Menu ENN Search
Language: English Français

Causality Analysis

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

» Post a reply

Anonymous 249

Normal user

16 May 2012, 12:13

Does someone know where I can get resources or information on experiences and standard design for conducting causal analysis of malnutrition?

Alexandra Rutishauser-Perera

International Medical Corps

Normal user

16 May 2012, 13:05

You might want to conduct Action Contre la faim (ACF) as they have quite a lot of expertise in this...

Mark Myatt

Frequent user

16 May 2012, 13:13

There is something by me and Mara Nyawo (UNICEF Sudan) is this edition of Field Exchange (p37). This method is being used (second time) in Sudan as i write this. It is quick and cheap (almost free if done with SQUEAC).

Rogers Wanyama

Emergency Nutrition Specialist

Normal user

16 May 2012, 15:06

Hav a look at causal analysis of causes of malnutrition in NBG conducted by ACF -US in 2011.

Check National reference material section.


Chantal Autotte Bouchard


Normal user

16 May 2012, 18:57

Hi ,

Julien his the referent for ACF about this question, you can contact ACF France the would help you probably.


Julien Chalimbaud


Normal user

16 May 2012, 20:04

ACF is regularly conducting causal analysis and I am currently conducting a project to develop field technical guide for such assessment. You can contact me at and I can send you our last version. We will finalise the work by september.

Anonymous 249

Normal user

17 May 2012, 06:56

This is very helpful. Thanks to everyone


Frequent user

17 May 2012, 09:22

A word of caution about the term « causal analysis ». One should be aware that this is just a way to look around and see what happens in the community which might explain the prevalence of malnutrition. So this is not determining the cause of the malnutrition the hard way, which is to remove the factors which are supposed to be involved and to see malnutrition disappear.

In the example quote in ENN Field exchange for instance p 37, the analysis suggests that promotion of ORS and hand washing would prevent SAM. I am not aware of any setting where this has been done successfully. This even failed in research settings.

See for example:

A major difficulty is to disentangle causes from consequences. In this example, one can argue that SAM children are more susceptible to diarrhoea and that this explains the association, without a causal effect.

Malnutrition always occurs in a contest of extreme poverty and it disappears with its elimination. Beyond that, difficult to identify a single biological mechanism as a cause. A sobering exercise is to ask scientists who have been working the hard way on the causes of malnutrition (especially stunting) for the last 30 y, usually they confess their ignorance of the most important cause, or those who think they have found it usually have diverging opinions on the topic or have a very limited evidence base.

So this is definitely a useful orientation tool, but one should be aware of its limitations.

Mark Myatt

Frequent user

17 May 2012, 09:57

Very good points André. I agree with all of this.

Causality can be a slippery concept ... We could have diarrhoea causing wasting (reduced intake, reduced uptake, increased energy requirement with fever) which increases susceptibility to diarrhoea (more frequent / longer episodes) which ... and so-on ... all the way to SAM and beyond. The cycle might start, however with reduced intake due to scarcity leading to wasting which increases susceptibility to diarrhoea (more frequent / longer episodes) which ... and so-on ... all the way to SAM and beyond. It is a "chicken and egg" situation. Nothing new here ... it is the "infection - nutrition cycle" model.

All we have with methods like the one described in the FEX article is association and association is not causality. It is only indicative of causality. I think the utility of these approaches is that they can, as André remarks, orient you in the "SAM landscape" so you know what is probably important and what is probably not important.


Frequent user

17 May 2012, 10:12

Thanks Mark for your comment.

There is also a risk with this approach of giving too much importance to factors that you can easily measure. For instance, it has been argued that aflatoxin exposure leads to stunting. Same can be argued for anti nutrients leading to poor micronutrient availability. All this is difficult to measure, and if it is not in your assessment, or if you cannot assess this properly, you will never pick it up, even if it is THE major cause of malnutrition in your settings.

Having said that, if you find a high prevalence of diarrhoea in your causal analysis, better quickly implement a programme to address this as it is a major health problem, independently of its impact on malnutrition levels.

Mark Myatt

Frequent user

17 May 2012, 11:19

WRT ease of measurement ... This is a risk.

Since the method described in the FEX was intended to be a quick and cheap "plug-in" to SQUEAC we did use off-the-shelf questionnaire components. The rationale for this was that it was quick and cheap but also because many have been validated in many settings as being useful and unbiased when compared to more expensive methods. When we could not find a standard set to cover our hypotheses we had to design one (or "make one up"). We did test these in role play and with carers of SAM children but testing was limited. So for some risk factors / markers we will very likely have had less than optimum question sets which may have introduced bias. That said, the design and testing was as thorough as would be done in a typical service epidemiology task such as an outbreak investigation in the UK.

As for anti-nutrients, I did a similar (i.e. matched case-control study) exercise with SC-UK in 2005. The summary results can be found in this FEX article. This causal analysis found some evidence for anti-nutrient effects. We did, of course, have to ask the appropriate question and this arose from the semi-quantitative first stage. The message here is that these methods require some intelligence as they rely on more than dumb replication of a standard questionnaire.

