European Journal of Clinical Nutrition (2000) 54, 555±562
ß 2000 Macmillan Publishers Ltd All rights reserved 0954±3007/00 $15.00
www.nature.com/ejcn
The use of the Mini-Nutritional Assessment (MNA) tool in
elderly orthopaedic patients
MC Murphy1*, CN Brooks2, SA New2 and ML Lumbers3
1
European Institute of Health and Medical Sciences, University of Surrey, Guildford, UK; 2Centre for Nutrition and Food Safety, School
of Biological Sciences, University of Surrey, Guildford, UK; and 3Centre for Food and Health Care Management, School of Management
Studies for the Service Sector, University of Surrey, Guildford, Surrey, UK
Objective: To assess the use of the Mini-Nutritional Assessment (MNA) in elderly orthopaedic patients.
Design: An observation study assessing the nutritional status of female orthopaedic patients.
Setting: The orthopaedic wards of the Royal Surrey County Hospital.
Subjects: Forty-nine female patients aged 60 ± 103 y; dietary records were obtained for 41 subjects and 36
subjects gave a blood sample for biochemical analysis.
Major outcome methods: MNA questionnaire, anthropometry, plasma albumin, transferrin, C-reactive protein
(CRP) levels and dietary analyses.
Results: The group as a whole had low mean values for body weight, albumin and transferrin and high CRP
levels. In addition, the group had mean energy intakes well below the estimated average requirement (EAR) and
mean intakes of vitamin D, magnesium, potassium, selenium and non-starch polysaccharides (NSP) were below
the lower reference nutrient intakes (LRNI). The MNA screening section categorized 69% of the patients as
requiring a full assessment (scored 11 or below), but for the purposes of the study the MNA was completed on all
patients. The MNA assessment categorized 16% of the group as `malnourished' (scored < 17 points), 47% as
`at risk' (scored 17.5 ± 23.5) and 37% as `well nourished' (scored >23.5). Signi®cant differences were found
between the malnourished and well nourished groups for body weight (P < 0.001), body mass index (BMI)
(P < 0.001), demiquet (P < 0.001) and mindex (P < 0.001). Mean values for energy and nutrient intakes showed a
clear stepwise increase across the three groups for all nutrients except sodium, with signi®cant differences for
protein (P < 0.05), carbohydrate (P < 0.05), ribo¯avin (P < 0.05) niacin (P < 0.05), pyridoxine (P < 0.05), folate
(P < 0.05), calcium (P < 0.05), selenium (P < 0.05), iron (P < 0.05) and NSP (P < 0.05) intakes. Stepwise
multiple regression analysis indicated that anthropometric assessments were the most predictive factors in the
total MNA score. The sensitivity and speci®city of the MNA was assessed in comparison with albumin levels,
energy intake and mindex. The sensitivity of the MNA classi®cation of those scoring less than 17 points in
comparison with albumin levels, energy intake and mindex varied from 27 to 57% and the speci®city was 66 ±
100%. This was compared with the sensitivity and speci®city of using a score of less than 23.5 on the MNA to
predict malnourished individuals. Using this cut-off the sensitivity ranged from 75 to 100%, but the speci®city
declined to between 37 and 50%.
Conclusions: The results suggest that the MNA is a useful diagnostic tool in the identi®cation of elderly patients
at risk from malnutrition and those who are malnourished in this hospital setting.
Sponsorship: Nestle Clinical Nutrition, Croydon, Surrey.
Descriptors: elderly; nutritional assessment; orthopaedic patients
European Journal of Clinical Nutrition (2000) 54, 555±562
Introduction
Recent ®ndings suggest that the prevalence of undernutrition in the free living elderly population is relatively low,
with the recently published National Diet and Nutrition
Survey (NDNS) of people aged 65 y and over indicating
*Correspondence: MC Murphy, European Institute of Health and Medical
Sciences, University of Surrey, Stag Hill, Guildford, Surrey GU2 5XH,
UK.
E-mail: [email protected]
Guarantor: MC Murphy.
Contributors: MCM was the main author, supervised the biochemical
analyses and carried out the statistical analysis. MLL's experience in this
patient group and the hospital setting initiated the research area and study
by writing the project protocol, and she also supervised the project. CNB
recruited the subjects and carried out the MNA interviews. SAN helped
write the proposal, set up the database and also supervised the dietary
analysis.
