Menu ENN Search
Language: English Français

Capture-recapture study duration

This question was posted the Assessment forum area and has 4 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 303

Normal user

19 Nov 2009, 09:31

I would like to hear your experience with the duration of capture-recapture study to assess the sensitivity of active case-finding for coverage survey.

I have found only one paper which mentions the duration of capture-recapture study: Implications of a Coverage Survey in Ethiopia by Simon Kiarie Karanja on Field Exchange n.31 (http://fex.ennonline.net/31/surveyinethiopia.aspx#). In that case a one day study was carried out.

This is my case. Based on this paper: http://www.brixtonhealth.com/CRCaseFinding.pdf I have found that we need a sample of 15 SAM children to perform the capture-recapture study. We will collect the sample using house-to-house and active case-finding methods. Therefore 1500 children should be measured in the case of the house-to-house method, supposing a 1% prevalence of SAM in the area. While using the active case-finding method the teams need to measure far less, say 200 children, to find the 15 with severe malnutrition.

Let's suppose that a team can take weight and height of up to 50 children per day using one of the two methods. And let's suppose that we have 5 teams. It will take almost one day for the active case-finding (200/50/5=0,8), and 6 days for the house-to-house screening (1500/50/5=6).

Am I wrong in my logic/calculation or is 7 days a normal duration for a capture-recapture study like this one?

What is your experience in this regard?

Thank you in advance.

Mark Myatt

Consultant Epideomiologist

Frequent user

19 Nov 2009, 16:26

Doing it all in one day seems a little quick to me. It seems likely that the case-finding procedure had been used locally before (a lot of CSAS surveys were done in Ethiopia during CTC development and roll-out) and the validation was just a quick check.

It is possible to do things this quickly if you are very experienced. In this case you would make a qualitative judgement about how well you case-finding procedure works. The "qualities" of a sensitive case-finding procedure are:

(1) The key-informant(s) clearly understand what you want.

(2) After introductions &c. you are taken immediately to a case of MAM or SAM.

(3) When you ask about the child the carer uses the same terms that you use in your case-finding question.

(4) When you ask the neighbours about the child they use the same terms that you use in your case-finding question.

(5) The carer and the neighbours of a case identify other cases.

(6) You find quite a few cases in the first 30 minutes.

(7) You visit all parts of the community (if not there is a good reason why not ... the mansions are there)

You get to "feel" it when you have a working case-finding method. I do not reccomend this for the first time round and I would, time permitting, do a capture-recapture validation.

I think we need to extend your question a bit. The capture-recapture study is usually the end-point of a process. I would take two or three days working out my case-finding procedure. This will be working out what to say, what to ask, who to ask, who to recruit as key-informants, &c. often your first stab will not work well (applying the qualitative test above). Only when you have something that you believe will work do you attempt a capture-recapture study. On one occasion I have had to start again after the capture-recapture study and so a social network analysis. This was in an IDP population and I was hoping I could get away with something simple.

Now to your question ... I will break it up ...

First ... I have to say "Weight-for-height in a CMAM program? How quaint!". I subscribe to the VALID doctrine (or is it that VALID subscribe to my doctrine?) that a CMAM program using W/H is a contradition in terms since W/H is a coverage killer. This is, however, a different issue.

Fifty children in one day is very few children. If you really must do W/H then a two stage screen is good enough for our purposes here. Take MUAC and if MUAC is below (say) 125 mm then measure W/H. You can decide the MUAC threshold using survey data. There is a tool to help with this at:

http://www.brixtonhealth.com/Screen100Setup.exe

Also see SC-UK's Emergency Nutrition Assessment manual.

Using MUAC with a reasonable threshold will increase throughput by a factor of eight or ten (i.e. 400 - 500 kids per day). Remember that you do not need to find every case so you can set the MUAC threshold quite low.

Let us not forget that we already have one list. That is the list of active cases already in the program coming from the study communities (this can be an empty set fro some communities). You only need to use your method to find cases and see if you find cases and compare the two lists. There is no need for a list to be complete (in ecology we never think that we have caught every fish in the sea ... we do capture-recapture simply because a complete list cannot be made).

Sample size ...

