Menu ENN Search
Language: English Français

Case-control study or something different?

This question was posted the Assessment forum area and has 17 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

3 Mar 2011, 16:57

Your help

We are thinking of conducting a study looking at the impact of our food security interventions (cash transfer, small scale agriculture ) on under five nutrition status. Because of the various livelihoods zones in the regions where we operate we wanted to do a one year prospective case control study (aware of the limitations) following children whose families is benefiting from our services. I wanted to know if there is anybody out there who has done such type of studies or those who can help up in defining the study protocol, and advising us in such way that we can manage to analyze all the possible constraints before engaging ourselves into this type of research

Thank your directions

Mark Myatt

Consultant Epideomiologist

Frequent user

3 Mar 2011, 19:01

I think that you may be confused regarding study designs. Case-control designs select subjects by outcome and investigate prior exposure. They are all retrospective. I think you intend to select subjects by exposure and then follow-up for a fixed period to investigate outcomes. This is a cohort design. These can be difficult to run. I suggest that you contact an epidemiologist or statistician to help with design and consider whether you will need to employ a study manager.

Kevin Sullivan

Professor

Normal user

3 Mar 2011, 19:20

There are conflicting definitions of "retrospective" and "prospective" in the epidemiologic literature. My preference for these terms is based on when the investigation is initiated vs. the occurrence of the outcome (see Kleinbaum et al., ActivEpi). If at the initiation of the investigation the outcomes will occur in the future, this is a prospective study. If at the initiation of the study the outcomes have already occurred, this is a retrospective study. If a study has elements of both prospective and retrospective then this is refered to as a "mixed" design.

What separates an unmatched case-control study from a cohort is that in the former, first those with the outcome are identified (the "cases") and a comparison group (the "controls") are selected, then previous exposures are investigated (called a"backward" design because you start with knowledge of the outcome status and inquire about previous exposures).

A cohort study first determines exposure status (exposed vs. unexposed), then determines at a later time who developed the outcome and who did not (called a "forward" design because exposure is determined first followed by the outcome).

Using these definitions and applied to study designs, it is possible to have:

* A prospective case-control study (at the initiation of the investigation, the cases will occur sometime in the future)
* A retrospective case-control study (at the initiation of the investigation the cases have already occurred)
* A prospective cohort study
* A retrospective cohort study (performed frequently for occupational exposures)

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

3 Mar 2011, 20:09

Dear Mark,

Thank you very much for your quick reply... to my question and as I said I am still confused when it come to type of studies but I put the above for suggestion and I believe you are right in suggesting me to look for an epidemiologist/specialist to decide on this type of study.

But let me add more on what I have written before...Why I did not talk about a cohort study? It is mainly because I wanted to follow two groups that will be different only from one of the support they are benefiting from the project (assumption is made that nutrition intervention will be available in both communities ...). Let me put it in this way. We are operating in two livelihood zones, one with potentialities for agriculture productions (for small gardening mostly) and the other is not. My understanding is that the second region is not viable for practicing agriculture but in these two regions we are providing the most vulnerable households with cash to improve both food consumption and sources of income.

At the end of this project we want to be able to answer few questions ... such as the possible impact if any of the intervention on children/households nutrition status but also be able to look at the differences between the two livelihoods zones following two or three groups: a) children/households that received cash in zone with agriculture potentiality b) household/children that received cash in zones with no agriculture potential and c) if needed households in agriculture zone who did not receive cash. In all of these regions we have a nutrition program that is being implemented so we would like to couple all this with a follow up at both health centres (looking at admission trends from the selected villages) couple with anthropometric measurements and HEA in the villages from the selected households ...

Sam Oluka

Nutritionist / Food Scientist

Normal user

4 Mar 2011, 08:33

Jeff, may I share with you regarding the above;

1) indeed you are planning to conduct a comparative cohort study (I perceive on nutrition status of pre-school children) among beneficiary and non-beneficiary households of your food security interventions.