Received 28 October 1999; revised 22 February 2000; accepted
6 March 2000
that only 3% of men and 6% of women had a BMI of less
than 20. However, the percentage of individuals classi®ed
as underweight increased to about 1 in 6 (16%) among the
institutionalized (Finch et al, 1998). The prevalence of
malnutrition in patients on admission to hospital has been
estimated to be much higher at about one-quarter to onehalf of all patients admitted (McWhirter & Pennington,
1994; Cohendy, 1998). In the elderly population, fracturing
a hip is one of the commonest reasons for being admitted to
hospital. Incidence has been estimated to be 54,000 cases
per year in England alone, with cases occupying one-third
of all hospital beds. The hospital costs alone of hip fracture
have been estimated to reach £360 million per year (Hollingsworth et al, 1995). Demographic trends indicate that
by the year 2021, 25% of the population will be over 65 y
and, since approximately 30% of people over the age of
65 y fall each year, increasing to 50% in those over 80 y
(Vellas et al, 1992), the risk of suffering an osteoporotic
The MNA in elderly orthopaedic patients
MC Murphy et al
556
fracture is high. Hip fractures are themselves associated
with poor outcome (Grimley-Evans, 1982). The fracture
may mark the end of a person's independent life, with
reports that only 50 ± 60% of patients return to their home
(Parker et al, 1996). It is known that nutritional status of the
individual affects clinical outcome (Lumbers et al, 1996;
Bastow et al, 1983).
The link between malnutrition in the elderly and recovery from illness and surgery has been highlighted as an area
which needs further investigation (Department of Health,
1992). However, the identi®cation and classi®cation of
nutritional status in the elderly is a dif®cult phenomenon
to study as no gold standard methodologies exist, although
several different approaches have been used. The MiniNutritional Assessment (MNA) has the bene®t of simplicity
and ease of use by relatively untrained personnel. The
original form of the MNA, which was divided into anthropometric measurements, global evaluation, dietetic assessment and subjective assessment, gave a maximum score of
30 and was extensively validated in Toulouse and New
Mexico in over 600 subjects (Guigoz et al, 1994; Guigoz,
1997). However, since the original validation (Guigoz et al,
1994) a new version of the MNA (Figure 1) has been
developed, consisting of an initial screening section followed by a more detailed assessment for subjects who
scored less than 11 on the initial screen.
The aim of the investigation was to evaluate the use of
the MNA in the UK in a group of female orthopaedic
patients recruited from the Royal Surrey County Hospital
(RSCH), Guildford. The project was designed to: (i) assess
the new version of the MNA and determine the sensitivity
and speci®city of the classi®cation of elderly patients;
and (ii) assess the comparability of the MNA with more
traditional assessments of nutritional status including
current dietary intake, biochemical and anthropometric
measurements.
Methods
Subject recruitment
All patients aged over 60 y admitted for emergency surgery
for a fractured neck of femur or elective surgery for a total
hip replacement into RSCH orthopaedic wards (namely
Ewhurst and Bramshot) were approached and asked if they
would be willing to take part in the study. The study was
approved by the South West Surrey District Local Research
Ethics Committees and all subjects gave written consent to
participate. Patients who agreed to participate were
assessed using the Abbreviated Mental Test (AMT).
Patients with mental impairment (as judged by a score
< 4 on AMT) in whom informed consent or reliable
communication was not possible were excluded.
Assessments of nutritional intake
The current nutritional intake of subjects was assessed
starting on day 5 after surgery by carrying out three
consecutive 24 h recalls using the hospital menu cards as
prompts to aid accurate recall. Nutritional Analysis was
made using Diet5 for Windows (The Robert Gordon
Institute, Aberdeen, UK). This method has previously
been used for the assessment of dietary intake in the elderly
(Lumbers et al, 1998; New et al, 1998).
European Journal of Clinical Nutrition
Anthropometry
In addition to weight, height, mid-upper arm (MAC) and
calf circumference (as required by the MNA), demispan
was measured for the calculation of demiquet
(weight=demispan2) and mindex (weight=demispan).
Biochemical analyses
Blood samples were collected by agreement with the
hospital phlembotomists and transported to the university
within 2 h of collection. Plasma was separated and stored
frozen until analysis. Analysis of two hepatic proteins with
varying half-lives gave an estimate of nutritional status
over different time points. Plasma albumin and transferrin
were analysed by standard Cobas Mira Kits (Ultimate 3
Albumin, Ultimate 3 Transferrin: Roche Diagnostics,
Welwyn Garden City, UK) and C-reactive protein (CRP)
by a kit supplied by Sigma (no. 371-A, Dorset, UK).
Albumin has a half-life of approximately 19 days and
therefore gives a long-term assessment of visceral protein
depletion. However, albumin values are also affected by
liver disease, sepsis and dehydration. Transferrin has a halflife of 9 days, but is affected by confounding iron de®ciency. C-reactive protein (an acute phase protein) is
measured to assess the level of in¯ammation or tissue
destruction as these affect albumin and transferrin levels.
CRP levels are raised immediately post-operatively but
normalize on recovery.
Statistical analysis
The data were analysed statistically using SPSS for Windows
version 8 (SPSS, Woking, UK, 1999). Statistical analysis of
the data involved examination of the classi®cation of patients
with the MNA (screening and the full examination) to
establish whether the malnourished, at risk and well nourished were discrete groups and whether there were any
signi®cant differences in nutritional intake, anthropometric
and biochemical data between them. The anthropometric and
biochemical data for each of the three groups were analysed
using One-way analysis of variance (ANOVA) followed by a
Tukey multiple comparisons test to locate the differences.
