Food insecurity, malnutrition and mortality in Maewo
and Ambae islands, Vanuatu
Andre MN Renzaho*†
Program Quality Advisor, Program Effectiveness, World Vision Australia and
Honorary Fellow, School of Health and Social Development, Deakin University, Australia
Submitted 4 April 2005: Accepted 17 November 2005
Context: This paper reports on findings from the ex-post evaluation of the Maewo
Capacity Building project in Maewo Island, Vanuatu, which was funded by World
Objectives: To examine the extent to which the infrastructure and systems left behind
by the project contributed to the improvement of household food security and health
and nutritional outcomes in Maewo Island, using Ambae Island as a comparator.
Setting: Two-stage cluster survey conducted from 6 to 20 July 2004, which included
anthropometric measures and 4.5-year retrospective mortality data collection.
Participants: A total of 406 households in Maewo comprising 1623 people and 411
households in Ambae comprising 1799 people.
Main outcome measures: Household food insecurity, crude mortality rate (CMR),
under-five mortality rate (U5MR) and malnutrition prevalence among children.
Results: The prevalence of food insecurity without hunger was estimated at 15.3%
(95% confidence interval (CI): 12.1, 19.2%) in Maewo versus 38.2% (95% CI: 33.6,
43.0%) in Ambae, while food insecurity with hunger in children did not vary by
location. After controlling for the child’s age and gender, children in Maewo had
higher weight-for-age and height-for-age Z-scores than children of the same age in
Ambae. The CMR was lower in Maewo (CMR ¼ 0.47/10000 per day, 95% CI: 0.39,
0.55) than in Ambae (CMR ¼ 0.59/10000 per day, 95% CI: 0.51, 0.67) but no
difference existed in U5MR. The major causes of death were similar in both locations,
with frequently reported causes being malaria, acute respiratory infection and
Conclusions: Project initiatives in Maewo Island have reduced the risks of mortality
and malnutrition. Using a cross-sectional ‘external control group’ design, this paper
demonstrates that it is possible to draw conclusions about project effectiveness where
baseline data are incomplete or absent. Shifting from donor-driven evaluations to
impact evaluations has greater learning value for the organisation, and greater value
when reporting back to the beneficiaries about project impact and transformational
development in their community. Public health nutritionists working in the field are
well versed in the collection and interpretation of anthropometric data for evaluation
of nutritional interventions such as emergency feeding programmes. These same
skills can be used to conduct impact evaluations, even some time after project
completion, and elucidate lessons to be learned and shared. These skills can also be
applied more widely to projects which impact on the longer-term nutritional status of
communities and their food security.
Food insecurity, defined as the inability of ‘individuals,
households and communities to acquire appropriate and
nutritious food on a regular and reliable basis, using
socially acceptable means’1, is a serious problem
threatening the lives of million of people worldwide.
The end product of food insecurity is malnutrition. Data
on the effect of malnutrition on child mortality from 53
countries indicate that 56% of all deaths among children
were attributable to acute malnutrition, but 83% of all
malnutrition-related deaths were attributable to mild-to-
moderate malnutrition2. These findings suggest that, even
though the relative risk of dying from malnutrition is
greater in children with severe malnutrition, the pro-
portion of deaths due to mild-to-moderate malnutrition,
which is the form most prevalent in cases of food
insecurity, is greater than that due to severe malnutrition.
q The Authors 2006*Corresponding author: Email email@example.com
†Correspondence address: World Vision Australia, 1 Vision Drive,
East Burwood, Victoria 3151, Australia.
Public Health Nutrition: 9(6), 798–807
Current estimates indicate that the number of people
suffering from chronic hunger is growing worldwide, with
a recent report by the Food and Agriculture Organization
of the United Nations estimating that 842 million people
were chronically hungry in 1999–2001: 798 million in the
developing world, 10 million in the industrialised
countries and 34 million in the transition countries3.
An annual trend analysis reported an increased rate of
5 million chronically undernourished people per year and
a malnutrition-related death rate in the under-fives of
6 million children per year4.
