Knowledge of an individual patient's height and weight are essential for modern intensive care management. Although lean body mass is a more useful indicator of pharmacokinetics than body mass itself  drug dosage is often calculated on a drug mass per kilogram body weight basis. Potent vasoactive drugs are frequently prescribed to correct or optimize haemodynamic or oxygen delivery parameters. These drug doses are calculated using body surface area (BSA) calculated from height and weight. Daily weighing can be a useful adjunct to the management of fluid balance in the intensive care unit (ICU) or other high dependency areas. It is also one of the most important measurements in assessing the nutritional status in non-ambulatory elderly patients .
Unfortunately, accurate measurements of height and weight are rarely available when patients are admitted into the ICU (with some exceptions - mainly in the case of planned admissions). In some ICUs, patients are weighed using weigh beds or bed scales, but this is far from universal. In our experience the accurately measured height is never known and weight rarely known. Instead both doctors and nurses visually estimate body weight and height, and clinical practice is based upon these figures. This study aims to determine the magnitude of the potential errors in such 'guesstimates'.
In order to gain an insight into whether estimation of height and weight is a common practice in other ICUs (and not to determine the national pattern of practice), we carried out a telephone survey in early 1999. The most senior nurse or doctor available in 20 southern English ICUs was asked which methods they used in order to determine the height and weight of patients.
We then studied 30 healthy volunteers from amongst the operating theatre staff, representing a wide range of body weights and heights. They were all attired in identical lightweight theatre dress 'blues' (they came to us from normal duties) to maintain modesty. They were not asked to change into a gown, as they would have been unlikely to co-operate. An independent observer coded each volunteer, then weighed each person using calibrated digital scales and measured their height with a spirit level and a fixed measuring scale attached to the wall. These values were termed the 'reference' height and weight of the subject. The independent observer played no part in the estimation process.
The subjects were then asked to lie on a bed to mimic a supine ICU patient. The panel of estimators consisted of an ICU consultant, two specialist registrars in anaesthesia and an experienced ICU sister - all of whom regularly make such estimations on our ICU patients, according to our normal practice. They independently made an estimate of each subject's weight, and then measured the height in a supine position with a tape measure. The subjects were not informed of these estimations, to avoid potential for embarrassment that might introduce bias.
Assuming that the height and weight of our subjects followed a normal distribution, statistical analysis was performed using Student's t-test. Data are presented as scatter plots (with Pearson regression coefficients calculated using Microsoft Excel for Windows) and also as Bland-Altman plots.
The results of the telephone survey revealed that in our sample of ICUs patient weight is infrequently measured. Only two of the 20 units regularly measured weight, while 'supine height' was measured with a tape measure in 15 of the 20 units (but estimated in the other five units). All use pulmonary artery catheters and compute haemodynamic variables.
The mean weight of the sample of 30 staff (age range 19-57 years) was 76.3 kg (SD ± 15.6 kg) with a range of 42.6-114.1 kg.
Table 1 shows the results for height and weight estimation for the four investigators. Three of the four observers' estimates differed significantly from the reference measurements (P < 0.05) for both weight and height. The difference between estimated and reference values for observer C was not significant (P > 0.05) for both parameters, suggesting an ability to estimate with greater accuracy that may be related to his greater experience. When the observers' estimates were pooled (by taking a mean of all estimates for each subject), the resulting values are closer to the reference value and fail to reach significance (P = 0.25).
These results are shown graphically as scatter diagrams with correlation coefficients (Fig. 1). In order to illustrate the degree of agreement, Bland-Altman plots  were constructed for each individual observer, as well as for the pooled mean estimations of all observers (Fig. 2).
The mean height of the sample was 170.7 cm (SD ± 7.0 cm) with a range of 157-186 cm. In this case, the difference remained significant even when the estimates were pooled (P < 0.05).
The telephone survey demonstrated that in our sample of ICUs in the UK educated guesswork is commonly practised in order to determine patient height and weight. Patients are infrequently weighed, and then only when special weigh beds are available. Few ICUs appear to routinely use these beds.
The results of this simple observational study demonstrate that considerable and random error may occur when experienced members of staff use 'educated' guesswork in order to establish the patient's weight. Critical illness is often associated with muscle wasting and fluid imbalance, with oedema. We suspect that errors in weight estimation may be greater in patients in comparison with our 'well' volunteers. This inaccuracy is not confined to estimates of weight - even when a tape measure is used to measure patient height, considerable error was found (Fig. 3). It is clearly more difficult to measure height in a supine subject with a simple tape measure than in the erect subject. Indeed, supine and erect heights may themselves differ.
