We use an events calendar to try to determine age, along with an appetite test, because of the risk of excluding eligible older infants from OTP.

I'll start with some pedantic points regarding language ... (1) It seems that you are asking about surveys. Stunting is a process (the "-ing" indicates this). You can only detect stunting using repeated measures on the same child. What you can detect (often with considerable error) is low H/A which might be termed "stuntedness". (2) Stunting usually occurs quite early in life. Longitudinal data suggests that after about two years the decline in H/A (i.e. stunting) ceases and the stunted child tracks the H/A growth curve at least until the pre-adolescent growth spurt. This means that you are measuring two things with the same indicator ... stunting and stuntedness. This has consequences for assessing interventions by survey since the bulk of the sample is older then two years and unlikely to respond to an intervention. In this case the survey should sample younger children. Some practical issues ... The problem with the ".. for age" indicators (e.g. W/A, H/A, MUAC/A) is that they are very sensitive to errors in age and age is often very difficult to measure accurately. Research suggests that any indicator that includes an age component requires that age be ascertained accurately and are more sensitive to errors in age than to random in anthropometry. Bairagi R, Effects of bias and random error in anthropometry and in age on estimation of malnutrition, Am J Epidemiol, 1986;123(1):185-91. This is not so much an issue for (e.g.) detecting "stunting" by longitudinal data since the initial error is replicated through the measurements and (if detected) can be fixed. It is, however, a problem with surveys. Error comes in many guises. One major source of error is that when age data are uncertain they are "corrected" by surveyors based on am expectation of age based on height. This leads to prevalence of low H/A being underestimated. Mike Golden gave some other examples of errors in age in [url=http://www.en-net.org.uk/question/413.aspx]this post[/url]. You may also find [url=http://www.en-net.org.uk/question/267.aspx]this post[/url] useful. There are ways to improve accuracy of age data. The use of calendars is one but I am sceptical as to whether these really work in survey conditions. My main concern about ad-hoc changes to survey eligibility criteria is that they get in the way of comparisons between surveys and between countries. I would stick with the 65cm - 110cm range in surveys. I would, however, be much more liberal in programming.

WHO Growth Standards recommends 67-110 for 6-59 months, doesn't it?

A reference can inform a recommendation. The recommendation is made by humans. Looking at the [url=http://www.who.int/childgrowth/en/]WGS[/url] : For boys : Median L/A at 6 months = 67.6 Median H/A at 59 months = 109.4 For girls : Median L/A at 6 months = 65.7 Median H/A at 59 months = 108.9 Average ("sex-combined"): Median L/A at 6 months = (67.6 + 65.7) / 2 = 66.7 Median H/A at 59 months = (109.4 + 108.9) / 2 = 109.2 That looks like 67 to 109 cm to me. Remember that the WGS is about ideal conditions. On EN-NET we don't often find ourselves dealing with populations that have or are experiencing conditions ideal for growth. That means that using a threshold of 67 cm will exclude a lot of 6 month old children. Far more that 50% because 67 cm will (by definition) exclude half of six months olds. We can expect "our" children to be a bit below the WGS. Let us say 1 SD below. What happens then? For boys : Median L/A - 1 SD at 6 months = 65.5 Median H/A - 1 SD at 59 months = 105.3 For girls : Median L/A - 1 SD at 6 months = 63.5 Median H/A - 1 SD at 59 months = 104.7 Average ("sex-combined"): Median - 1 SD L/A at 6 months = (65.5 + 63.5) / 2 = 64.5 Median - 1 SD H/A at 59 months = (105.3 + 104.7) / 2 = 105.0 That's where the 65 cm comes from. The 110 cm is a hang-over from the NCHS reference. I'm not to worried about that being quite a way above 105 cm because SAM is mostly a disease of far smaller and younger kids (unless you are using something weird like W/H in your SAM case-definition and fill you program up with healthy long-legged children). The initial question mentions 60 cm and 100 cm. These are, IMO, better than 65-110 cm and 67-110 cm for the contexts that we work in.

Mark, thanks for the excellent analysis. My conclusion from this is to NOT use length or height as a surrogate measure of age in population-based surveys. In populations with substantial stunting, which includes most populations in the world, not only do you fail to include in the survey sample some children who are older than 6 months of age, but you mistakenly include many children who are 60 months of age or older. As Mark has pointed out, younger children are much more likely to be wasted than older children, and older children are much more likely to be stunted than younger children. By using length or height to determine eligibility, you automatically bias your survey sample toward older children. This results in underestimation of the prevalence of wasting and overestimation of the prevalence of stunting. Please do not do this. Age is not necessary to determine the presence of wasting with either MUAC or weight-for-height, so a really accurate age is not that crucial. However, some estimate of age is necessary to determine eligibility for inclusion in the survey sample. It is not so difficult to construct a local calendar, and applying it at each household to determine a candidate child's eligibility is relatively easy. Even if an objective of the survey is measuring the prevalence of stunting, construction and application of a local calendar is much better than intentionally biasing the survey sample by using length or height as the criterion for eligibility.

Thanks for Mark and Brad for discussing this issue of using height as a proxy for age in detail in the context of nutritional surveys. I am in full agreement with your conclusion that height is a very poor proxy of age in a context of high prevalence of stunting. I think it is appropriate too to point out that this conclusion is also applicable in the context of individual screening. Apparently quite a few programmes still use height as a proxy of age to determine if low MUAC children are above 6 months and are eligible for therapeutic feeding with RUTF. As a result, they often exclude from treatment children who are above 6 months but are stunted. This is highly regrettable as there is now good evidence that these children have a high risk of death and do respond to treatment when given RUTF. Determining age of children when they are around 6 mo of age is usually not too difficult, as birth in this case is still a recent event and families do remember the date rather well. It would be regrettable to deny treatment to high risk children likely to respond to treatment just because those in charge of the programme don’t trust the age given by the mother.

There is some work on the effect of errors in age. For example: Bairagi R, Effects of bias and random error in anthropometry and in age on estimation of malnutrition, Am J Epidemiol, 1986;123(1):185-91 This shows that W/A and H/A are very susceptible to errors in age. Children grow fast and this means that small errors in estimating age may lead to large errors in indicator values. In famine, and situations where displacement and familial separation are common, field workers are often required to estimate the age of children based on little or no information. Estimates “by eye” are biased by assumptions about the relationship between height and age which are likely to be invalid in situations of nutritional stress. In these cases, indicator values will be subject to errors, probably systematic and upwards, that are products of random errors in estimating age and systematic errors in estimating age that may be influenced by growth failure. This is a rather long winded way of saying that I agree completely with Brad that if you are looking at stunting (or underweight) then you should take the time to make and test local calendars ... not just for eligibility but also for ascertaining age with some accuracy, Don't rely on maternal report as these are usually biased towards whole years. When I was a child I remember saying things like "I am six years, eight months, and elven days old" when asked my age (I'm sure we all did this). My mum would say "six" or "six and a half". That sort of rounding will (if widespread and typical) increase the estimate of the prevalence of stunting just by rounding error. I think that André has it right too (including regarding maternal report with the youngest children).