Skip to main content

A minimum set of nutrition indicators for agriculture surveys: attempting to bridge the agriculture-nutrition data gap

Katie Ricketts is the Program/Research Manager for Tata-Cornell Agriculture and Nutrition Initiative (TCi), a long-term research initiative based at Cornell University. 

For decades, agricultural surveys have focused on tracking household income and employment patterns, food supply and food prices, farm management and agronomy practices. Often, this has given economists and others a good sense of what and why certain types of foods are available or affordable to rural communities. However, understanding how human
nutrition status has changed in response to agricultural changes or
interventions has largely remained a mystery. Rarely is nutritional data
collected alongside mainstream, large-scale and long-term agriculture
datasets like, for example, those undertaken by the World Bank in Africa or the International Center for Research in the Semi Arid Tropics (ICRISAT) in India.
A better understanding of the links between agriculture and nutrition
will require new thinking around how agriculture and nutrition surveys
can collect the same, critical data needed for meaningful comparison and
analysis of nutritional trends.

Filling the agriculture-nutrition data void: enter the MNDA project by TCi

The Minimum Nutrition Dataset for Agriculture (MNDA) is a Tata-Cornell Agriculture and Nutrition Initiative
project looking to develop and agree upon the most essential metrics
for determining nutritional status. Along with interns and other TCi
staff, I am working to take the existing metrics and develop a short 1-2
page, generalizable survey that can easily be included in longitudinal
agriculture surveys. Such metrics include proxies and indicators that
can identify:

  • Dietary diversity,
  • Anthropometric and clinical indicators,
  • Biochemical markers,
  • Metrics around intra-household allocation, and
  • Metrics around early childhood care (e.g., breastfeeding).

Beginning with Dietary Diversity

In December 2013 we started the process of building the MNDA by
gathering together a global group of experts from nutrition, economics,
sociology, and natural resource management together at Cornell
University. During this workshop, we outlined the basic components and indicator categories (mentioned above) necessary for measuring human nutrition.

In early 2014 the dietary diversity module was selected as the first
module to develop and pilot test. Dietary diversity information can be
difficult to gather (recall periods and cutoff amounts are problematic
for determining what and ‘how much counts’), and expensive, and time
consuming to undertake. However, assessment of household and
individuals’ access to a diverse diet is a key metric that is useful for
identifying how food insecurity can contribute to malnutrition. Lack of dietary diversity
is a particularly severe problem among poor populations in the
developing world as starchy staples and grains—with little or no animal
products and few fresh fruits and vegetables—dominate diets (see a review on diets in the developing world here and here).
The MNDA dietary diversity module was designed to rapidly assess
dietary diversity, yield similar results to an intensive approach, and
require less than 30 minutes of a respondent’s time.

Pilot testing the first MNDA module: preliminary results

After a winter/springtime filled with designing the survey instrument, I took off in early summer with five TCi interns to India to pilot and validate the instrument in real field conditions. Prabhu Pingali,
our TCi Director, later joined us. Two villages in Telangana and
Maharashtra were surveyed and over 140 households participated (see
reflections from our interns who lived in our sample villages here and here).
Our participants, women between 18-45 years old, were randomly drawn
from a larger group that had previously participated in ICRISAT’s
intensive survey (which had included a dietary diversity component). Our
MNDA module asked women respondents to recall what they ate over the
past three days.

image

Amrita
Rao (far right) a 2014 TCi summer intern and a fellow ICRISAT
investigator (center) interview a woman in Telangana about her diet
during the recent pilot. Photo by Christian Di-Rado Owens, 2014 TCi
intern.

To obtain a woman’s dietary diversity score, the individual foods she
ate were collapsed into 9 possible food groups, yielding a score
between 1-9. All foods eaten by the household were collapsed into a
total of 12 possible food groups, yielding a score between 1-12. We
followed the individual and household food groups and scoring method
outlined in Food and Agriculture (FAO)/FANTA Dietary Diversity Guidelines.
The same FAO/FANTA food groups were used to derive the dietary
diversity scores yielded from the the ICRISAT intensive nutrition
survey. If the MNDA dietary diversity module captured the same dietary
information as that captured by the ICRISAT intensive survey—in less
time and with fewer questions—amongst the same group of respondents, we
could then conclude that the MNDA dietary diversity module offered
operational advantages and should be included in the larger MNDA metrics
set.

Findings from the 2014 pilot showed that the overall distribution of
MNDA dietary diversity scores, including the sample mean of the women
who participated, were not significantly different from the intensive
ICRISAT survey (figure below). Broken down by body-mass index groups
(BMI) that were categorized into low-BMI (underweight), average BMI
(normal) and high-BMI (overweight) sub-groups, no significant
differences emerged amongst the groups regarding distribution of scores
collected. In sum, the MNDA was able to collect similar dietary
diversity score information as the intensive nutrition survey in less
time (in an average of 27 minutes) and with fewer questions.

image

*MNDA=Scores
derived from the Minimum Nutrition Dataset dietary diversity module
(July-August 2014) **INS=Scores derived from the ICRSIAT intensive
nutrition survey (August-September 2013).

In addition to capturing similar dietary diversity scores effectively
and efficiently, the MNDA captures additional information useful for
contextualizing diets and better understanding how access might be
shaped by the overall food system. This additional information includes
atypical eating patterns (days where women fasted or went to a special
event), eating that occurred outside the home and identification of
where food items were procured. In fact, we found that in our pilot
testing, atypical eating patterns (in our case, fasting) were
confounding our results. Once I controlled for fasting we started to see
a strong connection between BMI and dietary diversity scores. Variables
such as atypical eating behaviors, eating outside the home, and food
procurement have been listed by the FAO dietary diversity collection guidelines as important factors to consider when evaluating diets through dietary diversity scores.

Moving ahead: building and piloting other MNDA modules

In collaboration with scientists and researchers at ICRISAT,
our interns did a great job on the data collection front for our pilot.
Right now I am continuing to analyze and improve the survey results,
given the positive preliminary findings we have thus far. Keep a lookout
for an upcoming paper Prabhu Pingali and I have with the New York Academy of Sciences
on the purpose of the MNDA as well as a more detailed analysis of the
work undertaken this past summer in practitioner and academic
publications.