Menu ENN Search
Language: English Français

Is capture-recapture method for CSAS coverage survey ethical?

This question was posted the Assessment and Surveillance forum area and has 6 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 303

Normal user

18 Nov 2009, 13:20

I am planning a coverage survey for a nutritional programme, and I am supposed to perform a capture-recapture study to assess the sensitivity of the active case finding.

I still have some doubts about the validity of this method (how confident can we be that the sensitivity calculated with the capture-recapture can be extrapolated to the entire area under survey? Interviewer bias: I suppose local staff will be more motivated knowing it's a pre-survey test, etc.) but I am mainly concerned about the delay in referral of cases identified in the first stage, i.e. the capture. The CTC field manual recommends not to do the two stages the same day.

Don't you think this is unethical in the case of severely malnourished children, which should be immediately referred to the proper service?

Mark Myatt

Frequent user

18 Nov 2009, 16:54

The purpose of the capture-recapture study is to check whether the method you use to find cases is likely to find all (or nearly all) cases of SAM in sampled communities. It is important that your case-finding procedure finds all (or nearly all) cases in sampled communities. If you use a case-finding method with a low sensitivity then there is a risk that your survey result will be biased. Since such a survey is likely to find the "easy to find" cases and these cases are likely to already be in the program, the likely effect of a low sensitivity case-finding procedure is for the survey to overestimate coverage.

I have done a number of CSAS surveys and, in some, the first attempt at a case-finding procedure has lacked sensitivity. This has been a particular problem in IDP populations.

It is good to be sceptical about the applicability of the sensitivity estimate to the entire survey area. You have to apply some common sense here. If your survey area is not highly variable (e.g. you have an area with one language and one ethnic group and the selected key informant(s) are ubiquitous) then you can be reasonably certain that the sensitivity will be similar in all communities. If you have a varied survey area then you will have to be careful to (e.g.) use local language terms. Also, you have to make sure that your teams are careful and well-trained. I usually involve team supervisors in the development and testing of case-finding procedures. Motivation is, in my experience, more about management, training, and remuneration than survey method.

One finding from CSAS surveys is that the development and testing of the case-finding procedure can inform useful changes to the program being assessed. In two CSAS surveys that I have done (in DRC and Niger) there was a big mismatch between program messages and local perceptions of SAM. In both cases the program was concentrating on "food security" issues whilst the population tended to define wasting in terms of infection. In both settings we found no cases when asking about "malnutrition" but many cases when we added the term "recovering from a recent illness especially diarrhoea and / or fever" (in Niger the term for "chronic diarrhoea" was most useful). In both programs, the change of emphasis regarding program messages informed by the development and testing of the survey case-finding procedure resulted in a considerable increase in patient numbers.

Your ethical concerns are correct. Most CMAM programs do not admit on a daily basis so any case referred on one day will not be enrolled in the program on the next day (probably not for the next few days) and you can time your first test to exploit this. Even if you cannot do this there should be no problem as cases on the first day will likely still be cases on the second day (provided the days are only a few days apart). Just make sure not to test / survey on CMAM clinic days. The exception to this is for children with complications who should be referred for stabilisation immediately (I would urge that you transport them yourself). Such cases will be quite rare.

Capture-recapture studies are notoriously difficult to do well and even then are done well are usually subject to a number of biases. You can find a more in-depth treatment of capture-recapture studies than that in the CTC manual at:

A note on sample sizes for CSAS surveys can be found at:

You might also find this:

useful. Software for analysing and mapping CSAS survey data is available at:

The CSAS method is good but can be expensive. A newer method (SQUEAC) has been developed by VALID. A ( slightly dated) description of SQUEAC can be found at:

I hope that this note addresses some of your concerns.

Anonymous 303

Normal user

18 Nov 2009, 18:22

Thank you for your prompt and very exhaustive reply.

I strongly agree with you about the need to improve sensitivity in estimating coverage.
And we also experienced the different perception of malnutrition between the community and the NGO's staff/health workers.

