Screening for Undernutrition in Geriatric Practice: Developing the Short-Form Mini-Nutritional Assessment (MNA-SF)

Transcription

1 Journal of Gerontology: MEDICAL SCIENCES 2001, Vol. 56A, No. 6, M366 M372 In the Public Domain Screening for Undernutrition in Geriatric Practice: Developing the Short-Form Mini-Nutritional Assessment (MNA-SF) Laurence Z. Rubenstein, 1,2 Judith O. Harker, 2 Antoni Salvà, 3 Yves Guigoz, 4 and Bruno Vellas 5 1 University of California at Los Angeles School of Medicine. 2 VA Greater Los Angeles Healthcare System, Geriatric Research, Education, and Clinical Center, Sepulveda, California. 3 Programe Vida Als Anys, Servei Català de la Salut, Barcelona, Spain. 4 Nestlé Research Center, Lausanne, Switzerland. 5 Service de Médicine Interne et Gérontologique Clinique, Hôpitaux de Toulouse, France. Background. The Mini-Nutritional Assessment (MNA) is a validated assessment instrument for nutritional problems, but its length limits its usefulness for screening. We sought to develop a screening version of this instrument, the MNA-SF, that retains good diagnostic accuracy. Methods. We reanalyzed data from France that were used to develop the original MNA and combined these with data collected in Spain and New Mexico. Of the 881 subjects with complete MNA data, 151 were from France, 400 were from Spain, and 330 were from New Mexico. Independent ratings of clinical nutritional status were available for 142 of the French subjects. Overall, 73.8% were community dwelling, and mean age was 76.4 years. Items were chosen for the MNA-SF on the basis of item correlation with the total MNA score and with clinical nutritional status, internal consistency, reliability, completeness, and ease of administration. Results. After testing multiple versions, we identified an optimal six-item MNA-SF total score ranging from 0 to 14. The cut-point score for MNA-SF was calculated using clinical nutritional status as the gold standard (n 142) and using the total MNA score (n 881). The MNA-SF was strongly correlated with the total MNA score (r.945). Using an MNA-SF score of 11 as normal, sensitivity was 97.9%, specificity was 100%, and diagnostic accuracy was 98.7% for predicting undernutrition. Conclusions. The MNA-SF can identify persons with undernutrition and can be used in a two-step screening process in which persons, identified as at risk on the MNA-SF, would receive additional assessment to confirm the diagnosis and plan interventions. N UTRITIONAL deficiency is common and serious in older adults. Up to 15% of ambulatory outpatients, 35% to 65% of elderly hospital patients, and 25% to 60% of institutionalized older adults have been reported as malnourished (1 4). Whereas some malnourishment stems from underlying illness, much is due simply to inadequate intake, which should be reversible if detected. Detecting nutritional problems in older persons has been recommended in population screening and as part of comprehensive geriatric assessment (CGA) (1,5 7). Thus, valid effective techniques are needed for both screening and diagnosis. To meet this need, several instruments have been developed by researchers and professional task forces (1,2,5 10). The Mini-Nutritional Assessment (MNA), a recent and extensively tested instrument, fulfills many criteria for both screening and diagnostic measures. It was developed and validated on large representative samples of elderly persons worldwide. It identifies persons at nutritional risk, provides information needed for intervention planning, and does not require laboratory data (4,11). However, its complexity and length impede its use as a brief screening tool. Several questions require special training (e.g., anthropometrics) or subjective judgments. Whereas components of a geriatric assessment typically take under 5 minutes to administer (12), the MNA takes approximately 10 to 15 minutes to administer (13). This is a reasonable length for a diagnostic test but is perhaps too long for a screener in a primary care setting. Shortening the MNA would be worthwhile if the short form retained the validity and usefulness of the original. In this study we have derived a short-form version (the MNA-SF) from the full MNA. We sought to create an assessment tool that preserves diagnostic accuracy while minimizing the time and training needed for administration and is therefore brief enough for widespread screening. METHODS Subjects To derive the MNA-SF, we used the database from the Toulouse-91, the original French population on whom the MNA was developed. This population, described fully by Vellas and colleagues (4), consisted of hospitalized geriatric patients (n 105) with a wide range of nutritional status (albumin at admission 3.0 g/dl, n 35; 3.0 to 3.5 g/dl, M366

