This is a very good question. Work is ongoing (led by André Briend and Claudine Prudhon) to investigate the best way to derive a correction factor. I suggest you contact them directly (I can put you in touch with then if needed). At present I think the advice is to use 2 to 3 (so ... 2.5). You should also factor in expected coverage into your calculations so you can get at "program needs" (e.g. how much RUTF will be needed). This is not easy to do. Rules of thumb are 5% for centres-based programs, 20% for OTP without extensive community mobilisation (this is pretty certain), > 60% for well-run CTC programs. It is all pretty approximate since the estimate of SAM prevalence will be imprecise (about 50% relative precision or worse with a standard survey), the estimate of the population size will have some error (a lot of error if there has been displacement and / or high mortality), expected coverage is a bit of a guess, and the prevalence to incidence correction factor will also be imprecise. All these errors will add up. I think, therefore, that it is important to monitor coverage closely and adapt needs estimates over time. Just my tuppence.

Estimation of incidence from prevalence is a hazardous exercise. Personally, I suggest to increase prevalence by 60% . This figure comes from the abstract below suggesting an average duration of untreated SAM of 7.5 mo. The idea is as follows: In stable conditions, with many hypothesis met, Incidence = prevalence /average duration of disease In this case, disease duration in year is 7.5/12 incidence = prevalence *12/7.5 = prevalence x 1.6 I heard that some NGOs use a higher correcting factor, up to 2 or even higher. It is quite possible indeed that the Garenne et al study overestimated SAM duration as it is based on study of 6 month intervals, censoring short episodes. In any case, a major challenge is to take into account the effect of seasonal vairations of SAM. The same prevalence will be associated with different incidences if it takes place just before or just after the hungry season. Important to triangulate your estimate with other informations re. food supply / harvest and so on. Indeed with Claudine Prudhon we are currently exploring this issue. Garenne M, Willie D, Maire B, Fontaine O, Eeckels R, Briend A, Van den Broeck J. Incidence and duration of severe wasting in two African populations. Pub Health Nutr, 2009. OBJECTIVE: The present study aimed to compare two situations of endemic malnutrition among <5-year-old African children and to estimate the incidence, the duration and the case fatality of severe wasting episodes. DESIGN: Secondary analysis of longitudinal studies, conducted several years ago, which allowed incidence and duration to be calculated from transition rates. The first site was Niakhar in Senegal, an area under demographic surveillance, where we followed a cohort of children in 1983-5. The second site was Bwamanda in the Democratic Republic of Congo, where we followed a cohort of children in 1989-92. Both studies enrolled about 5,000 children, who were followed by routine visits and systematic anthropometric assessment, every 6 months in the first case and every 3 months in the second case. RESULTS: Niakhar had less stunting, more wasting and higher death rates than Bwamanda. Differences in cause-specific mortality included more diarrhoeal diseases, more marasmus, but less malaria and severe anaemia in Niakhar. Severe wasting had a higher incidence, a higher prevalence and a more marked age profile in Niakhar. However, despite the differences, the estimated mean durations of episodes of severe wasting, calculated by multi-state life table, were similar in the two studies (7.5 months). Noteworthy were the differences in the prevalence and incidence of severe wasting depending on the anthropometric indicator (weight-for-height Z-score

Thanks Mark for your reply.You can put me in touch with André Briend and Claudine Prudhon

Roger, André has already replied. If you still want to contact him or Claudine directly then I will introduce you. Send me an e-mail (my address is "mark - AT- brixtonhealth - DOT - com").