We have :

N = ((M + 1) * (C + 1)) / (R + 1)

where:

M + C > N (i.e. the lists must have some overlap)

and:

R > 7

As with all sample size calulcations we need to make some guesses about what we are trying to find out. In this case we need to guess at two case-finding sensitivities. Let us assume that these are 50% and 80% respectively.

We also need to factor in the study population. Let us assume:

Number of communities in study = 8

Average community population = 300

Total population = 300 * 8 = 2400

Under 5 years population = 2400 * 0.2 = 480

Number of SAM cases assuming 1% prevalence = 480 * 0.01 = 4.8 (round up to 5).

With this study design we can expect:

M = 5 * 0.5 = 2.5 (round up to 3)

C = 5 * 0.8 = 4

M + C = 3 + 4 = 7 (greater than 5 so OK)

R = (3 * 4) / 5 = 2.4 (round down to 2, NOT greater than 7 so NOT OK)

In this case we need to increase the number of sampled communities so that we meet the constraint:

R > 7

We should have known this from the start since there are only 5 cases and this means R cannot be greater than 5 (so it cannot be greater than 7).

It is probably easiest to do the calculations in a spreadsheet (set up the calculations and alter N until the constraints are met). In this example we need 19 cases in the population. For there to be 19 cases at 1% orevalence we need there to be 19 * 100 = 1900 children in the study population. To get this number of children we will need to sample (1900 * 5) / 300 = 32 communities. This might be a very expensive study! You'd probbaly want to go with a smaller sample size and recognise that the study will probably underestimate sensitivity.

I would spend no more than two or three days on the capture-recapture study. In fact, I would probably do a SQUEAC coverage assesment.

You have to remember that the CSAS survey method and the capture-recapture method were designed primarily as research tools while CTC was being developed and spending a lot of time on the capture-recapature study was not a big issue. I think you will be safe with a small "quick and dirty" study provided you have put the effort into getting the case-finding procedure right.

Anonymous 303

Normal user

19 Nov 2009, 18:04

Dear Marc,

I am very impressed by your dedication to this forum, and would like to thank you for this exhaustive and clear reply.

Actually I had thought about using MUAC as pre-screen tool but was not sure about the applicability in capture-recapture study. I guess that is the best solution to save time and resources.

The survey is in a quite large area, with small villages and very low population density and low (0,9%) SAM rate. The 50 W/H measures per day was a rough guesstimate based on this context and not considering the MUAC pre-screening.

We would like also to assess the impact of the NGOs nutritional program by comparing quadrants where they support the nutritional centres with those not supported. With all the limitation of the case, I think that the SQUEAC method would be less sensible/sensitive in detecting inter-quadrants differences.

Thank you again and keep up the excellent work.

Mark Myatt

Consultant Epideomiologist

Frequent user

20 Nov 2009, 10:15

I am happy to help.

There should be no problem with using MUAC in this case. Since the lists do not need to be exhaustive you will be able to use a low MUAC threshold. This should not be so low as to exclude many W/H cases but does not need to be as high as (e.g.) 135 mm that you might use for standard screening activities. This should save a lot of time and effort.

If you have many small villages then you will need to sample from many villages. There is no strong requirment for a clever sampling strategy. I would select a cloud of villages and case-find in all of them. This will reduce travel time and, hence, vehicle and fuel costs.

For your application, the CSAS approach is probably best. The method has many strengths. Its main weakness is that it is resoitce intensive. That said, SQUEAC is resource intensive when you do it for the first time in any program. Savings for SQUEAC are considerable only in the second and subsequent rounds. This makes it ideal as an "audit" tool. We have aded new tools to SQUEAC which might be useful to teh problem you describe but, as yet, these are not very well documented and the tools required atre still in testing. We are currently well advances in seeking funding for further SQUEAC development.

Good luck with the survey. Let me know how you get on. If you need any further help do not hesitate to contact me via this forum or directly.

Mark Myatt

Consultant Epideomiologist

Frequent user

22 Sep 2015, 09:36

Did you ever do the capture-recapture study?

If you did, could you share the results (just the three numbers ... method 1, method 2, both methods)?

Back to top

» Post a reply