2 DEVELOPMENT OF MNA-SF M367 n 35; 3.5 g/dl, n 35), and a smaller group of healthy community-dwelling elderly persons (n 50). Overall, these subjects (N 155) had a mean age of 79 years, 66% were female, 56% had a Folstein Mini-mental State of 24 out of 30 (14), 43% had a Katz Activities of Daily Living (ADLs) score 4 out of 6 (15), 49% reported anorexia, and 45% were taking more than three medications. Deriving the MNA-SF We used several criteria to identify the best questions for the MNA-SF: (i) good correlation with the full MNA score; (ii) good individual diagnostic characteristics (i.e., high sensitivity, specificity, and overall accuracy) on the basis of independent clinical assessment of nutritional status as either adequately nourished or undernourished ; and (iii) high internal consistency for the item set. We avoided items that were redundant, required special training to administer, involved difficult subjective recall, or produced too many missing or don t know answers. Complete MNA data were available for most of the population (n 151). As described elsewhere (4), the MNA consists of 18 questions in 4 categories: anthropometric measurements (four questions), global assessment (six questions), dietary questions (six questions), and self-perception of health and nutrition (two questions). Individual questions have weighted scores. The full scale ranges from 0 to 30 and is interpreted as follows: 24 indicates well-nourished, 17 to 23.5 indicates at risk of malnutrition, and 17 indicates malnourished. Pearson correlations were examined between each item and the MNA total score. Items poorly correlated with total score were candidates for exclusion. Interitem correlations were also examined; none was high enough to indicate item redundancy. Next, we looked at internal consistency (coefficient alpha) using item analysis procedures (Systat 7.0, Chicago, IL and Crunch 4.1, Oakland, CA). The 18 items of the MNA are ordinarily treated as a single scale. Item analysis procedures report the alpha for all 18 items and the change in alpha if any item is excluded. If excluding an item would produce no change or an increase in alpha, thereby increasing overall internal consistency, that item was a candidate for exclusion. Successive calculations of internal consistency were done with the best remaining 15, 12, 9, 6, 5, and 4 items. Finally, we calculated sensitivity, specificity, and diagnostic accuracy for each item using physician judgment of clinical nutritional status as the gold standard. This judgment was made by two physicians who were trained in nutrition and who had no knowledge of the MNA score. Judgments were made independently with subsequent resolution of disagreements and were made on the basis of anthropometrics, weight loss, albumin, 1 month food intake diary, weighed food intake over 3 days, and medical record information. We identified and tested a best short form version against the full MNA to see if accuracy were lost or if items should be restored. Sensitivity, specificity, and diagnostic accuracy for the final MNA-SF were compared with those for the full MNA, measured against both physician-judged clinical nutritional status and low serum albumin. Threshold values for the MNA-SF were chosen using receiver-operating curves (ROCs) of diagnostic accuracy. Extension to Other Populations Two independent samples of patients who completed the full MNA were used to further test the MNA-SF. The Mataró, Spain sample included 114 elderly persons in a subacute convalescent unit, 89 elderly persons in a large boardand-care facility, and 199 independently community-dwelling elderly persons (16). The Albuquerque, New Mexico sample included 347 healthy elderly subjects from the New Mexico Aging Process Study (17). Table 1 compares these two samples and the Toulouse-91 sample. The primary difference among these samples was the proportion of community-dwelling participants: approximately one third of the Toulouse sample, half the Mataró sample, and all but one of the Albuquerque sample were community dwelling. Table 1. Sample Characteristics Characteristic France Spain New Mexico Combined Total n Mean age Female (%) n with complete MNA BMI mean (SD) 22.3 (4.4) 27.3 (5.0) 25.2 (3.7) 25.7 (4.8) Albumin (mg/dl) mean (SD) 3.30 (.70) 3.76 (.67) 4.09 (.31) 3.83 (.61) Below 3.0 (%) From 3.0 to 3.5 (%) Above 3.5 (%) Living location Community dwelling (%) Hospital/convalescent care (%) Nursing home (%) MNA results Mean MNA score (SD) 19.9 (6.9) 23.3 (5.0) 25.7 (2.4) 23.6 (5.0) Normal ( 24) (%) At risk ( ) (%) Undernourished ( 17) (%)