2) It is true you may need to stratify the study area by zone (e.g., agro-ecology-which I believe is by far, the largest determinant of livelihood strategy).

3) In agreement this is a cohort study,
i) You will at inception enter the community and recruit households (with under-5 children) before allocating intervention. Note the need for community entry dynamics)
ii) Obtain baseline data and food security status for the two groups using a suitable tool for under-five children. I suggest you could a tool that captures the entire household food security as well!
iii) After one yr of intervention, you then collect data again to assess the effectiveness of intervention for the two groups of households (with and with no intervention).

To Mark, would including households with no under-five children in such studies divulge any useful information save for funds?

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

4 Mar 2011, 12:01

Dear Oluka,

Thank you very much for your directions. I believe you have responded to many of my preoccupations already and as said you are really right about how i have to define the livelihoods zones.... i agree with both of you about the type of study.....

Re: tool, I know there are a lot of tools but any suggestion you may have will still guide us so that during the decision phase we know exactly what we are looking for. We would like to collect anthropometric data such as weight, height, MUAC and for the households food economy we wanted to use the IHEA (individual household economy assessment).

Another clarification and why we focus in households with children below five it is mainly because those are the target groups. We will focus on poor households and mostly those with children below the age of 5 years....

Sam Oluka

Nutritionist / Food Scientist

Normal user

4 Mar 2011, 12:44

My pleasure Jeff.

En-net is more than a learning but sharing platform. But like mentioned earlier, try to seek services of a statistician in this. I get you links to tools of choice-most on the FANTA website.

May I request Mark and others to help here with more recent and recommended tools

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

4 Mar 2011, 19:19

Dear Kevin,

Thank you very much for putting these clear details I'm my regarding the various types of studies. I don't know why I haven't seen your message until just few minutes ago and you have sent it since yesterday....

I would like to know your opinion once more about the idea i have developed further regarding the orientation of this evaluation

Thank you in advance

Mark Myatt

Consultant Epideomiologist

Frequent user

5 Mar 2011, 10:02

Just picking up on some issues ...

If the outcome of interest is anthropometry in children then HHs without children will add no data ... but ... HHs without children can, by (e.g.) adoption or birth, become a HHs with children. You'll have to decide what to do with these.

Tools : I assume you mean indicators. For a comparison between livelihood zones you should avoid W/H as this is strongly influenced by all sorts of genetic and environmental factors which determine body shape.

Age-effects : If you recruit children at the start for follow-up then you have to be careful of secular trend. Wasting is a disease with a non-uniform age-specific prevalence. Prevalence will drop (indicator values will rise) as the cohort ages. This is a problem with MUAC which increases with increasing age but is also a problem with W/H.

Design : I think you need to think hard about this. It seems to me that the unit of intervention is the livelihood zone so your sample size is n = 2. Confounding will be a huge problem. I think you may need to look at a community randomised trial with (e.g.) 100 or so villages in each arm. These are difficult designs (they have so many pratfalls and have numerous bias problems) and should, IMO, not be attempted without an experienced statistician.

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

5 Mar 2011, 11:06

Dear Mark,

Thank you for the above direction and i agree with you that we need to think very carefully ...It is great to know in advance all this challenging factors that need to be addressed first...yes we will need to call for a service of qualified statistician but before we involve ourselves fully in this we want to be clear about what we want or can achieve through the exercise then the design can be left to a statistician/epidemiologist... one clarification about the operation zone... I do not think we are operating in more that 75 villages in this specific region with our food security and livelihoods activities... is there any further suggestion based on this number of villages?

Thank you for you all

Mark Myatt

Consultant Epideomiologist

Frequent user

5 Mar 2011, 16:59

Sample size determination require more information ... population sizes, expected magnitude of effects in both arms, details of the study design, &c. Most of this is unknown at this stage. All I can say is that randomised community trials tend to require large numbers of communities in each arm and that small effects need large sample sizes. This could end up being very expensive.