The use of this multiple range test has the added advantage in
that it is based on 95% con®dence interval limits of each
estimate and is more informative than the simple t-test
because it identi®es which of the quartiles differ from each
other. The dietary data due to the fact that some of it was not
normally distributed was analysed by the nonparametric
Mann ± Whitney U-test. Association between the data was
analysed by Pearson's test for correlations. Values of
P< 0.05 were taken as the lowest level of signi®cance.
Stepwise linear multiple-regression analysis was used to
identify the questions in the MNA which best predicted the
MNA total and the screening score. In addition, the sensitivity and speci®city was examined to assess whether the risk
scores are associated with other indices of nutritional assessment according to the BDA Nutrition Screening Tools
Professional Development Committee Brie®ng paper (British Dietetic Association, 1999) as summarized is the table
below.
Sensitivity
Speci®city
Characteristics
Outcome
At risk patients are correctly
labelled
Those not at risk correctly
labelled
Those at risk not missed
Avoid inappropriate action
The MNA in elderly orthopaedic patients
MC Murphy et al
The prognostic signi®cance of the MNA model was
ascertained by calculating the sensitivity and speci®city of
the MNA scores. The value of a predictive model (in this
case the MNA) lies in its ability to detect patients who have
malnutrition (its sensitivity) and to exclude those who are
not suffering from malnutrition (its speci®city). The
number of subjects who had been correctly identi®ed by
the MNA to have low values (true positives, TP), those
correctly labelled by the MNA to have high values (true
negatives, TN), those with low values that were missed by
the MNA scoring (false negatives, FN) and the number of
false positives (FP) (where the MNA suggested malnutrition but the other variables did not) were counted for each
variable. Using these variables the sensitivity and speci®city was de®ned (according to Bland, 1997) as below:
Sensitivity ˆ TP=…TP ‡ FN†6100
Specificity ˆ TN=…TN ‡ FP†6100
The sensitivity or the true positive ratio is the proportion of
patients who are categorized correctly by both the MNA
and the other variables, whereas the speci®city or false
positive ratio is the probability that patients without low
variables will be classi®ed with low MNA scores. These
indices were calculated by identifying patients either in
group 1 alone (de®ned as malnourished by an MNA score
of less than 17) and in groups 1 and 2 combined (those who
scored less than 23.5 on the MNA). The numbers of
patients were counted whose malnutrition had been classi®ed according to the following parameters:
albumin < the lower reference limit (35 g=l);
energy intake < 1 s.d. below the mean intake
(< 2997 kJ=day);
mindex < 50% percentile (81.7 kg=m).
These three criteria were used as markers for biochemical,
dietary and anthropometric measurements and because they
were not parameters directly recorded in the MNA itself.
Results
A total of 49 patients were recruited with 42 patients
having had an emergency admission for a fractured neck
of femur and seven patients elective surgery for a total hip
replacement. The mean age of subjects was 79.5 y (range
61 ± 103). Descriptive statistics for the study group including MNA scores, age, anthropometry and biochemistry are
shown in Table 1. Current dietary intake data for the whole
group are presented in Table 2 with comparisons shown for
the dietary recommendations (Department of Health, 1991)
and the ®ndings from the recently published NDNS survey
in the elderly (Finch et al, 1998). For subsequent analysis,
the subjects were split into three groups according to their
total MNA score. Categories were as follows:
557
Group 1 Ð malnourished (scored < 17);
Group 2 Ð at-risk from undernourishment (scored
17 ± 23.5);
Group 3 Ð well nourished (scored >23.5).
The MNA scores, age, anthropometry and biochemistry
for each of the three groups are given in Table 3 and the
dietary data in Table 4.
Results of the MNA categorization
The total MNA scores suggested 16% (n ˆ 8) of the group
were malnourished, 47% (n ˆ 23) were at risk from being
malnourished and 37% (n ˆ 18) of the group were well
nourished. As expected, there were signi®cant differences
between the screening and the total MNA scores of the
three groups. In addition, there were signi®cant differences
between the malnourished group and the well nourished
group for body weight (P < 0.001), BMI (P< 0.001), demiquet (P < 0.001) and mindex (P < 0.001), but none of the
biochemical parameters. However, there was a weak but
signi®cant correlation between albumin and total MNA
score (r ˆ 0.34, P ˆ 0.05). There were also signi®cant
differences between the group classi®ed as being `at risk'
with the malnourished group for body weight (P ˆ 0.005),
BMI (P ˆ 0.004), demiquet (P ˆ 0.005) and mindex
(P ˆ 0.005). Using this classi®cation there were stepwise
increases in albumin levels, but this did not reach statistical
signi®cance. Many of the patients studied were found to
have raised CRP values, indicative of trauma associated
with the fall and surgery. In addition, the high CRP values
observed in some individuals, whilst not diagnostic, were
suggestive of sepsis.
