Menu ENN Search
Language: English Français

Nutrition Causal Analysis - inputs for the development of a methodology

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

Hannah Mattinen

Frequent user

25 Aug 2010, 14:05

Dear friends of en-net

ACF (Action contre la Faim), together with consultants from Tufts University and IRD-Nutripass, has started a research project on Nutrition Causal Analysis (NCA), aiming at producing a standardised and replicable methodology to identify causes of undernutrition at local level. Various organisations have conducted NCAs in the past1. The methodological approaches, however, have been different, leading to inconsistency of results and difficulties in the analysis. In addition, the most comprehensive studies have also been extremely costly and time-consuming, and therefore not easily replicable.

The current project aims to specifically address the following points:

- What is the most effective NCA methodology, in terms of comprehensiveness (addressing all/most components of undernutrition), replicability (cost effectiveness) and coverage (livelihood/district level)? What limits have been identified? How can these be addressed?

- How can undernutrition and its causes be analysed in a dynamic context, taking into account current and future risks (seasonality / trends / shocks)?

- Can the main causes be ranked by order of importance to better design interventions and/or policies?

We are currently designing the research protocol and defining the draft methodology to be tested in the field. The final guidelines will be published during the first half of 2011.

We would like to collect key material and get in contact with persons and organisations who have already done NCA studies, in order to incorporate the lessons learnt into the current project. Please, feel free to share this email within your network. If you are interested in the major developments and outputs of this project, please let us know so we can actively inform you.

For more details, please contact Julien Chalimbaud, head of project at

Thanks in advance and best wishes

1 : For example:
- UNICEF / WFP / FAO: "Nutrition Security and Food Security in seven districts in NTT province, Indonesia: status, causes and recommendations for response", 2010
- Institut du Sahel : "Analyse des Causes de la malnutrition dans trois pays du Sahel: Burkina Faso; Mali et Tchad »,2008
- GTZ « The nutrition situation in Laos: determinants of malnutrition and changes after four years of intensive interventions"; 2008
- OXFAM "Analyse Causale de la malnutrition » ; Mali 2008
- IFPRI « Overcoming Child Malnutrition in Developing countries: past achievements and future choices"; 2002
- ACF has since 2001 undertaken NCA surveys in Liberia, North Darfur, Somalia, Sierra Leone, Indonesia, Mali, Niger, Mauritanie and Burkina Faso


Nutritionist Epidemiologist / FAO

Normal user

20 Sep 2010, 20:32

Hi Hannah;

I´m from Colombia, here in 2005 was conducted a case - control research on risk factors associated with undernutrition related mortality in children less than 5 yo...but it is in Spanish... or maybe I can traslate to you the main results...just tell me..

Julien Chalimbaud


Normal user

21 Sep 2010, 09:27

Dear Santiago,

I am Julien, working with Hanna and coordinating the research on causal analysis.
I would be very interested to get your case control survey from colombia.No problem for the spanish, we have some spanish fellow around that would help me on this report.

On this study, I would be interested to know:

Have you been focusing on chronic or acute ?

Did you manage to have a dynamic approach of causalities ?

Which framework have you been using ?

Mark Myatt

Frequent user

21 Sep 2010, 11:55

Just a brief observation ...

The most efficient methodology for this is a case-control study. Some careful design is required to decide on (e.g.) the correct method for selecting controls. I'd want some qualitative work (basically case-finding and history taking) to inform instrument design. Almost any experienced epidemiologist could help you with this.

I would avoid approaches that use data collected in cross-sectional (e.g. SMART) surveys as they tend to lack power. The problem is, in part, due to GAM being relatively rare. With 5% prevalence you'd need to sample 1000 kids just to find 50 cases. Low case-numbers can be balanced by large numbers of controls but only to a point. When you have large sample sizes you tend to go for less in-depth questioning. the end result is an insensitive and shallow investigation. Also ... the PPS method tends to bias the sample two larger villages and towns.

Just my tuppence.

Tamsin Walters

en-net moderator

Forum moderator

22 Sep 2010, 09:27

Am in Somalia. and would be much happy if you could translate the main findings into English language.


MMH Gani


Nutritionist Epidemiologist / FAO

Normal user

22 Sep 2010, 17:28

Hi, Julien

I sent to you a copy of this study, but here is main results:

- The malnutrition variable was wasting (low weight for age) and acute (low weight for height).
- The malnutrition cases was incident cases: when a confirmed case was detected, It was selected two alive controls from the same area of the died child.

In a logistic regression model, the variables and risks were

- Illiterate mother or caregiver (OR 14,5)
- Gestation less than 37 weeks (OR 8,2)
- Breastfeeding difficulties (OR 6,5)
- Seek medical advice (OR 0.015)
- Malnutrition diagnosis 6 months before die (OR 178,8)
- No BCG Vaccine (OR 25,0)

Model´s specificity is 96,1% and sensibility is 77,4%, which means that with this model we can identify 77, 4% of the cases.

A child that seek for medical attention (just receive medical stabilization from acute situation, but doesn´t receive malnutrition treatment), with a illiterate caregiver (OR 14.5), with pregnancy duration less than 37 weeks (OR 8.2), who submitted difficulty breastfeeding (RD 6.5), has had a malnutrition diagnosis six months before (OR 178.8), and has not had BCG vaccination (OR 25.0) has a high likelihood of dying from malnutrition.

I hope this help to you, and Im interested to keep informed about your study.

Best regards,

Mark Myatt

Frequent user

23 Sep 2010, 11:22


These are interesting findings.

I think the unmatched case-control study approach with two or three neighbourhood controls is probably the most feasible approach.

How did you do the sample size calculation for this study?

The detailed results (e.g. the observed strengths of assocation) are likely to be context-specific. One thing that strikes me as odd in these results is the absence of infectious diseases as risk factors for wasting (perhaps that is in "seek medical advice" and you had good coverage / treatment seeking for primary healthcare services).

Back to top

» Post a reply