Menu ENN Search
Language: English Français

Is it valid to use similar cut-off points for prevalence of wasting using z scores for different countries and contexts?

This question was posted the Assessment forum area and has 16 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Melaku Begashaw

Normal user

10 Oct 2011, 10:13

Hi there,
II faced this situation and want to to hear your ideas. We have a programme running for long, but the GAM rates during the hunger gap based on WHO standards 2006 is unacceptably high in all surveys: for example the Gam estimate is around 18 (95% C.I. 15.7 – 20.8). When we see the GAM using MUAC it is consistently below 10 percent for all the surveys (half of the z score results). There is no coverage survey to see the coverage of the programme, but some of the results are in direct contrast with the situation on ground. One of this surveys was conducted in an urban setting with a very good program (programme indicators meeting the sphere minimum standard except default rate). Based on the WHO classification this level of GAM are critical, but based on my experience from Ethiopia this levels are observed only on near famine situations like the current Horn of Africa El nino droughts or in pastoral communities like Somalia. Therefore, I am wondering are this z scores really telling me the situation on the ground? Is this valid to say the situation ‘critical’ when you have a well-run CMAM program? Moreover, the program admission and discharge is done by using both MUAC and z scores (WHO). The NCHS results are similarly 'unacceptably' high. I have been reading the article '“Fluctuations in wasting in vulnerable child populations in the Greater Horn of Africa.” S Chotard, J Mason, N Oliphant, S Mebrahtu and P Hailery. linkFood and Nutrition Bulletin, vol. 31, no. 3 (supplement), 2010.' The research paper tryies to come up with different cut off points according to the local context, rather than having a single cut off points for all countries. I really want a discussion on this and an answer for my questions from you.

Rogers Wanyama

Emergency Nutrition Specialist

Normal user

10 Oct 2011, 13:59

Hi Anonymous 883
This has been discussed in different forums (Including e-net) with regard to the use of differential reference levels for different population groups or areas e.g. pastoralists in the Horn of Africa .Have a look on the Review of Nutrition and Mortality Indicators for the Integrated Food Security Phase Classification (IPC) - Reference Levels and Decision-making .The use of different reference levels for different population group is highlighted in the report . You can access the publication on : http://www.ipcinfo.org/pubs.php
Hope this helps

Melaku Begashaw

Normal user

10 Oct 2011, 16:29

Thanks a lot I will see in to the document.

Mark Myatt

Consultant Epideomiologist

Frequent user

10 Oct 2011, 17:21

I would guess that the population you are writing about are living at low altitude in a hot climate country and are probably pastoralists or recent pastoralists. Such populations tend to have a body-shape that is characterised by long limbs. This body shape will lead to low W/H. This means that W/H will overestimate prevalence in such populations. See:

http://www.brixtonhealth.com/MyattBodyShape.pdf

for further information.

A large proportion of linear growth in children from birth to about seven years is in the legs. This means that stunting results in short legs. A body shape with short legs has a high W/H. In your population it is likely that the healthiest and best nourished children will not be stunted. This means that they will have long legs and a low W/H. It could be that, in this population, W/H selects the healthier and better nourished children. If that is true then SFP activity may have been focussed on the healthiest with the least healthy (short limbed / higher W/H) being systematically excluded.

Populations that live at high altitudes and / or cold climates tend to have short limbs. This means that they have high W/H. In these populations we see pseudo-outbreaks of overweight and obesity using W/H and W/H related measures such as BMI.

The relationship between body-shape (or limb length) and W/H may explain the confused international picture regarding the predicative value (i.e. of near term death) of W/H. In some settings it is highly predictive but in others it is very weakly (if at all) predictive. MUAC has a much more consistent prognostic value and this is consistent better than W/H.

A simple diagnostic test is to look at prevalence by age-group. If you see prevalence increasing with increasing age then you probably have a body-shape biassing WHZ problem (sitting to standing height ratio increases with increasing age). In these setting MUAC is the better indicator.

Melaku Begashaw

Normal user

10 Oct 2011, 19:16

Thanks Mark for your explanation. Of course the population live in a low land areas with a significant pastoralist and agro-pastoralist population. I will do the test and look in the trends.Thanks a lot!

