Menu ENN Search
Language: English Français

clear definition of Small Scale SMART Survey

This question was posted the Assessment forum area and has 17 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 81

Public Health Nutritionist

Normal user

19 Aug 2014, 08:25

Now days, the term "Small scale SMART survey, SSSS" is becoming common. In principle, is it different from that of normal SMART survey procedure? To make it clear, does SSSS follow the same principle to that of normal SMART survey to calculate sample size ( estimated prevalence, precision, design effect, etc)? if not different why we say Small scale SMART survey? Or is it just to indicate geographic (small area) or volume of data/information?

Anonymous 730

Nutrition and Food Security Officer

Normal user

19 Aug 2014, 08:55

I will refer you to to a discussion which was on this forum earlier: http://www.en-net.org/question/1201.aspx

I will quote from the article:

'A multistage cluster sampling approach is used as for most anthropometric surveys and follow the SMART methodology approach. It has a 25 clusters by 12 households design. This sampling size though "small" yet ensures that variations of acute malnutrition of a minimum of 4% would be detected between two rounds of surveys using the CDC “2 surveys” calculator. Also, the CDC "probability calculator" can be used to present results / give a GAM threshold with 85% probability to be exceeded (relevant for recommendations purposes).
Those small scale surveys can be run 2 or 3 times a year (during key seasonal events) and apart from nutrition indicators, a set of key indicators about health, wash, food security, child care and feeding practices can be also collected for the purpose of early warning.'

Just a note to say that I have also seen 25 by 10 also being used rather than 25 by 12.
I would therefore say that the SSS is derived from the SMART survey concept, but is more useful for surveillance rather than for obtaining a valid estimate of the prevalence of malnutrition due to its lower precision. Let me know if any additional explanation is required.

Anonymous 81

Public Health Nutritionist

Normal user

19 Aug 2014, 09:21

Dear Blessing,
Thanks for your reply. Regarding 25 by 12. Is this apply only for Anthropometry? How about if Mortality is included? the other point is, how was the assumption done with regard to number of children 6-59 months per household? in some setting, the number under five per HH is much lower than 1 where as in other setting they are above 1 up to 1.5. so, don't you think this will affect the overall sample PARTICULARY if mortality is part of the assessment?

Anonymous 730

Nutrition and Food Security Officer

Normal user

19 Aug 2014, 09:30

I would probably say this was based on an assumption of 1 child/household, and so you would then adjust accordingly if you have less than 1 child/household.In SSS, I don't think you would do mortality, as the emphasis is on the anthropometry.Hope this helps.

Mark Myatt

Consultant Epideomiologist

Frequent user

19 Aug 2014, 09:48

There is very little information about the SSSS method. The term appears in recent ACF job adverts. As far as I can tell ...

Sample Design : The SSSS is based on the SMART methodology: PSUs are selected PPS, m = 25 PSUs are taken, n = 12 subjects are sampled from each PSU using a random sampling procedure, data-analysis can be done with ENA. I have seen a couple or survey reports and the "small scale" seems to refer to "small sample" rather than "small area".

Sample size : The sample size seems to not be calculated to estimate prevalence with a specified precision but rather to detect changes in prevalence between two survey rounds for surveillance without a time-series. One report that I have seen claims that the 25-by-12 sample will detect a change of >= 4% in the prevalence of GAM with 85% power (Blessing may have the same report). No details of the sample size calculation were shown but n = 300 seems to me to a bit small to detect a 4% difference between survey rounds (with p1=10%, p2=14%, p = 0.05, and n = 300 with DEFF = 2 ... I calculate power to be about 30% but I may be missing something)

Mortality : I don't think it is intended for this but a range of indicators such as WAHS, FIES, HHS, &c. could be collected. Precision (95% CI) would be about +/- 8% on a 50% prevalence (a little better than an EPI survey so good enough for most purposes) but this will vary with the design effect.

My initial impressions is that it is a safe but not particularly innovative method. It is unlikely to result in much reduced survey costs. In my experience survey costs are determined mostly by the number of PSUs. I'd guess the costs savings of SSSS over a standard SMART with 30 clusters would be about 17%. This makes it (IMO) unsuited to surveillance outside of well-funded emergences.

I'd like to know more.

Mark Myatt

Consultant Epideomiologist

Frequent user

19 Aug 2014, 10:36

I think I may have got that wrong. The 85% power seems to be about making an LQAS-like classification of prevalence as being above a given threshold. Not (IMO) so useful.