I agree ... no harm can be done in intervening against diarrhoea in the settings that the EN-NET community work in.

Tom Davis

Senior Specialist for Social & Behavioral Change

Normal user

24 May 2012, 00:45

We (in Food for the Hungry) have developed a methodology which is based on a case-control design which we have found helpful -- Local Determinants of Malnutrition. You can read the report on the methodology and our findings in several countries here:

While the methodology suffers from some of the same limitations as other causal analysis (e.g., what is a cause and what is an effect?), it has at least pointed us to some useful things to promote which are associated with better nutritional status which we may have overlooked, like finishing breastfeeding on one breast before proceeding to the other breast (rather than giving the foremilk in both breasts), salting of the child's food, and high levels of association with water purification (which encouraged us to do more in that regard).


Frequent user

24 May 2012, 12:57

A word of caution again re. causal analysis. The analysis discussed in the previous post is very interesting, but I doubt you can increase growth of children by salting infant foods. This will increase the osmotic load of the diet, which may be a problem in case of diarrhoea, and in places where kwashiorkor is highly prevalent, this may favour the apparition of oedema.

Results of "causal" analysis studies should be used for general orientation and their findings should not be translated directly into recommendations. Before that, they should always be examined critically and always taken… with a pinch of salt.

Mark Myatt

Frequent user

24 May 2012, 13:31

Yes. This is probably a finding due to inadequate control for confounding in the design or the analysis. The analysis seems to have been limited to the bivariate case. If stratified analysis or logistic regression had been used then some of these associations may have been discarded. There may also have been some "negative confounding" and borderline significant associations that were discarded may have been picked up. I think that there is much that is good in this method but the analysis seems (to me) to be incomplete.

In the method developed with UNICEF Sudan, we control for confounding in the design by using age-matched neighbourhood controls (age is a "grand confounder" in most epidemiological data and it is with SAM, neighbourhood controls (incompletely) for potential spatial / cultural / familial / economic factors). We used tabulation to identify "suspect" variables in a matched analysis (using p < 0.10 as an inclusion criteria). We then used conditional logistic regression (a method for finding independent associations between risk factors and outcomes in matched data) to find independent associations by repeated backwards elimination. There are strengths and weaknesses to this approach. The main strength is that we control for confounding in both design and analysis and reject most "spurious" associations. The main weakness is that we cannot investigate risks associated with our matching criteria (since the are the same in each set of cases and controls).

André is right ... whatever method we use (and I prefer pretty standard epidemiological methods such as the case-control approach) we need to be aware of its limitations.

Otieno K Musumba

M & E , IMC Kenya

Normal user

22 Mar 2013, 07:20

Is there any Universal Methodology for Nutrition Causal Analysis?

Mark Myatt

Frequent user

22 Mar 2013, 12:25

Not as far as I know. There is a lot of “Delphi method” (expert panel) coupled with a “problem tree” approach. These sorts of approaches have a mixed track record.

I prefer more “epidemiological” approaches to this problem. See this FEX article for a description of a pilot of nesting a matched case-control study within SQUEAC from Sudan. Another article describing experiences with this method isn Chad is in preparation.

You might want to try this search of FEX articles.

Anonymous 21857

Regional Humanitarian Nutrition Advisor/ SCI

Normal user

8 Jan 2019, 16:59


I am currently looking for tools/methodologies that could be used by nutrition teams to identify drivers of malnutrition in both humanitarian and protracted/development contexts.

I found this post quite pertinent and i was wondering if there was any update in the discussions held since 2013? I am familiar with ACF's work on the Link NCA methodology but would be interested in knowing if a listing/review/analysis of existing methodologies has been done.

Thank you for the support

Lenka Blanarova

Global Coverage Advisor / Link NCA Analyst

Normal user

9 Jan 2019, 20:44


as far as I know, a good quantity of recent nutrition causal analyses have been conducted according to the Link NCA methodology. Currently, it may be the most field-oriented methodology available and has recently been upgraded, reinforcing its qualitative and quantitative data analyses. For example, the introduction of bivariate or multivariate analyses in the quantitative phase of the Link NCA helps to highlight plausible relationships between different contributing factors and outline potential pathways for wasting and stunting separately - which was not possible with the previous version of the methodology.

You can access a library of guidelines and tools on a dedicated website I would recommend contacting Link NCA Technical Advisor at ACF-F HQ as she can help to adapt the methodology to suit your needs. The methodology was originally designed for rural, homogeneous settings, but has been successfully tested for urban and volatile post conflict settings, trans border operations as well as refugee camps.

While an existing body of Link NCA studies was independently reviewed some 18 months ago, I am not aware of a comparative analysis of existing causal analysis methodologies.

Hope this helps,

Back to top

» Post a reply