Stepwise linear multiple regression analysis was used to
identify which questions in the MNA best predicted the
screening or total scores. The screening score was signi®cantly predicted by all the questions in the screening
section but the questions could be rated in the following
order of decreasing signi®cance: question B (weight loss),
F (BMI rating), D (psychological stress), C (mobility), A
(loss of appetite) and E (neuropsychological problems)
(Table 5). Thus 55% of the variation in the screening
score was predicted by weight loss in the previous three
months. The MNA total was signi®cantly predicted by the
Table 1 Descriptive data for MNA scores, age, anthropometric data and biochemical data for the 49 patients
Whole group (n ˆ 49)
Reference 50% percentile (Pen Group, 1998)
Variables
Mean
s.d.
Min
Max
64 ± 74 y
75 ‡ y
Age (y)
Body weight (kg)
BMI (kg=m2)
Demispan (mm)
Demiquet (kg=m2)
Mindex (kg=m)
MNA screen (max score 14)
MNA total (max score 30)
80
60
23.7
72.5
112.6
82.4
10.1
22.1
9
12
4.3
3.9
24.6
18.3
2.3
3.4
61
38
15.5
65.0
69.7
52.3
5.0
13.0
103
89
32.5
83.0
164.1
116.5
14.0
29.5
Ð
63
24.7
73.6
116.3
84.8
Ð
Ð
Ð
59
23.6
72.5
112.2
81.7
Ð
Ð
4.7
0.73
40.6
29.0
1.00
3.1
46.0
3.99
120.0
Albumin (g=l)
Transferrin (g=l)
CRP (mg=l)
36.9
2.22
52.0
Reference range
35.0 ± 45.0
2.20 ± 3.80
< 10
European Journal of Clinical Nutrition
The MNA in elderly orthopaedic patients
MC Murphy et al
558
Table 2 Dietary energy and nutrient intakes of the subject group compared with the recommendations (estimated average
requirement (EAR) or the reference nutrient intake (RNI)) (Department of Health, 1991) and to data from the National Diet and
Nutrition Survey (Finch et al, 1998). Data are expressed as medians and upper (95%) and lower (5%) con®dence intervals (CI)
Whole group (n ˆ 41)
Comparisons
Intake per day
Median
Lower CI
Upper CI
EAR=RNI
NDNS study
Energy intake (kJ)
Fat (g)
Carbohydrate (g)
Protein (g)
Thiamine (mg)
Ribo¯avin (mg)
Niacin (mg)
Pyridoxine (mg)
Folate (mg)
Vitamin C (mg)
Vitamin A (mg)
Vitamin D (mg)
Calcium (mg)
Phosphorous (mg)
Magnesium (mg)
Sodium (mg)
Potassium (mg)
Iron (mg)
Selenium (mg)
Zinc (mg)
Non-starch polysaccharides (g)
4414
41
113
43
0.82
1.16
15.5
0.92
144
44
482
1.35
538
720
143
1834
1494
5.5
27.1
5.0
7
3771
36
109
38
0.70
0.98
12.7
0.85
120
46
61
1.24
481
665
130
1550
1410
5.1
23.7
4.4
6
4500
45
133
46
0.87
1.36
17.7
1.07
166
65
1735
2.00
584
789
158
1951
1710
6.5
32.4
5.5
8
7960 (60 ± 74),7610 (75 ‡ )
110
172
47
0.76
1.10
12.0
1.20
200
40
700
10.00
700
550
270
1600
3500
8.7
60.0
7.0
12 ± 18
5980
58
175
56
1.73
1.76
26.1
2.00
220
68
1073
3.44
491
707
141
1471
1588
6.2
Ð
4.9
11
Table 3 Classi®cation of the patient group by MNA scores; comparison of age, anthropometric data and
biochemical data
Malnourished (n ˆ 8)
Group 1
Age (y)
Body weight (kg)
BMI (kg=m2)
Demispan (mm)
Demiquet (kg=m2)
Mindex (kg=m)
MNA screen (max score 14)
MNA total (max score 30)
Albumin (g=l)
Transferrin (g=l)
CRP (mg=l)
At risk (n ˆ 23)
Group 2
Well nourished (n ˆ 18)
Group 3
Mean
s.d.
Mean
s.d.
Mean
s.d.
79.6
46.4*,**
18.7*,**
72.3
82.*,**
59.*,**
6.1**
16.1**
33.8
1.64
68.3
6.6
9.0
3.2
3.5
10.
5.
1.5
1.4
3.4
0.54
51.8
81.5
60.8*
24.0*
72.5
117.*
84.*
10.2**
21.7**
36.2
2.39
47.2
10.2
10.8
4.3
4.4
25.
16.
