MUAC indicates mid upper arm circumference; WHZ, weight-for-height z score. Large blue circles indicate the performance of
various cutoff points for MUAC
Berkley J, Mwangi I, Griffiths K, Ahmed I, Mithwani S, English M, Newton C, Maitland K. Assessment of Severe Malnutrition Among Hospitalized Children in Rural
KenyaComparison of Weight for Height and Mid Upper Arm Circumference. JAMA. 2005;294(5):591-597. doi:10.1001/jama.294.5.591
Author Affiliations: Kenya Medical Institute
Centre for Geographic Medicine Research (coast), Kilifi, Kenya (Drs Berkley,
Mwangi, Griffiths, Ahmed, Mithwani, English, Newton, and Maitland); Centre
for Clinical Vaccinology and Tropical Medicine (Dr Berkley) and Department
of Paediatrics (Dr English), University of Oxford, Oxford, England; Institute
of Child Health, London, England (Dr Newton); and Department of Paediatrics,
Imperial College, London (Dr Maitland).
Context Severe malnutrition has a high mortality rate among hospitalized children
in sub-Saharan Africa. However, reports suggest that malnutrition is often
poorly assessed. The World Health Organization recommends using weight for
height, but this method is problematic and often not undertaken in practice.
Mid upper arm circumference (MUAC) and the clinical sign “visible severe
wasting” are simple and inexpensive methods but have not been evaluated
in this setting.
Objectives To evaluate MUAC and visible severe wasting as predictors of inpatient
mortality at a district hospital in sub-Saharan Africa and to compare these
with weight-for-height z score (WHZ).
Design, Setting, and Participants Cohort study with data collected at admission and at discharge or death.
Predictive values for inpatient death were determined using the area under
receiver operating characteristic curves. Participants were children aged
12 to 59 months admitted to a district hospital in rural Kenya between April
1, 1999, and July 31, 2002.
Main Outcome Measure MUAC, WHZ, and visible severe wasting as predictors of inpatient death.
Results Overall, 4.4% (359) of children included in the study died while in
the hospital. Sixteen percent (1282/8190) of admitted children had severe
wasting (WHZ ≤−3) (n = 756), kwashiorkor (n = 778),
or both. The areas under the receiver operating characteristic curves for
predicting inpatient death did not significantly differ (MUAC: 0.75 [95% confidence
interval, 0.72-0.78]; WHZ: 0.74 [95% confidence interval, 0.71-0.77]) (P = .39). Although sensitivity and specificity
for subsequent inpatient death were 46% and 91%, respectively, for MUAC less
than or equal to 11.5 cm, 42% and 92% for WHZ less than or equal to −3,
and 47% and 93% for visible severe wasting, the 3 indices identified different
sets of children and were independently associated with mortality. Clinical
features of malnutrition were significantly more common among children with
MUAC less than or equal to 11.5 cm than among those with WHZ less than or
equal to −3.
Conclusions MUAC is a practical screening tool that performs at least as well as
WHZ in predicting subsequent inpatient mortality among severely malnourished
children hospitalized in rural Kenya. Visible severe wasting is also a potentially
useful sign at this level, providing appropriate training has been given.
Severe malnutrition has a high mortality rate among hospitalized children
in sub-Saharan Africa.1,2 The
World Health Organization (WHO) gives clear guidelines for the management
of children with severe malnutrition,3 and
where standardized management protocols have been followed, mortality has
been reduced.4,5 However, recent
reports suggest that there are problems in initially identifying severely
malnourished children at hospital admission.5,6 The
WHO defines severe malnutrition requiring hospital admission as weight-for-height
z scores (WHZs) of less than or equal to −3 or as less than or equal
to 70% of the reference median using US National Center for Health Statistics
(NCHS)/WHO reference values (severe wasting)3 or
symmetrical edema involving at least the feet (edematous malnutrition, kwashiorkor).3
There are several reasons why screening by WHZ is potentially problematic
and is frequently not undertaken in practice in sub-Saharan Africa. First,
height is difficult to measure accurately in children at any time but especially
so in a busy ward and when children are ill or distressed. The measurement
of weight depends on the presence of properly calibrated and functioning scales,
which often are not available.7 The actual
weight-for-height determination depends on correctly recording 2 separate
values and then looking up a third value on a chart, which must be readily
In contrast, mid upper arm circumference (MUAC) is a simple, low-cost,
objective method of assessing nutritional status. The MUAC is generally as
good as, or better than, other anthropometric measures in predicting subsequent
mortality in community-based studies (reviewed by Pelletier8).
