Menu ENN Search
Language: English Français

Quick and (not) dirty

This question was posted the Assessment forum area and has 5 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 432


Normal user

24 Sep 2010, 05:14

What would be the most adequate and practical rapid nutrition assessment tool to be carried out in a huge emergency e.g. Pakistan floods if circumstances and time do not allow to go for example for SMART?

Mark Myatt

Consultant Epideomiologist

Frequent user

24 Sep 2010, 23:14

This is something of a strange question since SMART is quick and dirty.

The problem with SMART is that it is very ill suited to wide-area assessments. It may be OK for small areas such as a single district with very little spatial variation in prevalence but the urban bias in the PPS sampling scheme and the homogeneity assumption render it useless for wide-areas with spatial heterogeneity.

There are quick and (not overly) dirty alternatives. You'll want to look at:

This might meet your requirement.

You'd probably want to MUAC rather than MUAC/H. You could use W/H but that would slow you down and add cost.

Note that mapping of the spatial distribution of malnutrition in similar detail has not been achieved using weight-for-height and two-stage cluster sampled (i.e. SMART surveys). This work makes SMART looks pretty "dumb".

For a wider area, I'd be tempted to go for a slightly more sophisticated sample and analsyis design. Something like ... A large (i.e. c. 60 - 90) PSU survey covering an area maybe 10 to 20 times that of a standard SMART survey (since SMART needs about 30 clusters per district the saving will be between 70% and 80%). The sample would start as an even spatial sample but would be less regular in practice since it would need to adhere to the underlying pattern of human settlement. Within-PSU sampling would be a modified EPI method such as EPI5 or QTR & EPI5 which yield a more representative sample than the simple EPI method used by SMART and add little extra work or cost. Analysis would use a triangulated irregular network (triangular estimation areas). The sample would aim for a mean triangle area of less than about 150 sq. km. (no child more than about 11 km or 12 km from a sampling location) with local prevalence estimates based on data from three PSUs. Wider area estimates could be produced by using population weighting. The aim would be for something simpler than the TCG Periodic Reviews for Myanmar. The method currently being developed by UNICEF / CONCERN / FMOH / EHNRI in Ethiopia for CMAM and GMP program coverage would, after some minor modifications, be suitable.

I hope this is of some use.

Anonymous 162

Normal user

29 Sep 2010, 17:04

Not sure if this is suitable on your context, but you may want to take a look at the Alternative Sampling Designs Guide produced by FANTA.

The guide outlines 3 different sampling designs, all of which are appropriate for emergency settings, where the time spent collecting data should be limited but must be sufficient to obtain the necessary information about the population. The three designs were developed to provide reliable methods for rapid assessment of the prevalence of acute malnutrition and useful measures of secondary indicators relevant to needs assessment and response planning, including child- and household-level indicators such as morbidity prevalence, vaccination coverage, household food security, and access to water and sanitation. The sampling designs (33x6, 67X3, sequential design) described in the guide and are each hybrid designs combing aspects of cluster sampling and Lot Quality Assurance Sampling (LQAS).

Mark Myatt

Consultant Epideomiologist

Frequent user

30 Sep 2010, 17:24

This is an interesting approach.

I have doubts regarding cost savings of this approach because we often spend a great deal of our time travelling to and from sampling locations and, once we arrive at a sampling location, meeting with local leaders for permissions and to recruit local guides. Whether we sample 7 children or 30 children with a modified EPI sampling scheme will often make little difference to overall costs. I do know that some trials of this method have shown considerable savings over 30-by30 and SMART surveys so I may being a little pessimistic here. The sequential sampling method will, in many situations, result in considerable savings ... if prevalence is very low or very high then sampling will stop early.

My main concern regards the use of cluster sampling. PPS is just a weighting scheme in which the weighting is done in advance of collecting data. This usually means that the sample is biased to communities with larger populations (urban bias). Locating the bulk of data-collection in the most populous communities may leave areas of low population density unsampled (i.e. those areas consisting of communities likely to be distant from health facilities, feeding centres, and distribution points). This may cause surveys to (e.g.) underestimate prevalence or evaluate coverage as being adequate even when coverage is poor or non-existent in areas outside of urban centres. The PPS process should result in a self-weighted sample but it cannot be relied upon to do so if estimates of cluster population sizes are inaccurate. Population estimates are usually derived from census data. In complex emergencies certain factors may lead to census data not being accurate (e.g. political manipulation, the absence of a functioning civil society, population displacement, and poor security). Population estimates are often corrected by the application of estimates of population growth which can seldom account for displacement, migration, or high mortality in the target population. In a situation, such as the Pakistan floods, where displacement is common, the PPS may direct you to empty communities. For these reasons. amongst others, I prefer a spatially stratified sample.

Mark Myatt

Consultant Epideomiologist

Frequent user

1 Oct 2010, 14:38

Something has been nagging at the back of my mind regarding the sequential sampling / early stopping variant of the method described in:

when used over wide areas.

It seems to me that there is a considerably increased risk of misclassification if there is heterogeneity (clustering) in the phenomena being surveyed. The obvious thing to do to save costs in any survey is to sample clusters conveniently so that (e.g.) you sample neighbouring PSUs on the same day. If you do this and (e.g.) start in a low prevalence area then you will likely get a low prevalence classification but the area you sampled before stoping may just have been a pocket of low prevalence in a wider area of high prevalence (or vice-versa).

If you adopt this approach I think you'll need to sample the PSUs in a random order. This will lose you some savings but there should still be some overall savings.

I hope this helps.

Tamsin Walters

en-net moderator

Forum moderator

3 Oct 2010, 15:22

Dear all,

Just a quick update from Pakistan; the Nutrition Cluster, Departments of Health in each affected Province, UNICEF with ACF-CA and CDC will be conducting SMART surveys in the flood-affected areas: Flood-Affected Nutrition Surveys (FANS); there will be a total of five SMART surveys conducted.

If you would like to be involved, please contact the Nutrition Cluster Coordinator for Pakistan, James King'ori at or Fawzia El Sharief, Nutrition Surveillance Specialist with UNICEF Pakistan at

Thank you in advance for your support,


Erin Boyd
Emergency Nutrition Specialist

Back to top

» Post a reply