Finding the right MUAC cut-off to improve screening efficiency
Author Koert Ritmeijer, MSF Holland
In
Hlaing Thayer township, Yangon, Burma, ORWs were spending a considerable
amount of time doing weight and height measurements on all children visited,
in order to refer less than 1% to the therapeutic feeding programme. MSF
undertook a sensitivity analysis to determine the most appropriate MUAC
indicator to use in a first stage screening. This was done to reduce the
time spent on the screening excercise without reducing sensitivity.
Introduction
In 1988, the government of Myanmar
(Burma) started a resettlement scheme for urban slum populations,
from the capital Yangon city to new areas around the city.
Hlaingthayar is one of these newly formed satellite townships with a population
of 157,000 inhabitants. A majority of those living in Hliangthayar are
very poor and rely on daily wages as day labourers, factory workers, food
vendors and other informal sector jobs in Yangon. Resettlement to the township
has increased their vulnerability. Essential infrastructure in these townships
like water supplies and sanitation works are poorly developed, and during
the monsoon rains large parts of the township are flooded. Basic health
care in Hlaingthayar is poorly developed and is characterised by chronic
shortages of staff, drugs and medical materials. The main health problems
include high maternal mortality, high childhood morbidity and mortality
and high incidence of STD/HIV and epidemic diseases (cholera and Dengue).
Nutrition surveys carried out by MSF in 1992, 1993 and 1996 in the township
revealed consistently high levels of both acute and chronic malnutrition
amongst children under five years of age.
MSF began implementing an urban health
care programme in Hliangthayar in 1993. The programme focuses on women
and young children and includes the running of independent MCH clinics,
feeding centres, community outreach programmes, AIDS/HIV awareness
programmes, water and sanitation initiatives, and supplying essential drugs
to public (government) health facilities.
Problem
At the start of MSF's programme, Outreach
workers (ORWs) conducted screening during their home
visits, referring children with a Mid Upper Arm Circumference (MUAC) less
than 135mm to the feeding centre for a weight for height assessment. At
the feeding centre children with a 'weight for height' < 75% of the
median in the reference population1 were admitted. The admission criterium
of 75% W/H was chosen for pragmatic programme capacity as well as 'economy
of scale' reasons. The feeding programme did not have the capacity to admit
both severely (<70% W/H) as well as moderately malnourished children
(70-79% W/H). However, on
the other hand, only admitting severely
malnourished would lead to feeding centres with only very limited case
loads (20-30 children per feeding centre). For efficiency reasons children
between 70 and 74% W/H were also admitted as a risk group "to fill up"
the capacity of the feeding centres.
Although this MUAC screening had a sensitivity2 of some 85% the positive predictive value3 was very poor. Only 5% of children
with a MUAC <135 had a weight-for-height <75% of the median.
Therefore most mothers and children referred to the feeding centre for
weight for height check ups were sent home again. This wasted a lot of
time for mothers, undermined credibility of the programme and led to loss
of motivation amongst the ORWs. At the time it was also felt that
while using a stricter MUAC screening cut-off would improve the positive
predictive value, it would also significantly reduce sensitivity, leading
to a situation where a high proportion of truly malnourished children would
be missed. As a result, MSF decided to let ORWs conduct weight-for-height
measurements of all children under five during their home visits in Hlaingthayar
so that they could refer malnourished children to therapeutic feeding programmes.
Currently all children under 75%
weight for height are admitted to the feeding centres. Roughly 3,900
children are screened every month, of whom 20-30 are referred to the feeding
centre (0.5-0.8%). As it takes on average 7.5 minutes per child to do a
weight for height measurement, the total time spent by ORWs each month
on nutritional screening to detect 20-30 cases amounts to almost 500 hours.
In order to increase the efficiency of the screening, MSF has recently
been considering reverting to MUAC measurements as a first screening tool,
so that more time could be made available for the ORWs to undertake other
priority tasks, e.g. health education.