In Uganda, for example, MUAC was the strongest predictor of subsequent death
of any anthropometric index among 3748 village children.9 Among
hospital admissions, MUAC was found to be as effective as other nutritional
indices in predicting death in a study of 352 children with diarrhea in Bangladesh.10 During famine, MUAC performed as well as body mass
index among adults admitted to a feeding center in Sudan.11
The clinical sign of “visible severe wasting” is a component
of the WHO Integrated Management of Childhood Illness evaluation that accounts
for the likelihood that measuring facilities are not available at the primary
care level in this setting. Given the challenges to the assessment of weight
for height at the secondary (referral) level, evaluation of malnutrition is
commonly performed by observation only, often without formal training. Despite
being inexpensive, convenient, and apparently effective in other settings,
MUAC and visible severe wasting have not been evaluated as predictors of mortality
among hospitalized children in sub-Saharan Africa.
Since 1998, we have conducted detailed clinical surveillance of all
children admitted to Kilifi District Hospital in Kenya. Using data collected
at admission and at discharge or death, we have previously reported that a
small number of clinical features, including neurologic status, respiratory
signs, temperature, and weight for age effectively predicted inpatient deaths
despite a variety of underlying etiologies.12 Our
principal aim in this analysis was to examine the predictive value for inpatient
death of MUAC compared with WHZ among children aged 12 to 59 months. We also
sought to determine whether there are any differences in the children identified
by the 2 methods. Furthermore, we aimed to evaluate both the clinical sign
of visible severe wasting as a predictor of subsequent inpatient death and
also MUAC as an indicator of the presence of the current WHO standard definition
of severe wasting, ie, WHZ less than or equal to −3.
The study was conducted at Kilifi District Hospital, located in a rural,
malaria-endemic area on the Kenyan coast. The hospital serves a population
of more than 200 000 people who are mainly rural farmers. Approximately
10% of women attending the hospital antenatal clinic were infected with human
immunodeficiency virus (HIV) in 2000.13 Antiretroviral
therapy was not in routine use at the time of the study. The Kenya Medical
Research Institute Centre for Geographic Medicine Research (coast) is located
at the hospital.
Data were prospectively collected from all pediatric admissions as part
of an ongoing surveillance study, and details of the clinical data collection
have been described elsewhere.12 Approval was
given by the Kenya Medical Research Institute national ethical and scientific
review committees. Parents or guardians of individual study participants gave
written informed consent in their own language. For this analysis we examined
data from all children aged 12 to 59 months admitted between April 1, 1999,
and July 31, 2002. Trained clinical assistants measured MUAC with a nonstretch
cloth measuring tape, weight with an electronic scale (Weylux, London, England)
that was checked for consistency weekly, and length using a measuring board
of standard design (for children younger than 2 years) or height using a wall-mounted
scale (for those aged 2 years or older). From September 1, 1999, onward, clinicians
were trained to recognize visible severe wasting, and data on this sign were
subsequently collected. Clinicians were taught to look for muscle loss, manifested
as a wasting of the gluteal area and as the presence of a bony prominence
over the chest wall.3,14 Children
with a clinical diagnosis of severe malnutrition were treated with broad-spectrum
antibiotics, vitamin and mineral supplements, F75 and F100 milk, and careful
attention to fluid and electrolyte intake. Other treatments were according
to WHO guidelines3 and local protocols. Inpatient
deaths were confirmed on clinical examination by research clinicians.
We calculated WHZ, weight-for-age z score (WAZ),
and height-for-age z score (HAZ) using the NCHS reference
standards3 with EpiInfo version 6.04 (Centers
for Disease Control and Prevention, Atlanta, Ga). Complete nutritional assessment,
especially for height, were sometimes omitted when children arrived in extremis.
Thus, to avoid bias, only children with complete data for all 3 indices were
included in the main analysis. To determine the predictive value for inpatient
death, we calculated the area under the receiver operating characteristic
(ROC) curves with 95% confidence intervals (CIs) using the roctab and roccomp (a χ2 test)
commands in STATA version 8.0 (Stata Corp, College Station, Tex). We also
examined the sensitivities and specificities of commonly used cutoff values
and examined the clinical data for differences between children identified
as severely malnourished by the MUAC and WHZ methods. Sensitivity was defined
as the number of inpatient deaths among children with anthropometric measures
equal to or below a cutoff value, divided by the total number of inpatient
deaths. Specificity was defined as the number of children discharged alive
with anthropometric measures above a cutoff value, divided by the total number
of children discharged alive.