I did not mention in my question that my ethical concern is referred to those programs which admit severe cases and give the first plumpynut ration every day, while distribution to those in the programme is done weekly. Even if we suppose a one/two day delay in the referral for the sake of the study, would that be acceptable? Please tell me if this question is out of the scope of this forum, and feel free to cancel this reply if you think so. In any case I will look for an ethical advisor at my NGO.

Thank you again for your help.

Mark Myatt

Frequent user

18 Nov 2009, 20:07

Perhaps I misunderstand but I see no great problem with the case that you describe since, in the capture-recapture study, we are only interested in case-status. We are not interested, at this time, in coverage status. We are only interested in case and coverage status simultaneously when we do the main survey. A "problem" might occur if the case is admitted to an inpatient unit. The usual way around this is to allow the case to count if the key-informant(s) identify the child as malnourished or in a program treating malnutrition. There may be some bias introduced into the capture-recapture study but I think this preferable to the alternative of delaying treatment. Also, I'm sure your realise, all cases found in the main survey should be referred immediately.

Anonymous 303

Normal user

19 Nov 2009, 09:10

Dear Mark thank you again for your time and patience, I'll try and reformulate my question for the sake of clarity, although you have answered to my question in your last post.

A CTC program is operative in the area and new severe non-complicated cases can be admitted every day and receive, the same day, a plumpynut ration and systematic medical treatment.

A coverage survey, including a capture/recapture study, is carried out. In the capture phase we detect say 10 severely malnourished children. We should refer the 10 children to the OTP to get medical and nutritional care the same day. Yet, the recapture phase is due the day after.

Therefore, if the malnourished children are referred as soon as they are identified during the capture phase, some of them will be absent the day of the recapture because they are at the OTP. And there might be other violation to the assumption of independence. If I understand well, due to these issues, in this Valid paper is said that "we have [...] issued referral slips only after the second sample has been taken so as to avoid introducing negative dependence."

In conclusion the conduit à tenir is to refer the cases to the OTP the same day of the capture. If during the recapture phase, the key informants identify some of the referred children, we consider them as detected by both the methods.

What confused me was the Valid paper statement, but it was likely referred to a program admitting new cases once per week.

Thank you again.

Mark Myatt

Frequent user

19 Nov 2009, 12:37

You are correct.

The program admitted on a weekly basis. Typically there day one and day two were only a day or two apart (sometimes we needed to avoid market days). This should not have introduced a delay in treatment delivery. In the example cited in the paper lists of cases with identifying and locating data were made on each day. This is needed to make the "overlap" list and to avoid the problem of common names ... an English example ... it is conceivable that two cases named "John Smith" might be found in the same community so we need location data (also mothers name and father's name) to differentiate them. On the second day, all cases found were referred as they were found. At the end of the day the two lists were compared and cases found on day one but not on day two were visited and referred (a copy of the first day's list was provided in a sealed envelope for this purpose). The exception to this was complicated SAM cases: Any child with complications was referred for inpatient stabilisation on the same day and, if the family agreed, transported to the stabilisation centre by the study team on that day.

In your case, I would NOT subvert the admissions policy (it is an excellent policy). You might want to consider introducing a delay between the two days. This will allow time for patients to get to the OTP site and be back in their community for the second study day. This will removes some negative dependance. Some children may be in stabilisation but that cannot be avoided and any attempt to delay admisssion into stabilisation would be most unethical. It would be sensible to avoid the weekly OTP day for data collection (and in the subsequent CSAS survey).

As the paper describes, capture-recapture studies seldom meet their assumptions even in the field of ecology where the method originated (e.g. animals may become shy of traps having been caught once introducing negative dependence and the closed population assumption is frequently violated except, perhaps, in study areas like ponds and other relatively isolated ecosystems). The "trick" is to identify potential biases and account for them when you interpret the sensitivity estimate (see Table 1 and Example 3 in the paper).

Mark Myatt

Frequent user

22 Sep 2015, 09:18

Would it be possible for you to share the results of you capture-recapture study?

Back to top

» Post a reply