While the nature of food insecurity tends to be similar in
developing and developed countries5, the environmental
setting in which it occurs differs considerably. In African
countries, drought, armed conflicts, inadequate agricul-
tural policies and poor governance have combined to
affect household livelihood, and this has been worsened
by the current HIV/AIDS epidemic (see Table 16,7). As a
consequence of the HIVepidemic, many farms have been
left uncultivated as economically active adults continue to
die, leaving orphans and elderly people at risk of hunger
and malnutrition. Thus, it is currently estimated that some
60 to 70% of farms in sub-Saharan Africa have suffered
labour losses as a result of HIV/AIDS3. In the Pacific
region, however, the causes of food insecurity have
included high population density, limited human
resources, lack of skilled labour due to high migration
within islands and overseas, limited access to markets due
to difficulty of community outreach and inter-island
transportation, declining soil fertility, land ownership
issues and natural disasters5,8. For example, it is currently
estimated that Vanuatu has nine active volcanoes, four of
which are submarine5.
Addressing food insecurity has varied depending on
the precipitating factors, ranging from food aid
programmes in the form of humanitarian assistance
such as in the Great Lakes region of Africa (Burundi,
Uganda, Democratic Republic
Rwanda)9,10, Lesotho and Mozambique11to agricultural
trade in Vietnam, teaching children how to plant, grow
and eat nutritious food in Panama, and development-
oriented projects in Kenya3.
However, while humanitarian assistance programmes –
a subsection of international aid programmes along with
development aid programmes – have successfully put in
place policies and practice for monitoring programme
outputs12,13, studies assessing the impact of international
aid programmes remain scarce12. The lack of an impact
evaluation framework for international aid programmes
has been referred to as ‘methodological anarchy’14. Many
factors have contributed to this situation, such as the lack
of consensus in the international aid sector on what
constitutes ‘impact’; non-governmental organisations’ lack
of skill, policies or practice required for an adequate
assessment of programme impact12; confusion over what
constitutes a hierarchy of evidence; and whether hierarchy
of evidence should take precedence over methodological
appropriateness when planning an impact evaluation15,16.
In addition, evaluation of international aid programmes
continues to be driven and motivated by the interests of
funding agencies as a tool to demonstrate accountability
rather than proving programme effectiveness. In Hansch’s
words: ‘aid agencies have been conditioned through years
of reporting to donors by accounting standards that focus
on what is known, not estimated. As a result, aid agencies
avoid reporting impact altogether and instead report softer
data, such as how many people were helped, or output
data, such as how many commodities were transported
In response to this gap, World Vision Australia has
developed a number of strategies to assess the long-term
impact of its programmes. The present paper examines an
ex-post evaluation of the Maewo Capacity Building
project, examining the impact of the project five years
after completion, with particular emphasis on the impact
on the well-being of children and the mobilisation of the
wider community. More broadly, the paper attempts to
demonstrate that it is not only possible but also extremely
useful to draw conclusions about project impact, even
some time after the project completion, where baseline
data are incomplete or absent.
The Maewo Capacity Building project
From 1997 to 1999, World Vision Australia implemented a
capacity-building project in Maewo Island, Vanuatu. The
project had three components. The basic health and
sanitation component aimed to reduce the incidence of
malaria, diarrhoea, skin and respiratory infections in the
implemented to achieve this objective included: training
the community in preventive health-care, nutrition and
sanitation practices; improving access to safe water by
repairing damaged water supply systems; providing first
aid/emergency medical supplies; initiating environmen-
tally sound rubbish disposal; and promoting the use of
validated traditional medicines to treat common ailments.
The community organising component aimed to
Table 1 Prevalence of HIV among adults and the food and nutri-
tion situation in selected sub-Saharan African countries
in need of
*Data from reference 6.
†Data from reference 7.
Food insecurity, malnutrition and mortality in Vanuatu 799
strengthen community structure and create capacity to
respond to community issues and concerns using
minimum external resources and assistance. Strengthen-
ing of community structures involved establishing a core
group of leaders in at least 17 of 23 villages with specific
roles and responsibilities and a clear vision and goal for
each community. The role of the established core group of
leaders was to develop a comprehensive village plan, to
organise monthly community meetings to discuss com-
munity projects and issues, and to put in place project
committees and community participation strategies. The
livelihood component aimed at ensuring that at least 80%
of communities in Maewo Island had diversified food
sources and/or increased income. Strategies included
engaging the community in crop diversification, construc-
tion of three coconut oil processing machines, training the
community in livelihood planning and management, and
the provision of seeds, seedlings and agricultural tools.