Since the WHO recommends that children be treated for severe malnutrition
if they have severe wasting or kwashiorkor, we also evaluated the positive
and negative likelihood ratios for death for each of MUAC, WHZ, and visible
severe wasting combined with (and/or) kwashiorkor. These were calculated as
follows: positive likelihood ratio = sensitivity/(1 − specificity),
and negative likelihood ratio = (1 − sensitivity)/specificity.15 We used the χ2 test, or the Fisher
exact test where appropriate, to compare categorical data. We used the Kruskal-Wallis
test to compare the distributions of age and length of history between groups,
which were not normally distributed. To determine the independent associations
of different nutritional indices with mortality, we used multivariable logistic
regression in a model that included age, sex, visible severe wasting, WHZ,
and MUAC. To determine the clinical features that differed between children
identified by MUAC and WHZ, we used backward stepwise multivariable logistic
regression. We evaluated the performance of the resulting models using the
Hosmer-Lemeshow goodness-of-fit test. All analyses were performed using STATA
version 8.0, and P<.05 was used to determine statistical
There were 8500 children aged 12 to 59 months admitted during the study
period (1999-2002). Anthropometric data were incomplete in 3.6% (310) of children,
who were mostly admitted in extremis: weight was missing in 0.2% (21), height
in 2.8% (235), and MUAC in 1.4% (126). Among these 310 children, the median
age was 27 months (interquartile range, 20-39 months), and 27.4% (85) died.
These 310 were excluded, leaving 8190 children in the main analysis. The mean
anthropometric values were −1.54 for WHZ, −2.15 for WAZ, −1.59
for HAZ, and 13.7 cm for MUAC. Severe wasting (WHZ ≤−3) was present
in 9.2% (756) and kwashiorkor in 9.5% (778); 15.6% (1282) had 1 or both of
these features. Severe wasting in admitted children was more common among
boys than girls (10.1% [446/4426] vs 8.2% [310/3764], respectively; P<.001), while kwashiorkor was less common among boys
(8.4% [370/4422] vs 10.8% [408/3763], P<.001).
Overall, 4.4% (359) of children included in the study died while in the hospital.
The distribution of nutritional indices within admissions and numbers
of deaths at commonly used cutoffs are shown in Table 1. The areas under the ROC curves for MUAC, WHZ, and WAZ in
predicting inpatient death did not significantly differ: 0.75 (95% CI, 0.72-0.78)
for MUAC and 0.74 (95% CI, 0.71-0.77) for WHZ (P = .39)
(Figure). The area under the ROC curve
for WAZ was 0.76 (95% CI, 0.73-0.79) and for HAZ was 0.69 (95% CI, 0.66-0.72).
The sensitivity and specificity for inpatient death were 46.2% (166/359) and
91.0% (7122/7829), respectively, for MUAC less than or equal to 11.5 cm, and
41.8% (150/359) and 92.4% (7233/7829) for WHZ less than or equal to −3.
The case fatality rate among admitted children with WHZ less than or
equal to −3 was 19.9% (151/756). The case fatality rate among admitted
children with MUAC less than or equal to 11.5 cm was 19.0% (166/873) and did
not significantly vary with age (Table 2).
Visible severe wasting was present in 9.0% (608) of 6727 children assessed.
The median age of children with visible severe wasting was 24 months (interquartile
range, 18-35 months) and did not differ significantly from that in children
without this sign. Of the 608 children with visible severe wasting, 22.5%
(137) died, compared with 2.5% (153/6117) without this sign (sensitivity,
47%; specificity, 93%). The positive and negative likelihood ratios for death
for WHZ less than or equal to −3 and/or kwashiorkor were 4.36 (95% CI,
3.95-4.84) and 0.47 (95% CI, 0.41-0.53), respectively; those for MUAC less
than or equal to 11.5 cm and/or kwashiorkor were 5.12 (95% CI, 4.49-5.84)
and 0.59 (95% CI, 0.54-0.65); and those for visible severe wasting and/or
kwashiorkor were 5.31 (95% CI, 4.71-5.97) and 0.46 (95% CI, 0.40-0.53). A
multivariable logistic regression model adjusted for age and sex showed that
MUAC, visible severe wasting, and kwashiorkor were all independently associated
with inpatient death (Table 3).
For detecting the WHO standard criterion for severe wasting (WHZ ≤−3),
the sensitivity and specificity of MUAC less than or equal to 11.5 cm were
65.1% (486/746) and 94.8% (7057/7444), respectively, and of visible severe
wasting were 52.6% (320/608) and 95.3% (5831/6119) (P<.001).
Of 608 children with WHZ less than or equal to −3, 29.3% (178) did not
have visible severe wasting or MUAC less than or equal to 11.5 cm (Table 4).
