I will use your numbering 1. Strictly speaking ... areas and populations are not categorised prior to phase II. Instead, testable hypotheses regarding coverage are proposed. These will include statements of the hypothesised level of coverage and the reason(s) for the hypothesised level of coverage. Phase II is all about testing these hypotheses. This will usually involve direct assessment of coverage but might also include checking the presence / absence of hypothesised reasons. Phase I is, amongst other things, about creating these hypotheses. Phase II is about testing these hypotheses (this is where "categorisation" is done). 2. How do you know admissions are high? It is a common misconception that high numbers means high coverage. It may mean high prevalence / incidence with low to moderate coverage. Large numbers suggest high coverage but do not confirm high coverage. Population is an issue but be careful not to confuse population with need (it is one component of need). It is very possible (e.g.) that SAM prevalence is very low in urban centres but very high in rural areas. In that case we would expect to see low numbers in urban settings despite high population density. What matters is need: need = population * prevalence If I see few cases coming from a high population area I would check to see if this was because there were no cases in that area (i.e. I'd go look for cases in that area). 3. Late presenters : If you admit on MUAC < 115 mm then any admission with MUAC much below this is a late presenter. We tend to look at the shape of the plot of admission MUACs. If we see (e.g) a plot like this: | | * | * | ** | *** | **** | ***** | ******* * | ************ ** * * + +----+----+----+----+----+ 115 110 105 100 95 90 Then we'd probably be pretty happy with treatment seeking overall. We might then treat cases with MUAC < 100 mm at admission as critical incidents and pull records and take histories &c. If we see (e.g) a plot like this: | | | | * | ** | *** | ***** ** | ********* *** | * **************** * ** + +----+----+----+----+----+ 115 110 105 100 95 90 Then we'd be concerned about treatment seeking. We might still treat cases with MUAC < 100 mm as "cases of interest" but the problem looks more systemic than a handful of late presentations. I'd be tempted to look at the "positive deviants" (i.e. MUAC >= 110 mm at admission) and the "negative deviants" (i.e. MUAC < 100 mm at admission) and compare and contrast ... BUT ... this plot shows a general problem with case-finding / recruitment. In the second plot there does seem to be a "hump" below 100 mm. The pattern suggests that we have poor case-finding and recruitment generally but that it may be worse in certain places / populations. A cut-point of 100 mm makes sense as a basis for further investigation. Here the cut-point is chosen by examination of the plot (i.e. the data). NOTE : In the second plot there is little "good" only "bad" and "worse". It is better, I think, to select a cut-point from the plot. Another approach might be to look at the bottom 10%. 4. Which guidelines are you referring to? The "handbook" says Summary measures may be calculated but visual inspection and interpretation of the plot is usually more informative. I find no reference to the use of median admission MUAC in the the "handbook". 5.The classes in the handbook have been found useful for SQUEAC coverage assessments. They are not intended to be used for other purposes. Be careful of confusing contexts. In the SQUEAC context we are looking to describe the problem of defaulting as it relates to coverage. The "minimum of two months" is about something different. BTW ... I have seen programs with median LOS well below 2 months. To my mind a two month threshold for "early defaulters" would only apply in a program with very poor rates of recovery. Any help?