It is not, I think, a good idea to pick one or the other (i.e. point or period coverage). The best choice will depend on context ...

If the program has good case-finding and recruitment, short lengths of stay, and low levels of defaulting then the period coverage estimator is likely to be the most appropriate estimator.

If the program has poor case-finding and recruitment, long lengths of stay due to late presentation and / or late admission, and high levels of defaulting then the point coverage estimator is likely to be the most appropriate estimator.

Point and period estimators measure different things ...

Point coverage reflects the ability of a program to find and recruit cases. The point coverage estimator does not account for recovering cases and so does not directly reflect the program's ability to retain cases from admission to cure. This means that it may give a misleading (i.e. downwardly biased) picture of program performance. This is a particular problem with programs that recruit cases very soon after they meet program admission criteria and are likely to be treating large numbers of recovering cases. In these programs the point coverage estimator will “penalise” good performance.

Period coverage is intended to reflect the ability of a program to find, recruit, and retain cases. The period coverage estimator does directly reflect the program's ability to retain cases from admission to cure but tends to overestimate program performance because the denominator does not include recovering cases that are not in the program. The period coverage estimator also overestimates program performance when cases are retained in a program after meeting the criteria for being discharged as cured.

The single coverage estimator is an attempt to correct these biases while maintaining compatibility with existing methods and allowing historical comparisons to be made. It relies on some estimate of the ratio of the mean length of an untreated episode to the mean length of a treated episode.

For SAM we have relied on published estimates but methods for estimating caseloads have been developed in recent years that allow us to estimate the  mean length of an untreated episode. We can use both approaches for MAM.

SURPRISE!!! Here are some published estimates of the duration of an untreated MAM episode.

I hope this is of some use.

Dear Mark,

Many thanks for this explanation. Looking at the paper you provide a link to the length of untreated MAM is given for W/H (75-81 days) and MUAC (101-116 days).

The data we have for length of stay of a treated case (42 days) is not disaggregated by W/H and MUAC. So what value would you recommend using to as our correction factor to calculate Rout?

Many thanks

Ben

If you have admissions disaggregated by type then you could use a weighted average, if (e.g.) you have 70% admitted by MUAC and 30% admitted by WHZ then you could use something like:

    LOS = 0.7 * (101 + 116) / 2 + 0.3 * (75 + 81) / 2 = 99.35

Note that the two case-definitions will overlap (usually to a considerable degree). MUAC is usually the mode severe case-definition because it tends to be associated with low HAZ, WAZ, WHZ, and low peripheral tissus mass. MUAC is also more strongly associated with mortality than WHZ. I think that I would treat MUAC < 115 mm and WHZ < -3 as MUAC admissions.

If you do not have disaggregated admissions data then you could do a small records pull to get estimate proportion in each class. You could pull data to get average LOS for each admission type but having that will probably lead to a more complicated analysis of coverage.

I hope this is of some use. 

Thank you Mark for this, very helpful and much appreciated.