Following my previous post, I received the following comment from Hedwige de Coninck (Fanta 2). With her persmission, I reproduce it below. I think these are interesting considerations and illustrate the difficulty of the issue. I read with interest your answer to the ENN question on SAM incidence, and before exposing this discussion on ENN: Quote: You very clearly defined incidence based on evidence, but I think that people want to hear your opinion on calculating case load, to go one step further than incidence, thus give advice on how to calculate case load. I have learned that this is not an easy step for many. A suggestion could be: For estimating SAM case load for planning purposes, for a 12-month period we base the estimations on: case load = prevalence (take prevalent cases at start of program) + incidence (add new cases expected over 12-month period, based on), and incidence = prevalence / duration of illness (with duration of illness estimated at 7.5 months or 7.5/12) thus we suggest to use: case load= prevalence + incidence, or case load= prevalence + prevalence x 1,6. Next step should account for e.g., expected coverage Example of planning for treatment of SAM for the year 2010 in a population of children 10,000: - If the estimated SAM prevalence rate from a survey done in December 2009 is 1.2 percent On January 1, 2010, there are 120 kids with SAM - The number of new cases that will be expected to develop during the year, or 12-month incidence = prevalence/duration of disease= prevalence x 12/7.5 Incident cases are then expected to be 1.2 x 1.6 = 1.92 or 192 kids - Then for a 12-month program we plan to treat the prevalent cases of 120 kids and add the incident cases over 12-month period of 192 kids, and plan for treating 312 kids -- if coverage is 100% including, e.g., multiple other caveats on seasonality vs stability, on precision of the prevalence estimate, on indicators used for prevalence vs admission or a combination of several Usually the case load is a number higher then prevalence x3, and to account for a coverage lower than 80% we often end up using prevalence x2. It still remains a rough estimate, but at least one learns about prevalent and incident cases and the other assumptions to take into account.

The discrepancy between the incidence and the prevalence can be attributed to the classification of the SAM cases and the MAM cases in the survey reports, as the cases are reported in the surveys as MAM based on the classification of 70-80% median (the surveys here are the tools which can speak on the prevalence), while the admission criteria for the feeding centers for the SAM is less than 75% median according to the new WHO gross chart (the admission criteria here is the tool for indicating the incidence ) , there is 5% of the cases in the prevalence considered as MAM cases (70-75%) while they are at the treatment level appearing as SAM cases which makes the figures doubled or troubled, unifying the measurements tools of the prevalence expression and the admission criteria will make the planning for the supplies requirements easier. New classification for the SAM and MAM in the surveys according to the admission criteria is required.

Indeed, as mentioned in the previous post, it is important when estimating CMAM needs to use anthrometric surveys using the same SAM definition as used for admission for treatment. In this regard, there may be a problem when WFH is used to estimate SAM prevalence in areas where CMAM programmes use MUAC as admission criteria. To avoid this problem, WHO now recommends to measure MUAC (along with weight and height) in anthropometric surveys in areas where a CMAM programme uses mUAC as admission criteria. And also the same WFH definition of SAM for surveys and admission criteria (WFH < -3). See: http://www.who.int/nutrition/publications/severemalnutrition/9789241598163/en/index.html

Hi, While caseload has been described as: prevalence + incidence (which is prevalence x 1.6) x coverage I wanted to check our workings out as when you bring in the length of the programme the statistics are confusing. Please see below for a worked example - please can people comment if it is correct? Would be REALLY glad for any thoughts. If rural pop is 200,000 and 20% are under 5, GAM is 15% , SAM is 2% (MAM is then 13%) and coverage is 50% and proposal funding period is 9 months then: To work out SAM caseload for our proposal: 20% of 200,000 is 40,000 under 5 years 2% of 40,000 has SAM = 800 Incidence is 800 x 1.6 = 1280 Over a 9 month period this is 1280 divided by 12 months = 106.6 to get number per month and then multiplied by number of 9 months = 960 So caseload is 800 + 960 = 1760 Coverage is 50% so divide it by 2 = 880 You would expect 880 as a caseload over 9 months..

Let me just work through your figures again: Number of kids = 200,000 * 0.2 = 40,000 Prevalent cases = 40,000 * 0.02 = 800 Incident cases over 9 months = 800 * 1.6 * (9/12) = 960 Estimated case-load = (800 + 960) * 0.5 = 880 Which agree with your figures. A little depressing that you aim only to reach the SPHERE minimum for coverage when we know that CMAM is capable of so much more than that.

For planning SFP within CMAM, is there any correction factor if the MAM children are admitted on MUAC but the only available data is W/H?