1.3
1.5
3.9
0.85
35.4
78.0
65.6**
25.7**
72.8
122.**
90.**
11.7**
25.3**
39.3
2.15
53.4
7.1
10.7
3.0
3.3
17.
14.
1.5
1.5
5.4
0.46
47.9
Statistical differences between groups are indicated by similar symbols (*P < 0.01, **P < 0.001).
following variables: question F (BMI rating), B (weight
loss), C (mobility), O (self-view of nutritional status), J
(number of full meals eaten daily) and E (neuropsychological problems) in order of decreasing signi®cance (Table
6). This indicates that 53.5% of the variation in the MNA
score was explained by the variation in BMI rating. In
addition, this technique was used to assess if any of the
anthropometric, biochemical or dietary data signi®cantly
predicted the MNA screening or total scores. The MNA
screening score was predicted primarily by mindex
(adjusted r2 ˆ 0.36, P < 0.001) whereas the MNA total
scores were predicted primarily by body weight, (adjusted
r2 ˆ 0.42, P < 0.001) when all factors were considered.
Interestingly, neither weight for height nor demispan
were independent predictors, nor were any of the biochemical indices.
Dietary intake results
None of the subjects consumed the EAR for energy intake
(Department of Health, 1991), nor did they reach the
European Journal of Clinical Nutrition
NDNS mean energy intake in free-living individuals
(Finch et al, 1998). This was because the mean intakes of
fat, protein and carbohydrate were all extremely low. Mean
intakes below the RNI were noted for pyridoxine, folate,
vitamin D, calcium, magnesium, potassium, iron, selenium,
zinc and NSP. In addition, mean intakes of vitamin D,
magnesium, potassium, selenium and NSP were below the
LRNI. Using the MNA to classify patients into malnourished, at risk and well nourished groups appeared to
separate subjects into levels of dietary intake in that all
nutrients except sodium followed a stepwise increase
across the three classi®cation groups. There were signi®cant differences between the undernourished group and the
at risk group for carbohydrate (P ˆ 0.03) and ribo¯avin
(P ˆ 0.04) intakes and between the at-risk and the wellnourished group for selenium intake only (P ˆ 0.01). However, as might be expected, the majority of signi®cant
differences were between groups 1 (malnourished) and
group 3 (well nourished). These were for carbohydrate
(P ˆ 0.008), protein (P ˆ 0.007), ribo¯avin (P ˆ 0.003),
The MNA in elderly orthopaedic patients
MC Murphy et al
559
Table 4 Classi®cation of the patient group by MNA scores; comparison of macro- and micronutrient intakes between groups. Data are expressed as
medians and upper (95%) and lower (5%) con®dence intervals (CI)
Malnourished (n ˆ 8)
Group 1
Intake per day
Energy (kJ)
Fat (g)
Protein (g)
Carbohydrate (g)
Thiamine (mg)
Ribo¯avin (mg)
Niacin (mg)
Pyridoxine (mg)
Folate (mg)
Vitamin C (mg)
Vitamin A (mg)
Vitamin D (mg)
Calcium (mg)
Phosphorus (mg)
Magnesium (mg)
Sodium (mg)
Potassium (mg)
Selenium (mg)
Zinc (mg)
Iron (mg)
Non-starch polysaccharides (g)
At risk (n ˆ 23)
Group 2
Well nourished (n ˆ 18)
Group 3
Median
Lower CI
Upper CI
Median
Lower CI
Upper CI
Median
Lower CI
Upper CI
2947
30.0
32.7*
80.*,**
0.75
0.81*,**
11.5**
0.65**
103.**
49.9
287.
0.97
416.**
582.**
121
2099
1346
18.0*
3.9
4.3**
4.4*
1087
4.0
10.7
47
0
0.62
6.4
0.58
73
18.9
0
0.50
171
284
60
763
669
8.0
1.4
1.6
0.9
5068
66.1
53.8
108
1.03
0.92
16.0
0.76
141
78.1
897
1.4
634
826
146
2593
1720
25.5
5.7
5.7
6.8
4389
42.2
42.7
111.*
0.79
1.20*
14.5
0.95
145
40.5
520
1.23
529
717
126
1772
1465
27.0*
5.2
5.7
6.6
3475
32.0
35.5
101
0.65
0.97
10.7
0.77
93
38.1
402
0.81
431
595
117
1324
1292
18.3
4.2
4.5
5.4
4623
45.3
48.7
136
0.89
1.30
18.8
1.11
151
66.1
581
1.92
604
802
158
1856
1720
30.9
5.7
6.3
7.9
4438
44.0
44.0*
130.**
0.82
125.**
17.5**
0.96**
149.**
53.3
542
1.79
560.**
757.**
149
1790
1614
37.0*
5.3
6.4**
7.6*
3877
36.6
39.0
110
0.71
0.83
14.4
0.84
119
39.8
0
1.22
483
699
139
1477
1387
28.5
4.1
5.5
6.5
4955
50.0
47.3
153
1.00
1.92
20.3
1.20
222
79.2
4326
2.83
616
856
185
2275
1949
44.3
6.4
7.8
10.3
Statistical differences between groups are indicated by similar symbols (* P < 0.05, ** P < 0.005).
