The question is about a refugee camp setting. I think we can assume that this is a more "controlled" or "regulated" population than a free-living population and, to some extent, IDP camp populations. By this I mean that nutritional needs tend to be met by a general ration, health needs by a system of primary and secondary service delivery units, minimal public health needs are met by (e.g.) sanitation programs, &c. In short, we can reasonably expect basic needs to be met by planned provision. In such a setting, surveillance of processes may be as useful as surveillance of outcomes. For nutrition, coverage of rations, adequacy of rations (in terms of energy and other components), and delivery of a complete ration by (e.g.) small audit surveys of baskets) can be done. The idea is to check that the systems intended to deliver nutritional "goods" are functioning as intended. A clinical audit approach could be adopted in order to steer systems by continual moderate reform towards acceptable standards of delivery. Get this right and malnutrition is unlikely to be a problem. Of course, things are not quite as simple as that and many things will need to be monitored. Many of these things can be monitored from within existing structures (e.g. diarrhoeal disease can be monitored by clinic workload returns). The question of surveillance of nutrition status (e.g. cases of wasting) can follow the same pattern (i.e. using clinic workload returns,GMP program returns, &c.). An annual cross-sectional nutritional survey provides only limited information for a considerable outlay. It provided snapshots of prevalence at points in time separated by a year. A lot can happen in that year between surveys. An extreme case would be "famine" conditions at the mid-point between two surveys in which a lot of children die leading to an increase in nutritional status. A system of surveillance on (e.g.) a monthly basis could be done using (e.g.) block leaders to report population, births, deaths and the number of children falling below a MUAC threshold on a monthly basis. An alternative approach might be to use small surveys using either MUAC or W/H using some form of systematic or random sampling (beneficiary lists allow for a random sample) of households. LQAS decision rules could be used to classify prevalence with reasonable levels of error with sample sizes of 100 or fewer children. Using averages of an anthropometric index rather than proportions falling below a case-defining threshold will also allow smaller sample sizes to be used. The purpose of surveillance here is NOT to provide an accurate or reliable prevalence estimate but to detect a potential problem which once detected will lead to further actions which might include a full-blown cross-sectional survey.

Thanks Mark. Indeed the purpose of surveillance is to detect a potential problem which once detected will lead to further actions. You have indicated that an alternative approach of conducting small surveys using either MUAC or W/H using some form of systematic or random sampling would be appropriate. My question again is how frequent should the small surveys be conducted.(Quartely or biannually e.t.c) Would you please clarify on Using averages of an anthropometric index rather than proportions falling below a case-defining threshold would allow smaller sample sizes as you have indicated in your response.

Frequency : This is a dificult question. In a setting in which you expect to have a problem once per year (e.g. in the "lean period") then it might be sufficient to undertake a large sample (e.g. 30-by-30) survey once per year at the appropriate time. Usually you will want more frequent data than that. Surveillance systems can use smaller sample sizes and apply some from of smoothing to the data at each timepoint to filter out some of the sampling variation. This means that you can get more frequent data quite cheaply. I'd try to match data collection times to the agicultural and disease calendars and make sure that they line up with both good and bad time starting with quarterly (every three months) as a starting point. Soory not be more specific. As for "Would you please clarify on using averages of an anthropometric index rather than proportions falling below a case-defining threshold would allow smaller sample sizes as you have indicated in your response" : With surveillance you want to detect change. Using mean weight, mean weight change, mean MUAC, or mean W/H as the "indicator" you can usually detect changes with small sample sizes. Also, you can detect changes that reflect a worsening of the nutrition situation but are not sufficiently large to be detectable as changes in prevalence without using a large sample size.