Objective and Method
The task faced by MSF was to select
the most appropriate MUAC cut-off point for initial (first stage) screening
of children under five, so as to reduce the number of unnecessary weight
and height measurements. This required analysis and calculation of
the sensitivity, specificity and positive predictive value of different
MUAC cut-off points, in identifying children less than 75% weight for height
for admission to therapeutic feeding. However, as comprehensive
MUAC data for the Hlang Thayar population were not available, nutrition
survey databases from other populations with a similar global prevalence
of malnutrition were used. These data were from Rohinga refugees in Bangladesh
between 1992-3 and the urban population n Malange, Angola in 1994.
| Table 1.Comparison of prevalence using
weight for height cut-offs from nutrition survey databases. Survey Database |
| Survey Database |
Hlaing
(1996)
|
Bamgladesh
(1992/93)
|
Angola
(1994)
|
| sample size |
938
|
1715
|
1147
|
| (6-59 months) |
|
|
|
global malnutrition rate
(W/H <-2 Zsc) |
12.8% |
12.2% |
11.9% |
severe malnutrition rate
(W/H <-3 Zsc) |
1.5% |
0.4% |
1.5% |
'admission Criteria'
(W/H <75% Med.) |
3.4% |
1% |
2.2% |
| MUAC < 135 mm |
|
28.7% |
41.9% |
| MUAC < 130 mm |
|
15.9% |
27.6% |
| Table 2. Sensitivity specificity and
positive predictive value analysis |
|
Bangladesh
(Rohingas)
|
Angola
|
Admission Criteria
(W/H <75% Med.) |
MUAC<135mm |
MUAC<130mm |
MUAC<135mm |
MUAC<130mm |
| Sensitivity |
88.2% |
82.4% |
92.0% |
84.0% |
| Specificity |
71.8% |
84.7% |
59.2% |
73.7% |
| Positive predictive value |
3.0% |
5.1% |
4.8% |
6.6% |
Analysis of the data
In multiple database analyses there
was a limited correlation between MUAC and weight-for-height (r=0.6)
and a generally poor positive predictive value of MUAC cut-off points of
135 and 130 mm in identifying children < 75% weight for height (see
table2). This confirmed that MUAC on its own was an inappropriate screening
tool and would need to be used in combination with weight for height measurements.
However, the data also showed that MUAC cut-offs of 130 and 135 mm may
be considered sufficiently sensitive to identify children under 75% weight
for height, but that the specificity of a MUAC cut-off of <130
mm is significantly higher than a MUAC of <135 mm. Although the
positive predictive value of a MUAC cut-off of 130 and 135 mm for a weight
for height of <75% remains poor, considerable efficiency gains can be
made by using MUAC as a first screening stage. The total time saved on
the whole screening procedure can amount to 65% using a 130 mm cut off
and 52% using a 135 mm cut off. The number of false negatives by MUAC screening
(children with a weight for height <75% but not selected by MUAC screening)
remains acceptable. Of the 20-30 children currently referred each month
using weight for height screening, an estimated 3-5 children (17%) would
be missed each month when using a MUAC of < 130 mm. Further analysis
shows that these false negatives do not include severely malnourished children
(weight for height < 70%) but rather "borderline" cases with a weight
for height between 72-75%.
|
Table 3 Cost-efficiency comparison of
different nutritional screening strategies (estimations, based on extrapolation
of average approximations |
Screening strategy
(3900 children per month) |
Total monthly time sent on screening
(hrs.)
|
Time saved |
Children missed (false negatives by MUAC) |
|
MUAC
(1min/chid) |
MUAC
(7.5min/child) |
Total
(MUAC+ W/H) |
|
|
| 1. W/H screening of all children (=3900
children) |
0hrs |
488 hrs |
488hrs |
0 |
0 |
| 2. MUAC of all children + W/H screening
of children of children with MUAC <135mm (3900x0.35=1365 children) |
65hrs |
171hrs |
236hrs |
52% |
10% |
| 3. MUAC of all children + W/H screening
of children with MUAC <130mm (3900x0.22=860 children) |
65hrs |
107hrs |
172hrs |
65% |
17% |
Recommendations
Our final recommendation has been to revert
to a two-stage nutritional screening strategy involving;
1. MUAC screening of all children
under five.
2. Weight for height screening of
all children with a MUAC <130 mm.
1For explanation of terms
see 'Z-scores' article in Field Exchange Issue I
2The term sensitivity means
the ability of the indicator (in this case a MUAC of < than
135 mm) to identify correctly children less than 75% weight for height.
In other words a sensitivity of 85% means that 85% of children with a weight
for height less than 75% would also have a MUAC of less than 135 mm.
3The positive predictive
value of the MUAC screening test is the true positives as a percentage
of all those tested positive by the screening method., i.e. those with
weight-for-height < 75% as a percentage of all children tested
with a MUAC < 135mm
Taken from Field Exchange Issue 4, June 1998
http://fex.ennonline.net/4/finding.aspx