Although the sensitivity and specificity for death of MUAC less than
or equal to 11.5 cm, visible severe wasting, and WHZ less than or equal to
−3 appeared roughly similar, the sets of children identified by these
nutritional indices differed, with only partial overlap (Table 4). Comparing children with MUAC less than or equal to 11.5
cm with children with WHZ less than or equal to −3 by univariate analysis,
children with MUAC less than or equal to 11.5 cm were more likely to be stunted,
female, and to have a longer history of illness, cough, diarrhea, subcostal
indrawing, visible severe wasting, kwashiorkor, moderate anemia, and bacteremia
(Table 5). They also were less likely
to have a history of seizures or to be unable to localize a painful stimulus.
Although the median ages appeared similar, the distribution of ages significantly
differed between the 2 groups (Table 5).
Multivariable analysis showed that skin/hair changes associated with recent
kwashiorkor, bipedal edema associated with current kwashiorkor, stunting,
subcostal indrawing, no history of seizures, female sex, and younger age were
independent associations of having MUAC less than or equal to 11.5 cm rather
than WHZ less than or equal to −3 (Table
In a large study of children admitted to a district hospital in sub-Saharan
Africa, MUAC performed as well as WHZ in predicting inpatient mortality. Since
MUAC is inexpensive, more commonly available, does not require a chart to
calculate, and is easier to measure than WHZ, it may be a useful screening
tool for such children. However, there were differences in the groups of children
identified by these methods, and they independently predicted inpatient death.
The observation that the sets of children identified by these methods of assessment
do not entirely overlap has been previously reported in 2 African studies,
both conducted outside the hospital setting and involving only children.16,17 In both of these studies, the differences
were attributed to age, with low MUAC values identifying younger children
but not older children with a low WHZ. We found that there were statistically
significant, independent associations of age and sex and identification by
MUAC alone compared with WHZ alone when adjusted for the effects of other
variables (Table 6). However, the case
fatality rate for MUAC less than or equal to 11.5 cm was consistently high
(19.0%) at all ages (Table 2), suggesting
that an unadjusted MUAC may be clinically useful in this setting.
There are 2 other potentially important reasons why MUAC, visible severe
wasting, and WHZ might identify different children. First, while all of these
measures reflect bone, fat, and muscle mass, WHZ is also influenced by total
body water. The WHZ may therefore be potentially lowered by acute dehydration.
While this is unlikely to be a significant problem in community-based studies,
children admitted to hospitals in sub-Saharan Africa are commonly dehydrated.
Dehydration does not only occur with gastroenteritis4,18 but
also may occur in other common conditions such as severe malaria.19,20 To evaluate the potential for severe
dehydration to influence WHZ, we examined the changes in z score with respect to NCHS standards for a hypothetical 1-year-old
child with severe dehydration (10% loss of body weight). If such a child’s
usual WHZ was zero, then 10% dehydration would have reduced the WHZ by approximately
1 z score. If the child’s usual WHZ was −3,
then the WHZ would be reduced by approximately 0.6 z scores.
However, examining the clinical features of the groups of children with MUAC
less than or equal to 11.5 cm only and WHZ less than or equal to −3
only, there were no significant differences in the prevalence of deep breathing
(indicating severe metabolic acidosis), sunken eyes, or malaria parasitemia.
Furthermore, diarrhea was more commonly present among those with MUAC less
than or equal to 11.5 cm only than in those with WHZ less than or equal to
−3 only. It is possible that some children with very severe hypovolemia
were excluded from the analysis because they were in extremis on admission.
However, we have previously reported that low weight for age did not predict
deaths within the first 4 hours after admission but was predictive of later
inpatient deaths.13 Thus, our findings suggest
that dehydration does not have a significant confounding effect in the hospital
Second, although small physical size due to stunting may contribute
to a low MUAC measurement, stunting is not reflected by a low WHZ value. The
prevalence of stunting in children with MUAC less than or equal to 11.5 cm
was greater than in those with WHZ less than or equal to −3 (Tables 5 and 6).
Although wasting has generally been the focus of acute inpatient management,
stunting has been reported to be associated with a poor outcome from diarrhea
and lower respiratory tract infection. However, the evidence is relatively
weak, since it is based on a very small number of studies (reported by Rice
et al21). Our data show that even mild stunting
(HAZ −1.99 to −1) is associated with inpatient death (age and
sex–adjusted odds ratio, 1.45 [95% CI, 1.02-2.07]) (Table 1) compared with no stunting (HAZ >−1), having excluded
those in extremis on arrival. Since stunting is not altered by acute illness,
it is the most likely to be an actual risk factor rather than simply an association.