3 M368 RUBENSTEIN ET AL. Extensive nutritional and laboratory data were collected from all three samples, but only the Toulouse-91 had an independent physician rating of clinical nutritional status. Thus, for the New Mexico and Mataró samples, the full 18- item MNA was used as the gold standard to evaluate the MNA-SF. In addition, we examined the relationship of the full MNA and the MNA-SF with serum albumin. There were clear differences among these three samples, including language, nationality, and proportions recruited from institutions. By pooling the samples, we took advantage of this wide diversity to see how effectively the MNA- SF identified persons in the combined data set who were nutritionally at risk. We also examined diagnostic accuracy of the MNA-SF in the pooled sample separately for institutionalized and community-dwelling subjects. Whereas data pooling for purposes of regression modeling or hypothesis testing (e.g., in multisite data used in multiple regression or factor analysis) ordinarily requires tests demonstrating homogeneity across samples or methods to control for sample differences, our intent was to create a large heterogeneous sample with good representation of both frail and healthy older adults. To further validate the MNA-SF items, stepwise discriminant analysis was performed on all 18 items using a 50% random sample of 902 cases from all three sites and cross validated on the remaining 50%. Our purpose was to identify items that best discriminated between the malnourished, at-risk, and normal groups on the full MNA. If the items selected by this procedure were the same as those selected by clinical nutritional status in the Toulouse-91 sample, we would have confirmation that we had selected the best items for the MNA-SF. RESULTS Deriving the MNA-SF Table 2 shows the Pearson correlations of each MNA item with the total MNA score and each item s diagnostic characteristics (sensitivity, specificity, and diagnostic accuracy), using clinical nutritional status as the gold standard. For the latter calculations, subjects with a clinical status rating of uncertain (n 13) were excluded. For each item except Item 5, both Pearson r and chi square analyses were significant (p.05), indicating its relation to the total MNA score and to clinical status. For items with more than two response categories, classification cut-points were selected to maximize sensitivity. For example, with Item 1, we tested body mass index (BMI) values of both 21 and 23 as indicators of malnourishment. Individual items varied in their ability to predict nutritional status. Some items were quite good (i.e., Item 1 [BMI], Item 4 [weight loss], and Item 7 [illness/stress]), and others were not good (e.g., Item 2 [mid-arm circumference], Item 5 [institutionalized], and Item 6 [takes 3 medications]). Six items with the highest sensitivity and overall accuracy were selected for the working draft of the MNA-SF. Three other items fairly close in test characteristics had some problems. Item 17 (self-rated malnourishment) had good apparent accuracy but was excluded because of a large proportion (32/155: 21%) of don t know answers. Item 3 (calf circumference) required extra examination time and calculation. Item 18 (self-rated health) seemed to have less face validity. Successively deleting items in the item analysis yielded the same six-item scale with good internal consistency (al- Table MNA Items: Correlation With Total MNA Score and Diagnostic Characteristics Relative to Clinical Nutritional Status (n 142) Item Content Pearson r With MNA Total Score Sensitivity Specificity Accuracy 2 1 BMI Mid-arm circumference Calf circumference Recent weight loss 1 kg Lives in nursing home or hospital Rx meds 3/d Acute illness or stress Housebound Dementia or depression Pressure sores or ulcers meals/d protein sources/d servings of vegetables or fruit Appetite loss or eating difficulty cups/d fluid Self-fed with difficulty/unable Moderate or major malnutrition Health worse/don t know MNA 18-item MNA-SF 6-item For dichotomous variables the point-biserial correlation technique was used, which is, in fact, a special case of the Pearson r. Selected for MNA-SF.