While we have shown that the differences between estimated and reference values were statistically significant for three of four observers for weight and height, this does not necessarily prove clinical significance, which is a question of judgement. However, it seems to be a sound principle that the determination of values that are essential for intensive care practice should within reason be as accurate as possible, in order to avoid the derivation of potentially misleading cardiovascular values (often used as therapeutic goals) and errors in drug dosage. Furthermore, these data suggest that the error may indeed be clinically important. The Bland-Altman plots illustrate this (Fig. 2). They demonstrate a degree of lack of agreement that we would argue to be clinically unacceptable. For example, the mean difference between the estimated and reference weights for observer B was 5.1 kg (SD ± 8.1 kg). The clinical importance of this inaccuracy is illustrated by an example taken from our data. The range of estimated weights for one subject of 72.3 kg was 62-82 kg, leading to a potential error in drug dosage of up to 14%. A tendency to underestimate at lower weights and overestimate at higher weights is apparent (Fig. 2). This is also reflected in the pooled values. Interestingly, this is contrary to the findings of Coe and his colleagues of overestimation at low actual weight .
We recommend that either efforts should either be made to obtain more accurate values or values derived from these parameters must be used with great care. The apparently random nature of the error precludes the use of constants to correct inaccuracy. There are various strategies by which the accuracy of height and weight assessment may be improved.
First, we could improve the performance of conventional direct measurement of height and weight. Height may be easily and accurately measured by the use of a large calliper consisting of a long rule with one fixed and one movable cursor. A hospital works department could easily manufacture this item. There seems to be no good reason to guess height, because this apparatus is inexpensive, simple and accurate. However, measurement of weight is more problematic. Efforts should be made to ensure that all patients routinely admitted into the hospital are weighed on calibrated scales. However, this is rarely possible for emergency admissions and does not solve the problem of reassessment of weight during a patient's stay. Personnel in ICUs should consider using weigh beds, but they are expensive and not always easy to use. A unit may need several such beds to guarantee availability for every patient. Bed scales are an alternative option, but their use is also associated with problems of practicability and cost. Neither hydrostatic nor air weighing are of practical use in the ICU, although they have been advocated in other circumstances .
Second, we could attempt to increase the accuracy of indirect assessment of weight. Several methods of extrapolation of body weight from other more easily measured parameters have been described in both human beings [2,6,7] and animals [8,9].
The weight of horses is highly correlated with height at the withers, a condition score, heart girth and length . A formula connecting these variables accurately predicts weight. In a similar way, body surface area can be predicted in Indian elephants by a regression equation, which utilizes forefoot pad cir-cumference and height at the shoulders .
Nomograms have been devised to predict body weight in geriatric human patients from measurements of arm circumference and chest girth in males and waist skin-fold thickness plus thigh circumference in women . However, they are probably not valid for younger patients. Multiple regression analysis has been used to estimate body weight from body height, knee width and wrist width in young adults . However, these methods all have the disadvantage of being complicated and derived from a selected patient population. It is unlikely that a unified relationship can be found which applies to a wide range of intensive care patients both in terms of age and body composition.
A similar study of visual estimation performed in the operating theatre by anaesthetists and operating department assistants (ODAs) also found significant inaccuracy . Our study confirms their findings. However, we have demonstrated that considerable improvement can be achieved using mean values of multiple individual estimations.
In conclusion, this study has shown that the accuracy of weight estimation by experienced ICU medical and nursing staff is improved if multiple estimators are used. In the light of these findings, we recommend that if the highest possible accuracy is sought weigh beds or scales should be used, because these are the current 'gold standard'. If they are not available or their use is not feasible (and thus weight must be determined by estimation), an effort may be made to identify the best 'guesstimator' by means of a similar study to ours. This is probably impractical and the availability of 'good' estimators cannot be guaranteed at all times. However, because this study has shown that averaging the estimates of four observers resulted in greater accuracy than the best that any individual achieved, the best compromise seems to be to use mean values from several experienced observers.
The authors thank Sister Sue Shipton for her help with this study, and Dr Brian Tom for statistical advice.
1 Morgan DJ, Bray KM. Lean body mass as a predictor of drug dosage. Implications for drug therapy. Clin Pharmacokinet
2 Altea JA, Haboubi NY, Hudson PR, Sastry BD. Body weight estimation of elderly patients by nomogram. J Am Geriatr Soc
3 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
4 Coe TR, Halkes M, Houghton K, Jefferson D. The accuracy of visual estimation of weight and height in pre-operative supine patients. Anaesthesia
5 Dempster P, Aitkens S. A new air displacement method for the determination of human body composition. Med Sci Sports Exerc
6 Latin RW, Johnson SC, Ruhling RO. An anthropomorphic estimation of body composition of older men. J Gerontol
7 Baecke JA, Burema J, Deurenberg P. Body fatness, relative weight and frame size in young adults. Br J Nutr
8 Carrol CL, Huntingdon PJ. Body condition scoring and weight estimation of horses. Equine Vet J
9 Sreekumar KP, Nimalan G. Estimation of the total surface area in Indian elephants (Elephas maximus indicus). Vet Res Commun