At the end of the 3-year project, an end-of-project
evaluation was undertaken17. However, the end-of-project
evaluation was not objective-oriented as it was judged that
three years of project implementation was not enough to
yield the intended long-term results, which were
reduction in mortality and malnutrition prevalence, and
sustained household food security. Additionally, the
project did not establish minimum baseline data against
which to assess the project’s progress and long-term
impact on poverty alleviation. In the absence of baseline
data, the objective of the evaluation was to examine the
extent to which the infrastructure and systems left behind
by the project, as documented in the end-of-project
evaluation report17, contributed to improving and
sustaining household food security, and improving health
and nutritional outcomes in Maewo Island. Two questions
were posited: (1) Did the project actually improve health
and nutritional outcomes in the target population? and (2)
Could observed health and nutritional outcomes be
explained by other factors? Thus, the purpose of the
study was to examine the extent to which the
infrastructure and systems left behind by the project
contributed to the improvement of household food
security and health and nutritional outcomes in Maewo
Island, using another island (Ambae) of similar character-
istics as a control group. It was therefore hypothesised
that, as a result of the infrastructure left behind by the
Maewo Capacity Building project, Maewo would have
better health and nutrition outcomes than Ambae.
Study design, sample and procedure
The study used a cross-sectional ‘external control group’
design16. Ambae Island was used as a control group
because it was not exposed to the project, its proximity to
Maewo and was the only island with sociodemographic
and economic characteristics similar to those observed in
Maewo18. In addition, since the completion of the Maewo
Capacity Building project, the island has not been
beneficiary of any international assistance.
The sample size calculation was related to the study’s
long-term outcome variables. The primary study’s out-
come specified that, at project completion and beyond, at
least 80% of households in the target communities would
have adequate food security (diversified food sources
and/or increased income). The end-of-project evaluation
carried out when the project phased out in 1999 estimated
that 70% of the project beneficiaries had access to
diversified food sources. Both the effect of the time
elapsed since project completion and the effect of cyclone
Ivy that devastated the two islands in 25–27 February 2004
were taken into account when calculating the sample size.
Thus, it was hypothesised that, at the time of the
evaluation, 65% (instead of the projected 70%) of the
target households in Maewo would have adequate food
security. Using the usual formula for sample calculation to
compare two proportions from two populations19–21, it
was estimated that a sample size of 666 households (333 in
each island) would be sufficient to detect a difference of
15% between locations in the proportion of households
with adequate food security with 80% power, a 5%
significance level and a design effect of 2. This 15%
difference represents the difference between 65% of
households with adequate food security in the population
exposed to the project (Maewo) and 50% in the non-
exposed population (Ambae).
Data were obtained on 406 households in Maewo and
411 households in Ambae. Nine data collectors in each
island, selected from the studied communities, were
trained and collected the data between 6 and 20 July 2004
using a two-stage cluster sampling strategy10,22–24. This
was the most efficient method of recruiting households, as
villages were structured in such a way that geographic
units were not generally sufficiently well organised to
allow for systematic sampling10,24and an accurate list of
the population was non-existent.
At the first stage, tables listing villages per island were
drawn up and their respective number of households
estimated with the help of village chiefs. A mapping
exercise with community leaders and staff at World Vision
Vanuatu indicated that there were approximately 28
villages in Maewo and 25 villages in Ambae. Thirty clusters
were randomly allocated in each island using a sampling
interval such that the total number of clusters in each
village was proportional to the total number of households
in that village. This was achieved by using a sampling
interval obtained by dividing the total number of
households in each island by 30. The first cluster in each
island was determined by randomly drawing a number
within the first sampling interval using a random number
table10. The sampling interval was then added to this
number until 30 clusters were obtained in each island.