4 DEVELOPMENT OF MNA-SF M369 Figure 1. Scatterplot for the MNA-SF scores versus full MNA scores for 142 persons clinically judged as having normal nutrition or as malnourished/at risk. Cut-points are indicated by dashed lines and are labeled on the top and right side. Pearson correlation.969. Figure 2. Scatterplot for MNA-SF score versus serum albumin (mg/dl); n 151. Dashed lines indicate cut-points for normal ( 3.5) and below normal ( 3.0) albumin, and for normal nutrition (MNA- SF 10), as labeled on the top and right side. Pearson correlation.679. pha coefficient.843) compared with.865 for the full MNA. Reducing the item set to five and four items reduced internal consistency. As shown in Figure 1, the correlation between the MNA- SF and the full MNA is high (r.969). We used a score of 10 or less (out of 14) on the MNA-SF to indicate possible undernutrition and a score of 11 or more to indicate normal nutrition. In the original sample, all persons identified on the MNA as malnourished (MNA score 17) and all but two at risk persons (MNA score 17 23) would have been detected on the MNA-SF as undernourished (sensitivity 97.9% [96/98], specificity 100% [53/53], and diagnostic accuracy 98.7%). Table 2 shows that the diagnostic accuracy of the full MNA (97.2%) is indistinguishable from the MNA-SF (96.5%) in detecting persons clinically judged undernourished. The Pearson correlation (r) between serum albumin and the MNA-SF is.679 (shown in Figure 2), which is similar to that between albumin and the full MNA (r.699). Moreover, all but two persons with subnormal albumin had MNA-SF scores 10 out of 14 with a diagnostic accuracy of 98.7%. Extension to Other Populations Because clinical nutritional status was not available for the two validation populations, we tested diagnostic characteristics of each of the 18 MNA items against the full MNA score to determine if those chosen for the MNA-SF remained the most sensitive and accurate. In the combined sample, six of the seven items with high sensitivity and accuracy were the MNA-SF items. (The seventh, health status, had 12% don t know responses.) All six MNA-SF items had accuracy above 72%. Thus, the choice of the best six items for the MNA-SF appears to be confirmed in the larger sample. Figure 3 shows the relationship between the MNA and MNA-SF scores using the combined sample of 881. Pearson correlation was.945 (p.0001). As with the original sample, all persons classified as malnourished on the full MNA were also classified as at risk for undernutrition on the MNA-SF. The only misclassifications were in the MNA at risk category, with 49 false negatives (8.6%). Using a cutpoint of 11 as normal nutrition, the MNA-SF has a sensitivity of 97.9%, a specificity of 100%, and an overall diagnostic accuracy of 98.7% for predicting malnutrition on the full MNA; 64.7% screened out as normal. Cut-point Selection Choice of threshold (cut-point) is crucial in determining diagnostic characteristics of any screening test. Generally, Figure 3. Scatterplot for MNA-SF versus full MNA scores for the combined sample (n 881). Cut-points are indicated by dashed lines and are labeled on the top and right side. In contrast to Figure 1, this figure illustrates a more sensitive cut-point for normal ( 11). Pearson correlation.943.

5 M370 RUBENSTEIN ET AL. Figure 4. Receiver operating characteristic (ROC) curve for the combined sample (n 881) for the MNA-SF (as compared with the full MNA). Area under the curve is an inverse relationship exists between sensitivity and specificity; as the threshold for normality is raised (i.e., classifying more persons as abnormal ), sensitivity increases while specificity decreases. As indicated in Figure 1, the optimal threshold score for normal nutrition in the MNA-SF appears to be 11. However, increasing sensitivity by raising the threshold for normal to 12 could be considered (see Figure 3), because failing to identify a malnourished person may be of greater concern than requiring additional evaluation for people with normal nutrition who mistakenly screen in. Using this more sensitive threshold ( 12) in the combined sample, there were only 24 false negatives (4.9%); however, only 56% of the people tested were screened out as normal, raising the false positive rate. Figure 4 shows the ROC curve for the combined sample, illustrating the sensitivity specificity trade-off in setting the cut-point at 11 or 12. The area under the curve is Discriminant Analysis Table 3 compares the stepwise discriminant analyses, which were performed in the two 50% random samples of the combined population. Although the variable lists were slightly different, each analysis selected 14 variables that produced significant discrimination (at p.0001) among the three MNA categories (on the basis of Wilks s Lambda and Mahalanobis distance), and four variables were excluded. Notably, in both random samples, five of six of the MNA-SF items were selected among the first seven to