assessing nutritional status which has the benefit of being based on a system that is already well understood by many nurses. We suggest that the Burton score provides the assessment tool that was recommended by the King's Fund report in 1992. Hospital induced malnutrition is a common occurrence: it can be recognised and prevented. Nurses, doctors and all members of the multi-disciplinary team can play a key role in its recognition [1].

Conclusions