I think estimation is possible. With n = 300 and DEFF = 1.5 an estimate of 10% would have a precision a little better than +/- 4.5% (something like 6.3% - 15.0%) on a classical estimate. This might drop to +/- 3% or better if A PROBIT estimator were used.

Victoria Sauveplane

Senior Program Manager, Action Against Hunger CA

Normal user

19 Aug 2014, 13:17

A small-scale SMART was derived from the SMART methodology for contexts of emergencies where a rapid estimation of GAM based on WHZ and MUAC for children aged 6-59 months is needed for emergency programming or when time for collecting the information is limited (high insecurity, limited accessibility of the survey area). Small-scale SMART (SS SMART) surveys primarily focus on GAM and no other indicator. If information on IYCF, mortality, WASH, etc. are required, then it is best to carry out a full SMART survey. A SS SMART aims for fast analysis and dissemination of survey results to inform emergency response and cannot be used for long-term programming.

Data collection can last from a few days to one week depending on the accessibility of the zone, size of the team and the level of information available. A SS SMART should be conducted only when the geographic area can be clearly delimited (group of villages, camps, settlements) and when the target population is homogenous (shares the same living conditions, agro-ecological zone and with similar socio-cultural patterns). The results of a SS SMART cannot be extrapolated to neighbouring areas that were not included in the sampling frame.

The number of clusters in a SS SMART is fixed at 25, with generally 5 survey teams composed of 2 surveyors. The minimum clusters per day and per team from should 2 so that data collection is completed within 3 days. If the circumstances do not allow minimum 2 clusters per team and per day, then it is best to go with a full SMART survey.

Regarding sample size, if you carrying out the survey in one settlement/village/camp, then a sample size of 150 children using simple or systematic random sampling would be sufficient. If there are more than one settlement and if the population is dispersed, then cluster sampling should be used. If the % of children under 5 years in the target population is below 15%, then 25 Clusters of 12 househols have to be selected. If the % of children under 5 years is above 15%, then 25 Clusters of 10 households have to selected. Clusters will be selected based on the Probability Proportional for Size (PPS) method and the households will be selected at random using either simple or systematic random sampling.

Furthermore, the overall validity of a small-scale SMART is assessed based on the representativeness, accuracy and precision of the results with the SMART Plausibility Check, similarly to a full SMART survey.

Being already been field tested in Afghanistan, Madagascar and in India, numerous SS SMART surveys are currently being done in South Sudan given the emergency situation and the urgency of nutrition data and once their lessons learned have been incorporated, the SS SMART methodology will be readily available for download off of the SMART website in the coming months.

Anonymous 81

Public Health Nutritionist

Normal user

19 Aug 2014, 13:39

Dear Victoria,
Thanks for your reply. currently, one of the key challenge in South Sudan is the absence of mortality data. Given the humanitarian situation, there is pressure from all corners about the Mortality information. I have seen about three Small scale surveys (sample size 200, 410 and 425 HHs) conducted very recently in South Sudan and none of them have mortality data. how should we address the mortality information if everyone is tending to us SSS? there has to be clear message on this regard whether it is only surveillance purpose or for everything.

Anonymous 730

Nutrition and Food Security Officer

Normal user

19 Aug 2014, 13:40

Thanks Victoria,

Very interesting. I just have a question on the utility of the method to be able to detect changes in the level of malnutrition.Are we able to analyze changes between one SS SMART and the other in terms of statistical difference just in the same way as we use the CDC calculator for the full SMART survey?

Victoria Sauveplane

Senior Program Manager, Action Against Hunger CA

Normal user

19 Aug 2014, 14:00

Regarding the mortality issue in South Sudan, indeed this is concern given the emergency situation and although there are SS SMART surveys being done in certain states, full SMART surveys which include mortality are also being done when access and security permits. Overall, a full SMART survey is still the method of choice and to my knowledge, these recommendations have been put forward for upcoming assessments in South Sudan.

Regarding your question Blessing, once again a SS SMART is to inform emergency programming and if you aspire to detect changes in the levels of malnutrition, then a full SMART survey should be your method of choice as you will have a much larger sample size and more precision in your results than in a SS SMART.