In practice, measurement and calculation of HAZ shares some of the same disadvantages
as WHZ. However, MUAC measurement incorporates elements of both stunting and
wasting. A number of other clinical characteristics, including bipedal edema
associated with current kwashiorkor and skin/hair changes associated with
recent kwashiorkor, among children with MUAC less than or equal to 11.5 cm
suggest MUAC may be a better indicator of severe malnutrition than WHZ in
Because of the recognized difficulties of measuring weight for height,
the WHO Integrated Management of Childhood Illness program for primary-level
care makes use of the clinical sign of visible severe wasting.14 Our
results suggest that there is little difference in the sensitivity and specificity
of visible severe wasting and the single cutoff points of MUAC less than or
equal to 11.5 cm and WHZ less than or equal to −3. When combined with
the identification of kwashiorkor, the positive likelihood ratio for death
of visible severe wasting was actually greater than for WHZ less than or equal
to −3. This supports the use of visible severe wasting as a bedside
test for acutely ill children at the district hospital level, provided appropriate
training has been given. The observation that clinical assessment may be as
useful as anthropometry in hospitalized patients in developed country settings
has previously been reported.22 However, visible
severe wasting is a subjective, binary assessment. Where possible, an objective
measure with more than 1 cutoff point, such as MUAC, is preferable to allow
standardization between centers and classification of the degree of malnutrition.
Furthermore, as a predictor of the presence of WHZ less than or equal to −3,
visible severe wasting did not perform as well as MUAC less than or equal
to 11.5 cm, suggesting that MUAC may be better in identifying less–severely
ill children in need of nutritional rehabilitation.
The main limitations of this study are that it was performed at only
1 site, and data from areas of differing malaria transmission and HIV prevalence
would be valuable. There was a lack of individual data on HIV infection, which
may potentially cause wasting and influence fat distribution. In addition,
no laboratory assessment of nutritional status, such as measurement of serum
albumin levels, was routinely performed. Finally, no systematic follow-up
was made to identify deaths occurring after hospital discharge.
In summary, MUAC and visible severe wasting performed as well as the
WHO-recommended assessment method in predicting inpatient death. Our findings,
as well as considerations of cost and practicality, suggest that MUAC may
be more appropriate than WHZ for identifying severe malnutrition in children
aged between 1 and 5 years who are admitted to an African district hospital.
However, an assessment that includes MUAC, WHZ, and visible severe wasting
increases the number of at-risk children who are identified on admission and
highlights those in overlapping groups who are at the greatest risk of dying.
Further studies are needed to evaluate MUAC and visible severe wasting in
other operational settings and other situations in which anthropometric assessment
is difficult to perform.
Corresponding Author: James Berkley, MD,
Kenya Medical Institute Centre for Geographic Medicine Research (coast), PO
Box 230, Kilifi, Kenya (firstname.lastname@example.org).
Author Contributions: Dr Berkley had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Berkley, English,
Acquisition of data: Berkley, Mwangi, Griffiths,
Ahmed, Mithwani, English, Maitland.
Analysis and interpretation of data: Berkley,
Drafting of the manuscript: Berkley, Ahmed,
Critical revision of the manuscript for important
intellectual content: Berkley, Mwangi, Griffiths, Mithwani, English,
Statistical analysis: Berkley, Mithwani, Maitland.
Obtained funding: Berkley.
Administrative, technical, or material support:
Griffiths, Mithwani, English.
Study supervision: Mithwani, English, Newton,
Financial Disclosures: None reported.
Funding/Support: This study was supported by
the Wellcome Trust. Dr Berkley was a Wellcome Trust Training Fellow in Clinical
Tropical Medicine (053439), Dr Newton is a Wellcome Trust Senior Clinical
Fellow (050533), and Dr English is a Wellcome Trust Career Development Fellow
Role of the Sponsors: The Kenya Medical Research
Institute (KEMRI) and the Wellcome Trust had no role in the design and conduct
of the study; the collection, management, analysis, and interpretation of
the data; or the preparation, review, or approval of the manuscript. This
article is published with the permission of the director of KEMRI.
Acknowledgment: We thank the Medical Officer
of Health, the Medical Superintendent, the staff of Kilifi District Hospital,
and Norbert Peshu, MPH, Kenya Medical Research Institute Centre for Geographic
Medicine Research (coast), for their support. We are grateful to all the KEMRI
and nursing staff for their assistance. We also thank Steven Allen, PhD, Department
of Paediatrics, University of Wales Swansea, for his helpful advice.