Table 5 Questions in the MNA which most signi®cantly predicted the
screening score
MNA
question
B
F
D
C
A
E
Descriptor
Adjusted
r2 values (%)
P value
Weight loss over previous 3 months
BMI rating
Psychological stress
Mobility rating
Loss of appetite
Neuropsychological problems
55
17
12
9
3
3
0.001
0.001
0.001
0.001
0.001
0.001
Table 6 Questions in the MNA which signi®cantly predicted the total
score
MNA question
F
B
C
O
J
E
Descriptor
BMI rating
Weight loss over previous 3 months
Mobility rating
Self-view of nutritional status
Number of meals eaten per day
Neuropsychological problems
Adjusted r2
values (%) P value
53.5
9.9
2.7
4.1
3.5
1.9
0.001
0.001
0.003
0.004
0.007
0.049
niacin (P ˆ 0.004), pyridoxine (P ˆ 0.004), folate
(P ˆ 0.005), calcium (P ˆ 0.05), phosphorus (P ˆ 0.04),
iron (P ˆ 0.006) and selenium (P ˆ 0.04) intakes.
Sensitivity and speci®city
Since none of the subjects were eating the EAR for energy
(Department of Health, 1991) at 7610 kJ=day, this cut-off
point was not thought to be appropriate. Using albumin as
the comparison, the sensitivity of the MNA classi®cation of
group 1 was 27% and the speci®city was 66%, using energy
intake the sensitivity was 57% and the speci®city 94% and
using mindex the sensitivity was 33% and the speci®city
100%. This suggests that, using an MNA score of less than
17 to identify subjects, in general the MNA was not so
sensitive in picking up low values but fairly speci®c, in that
all the individuals in group 1 also had low mindex scores,
for example (ie there were no false positives).
When an MNA score of less than 23.5 was used to
identify patients compared with albumin, the sensitivity of
the MNA classi®cation was 75% and the speci®city 50%,
using energy intake the sensitivity was 100% and the
speci®city 37%, and using mindex the sensitivity was
81% and the speci®city 47%. Therefore using groups 1
and 2 to identify patients the sensitivity was improved, ie
more patients are picked up but the speci®city declines as
there are more false positives. The cut-off value for energy
intake (< 2997 kJ=day) was considered to be extremely low
and therefore the analysis was repeated at a higher cut-off
point of mean intake (4054 kJ). This resulted in a sensitivity
of 72% and a speci®city of 32%, therefore the sensitivity
and speci®city had declined, as there were more cases of
misclassi®cation.
Discussion
According to the MNA classi®cation, 16% of the group
were diagnosed as malnourished, and 47% were at risk of
undernourishment which compares well with the ®gures
quoted by McWhirter and Pennington (1994) and Cohendy
(1998) at about one-quarter to one-half of all patients
admitted. Although the mean values for demiquet,
mindex, albumin and transferrin were within the reference
ranges, a high percentage of individual values fell below
the reference ranges. The low albumin levels observed in
elderly FNF patients in the present study con®rms the
®ndings of Huang et al (1996) and is linked to the
trauma associated with hip fracture and the observed
raised CRP levels (Gersovitz et al, 1980, Pepys, 1987).
The signi®cant correlation found between total MNA score
and albumin level also con®rmed those of Vellas et al
(1999) who show signi®cant correlations between MNA
European Journal of Clinical Nutrition
The MNA in elderly orthopaedic patients
MC Murphy et al
560
Figure 1 The Mini Nutritional Assessment (MNA).
European Journal of Clinical Nutrition
The MNA in elderly orthopaedic patients
MC Murphy et al
score and BMI, albumin, prealbumin, prognostic in¯ammatory and nutritional index (a scoring system which
includes CRP measurements) and serum zinc. According
to the dietary intake data, particularly low intakes were
observed in this group. None of the subjects were consuming the EAR for energy and therefore the intake of most
other nutrients was also compromised. The mean intakes of
pyridoxine, folate, vitamin D and calcium, magnesium,
potassium, iron, selenium, zinc and NSP were below the
RNI and of even greater concern was that mean intakes of
vitamin D, magnesium, potassium, selenium and NSP were
below the LRNI. Low vitamin D and calcium intakes were
also highlighted in the NDNS report ®ndings (Finch et al,
1998). Low intakes in this group may not be surprising in
that the intake measurements were taken when the subjects
were recovering from an operation and many were nauseous. However, it would be extremely interesting to examine
how long poor intakes persist in this recovery phase and the
affect on clinical outcome.