enter the stepwise analysis, and all six MNA-SF items were significant. The advantage of the combined sample was that it contained a good case mix from very frail to healthy independent persons. However, it contained the original Toulouse subjects used in developing the MNA-SF. Therefore, an additional discriminant analysis to test the MNA-SF items was performed using only data from Spain and New Mexico (n 729, excluding those with missing data). Again, all six items of the MNA-SF entered the discriminant analysis at p.0001, and 87.9% of the cases were correctly classified according to their full MNA score category. Thus, the original MNA-SF items were confirmed as appropriate choices. DISCUSSION Previous work shows that the MNA is an accurate assessment tool for nutritional problems, and is highly correlated with both clinical assessment of nutritional status and objective indicators such as serum albumin; however, the MNA may be too long for routine screening. Thus, a brief nutritional screening test was needed that would be accurate and useful in the clinical setting. In this analysis, we shortened the MNA while preserving its accuracy. The MNA-SF has 6 questions instead of 18, eliminates time-consuming and subjective items, and can be administered in approximately 3 minutes. Whereas height and weight measurement do involve time and training, particularly with bed-bound persons or amputees, these measures are relatively easy to obtain and are often available in patient records. The MNA-SF has high diagnostic accuracy relative to clinical nutritional status, high correlation with Table 3. Discriminant Analyses to Validate the MNA-SF First 50% Random Sample (n 438) Second 50% Random Sample (n 442) (Cross-Validation) Analysis Step Number Variable to Enter Wilks s Lambda Variable to Enter Wilks s Lambda 1 Mobility (number 8).441 Mobility (number 8) Stress or illness (number 7).337 Stress or illness (number 7) Nutrition assessment (number 17).259 Weight loss (number 4) BMI (number 1).228 Nutrition assessment (number 17) Meals/d (number 6).202 BMI (number 1) Dementia or depression (number 9).178 Dementia or depression (number 9) Weight loss (number 4).160 Number of Rx meds , 12: Loss of appetite (number 14).128 Loss of appetite (number 14).129 Variables not entering analysis Variables not entering analysis Midarm circumference Midarm circumference Number of Rx meds Nursing home/hospital Vegetables Meals/d Fluid intake Fluid intake Notes: All items in analyses at p BMI body mass index. Item in MNA-SF.

6 DEVELOPMENT OF MNA-SF M371 medical conditions. To answer this question, we need controlled intervention studies (2,9). However, many clinicians are not waiting for such study results, as the argument for intervening in persons with clear nutritional problems is so convincing (1,3 5,7). Whereas treatment could normalize only those persons with reversible malnutrition, this subgroup is likely large enough to justify a systematic screening process. Acknowledgments This research was supported by the U.S. Department of Veterans Affairs and by an unrestricted educational grant from Nestlé, S.A., Nutrition Division. Address correspondence to Laurence Z. Rubenstein, MD, Director GRECC (11E) VA Greater Los Angeles Healthcare System, Plummer St., Sepulveda, CA lzrubens@ucla.edu Figure 5. Nutritional screening algorithm for 650 communitydwelling persons in the combined sample, using the more sensitive cut-point of 12 or greater to indicate normal nutrition. False-negative rate 3.4%; false-positive rate 34.3%. the full MNA (r.945), and is as good as the MNA in predicting serum albumin. A rational strategy for nutritional screening in the elderly population is needed. Data presented here suggest that nutritional screening can be performed efficiently and effectively using the MNA-SF in a two-step screening process as proposed in Figure 5. In the first step, the MNA-SF would be administered to patients undergoing CGA or periodic health risk appraisal. The second step entails confirming the diagnosis for persons identified as at risk by the MNA-SF and planning needed interventions (e.g., nutritional supplementation and specific assistance with meals) by administering the full MNA, which provides additional clinically useful information, and/or performing further assessments (e.g., laboratory studies and evaluation by a dietitian). Following further validation of the MNA-SF on different populations, this two-step strategy must be tested in clinical trials. If the population screened were similar to the community-dwelling subgroup of our pooled sample, this strategy would classify 77% of the population as normal and thus reduce the need for further evaluation. Of the 23% of subjects classified as at risk (MNA-SF 11), approximately 20% would be false positives and 80% would be confirmed by the full MNA as needing further assessment and intervention. In a frail population, fewer individuals would be expected to screen out by the MNA-SF, resulting in smaller savings in screening resources. For example, most nursing home residents would likely be classified as undernourished by the MNA-SF. Thus, administering the full MNA to nursing home residents might be simpler. A major assumption underlying a nutritional screening strategy is that patients discovered to be malnourished or at risk will benefit from being detected. It is unclear what proportion of malnourished persons could really benefit from treatment and what proportion has underlying nonreversible References 1. Posner BM, Jette AM, Smith KW, Miller DR. Nutrition and health risks in the elderly: The Nutrition Screening Initiative. Am J Public Health. 