I suggest that you take some time to work out what you want the study to test and consult a statistician or epidemiologist. It may turn out that you need something simpler and quicker. It may also turn out that a more qualitative analysis would serve your needs. If you do go for a randomised community trial than be sure to take advice from someone who has done something similar as these sorts of trials are very easy to get wrong.

Sorry not to be of more help.

Sonya LeJeune

Normal user

5 Mar 2011, 19:25

hi
I just wanted to support Mark's suggestions that a well thought out qualitative assessment may give you some of the answers that you need. If the study design becomes too complex then it could be difficult to interpret the results. And, not wishing to complicate things, but were you planning to consider socio-eocnomic grouping and seasonality in addition to livelihoods?
good luck with it however and do let us know how you get on!

Jeff Matenda

Nutrition Coordinator, IRC

Normal user

6 Mar 2011, 12:00

Hi Dear Mark and Sonya...thank you for your feedback in this and again i agree with you about the study design and all the rest...but looking at what is suggested by Sonya ..maybe it will be easy to do a simple qualitative analysis at the end as we are not thinking of doing a very big study at this point...we can share the population figures in the targeted area with you... and that of the programme participant (one socio economic group only will be consider.....poor households with particular attention to those with children below 5 years).... Simply we wanted to look at the outcome (change in food diversity of the household and/or change in nutritional status which i beleive it is difficult to measure). Yes Sonya we will look at seasonality as the program will be initiated during the lean season (dry season) and that why we were trying to focus on small scale agriculture as we know in some villages this will never be an option as they do not have such opportunity but they can invest in other things such as income generating activities with the cash they may receive...that is exactly what we would like to achieve and we wanted to shape it further and see if it will make sense...or bring any added value...

We will be looking at the service of an experienced epidemiologist/statistician in the next phases...

But we want it to be very small but with outcome that open ways for further studies....

Once more thank you very much..........

Mark Myatt

Consultant Epideomiologist

Frequent user

6 Mar 2011, 12:34

I think an initial more qualitative investigation is a good approach as it will yield an indication of whether the intervention is having the desired effect and, if it is, give you (a) Some idea of the magnitude of effect (essential for informing a fuller quantitative study) and (b) identify boosters and barriers to coverage and uptake that will enable you to improve the intervention (i.e. you will really get to know how your program operates(.

In addition you could use this first investigation to design and test instruments (e.g. food frequency questionnaires, pictorial diaries, diversity scores, &c.) that may be useful in a more quantitative study.

Esther Busquet

Nutrition Adviser - Save the Children UK

Normal user

22 Mar 2011, 13:05

Hi Jeff,

I think the research you want to do is very interesting, and will be very useful for many of us! I would like to react on the fact that you plan to use HEA (or IHEA) to look into the household's economic situation. However, if you want to see if the household can afford a nutritious diet, the HEA might not give you sufficient information. The HEA only looks into affordability of energy requirements within households, while it is known that the costs of foods providing other nutrients (especially iron, zinc and vitamin A) are higher, and thus the first foods excluded from the diet..if you want to look at families being able to afford a nutritious diet, you should consider using another tool, such as the Cost of Diet tool of Save the Children UK. Please let me know if you would like to know more about this tool...(I am focal point for use of this tool, so should be able to give you more information if needed).
Esther

Mark Myatt

Consultant Epideomiologist

Frequent user

22 Mar 2011, 14:23

Esther,

Can you post a link to a manual or something on this tool?

Mark

Esther Busquet

Nutrition Adviser - Save the Children UK

Normal user

22 Mar 2011, 17:07

Mark,

Sorry, I meant to include the link for some further reading on the Cost of Diet tool in my reaction, so here it is:
http://www.savethechildren.org.uk/en/54_9288.htm

A good impression of the joint outcomes of an HEA and CoD assessment is the following example from DRC:
http://www.savethechildren.org.uk/en/54_13367.htm

Esther

Mark Myatt

Consultant Epideomiologist

Frequent user

23 Mar 2011, 10:10

Thank you. Most useful.

Back to top

» Post a reply