Splitting the group by their MNA score clearly separated
the group according to their body weight and the other
anthropometric measurements. Anthropometric measurements appear to be very important as the question relating
to BMI in the MNA assessment and mindex, when all the
measurements were considered, were the most predictive of
the MNA total. In addition, the MNA categorization
appeared to separate the group into levels of energy and
nutrient intake. Although not signi®cant in every case, due
to the spread of data and the low numbers of subjects, there
were clear stepwise trends across the groups for the
majority of nutrients.
The study examined the use of the MNA in hip fracture
patients on day 5 of their hospital stay. At this time the
patient is traditionally viewed as in the recovery phase, and
their blood biochemistry and food intake are returning to
pre-admission levels. However, the raised CRP levels
might suggest that the effects of the trauma were still
persisting. There was some dif®culty implementing the
MNA due to the changing circumstances of the patients.
Some of the questions on the MNA relate to previous habits
and others to the present time and therefore the authors
would advise the development of some detailed guidance
notes to accompany the MNA, particularly for health
professionals using the MNA in acute clinical situations.
The MNA has been described as taking less than 10 min to
complete (Vellas et al, 1999), but in our experience, with
hospitalized elderly, it takes at least 30 min, especially to
obtain the anthropometric data especially weight, that is
required. In practice, regrettably, it is common to ®nd that
patients do not have body weight recorded in their notes
and little or no attempts at nutritional assessment. The
importance of nutritional assessment must be fully understood and accepted if nurses or other staff are to be willing
to commit the time to complete the MNA assessment
properly.
The stepwise multiple regression analysis highlighted
the questions in the MNA which were the most predictive
to the total score. This assessment identi®ed BMI and
weight loss as the most important factors, but also identi®ed mobility, self-view of nutritional status, number of
meals eaten per day and neuropsychological problems as
signi®cant but less important predictors.
The sensitivity and speci®city of the MNA were compared with albumin levels, mindex and energy intake. As
stated previously, there are no gold standards with which to
compare the MNA and it must be clearly stated that none of
the comparisons are ideal. Albumin measurements in this
group are compromised due to the raised CRP levels. The
use of albumin levels as a marker of nutritional status is
controversial but widely used. Mitchell and Lipshitz (1982)
found albumin to be the best predictor of malnutrition in
any age group, whilst Friedman et al (1985), disputed its
use as a marker of nutritional status. It is well established
that the levels of both albumin and transferrin are affected
by many other factors including dehydration, trauma and
sepsis (Gibson, 1990). CRP is an acute phase protein which
is produced in response to a wide range of stimuli including
microbial invasion, tissue injury, immunologic reactions
and in¯ammatory responses; however, it is not diagnostic
of any particular condition (Pepys, 1987). This highlights
the need for more suitable biochemical indicators of undernutrition in elderly patients. Further, the use of any single
factor such as albumin, energy intake or mindex against
which to compare the MNA is not entirely satisfactory as
the MNA encompasses many other factors which are
known to affect undernutrition in the elderly. The ®nding
that the MNA score had a high predictive capability
compared with other parameters not directly assessed in
the MNA, such as albumin levels, does lend weight to its
use as a diagnostic tool. However, the small number of
subjects in each category suggests that caution should be
used when interpreting the results. The patients with an
MNA score between 17 and 23.5 had normal body weight,
BMI and albumin levels above the lower end of the
reference range, but had very poor nutritional intake. This
®nding is in agreement with the ®ndings of Vellas et al
(1999), who highlight the importance of this ability of the
MNA to identify people at risk of malnutrition before
detectable changes in body weight or plasma albumin.
In practical terms the MNA could be used to identify
patients who would most bene®t from nutritional intervention. The uses of the sensitivity and speci®city data depend
on the nature of the treatment. If the treatment had many
adverse side effects, it would be advisable to treat only
those who had the disease, and therefore the method used to
detect patients must have a high speci®city (so there would
be no false positives but some people with the disease may
be missed). In the case of nutritional support, which has no
adverse side effects, it would be preferable to use a screening method with high sensitivity, thus the method does not
miss a malnourished patient (there are no false negatives)
but there is a risk of over-treatment. Costs accepted, giving
nutritional support to everyone with an MNA score of less
than 23.5 would treat 75% patients with a low albumin,
100% patients with energy intakes < 2997 kJ=day, 72%
with energy intake < 4054 kJ=day and 81% patients with a
low mindex. Further work is required to assess whether the
costs of nutritional interventions to this identi®ed group are
cost effective in terms of bene®cial clinical outcome,
reduced time under hospital care and complication rate.
561
Conclusion
These data suggest that the MNA is a useful diagnostic,
tool for the identi®cation of elderly patients at risk from
malnutrition and those who are already malnourished in
this hospital setting. However, further studies are required
to determine whether it is effective as a tool for monitoring
response to treatment, such as nutritional support.
European Journal of Clinical Nutrition
The MNA in elderly orthopaedic patients
MC Murphy et al
562
References
Bastow MD, Rawlings J & Allison SP (1983): Undernutrition, hypothermia and injury in elderly women with fractured neck of femur: an injury
response to altered metabolism? Lancet i, 143 ± 145.