Thanks again,
Victoria

Anonymous 730

Nutrition and Food Security Officer

Normal user

19 Aug 2014, 14:22

Many thanks Victoria

Anonymous 81

Public Health Nutritionist

Normal user

20 Aug 2014, 08:49

Dear Victoria,
Looking at your responses, SS SMART is a sort of rapid assessment tool needed when urgent estimation of acute malnutrition is required. Moreover, as per your reply, SS SMART result is not comparable with the standard SMART result. If this is so, do you think it is good idea to name this tool as "SS SMART"? To avoid any confusion, I think it is good if ACF Canada reconsider naming SS SMART before releasing through SMART websites. it is better to work with broader consultation with key experts.

Mark Myatt

Consultant Epideomiologist

Frequent user

20 Aug 2014, 09:40

I don't really see the problem using this method to detect changes. When you take a continuously distributed variable such as WHZ and reduce this to a Yes/No case definition you lose information. Losing information in this way is the same as squandering sample size. If you leave the data as WHZ then you can use a more powerful test (e.g. a t-test or a non-parametric equivalent) to detect a difference in mean (or median) WHZ. You'd still have to take into account the sample design when analysing the data but this can be done in most statistical software and could be easily incorporated into the SMART software. Moving from means and SDs to prevalence is straightforward using a PROBIT estimator (this alone would increase precision of a single survey).

There are a number of other approaches that could be used to make better use of the sample that SSSS seems not have implemented. The simplest would be to use a Bayesian approach to the problem to bring prior knowledge about how the variable tends to behave in terms of mean and SD to the problem (i.e. in a normal-normal conjugate analysis). This can increase precision for a give survey (likelihood) sample size. A similar approach is used in SQUEAC (i.e. a beta-binomial conjugate analysis).

I hope this helps.

Victoria Sauveplane

Senior Program Manager, Action Against Hunger CA

Normal user

20 Aug 2014, 17:49

Indeed the SS SMART methodology is still under the works and not yet finalized, including its nomenclature. Nevertheless, the SS SMART does follow the same principles of households selection and data quality checks as a SMART survey, with some slight differences in the sampling methodology and therefore the reason of naming it a small-scale SMART.

For the comparison of results, the SS SMART is an alternative methodology for nutrition surveys when results are urgently needed for emergency programming or in situations with very limited accessibility due to insecurity. You can use its results for programming purposes but as it primarily focuses on GAM and not any other indicator, it is most often not the method of choice for longer-term programming or to complement any surveillance system versus carrying out a full SMART.

Furthermore, perhaps I was unclear in my previous answer to Blessing regarding the detection of statistical differences. As numerous SS SMARTs have not yet been undertaken in a periodic manner in the same area, it is not clear how well these statistical differences can be measured; and hence a full SMART would be a better method of choice if that was the objective.

AS

Nutrition manager

Normal user

22 Jan 2016, 05:36

Hi & Thanks for useful information. I need a bit more clearance on small scale smart survey (SSSS). In which situation small scale smart survey is feasible to be conduct??

Regards

Kennedy Musumba

SMART Program Manager

Normal user

22 Jan 2016, 14:11

The small scale SMART surveys presently referred to as Rapid SMART Surveys are ONLY applicable in the following contexts;

1. Where there is need for rapid estimate of the nutritional status in a representative and accurate manner in emergency contexts

2. Where there is limited time for collecting the information in terms of accessibility due to high insecurity

Access further information from

http://smartmethodology.org/survey-planning-tools/smart-methodology/

Sameh Al-Awlaqi

Public Health and Nutrition Consultant

Normal user

22 Jan 2016, 14:28

Thanks Kennedy,

I was wondering if SSSS and Rapid SMART are synonyms-Because I am quite familiar with the term rapid SMART methodology. I am sharing a dropbox link for a recent rapid SMART presentation (in power point)- shared by GNC/ACF Canada while I was in Yemen in 2015 to give you a brief overview about the methodology. The tools on SMART website are detailed and could help further reading.

Here is the link for the presentation:

https://www.dropbox.com/s/py3w7zina6uywsl/Rapid%20SMART%20_Yemen%20Webinar.pptx?dl=0

Please let me know if the link is not working.

Regards,

Sameh

AS

Nutrition manager

Normal user

23 Jan 2016, 17:57

Thanks...Kennedy Musumba & Sameh for valuable inputs on my question..Its clear my confusion, because I was consider SSSS as on Small geographical units. And was familiar with Rapid Smart. Thanks..

Back to top

» Post a reply