A cluster was defined to include the minimum number of
households in order to obtain the required sample size
when 30 clusters were selected. Thus each cluster had a
minimum of 11 households. A household represented an
aggregate of persons who lived together either under the
same roof or in different units in the same compound and
who ate together or shared in common the household
The second stage was a selection of households that
made up each cluster. In step 1, the team of data collectors
went to the centre of the village and spun a bottle to
determine a random direction. The data collectors went in
the direction of the bottle’s neck from the centre to the
edge of the village (villages are quite small) counting the
number of households. In step 2, a number between 1 and
the total number of households counted was chosen using
a currency note (Vatu). The number chosen represented
the first sampling point and the household chosen was
interviewed. Once the interview of the first household was
completed, the subsequent households to be interviewed
were chosen by proximity, i.e. households physically
closest to the selected household, until the required
number of households per cluster was attained. In the case
that data collectors reached the border of the village
before the number of households required for a cluster
was reached, they went to the centre of the village and
repeated steps 1 and 2. In the last step, where household
members were absent at the time of the interview, data
collectors returned there later if household members were
likely to return or otherwise the household was replaced
with the next household. Data on household food
security, anthropometry among 6–59-month-old children
and mortality data were collected by an interviewer-
administered questionnaire. The interview took place in
the participants’ home, and informed consent was
A 12-item food security scale was generated based on the
Radimer–Cornell hunger scale25and the US national
measure of food security26(Appendix). Although devel-
oped as a tool to assess food security in developed
nations25,27, the Radimer–Cornell hunger scale has been
used successfully as a vulnerability assessment and early
warning tool to assess the level of deterioration in the
quality and diversity of the diet of a given population in
emergency settings, such as during the economic crisis in
Russia28and in Indonesia29. The instrument was found be
reliable and valid27,29,30. The Radimer–Cornell hunger
scale25as a measure of food security was chosen for
several reasons: the instrument has been found to be quick
to apply at household level as it does not require the
physical audit of food stores; and food consumption-
based measures of food security such as individual
intakes, household energy acquisition or dietary diversity
are too detailed and labour-intensive, and above all
require highly skilled data collectors who can measure
accurately food quantities in terms of availability and
consumption31,32. With more than three-quarters of the
population in Maewo and Ambae estimated never to have
attended school or be educated only to primary level18,
the implementation of a food consumption-based food
security assessment methodology was not possible.
Prior to the proper implementation of data collection,
the instrument was administered to community represen-
tatives, community workers and staff at World Vision
Vanuatu for thematic analysis with respect to the clarity,
relevance of the items to food insecurity and cultural
sensitivity of the questions being asked. The trial phase
established that, because the majority (.75%) of the
population on both islands are subsistence farmers, the
focus on the lack of money as a reason for not consuming
or accessing food could lead to misleading results. It was
suggested that all questions relate to both food production
(e.g. own harvest or food bartering) and food purchasing
capacity. Also, the consultation found that breadwinners
(mainly the head of household) are fed first, meaning that
in case of dire situations they would be the last to be
affected at the expense of children. This contrary to the
current assumption in the developed world, that in cases
of food insecurity there is a managed process whereby
children are protected until the severe later stages25. Other
issues identified included the confusion related to the
cultural construction of concepts such as ‘healthy and
balanced diet’ or the effect of the planting–harvest cycle in
terms of food availability. All the identified issues were
taken into account to produce the final instrument used in
the survey, and specific questions were formulated in a
manner that could detect reduced intake in children.
Three types of malnutrition were considered in this
study: underweight measured by weight-for-age, stunting
measured by height-for-age and wasting measured by
weight-for-height. Z-scores as indicators of the nutritional
status in children were used, with wasting defined as
weight-for-height Z-score (WHZ) ,22, stunting defined
as height-for-age Z-score (HAZ) ,22 and underweight
defined as weight-for-age Z-score (WAZ) ,2233,34.
Mortality rate was computed as follows35:
Recallperiodindays £ Mid-pointpopulation
where mid-point population ¼ number of living þ 1/2
joining (live births, reunifications after an absence of at
least 2 years) during recall period 2 1/2 people leaving
(deaths, prolonged absence of at least 2 years) during the
Thus, the recall period for assessing mortality covered a
4.5-year period, i.e. from 1 January 2000 to the survey date
(6–20 July 2004), averaging 1648 days. A calendar of
events was constructed and used as a reference. All dates
Food insecurity, malnutrition and mortality in Vanuatu801
for newborns and deaths during the recall period were
recorded. After consulting community representatives, for
each death identified respondents were asked which of
the following was the most likely cause of death: malaria,
diarrhoeal diseases, acute respiratory infection, measles,
malnutrition, domestic violence, cyclone or other.