1993;83: Reuben DB, Greendale GA, Harrison GG. Nutrition screening in older persons. J Am Geriatr Soc. 1995;43: Rudman D, Feller AG. Protein caloric undernutrition in the nursing home. J Am Geriatr Soc. 1989;37: Vellas BJ, Guigoz Y, Garry PJ, Albarede JL, eds. The Mini Nutritional Assessment: MNA. 3rd ed. Paris: Serdi Publishing; Antonelli Incalzi R, Landi F, Cipriani L, et al. Nutritional assessment: a primary component of multidimensional geriatric assessment in the acute care setting. J Am Geriatric Soc. 1996;44: Vellas BJ, Guigoz Y. Nutritional assessment as part of the geriatric evaluation. In: Rubenstein LZ, Wieland D, Bernabei R, eds. Geriatric Assessment Technology: The State of the Art. Milano, Italy: Kurtis Publishers; 1995: Schoenberg NE, Coward RT, Gilbert GH, Mullens RA. Screening community-dwelling elders for nutritional risk. J Appl Gerontol. 1997; 16: Galanos AN, Pieper CF, Cornoni-Huntley JC, Bales CW, Fillenbaum GG. Nutrition and function: is there a relationship between body mass index and the functional capabilities of community-dwelling elderly? J Am Geriatr Soc. 1994;42: Rush D. Nutrition screening in old people: its place in a coherent practice of preventive health care. Ann Rev Nutr. 1997;17: Wolinsky FD, Coe RM, McIntosh WA, et al. Progress in the development of a nutritional risk index. J Nutr. 1990;120: Garry PJ, Vellas B. Practical and validated use of the Mini-Nutritional Assessment in geriatric evaluation. Nutr Clin Care. 1999;2: Rubenstein LZ, Wieland D, Bernabei R, eds. Geriatric Assessment Technology: The State of the Art. Milano, Italy: Kurtis Publishers; Guigoz Y, Vellas B. The Mini Nutritional Assessment (MNA) for grading the nutritional state of elderly patients: presentation of the MNA, history and validation. In: Mini Nutritional Assessment (MNA): Research and Practice in the Elderly. Nestlé Nutrition Workshop Series, Clinical & Performance Programme; 1997: Folstein MF, Folstein SE, McHugh PR. The Mini-mental State: a practical method. J Psychiatr Res. 1975;12: Katz S, Downs TD, Cash HR, Grotz RC. Progress in development in the index of ADL. Gerontologist. 1970;10: Salvà A, Bolibar I, Muñoz M, Sacristán V. Un nuevo instrumento para la valoración nutricional en geriatría: el Mini Nutritional Assessment (MNA). Rev Gerontol. 1996;6: Guigoz Y, Vellas B, Garry PJ. Mini Nutritional Assessment: a practical assessment tool for grading the nutritional state of elderly patients. Facts and research in gerontology, supplement no. 2. Nutrition Received January 23, 1999 Accepted June 20, 2000 Decision Editor: John E. Morley, MB, BCh

7 M372 RUBENSTEIN ET AL. Appendix Mini Nutrition Assessment A. Has food intake declined over the past three months due to loss of appetite, digestive problems, chewing or swallowing difficulties? 0 severe loss of appetite 1 moderate loss of appetite 2 no loss of appetite B. Weight loss during last three months 0 weight loss greater than 3 kg (6.6 lbs) 1 does not know 2 weight loss between 1 and 3 kg (2.2 and 6.6 lbs) 3 no weight loss C. Mobility 0 bed or chair bound 1 able to get out of bed/chair but does not go out 2 goes out D. Has suffered psychological stress or acute disease in the past three months 0 yes 2 no E. Neuropsychological problems 0 severe dementia or depression 1 mild dementia 2 no psychological problems F. Body Mass Index (BMI) (weight in kg)/(height in m) 2 0 BMI less than 19 1 BMI 19 to less than 21 2 BMI 21 to less than 23 3 BMI 23 or greater Screening score (subtotal max. 14 points) 12 points or greater: Normal no need for further assessment 11 points or below: Possible malnutrition continue assessment Note: If greater specificity is desired consider 10 points or below as possible malnutrition. Alternative height calculations using knee to heel measurements: with knee at 90 angle (foot flexed or flat on floor or bed board), measure from bottom of heel to top of knee. Men (2.02 knee height, cm) (0.04 age) Women (1.83 knee height, cm) (0.24 age) Body weight calculations in amputees: For amputations, increase weight by the percentage below for contribution of individual body parts to obtain the weight to use to determine Body Mass Index. Single below knee 6.0% Single at knee 9.0% Single above knee 15.0% Single arm 6.5% Single arm below elbow 3.6%