Melaku Begashaw

Normal user

10 Oct 2011, 20:05

The documents and ideas presented to the question are useful. It is important to use as many qualitative and other contextual information as possible to decide on the situation and appropriate responses. GAM rates are wrongly understood and used by many programmes. The SMART recomends to use contextual information to understand nutrition surveys. Unfortunately, the attention is on the GAM estimate with out any other consideration. CMAM programmes are useful and save the dear lives of childeren. However, the current international economic crisis, soverign debts and the burden of austerity measures by western govenments on their population signal to us in the future getting funds may not be easy. Therfore, it will be very important to carefully allocate resoureces to their best use with a very cautious assessement and design. Savings in nutrition surveys and assessemnts are one considerable potential area. The field exchange on its exclusive coverage on Ethiopia asked whether it is important to do 600+ nutrition surveys in the country in the past decade. I have a chance to attend one nutrition cluster meeting that discuss the issue. Their are substantial evidences that GAM estimates can be reduced with carefull assessements. I strongly hope that we will help programme managers as well as host governments to come up with better methods to determine action thresholds to inform decision. For instance, in the current horn of Africa drought a lot of surveys have been conducted to direct the attention of all key stalkholders to the crisis. However, FEWS net has unmistakingly predicted the La nina situation as well as other aggravating factors. For the whole affected area a carefully planned few surveys were enough to show the develoment of the drought. However, many organizations has been running independently to conduct nutrition surveys and intiate a response. This is partly due to eagerness to help the affeceted population or misunderstanding the basic use of the survey results or misunderstanding the rules of the host government. Hope we will predict crisis situations without a lot of expenditure in the future.

Mark Myatt

Consultant Epideomiologist

Frequent user

11 Oct 2011, 08:29

I should have said ...

If you see prevalence by WHZ increasing with increasing age then you may have a problem with a body-shape associated downward bias in WHZ (and an upward bias in prevalence).

If you look at prevalence by age for MUAC you should see peak prevalence at typical weaning ages (lower prevalence in younger children and much lower prevalence in older children).

Melaku Begashaw

Normal user

13 Oct 2011, 01:31

Hey Mark,

These are the results of prevalence estimates by different age groups for the three latest surveys. The prevalence estimates for MUAC and z scores are presented. Based on the sample the prevalenece rates decrease for each age increases. The corrosponding MUAC results are presented too. The difference in prevalence rates between the two criterias is visible. Other results of the survey including mortality rate show a normal situation. The results are latest survey estimates for the last 4 months. I just want to show the results and the wide difference in GAM based on MUAC and z scores.

Survey 1

Prevalence of acute malnutrition by age based on weight-for-height z-scores and/or oedema (n=845)

                        Severe wasting  Moderate wasting  Normal
Age (mths)   Total no.	No.  %	        No.  %	          No.  %
6-17         167        11   6.6        41   24.6         115  68.9
18-29        172        10   5.8        29   16.9         133  77.3	
30-41        217        2    0.9        21   9.7          194  89.4	
42-53        194        5    2.6        17   8.8          172  88.7	
54-59        95         1    1.1        22   23.2         72   75.8	
Total        845        29   3.4        130  15.4         686  81.2	

Prevalence by MUAC

MUAC analysis N=832 (children 6 month to 59 months)

Status                                                 Number of children   Percentage
Severe Acute malnutrition (MUAC <115 mm)	       8                    1.0%
Moderate Acute malnutrition (MUAC >=115 to <125 mm)    47                   5.6%
Risk (MUAC >=125 to <135 cm)                           176                  21.2%
Normal, well nourished (MUAC >135 mm)                  601                  72.2%
Total                                                  832                  100%

Survey 2
Prevalence by z scores aggregated by age

				Severe wasting  Moderate wasting   Normal
Age (mths)	Total no.	No.	%	No.	 %	   No.	 %
6-17		149		1	0.7	27	 18.1	   121	 81.2
18-29		199		2	1.0	34	 17.1	   163	 81.9
30-41		160		1	0.6	19	 11.9	   140	 87.5
42-53		145		1	0.7	18	 12.4	   126	 86.9
54-59		66		0	0.0	10	 15.2	   56	 84.8
Total		719		5	0.7	108	 15.0	   606	 84

Prevalence by MUAC
MUAC analysis N=724 (children 6 month to 59 months)

Nutritional Status 	                             Number of children	 Percentage
Severe Acute malnutrition (MUAC <115 mm)	     8	                 1.1%
Moderate Acute malnutrition (MUAC >=115 to<125 mm)   51	                 7%
Risk (MUAC >=125 to <135 cm)	                     130	         18%
Normal, well nourished (MUAC >135 mm)	             535	         73.9%
Total                                                724	         100%

Survey 3


Severe wasting Moderate wasting Normal
Age (mths) Total no. No. % No. % No. %
6-17 189 7 3.7 21 11.1 161 85.2
18-29 191 4 2.1 25 13.1 162 84.8
30-41 232 4 1.7 22 9.5 206 88.8
42-53 196 1 0.5 32 16.3 163 83.2
54-59 112 1 0.9 24 21.4 87 77.7
Total 920 17 1.8 124 13.5 779 84.7

prevalence by MUAC
MUAC analysis N=920 (children 6 month to 59 months)

Nutritional Status                                   Number of children	 Percentage
Severe Acute malnutrition (MUAC <115 mm)	     5	                 0.6%
Moderate Acute malnutrition (MUAC >=115 to<125 mm)   50	                 5.6%
Risk (MUAC >=125 to <135 cm)	                     183	         20.3%
Normal, well nourished (MUAC >135 mm)	             661	         73.5%
Total                                                899	         100%

Mark Myatt

Consultant Epideomiologist

Frequent user

13 Oct 2011, 11:23

The differences are very large. The tables got a bit mangled by the forum software. I think we have:

    Survey      WHZ     MUAC
    ------    -----     ----
         1    18.8%     5.6%
         2    16.0%     8.1%
         3    15.3%     6.3%
    ------    -----     ----

is that right?

