Monitoring with MUAC on admission

Guidance on this is included in the FANTA SQUEAC / SLEAC Technical Reference on pages 18-21.

If you want to use a summary statistic then you could use the median MUAC at admission. This should be close to the admission criteria. If (e.g.) you admit on MUAC < 115 mm then the median MUAC at admission should be 114 or 113 mm as in Figure 14, 15, and 16A in the FANTA SQUEAC / SLEAC Technical Reference. Another approach would be to pick a threshold (e.g. 110 mm) and report the proportion of admissions with (e.g.) MUAC >= 110 mm. This should be high (e.g. >= 75% or >= 80%).

I have found the "distribution" approach (as in the FANTA SQUEAC / SLEAC Technical Reference) most useful.

Dear Casie,

in coverage surveys, we usually calculate MUAC median at admission to estimate early/late seeking behaviour/timely identification. The following cut-off generally used are:
11.0 – 11.4 early
10.5 – 10.9 medium
<10.5 late
I personally consider "critical cases" when MUAC is <10.0
Research data indicates that mortality risk significantly increases from MUAC <11.0. However, the decision to use 0.5 cm between each category is more a question of "agreement" between implementation partners and a way of monitoring the progress of a programme. When a CMAM program is set, it is expected that median is low but with the uptake of the activities, including active case finding and self-referral, the median should increase and children should be identified very quick. I am sure Mark Myatt can give you more information about the cut-off but, as far as I know, it is more a programmatic monitoring cut off decision. When MUAC is <10.0, the mortality risk is very high and the reason should be investigated

It has just occurred to me ... Another reason to monitor MUAC at admission particularly in innovative programs (e.g. IRC's CMAM by community health workers with poor letter and number skills) is that it is a key component in estimating expected mortality in the patient cohort. This is essential for cost-effectiveness analysis using DALYs. As CMAM program models proliferate (e.g. NGO CMAM, integrated CMAM, COMPAS, CAM/Surge, CHW delivered CMAM, IRC's community delivered CMAM) it will be important to have some idea of what works and at what cost.

MUAC monitoring is essential but we do need a simple and standard way of doing CEA fro CMAM programs. UNICEF Nigeria and Concern Worldwide are working on this.

I hope this is of some use.

I don't know if the threshold has a detailed evidence base. It has an evidence base (i.e. low MUAC = high severity = high mortality). A review of the original CTC literature might show the reasoning. My guess is that it comes from accumulated clinical experience. I know (e.g.) that Lio and Paul have extensive CMAM experience.

I have a few OTP datasets to hand. Just looking at MUAC in the two groups in one program:

    MUAC in died or transfered to inpatient care
       Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
      80.00   91.00   96.00   97.91  105.50  112.00 
    -----------------------------------------------
    MUAC in others
       Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
       82.0   105.0   110.0   108.4   114.0   115.0 

The difference is significant (p < 0.0001) even though the number of deaths and transfers to stabilisation is small.

It should be possible to collect several OTP datasets (so we have a good number of the events of interest) and analyse the collected data using ROC analysis. This would not be a very difficult analysis. Most of the work will be in assembling the study database.

I have data from a handful of programs that I can seek permissions for such an analysis. Is anyone out there interested in providing data for such an analysis or doing the analysis or writing up the analysis?

Franck,

Thanks for volunteering to do the analysis / share data. I have some data too.

Let us wait for a few days to see if anyone else wants to join the fun.

Hi Mark and everyone

Do you have data on any symptoms and clinical signs (including those that would not qualify the child as ‘complicated’? The reason for asking is that we have started looking at our ‘complicated’ SAM databases and looking at signs to see just how complicated children are, a work in progress, but we are seeing a very strong relationship with mortality. Children at lowest level of being ‘complicated’ have a low mortality risk during the next 6 months, around 1-2% as might be seen at an OTP.

Cheers

Jay

Jay,

From my experience ...

Most OTP datasets will be limited to routine program data (i.e. admissions and exits by type over time). In some settings this might be expanded to include admission criteria. Data collected by SQUEAC coverage assessments will usually include MUAC / WHZ at admission in addition to routine programming data. Some NGOs collect wider datasets but these are usually limited to anthropometry at admission and throughout the treatment episode.

This means that we are limited to research datasets. I think the richest datasets will be from the VALID's CTC research program. They collected full clinical data for all cases. I guess that ACF may have similar data from their alternative to CTC (similar to CTC/CMAM but with a short inpatient stay for all children). VALID Nutrition have tested quite a number of RUTF formulations over the past decade or so and may also have useful data.

Maybe you know all this.

I have data from a study in Malawi which records week number, date, weight, height, MUAC, oedema, TSF (2 measurements), diarrhoea, fever, vomiting, cough, RUTF ration, and outcome for each visit.

I also have data from Nigeria which was a records pull, which records routine admission medications give, admission weight, admission height, admission MUAC, admission oedema, diarrhoea at admission, fever at admission, vomiting at admission, cough at admission, diarrhoea after, fever after admission, vomiting after admission, cough after admission, and some other data. This has the usual issues with pulled records but the sample size is > 100,000. The sample size brings false discovery issues.

I think it should be possible to share these datasets with you. Let me know if they are of any interest.