Estimations de cas pour lesquels les critères de MUAC ont changé

Most anthropometric measurements follow a normal-like distribution. This allows us to use good old fashions mathematical statistic sto estimate things like this. If, for example, we have a distribution of MUAC with a mean of 140 and an SD of 13 then prevalence at MUAC = 110 mm will be about:

    pnorm(110, 140, 13) * 100 = 1.05%

At MUAC = 115 we expect:

    pnorm(115, 140, 13) * 100 = 2.72%

So prevalence estimates will be about:

    2.72 / 1.05 = 2.6

times higher. The ratio changes with changing means and SDs.

You can do the calculation with survey data. I did a quick check with dome data I have here from Kenya (86,018 children), Tanzania(5,290 children), and Uganda(54,236 children) and got mean = 143.32 and SD = 12.94. These give:

    pnorm(110, 143.32, 12.94) * 100 = 0.50%
    pnorm(115, 143.32, 12.94) * 100 = 1.43%

The ratio is:

    1.43 / 0.50 = 2.9

I also looked at classic prevalence estimates in the same data:

    P(MUAC < 110) = 0.54%
    P(MUAC < 115) = 1.41%

the ratio is:

    1.41 / 0.54 = 2.6

These estimates are broadly in line with WHO/UNICEF numbers:

    3.27 / 1.49 = 2.2

from a global database of surveys.

The shift from NCHS to WGS references for WHZ was to give a 2.18 fold increase in prevalence.

See:

WHO / UNICEF. (2009). WHO child growth standards and the identification of severe acute malnutrition in infants (JOINT STATEMENT), WHO, 1–12

Not that this is about prevalence not caseloads which are influenced by incidence and coverage. The "wild card" here is coverage. If your programs lack good sensitisation, mobilisation, ease of access giving good spatal and temporal coverage of case-finding and recruitment the the increases in caseloads are likely to be small than estimated above.

I hope this is of some use.

Thank you, Yared, for your kind comments.

The main difference between (e.g.) WHZ and MUAC is that WHZ is a measure of comparative thinness against an urban haute bourgeoise population and MUAC thresholds are "functional" in the sense that they predict near term mortality. Ter is a lot of confusion about this.

The fact that MUAC picks up young and stunted children is a good thing as these children are at high risk of death and are likely to be missed be WHZ. They are not "false positives". I think you have this in "I believe it will benefit a lot of smaller children at risk of death".

Prevalence in the shortest quarter will increase as MUAC picks up the younger and/or stunted kids. Prevalence in other height/age groups will likely be lower. The effect will be most marked in pastoralist population where body shape causes WHZ to pick the older healthiest kids with good limb growth. These are false-positives.

It is a sign of how MUAC is seen by some that a doubling of case numbers is seen as problematic when a similar increase in case-numbers caused by the introduction of the WGS reference is seldom raised as an issue. Maybe this is because WHZ has coverage issues so caseloads do not change much.

Historical note : The impetus for moving MUAC thresholds from 110 mm to 115 mm was, in part, to keep MUAC and WHZ case numbers comparable (as they were with NCHS WHZ < -3 and MUAC < 110 mm).

I agree that desk exercises (as we have done here) are useful and pilots are also needed. It is usually better to start small and iron out problems than to go large and untested. I think the main thing to note is that caseloads will depend on coverage. It is possible to double case-numbers by doubling coverage. A simpel change of case-definitions may not have a similar effect.

Not sure if this helps.

What interesting questions!

There are a number of things to unpick.

My experience has been that IMAM (and CMAM) tend to be restricted to SAM treatment. Mostly I have seen separate or "semi-detached" MAM programming. I think there is an issue of incompatibility between SAM and MAM treatment protocols and with the support for these programs. I have seen programs in which inpatient care (TFP) was supported by the WHO with no co-location and integration with CMAM programming (e.g. no referral from TFC to OTP or back), OTP supported by UNICEF with no co-location with SFP, and SFP supported by WFP with no referral of non-responders to OTP or TFP. Integration between TFP and OTP does seem to be more common and better than between OTP and SFP. This is a difficult issue. Recent and ongoing work on unified SAM/MAM protocols may help.

I think we can be very bad at "horizontal" programming. Linkages between (e.g.) IYCF counselling programs and CMAM often do not exist even though both use community-based volunteers and are child nutrition programs. Getting MUAC screening into EPI programs and GMP programs can be difficult. Even the term "integration" is confused. In some places this can mean treatment in general health facilities (i.e. not in NGO centres or specialised nutrition units) with some NGO support rather than "joined-up" programming. This all sounds a bit desparate but there are examples of "joined up" programming that work. I think we need to identify and replicate this style of programming.

We are also not great at the numbers and costs issues. First the easy bit (using the data give above). For the 110 mm / 125 mm case in a population of 10,000:

  SAM caseload = pnorm(110, 143.32, 12.94) * 10000 = 50
  Overall caseload = pnorm(125, 143.32, 12.94) * 10000 = 784
  MAM caseload = Overall caseload - SAM caseload = 734

For the 115 mm / 125 mm case in a population of 10,000:

  SAM caseload = pnorm(115, 143.32, 12.94) * 10000 = 143
  Overall caseload = pnorm(125, 143.32, 12.94) * 10000 = 784
  MAM caseload = Overall caseload - SAM caseload = 641

We have no more cases. The case-mix has altered to more SAM and less MAM.

I think, however, your point is more about treating MAM at health facilities when the number of prevalent cases increases from 50 (or 143) to 734. That requires some thought. Costs can be kept down by reducing the size of the RUTF ration for MAM cases. Crowding can be controlled in a variety of ways by (e.g.) queue control methods. use of nurse-practitioners, and by reducing frequency of contact for MAM cases. Taking steps to increase coverage can increase patient numbers but can be cost-neutral down by finding and recruiting cases early so they can be treated quickly and the number of SAM cases treated in OTP or TFP reduces. This is an important point for the 110 - 114 mm group. If we can manage to get most SAM cases early (e.g. get a median admission MUAC of 112 mm or better) then treatment should be short and successful and costs reduce.

The reduction of the case-defining threshold to 120 mm is an option. It has been used in several settings. It may not have much effect on patient numbers in low/moderate coverage programs because most cases admitted will usually have MUACs below 120 mm.

Another issue with numbers is that we frequently ignore coverage. Average coverage of OTP programs is a little below 40%. We can assume that this will be lower for MAM (say 20%). This gives

  SAM caseload = pnorm(115, 143.32, 12.94) * 10000 * 0.4 = 57
  Overall caseload = pnorm(125, 143.32, 12.94) * 10000 * 0.2 = 157
  MAM caseload = Overall caseload - SAM caseload = 100

This look more manageable.

BUT ... We have to be very wary about predicting caseloads from prevalence and coverage alone as it ignores incident cases. Estimating incidence from prevalence is a matter of ongoing research (UNICEF has an ongoing project) and it seems that we should inflate case-load predictions by between 2 and 17 times depending on location (the bigger numbers seem to be for West Africa).

The nutrition community is also, IMO, quite weak about costs. Much work in done on budgets and these are often large. Much less work is done on what is being bought for the money spent. I have looked at this in Nigeria CMAM programs. The supply side cost per life saved is about US$275. This cheap. The cost per DALY averted (using discounting and age-weighting) is below US$10. This is very cheap. I think we have to realise (and then keep proving it and telling people) that our programs represent exceptional value for money. They reamin cheap at 10 times the cost.

I hope this is of some use.