These are similar to the differences seen in (e.g.) Somali pastoralists and agro-pastoralists.

For all surveys we see prevalence remaining high in older children and increasing with increasing age in children >= 30 months. This is not what we expect or observe in populations with higher sitting to standing height ratios.

It would be useful to be able to see age-specific prevalence by both indicators. Here (e.g.) are some data from Somalia:

    Year-centred
       age-group       WHZ    MUAC
    ------------    ------  ------
               1     5.39%   9.58%
               2    14.44%  11.67%
               3    15.42%   4.21%
               4    12.92%   2.81%
               5    17.19%   1.04%
    ------------    ------  ------

The expected distribution of the prevalence of wasting by age is for prevalence to be highest in the first and second years with increasingly lower prevalences observed in older children. In the Somali data (above), this pattern is observed in the prevalence estimates using the MUAC case-definition but is not observed in the prevalence estimates using the WHZ case-definition.

Mark Myatt

Consultant Epideomiologist

Frequent user

13 Oct 2011, 14:17

Just a note to say that EN-NET have addressed the issue of having text tables like this:

            col1 col2 col3
    row1      1     2    3
    row2      4     5    9

Just put the data between "{code}" and "{/code}" tags suing square not curly brackets. See:

    http://en.wikipedia.org/wiki/BBCode

for details of how to format messages here.

Melaku Begashaw

Normal user

13 Oct 2011, 17:00

Thank you very much and it was a learning expierience.

Kathleen Lamaute

Professor- Division of Nursing , Molly College

Normal user

13 Oct 2011, 17:04

I am trying to purchase MUAC tapes but I am unable to get them. Can anyone help?

Sonya LeJeune

Normal user

13 Oct 2011, 17:26

Have you tried to contact TALC for MUAC tapes?
I think that their contact details are given elsewhere on this forum.

Melaku Begashaw

Normal user

15 Oct 2011, 07:29

Thank's Mark and Roger. I have read Mark's suggested paper and it is really true. Most surveys in Ethiopia in Highland and agrarian communities give a more or less close estimate of prevalence rates for GAM and SAM. This was one of the problems I had with the results of the survey. The paper 'The effect of body shape on weight-for-height and mid-upper arm circumference based case definitions of acute malnutrition in Ethiopian children' has helped me understand the issue. As a matter of fact, recently there was an intiative to do similar studies and I will inform the participating agencies to read this document before going to do the surveys.

Similarly, the Integrated food security classification technical paper has adressed the concerns I have. Thank you very much for your assisstance.

Melaku Begashaw

Normal user

19 Oct 2011, 14:59

I find this useful for the questions I had:

''Weight-for-height and MUAC for estimating the prevalence of acute undernutrition?
A review of survey data collected between September 1992 and October 2006''
Mark Myatt
Arabella Duffield
University College London Save the Children UK

Melaku Begashaw

Normal user

26 Oct 2011, 06:07

''The cut-off point of >+2 SD classifies high weight-for-height as overweight in children.The use of
-2 Z-scores as a cut-off implies that 2.3% of the reference population will be classified as malnourished even if they are truly "healthy" individuals with no growth impairment. Hence, 2.3% can be regarded as the baseline or expected prevalence. To be precise the reported values in the surveys would need to subtract this baseline value in order to calculate the prevalence above normal. It is important to note, however, that the 2.3% figure is customarily not subtracted from the observed value. In reporting underweight and stunting rates this is not a serious problem because prevalences in deprived populations are usually much higher than 2.3%. However, for wasting, with much lower prevalence levels, not subtracting this baseline level undoubtedly affects the interpretation of findings.''
http://www.who.int/nutgrowthdb/about/introduction/en/index5.html

Mark Myatt

Consultant Epideomiologist

Frequent user

26 Oct 2011, 08:24

As with the left-hand tail of the distribution (i.e. < -2) there is a problem with WHZ > +2 as a case-definition for "overweight". It is the body-shape issue again. In cold climates or high altitudes body-shape tends to short limbs and large "barrel" chests. This means high W/H. Using W/H will make us see "obesity epidemics" in (e.g.) Latino populations when no such epidemic exists.

Back to top

» Post a reply