Lilia, I am not quite sure what you are asking. What would be the focus / purpose / outcome of the investigation? What question(s) do you want answered. Sorry not to be of more help. Mark

Lilia contacted me directly as s/he has experienced some problems posting on EN-NET: The ENN forum email does not seem to accept my email; thus, am writing on your internet-published email, to clarify on my earlier inquiry. and writes: The purpose of the investigation would be to document the event (cases of malnutrition in few village) and (possibly) explore potential causes. This being done a in manner "quicker" than a full fledged SMART survey. Carrying a full fledged survey is time and resource consuming, including, preparations. This in settings like Lao, of low capacity for SMART surveys is extremely low. Lao has a team of people carrying population-based survey, yet, those are statisticians + health workers used at periodic surveying, who cant be used ad-hoc, when "emergency" (outbreak) is suspected. In contrast, a "field investigation" should be more of an "intuitive" exploration to which district officers here would be accustomed from older/earlier practice of infectious disease control supported by WHO, and therefore, their pool, instead of nationally-based or not readily available pool, could be used. A field investigation could be quicker (?..though "dirtier?" but more accurate given we would screen 100% of village child population). It would be carried by local health officers/staff and ensure "ownership", local learning and empowering local staff to explore/analyze and propose solutions. Depending on the outcome, it can evolve a full fledged survey but, foremost, it will offer immediate tool and analysis. I wondered if (aside existing protocol for, say MUAC screening) a light tool (questionnaire) exists, looking into immediate (underlying?) causes of malnutrition cases in village. The first thing to note is that cross-sectional surveys (e.g. SMART) are very poor tools for investigations of this kind because the sample size of cases will usually be too small. If (e.g.) SAM prevalence is 1.5% and the sample size in n = 900 (a big sample for SMART) then there will be about 900 * 0.015 = 14 SAM cases. The large number of non-cases does contribute some power but not much. The contribution of non-cases falls away after their number reaches 4 or 5 times the number of cases. In the n = 900, p = 1.5% example, the effective number of non-cases is about 50. Cross-sectional surveys are good for estimating prevalence of disease or risk factors in a population but are not the proper tool for a "causal analysis". The better tool to use is the case-control study. Since SAM is a rare condition we will still not have many cases to work with. We can increase power using a matched design. Matching will typically be on the "grand confounders" of age and sex. I have tended to also match on neighbourhood. For each case found we find non-cases (controls) living in the same part of the village of the same sex and similar age (e.g. +/- 3 months) of the case. It is common to take up to five controls per case. Cases and matched controls are numbered so that we know which controls are matched to which case. Data analysis is by Mantel-Haenszel stratified analysis techniques or conditional logistic regression. Prior to starting recruitment a qualitative investigation is carried out to identify likely risk factors. In some cases you may not want to go further than this. I have had good results with using the SQUEAC toolbox for the qualitative work and recruiting SAM cases from clinics and in the community. I have avoided using clinical controls (to avoid odd biases) and selected control from the community. This method has been taught to MoH / NGO / UNO staff and proved feasible in African settings. Local health officers (e.g. district medical officers) may already be familiar with case-control studies which are a standard tool in outbreak investigation. Field exchange articles [url=http://www.ennonline.net/fex/42/causal]here[/url] and [url=http://www.ennonline.net/fex/45/determining] here[/url] describe experiences with the SQUEAC toolbox + matched case-control study approach. These EN-NET and FEX links: [url]http://www.en-net.org/question/1383.aspx[/url] [url]http://www.en-net.org/question/847.aspx[/url] [url]http://www.ennonline.net/fex/46/causal[/url] [url]http://www.en-net.org/question/729.aspx[/url] may also be useful. I hope this helps.