From Vicky Sibson: Dear Mark Thanks for your advice. We in fact aim to surpass Sphere standards, but it can be very difficult. We are currently undertaking a synthesis of recent CMAM evaluations within SCUK to capture common challenges and look at how to address them. In fact we find it hard in many contexts to reach Sphere minimum standards for coverage. And this it not because the programmes are poorly managed or run. There are many reasons coverage might be challenged that are extremely hard to tackle; e.g. population mobility, lack of funding from the donor for sufficiently decentralised service or sufficient community mobilisation - deemed as inappropriate for 'development' contexts. We can pick this conversation up once our review is done. Best wishes Vicky (nutrition adviser SCUK)

The above discussion thread on estimating SAM incidence cases is helpful. We are in a planning phase to address MAM and it is likely that treatment will be based on WFH measurements though MUAC measurements may be used to screen children at community level before referring to the health facilities. It has been observed from survey data that MUAC identifies similiar or slightly higher proportion of children with acute malnutrition in comparision to WFH. The question being is it the same pool of children identified by both indicators or are there different children also with some overlaps? Also, would like to confirm if we could also follow the same method while estimating total MAM caseloads - i.e. prevalence + prevalence x 1.6. Thank you.

If you are identifying children as moderately malnourished by MUAC in the community why are you not also admitting them on MUAC? It will be damaging to the programme and you will not be able to achieve good coverage if you use different screening and entry criteria. MUAC and WFH are 2 different tools to measure malnutrtion, but if you are doing a community based programme the tool of choice is MUAC. If it is because National Guidelines dictate the use of WFH then in most circumstances it is easy to agree a compromise of entry criteria of MUAC or WFH. In practice the vast majority of children will be on MUAC as that is how they were identified and referred in the first place. Apologies, this is not answering your prevalance question! Andre is best placed to answer so hopefully he will have achance to reply.

Thank you for the reply. A clarification from my side - children identified with MUAC measurements will be enrolled into the programme but WFH will also be taken as part of detail examination. Anyhow we are in a very preliminary planning stage and implementation details are yet to be properly worked out.

Dear Sophiya, Children identified by low MUAC and low WFH are not the same, although there is considerable overlap. Children identified by a low MUAC tend to be younger, as MUAC increases with age. Those identified by low WFH tend to have longer legs. Children identified with MUAC tend to have a higher mortality for two reasons. First, young children have usually a higher risk of death. Second, MUAC is closely related to muscle mass, a key determinant of survival. Both factors may act together to select high risk children, as young children have comparatively a low muscle mass, which presumably explains their vulnerability when malnourished. Also, children with long legs are likely to be in better health and also presumably to have more muscle, so targeting them on the basis of a low WFH them might be somehow counterproductive. I hope this helps,

Thank you Andre, very helpful and interesting. I guess those children who are indentified only by WFH will later on get picked up by MUAC also if their acute malnutrition worsens. Also, when you say 'considerable' overlap, would you have any rough estimate of what proportion? Or is this something that varies between population?

Dear Sophiya, A discussion on the overlap between low MUAC and low WFH is available in the paper : Myatt M, Khara T, Collins S. A review of methods to detect cases of severely malnourished children in the community for their admission into community-based therapeutic care programs. Food Nutr Bull. 2006 Sep;27(3 Suppl):S7-23. See fig 5. This paper is freely available on the Food and Nutrition Bulletin website. A discussion on the effect of body size of selection of children with low WFH and low MUAC is available in: Myatt M, Duffield A, Seal A, Pasteur F. The effect of body shape (and leg lenth) on weight-for-height and mid-upper arm circumference based case definitions of acute malnutrition in Ethiopian children. Ann Hum Biol. 2009 Jan-Feb;36(1):5-20.