Bland M (1997): An Introduction to Medical Statistics; 2nd edn, Oxford:
Oxford Medical Publications. pp 273 ± 276.
British Dietetic Association (1999): Nutrition Screening Tools. Professional Development Committee Brie®ng Paper no 9.
Cohendy R (1998): The MNA in preoperative nutritional evaluation: a
study of 419 surgical patients. In: Mini Nutritional Assessment (MNA)
Research and Practise in the Elderly. Nestle Nutrition Workshop
Series, pp 24 ± 25. Vevey, Switzerland: Nestec Ltd.
Department of Health (1991): Dietary Reference Values for Food Energy
and Nutrients for the United Kingdom. Report on Health and Social
Sciences, Vol 41. London: HMSO.
Department of Health (1992): The Nutrition of Elderly People. Report on
Health and Social Sciences, Vol 43. London: HMSO.
Finch S, Doyle W, Lowe C, Bates CJ, Prentice A, Smithers G & Clarke PC
(1998): National Diet and Nutrition Survey: People Aged 65 Y and
Over, Vol 1. London: HMSO.
Friedman P, Campbell A & Caradoc-Davies T (1985): Hypoalbuminaemia
in the elderly is due to disease not malnutrition. J. Clin. Exp. Gerontol.
7, 191 ± 203.
Gersovitz M, Munro HN, Udall J & Young VR (1980): Albumin synthesis
in young and elderly subjects using a new stable isotope methodology:
response to a level of protein intake. Metabolism 62, 1075 ± 1087.
Gibson R (1990): Assessment of protein synthesis. In: Principles of
Nutritional Assessment, pp 307 ± 320. Oxford: Oxford University Press.
Grimley-Evans J (1982): Epidemiology of proximal femoral fractures.
Rec. Adv. Geriatr. Med. 2, 201 ± 214.
Guigoz Y (1997): Recommended dietary allowances (RDA) for the freeliving elderly. In: Facts, Research and Intervention in Geriatrics, pp
113 ± 144. Paris: Serdi.
Guigoz Y, Vellas B & Garry PJ (1994): Mini Nutritional Assessment
(MNA): A Practical Tool for Grading the Nutritional State of Elderly
Patients. Facts and Research in Gerontology, Supplement 2, pp 15 ± 59.
European Journal of Clinical Nutrition
Hollingsworth W, Todd CJ & Parker MJ (1995): The cost of treating
hip fractures in the twenty ®rst century. J. Publ. Health Med. 17, 269 ±
276.
Huang Z, Himes JH & McGovern PG (1996): Nutrition and subsequent hip
fracture among a national cohort of white women. Am. J. Epidemiol.
144, 124 ± 134.
Lumbers M, Driver LT, Howland RJ & Williams CM (1996): Nutritional
status and clinical outcome in female surgical patients. J. Clin. Nutr. 15,
101 ± 105.
Lumbers M, Murphy MC, Pither CAR, Creedon MH, Older MWJ & New
SA (1998): Differences in dietary intake and food selection between
hospital patients and day-centre visitors: implications for supplementation. Proc. Nutr. Soc. 57, 56.
McWhirter JP & Pennington CR (1994): Incidence and recognition of
malnutrition in hospital. Br. Med. J. 308, 945 ± 948.
Mitchell C & Lipschitz D (1982): The effect of age and sex on the
routinely used measurements to assess the nutritional status of hospital
patients. Am. J. Clin. Nutr. 36, 340 ± 349.
New SA, Murphy MC, Pither CAR, Creedon MH, Older MWJ & Lumbers
M (1998): Assessment of risk factors for osteoporosis: comparison
between hip fracture patients and healthy age-matched controls. Proc.
Nutr. Soc. 57, 72.
Parker MJ, Twemlow TR & Pryor GA (1996): Environmental hazards and
hip fractures. Age Ageing 25, 322 ± 325.
Pen Group (1998) A Pocket Guide to Nutrition, 2nd edn, ed. VE
Todorovis & A Micklewright. Birmingham: British Dietetic
Association.
Pepys MB (1987): The acute phase response and CRP. In: Oxford Textbook of Medicine, 9th edn, ed. DJ Weatherall, J Ledinghan & DA
Warrell, pp 157 ± 164. Oxford: Oxford University Press.
Vellas B, Baumgartner RN, Wayne SJ, Conceicao J, Lafont C, Albarede
JL & Garry PJ (1992): The relationship between malnutrition and falls
in the elderly. Nutrition 8, 105 ± 108.
Vellas B, Guigoz Y, Garry PJ, Nourhashemi F, Bennahum D, Lauque S &
Albarede J-L. (1999): The Mini-Nutritional Assessment (MNA) and its
use in grading the nutritional state of elderly patients. Nutrition 15,
116 ± 122.