Since W/H varies with body shape and body shape varies with climate (hotter = longer legs), altitude (higher = shorter legs, diet (better = longer legs. more milk = longer legs), and genes it follows that the overlap will differ depending on where you happen to be (i.e. because in some settings WFH will select healthy older kids with long legs). I tried this on a database of 538 nutritional anthropometry surveys. I censored all cases with oedema or WHZ < -5 or WHZ > -5 or age < 6 months or age > 59 months. I then applied the case-definitions WHZ (WHO) < -3 and MUAC < 115 mm and looked at the number of cases selected by each case definition. I summarised this by country. I got: Country N MUAC WFH Either Both Notes ----------------- -- ---- ---- ------ ---- ---------- Afganistan 35 1165 822 1579 408 MUAC > WFH Angola 17 510 248 621 137 MUAC > WFH Burundi 14 255 174 342 87 MUAC > WFH Chad 31 1009 1280 1825 464 MUAC > WFH DRC 33 642 508 945 205 MUAC > WFH Ethiopia 45 951 681 1349 283 MUAC > WFH Ethiopia (Somali) 8 80 228 270 38 WFH > MUAC Haiti 27 249 159 340 68 MUAC > WFH Kenya (Somali) 7 61 252 282 31 WFH > MUAC Liberia 30 653 557 958 252 MUAC > WFH Malawi 9 301 236 466 71 MUAC > WFH Mozambique 9 79 46 103 22 MUAC > WFH Myanmar 8 261 228 383 106 MUAC > WFH Niger 4 169 201 259 111 WFH > MUAC Pakistan 9 173 181 274 80 WFH ˜ MUAC Rwanda 13 390 276 528 138 MUAC > WFH Sierra Leone 38 1641 1164 2157 648 MUAC > WFH Somalia 17 770 571 1119 222 WFH > MUAC Sir Lanka 3 17 66 74 9 WFH > MUAC Sudan (BOTH) 141 3683 5327 7379 1631 WFH > MUAC Tajikistan 5 220 157 310 67 MUAC > WFH Tanzania 6 71 42 97 16 MUAC > WFH Uganda 29 529 291 649 171 MUAC > WFH ----------------- -- ---- ---- ------ ----- ---------- N : Number of surveys MUAC : Number of cases with MUAC < 115 mm WFH : Number of cases with WHZ < -3 Either : Number of cases meeting EITHER case-definition Both : Number of cases meeting BOTH As you can see the degree of overlap varies from place to place. The issue here is, I think, whether, WHZ is a useful for CMAM. There has already been much discussion of this on these forums.

BTW ... You can see mu body shape paper at: [url]http://www.brixtonhealth.com/MyattBodyShape.pdf[/url]

Thanks for such interesting analysis by Mark Myatt. In ACF we are very much interested by this analysis but in Asian countries. But I see that only Pakistan is included in Mark's analysis. Does anyone similar information/analysis concerning asian countries? At ACF-Spain we are planning a SMART survey in the Philippines and we are evaluating possibility to measure the "sitting height" to evaluate the "body shape" through SSR as in previous surveys we found MUAC having a very low correlation with W/H. Has anyones experience on doing that and could share it with us?

Look again ... Afghanistan, Myanmar, Sri Lanka, and Tajikistan are Asian countries too. See: [url]http://www.brixtonhealth.com/MyattBodyShape.pdf[/url] For a description of a study looking at WHZ, MUAC, and SSR. If you already have data on just MUAC and WHZ than you could look at prevalences by age as outlined in: [url]http://www.en-net.org.uk/question/543.aspx[/url] on this site.

This forum has discussed caseload estimation previously with caseload = prevalence + incidence and taking coverage into consideration. However, I am not clear if this calculation is for NEW (start-up) programmes only, or whether it is also for ON-GOING programmes (where you are going for further funding), so you are already treating people. In this case to calculate future caseload do you just take prevalence to be your current admissions? Or do you ignore prevalence and only estimate your future caseload on incidence?

This is a very simple and very approximate approach to estimating caseload. Using this method the caseload in the first year (i.e. your "NEW (start-up) programmes") will be: (prevalence + incidence) * coverage assuming that you have done the needed calculations to turn prevalence and incidence into numbers. We can usually only guess at incidence and estimate it as something like: 1.6 * prevalence So .. the expected case number is something like: (prevalence + prevalence * 1.6) * coverage For a NEW program. Note that the 1.6 applies to a whole year. For ONGOING programs after a period of about 7.5 months after start-up the case-load might be estimate as: prevalence * 1.6 but you will probably get a better idea of what to expect from the experience over a year or from something like: cases treated since startup - (prevalence * coverage) You should have a better idea of what to expect in an ongoing program. Does this make sense?

what will be yearly target (% point drop in prevalence of stunting and wasting from baseline) for a two year project. What other factor we have to look for in setting the such targets. what are the scientific evidence for it?