Data were entered using SPSS for Windows, version 10.0
(SPSS Inc., Chicago, IL, USA) and analysed in Stata version
7.0 (Stata Corporation, College Station, TX, USA). To
obtain the prevalence of food insecurity, interviewed
households were classified into four categories: food-
secure, food uncertainty, food insecurity without hunger,
and food insecurity with hunger in children. Households
classified as food-secure were those which reported no
food-insecure conditions, i.e. no affirmative answer to all
12 items plus those with one or two affirmative answers to
the four items depicting inadequate access to food
(questions 1, 2, 3 and 6), but not to any other items.
Food uncertainty included households with three or four
affirmative answers to items about inadequate access to
food (questions 1, 2, 3 and 6) but without interrupted
eating pattern (questions 4, 5, 7, 10 and 11) and/or
reduced food intake (questions 8, 9 and 12). Food
insecurity without hunger encompassed households with
one or more affirmative answers to the items depicting
distorted eating pattern at household level and/or among
adults, but not in children (questions 4, 5, 7, 10 and 11).
Food insecurity with hunger in children included house-
holds with one or more affirmative answers depicting
distorted eating pattern in children or reduced intake
(questions 8, 9 and 12) and at least two affirmative
responses to items indicating distorted eating patterns at
household level and/or among adults, but not to any items
related to inadequate access to food. After this classifi-
cation, the proportion of food insecurity and its 95%
confidence interval (CI) were computed.
In addition, the proportion of malnutrition and its 95%
CI were computed and mortality data were expressed per
10000 per day. The difference and its 95% CI in outcome
measures of interest were computed to compare Maewo
and Ambae. To adjust for confounding factors when
comparing Maewo and Ambae on anthropometric out-
comes, standard multiple regression was used. The
SVYSET command in Stata was used to specify clustering
within the household (i.e. to account for greater
similarities on many attributes of children from the same
household), stratification and weighting prior to analysis.
The demographic characteristics of the surveyed house-
holds are summarised in Table 2. A total of 817 households
(406 households in Maewo and 411 households in Ambae)
were surveyed, with an average size of 5.3 (95% CI: 5.1,
5.5) people per household and an average income of
4090.5 Vatu ($US 37) per month after controlling for
seasonal variation in food availability. Approximately one
in 18 households (5.5%, 95% CI: 3.3, 7.7%) was found to be
female-headed. Overall, 14.1% of the population in the
surveyed households was found to have never attended
school, while 37.2% did not complete primary school and
13.7% dropped out of school at secondary level.
Table 3 summarises data on food insecurity in both
islands. The proportion of households classified as ‘food-
secure’ was lower by 1.5 percentage points (95% CI: 27.0,
3.9%) in Maewo than in Ambae, but an examination of the
95% CI indicated that this difference was not statistically
significant. However, Ambae recorded a significantly
Table 2 Summary of demographics and anthropometric measurements: Maewo versus Ambae
Household (HH) characteristics
No. HHs visited
Average HH size
Population ,5 years of age (%)
Female-headed HHs (%)
Child (6–59 months) characteristics
Gender ratio (boys–girls)
Mean age (months)
Mean weight (kg)
Mean height (cm)
5.7 (5.4, 5.9)
15.1 (14.3, 15.9)
0.91 (0.71, 1.15)
3.0 (0.7, 5.3)
5.0 (4.7, 5.3)
14.9 (14.0, 15.8)
0.92 (0.75, 1.12)
7.6 (5.5, 9.7)
5.3 (5.1, 5.5)
14.9 (14.3, 15.5)
0.91 (0.62, 1.34)
5.5 (3.3, 7.7)
0.78 (0.66, 0.90)
1.38 (1.24, 1.52)
1.11 (1.00, 1.21)
38.2 (35.3, 41.0)
14.3 (13.7, 14.8)
92.4 (90.9, 94.0)
2 0.37 (20.76, 0.03)
0.40 (0.15, 0.64)
2 0.06 (20.39, 0.26)
1.31 (1.17, 1.45)
1.21 (1.05, 1.37)
1.25 (1.13, 1.37)
37.2 (34.5, 39.9)
12.9 (12.4, 13.5)
88.5 (86.3, 90.7)
2 1.30 (21.66, 20.94)
0.17 (20.02, 0.37)
2 0.98 (21.24, 20.71)
1.03 (0.93, 1.13)
1.29 (1.19, 1.39)
1.18 (1.10, 1.26)
37.7 (35.7, 39.6)
13.6 (13.2, 14.0)
90.4 (88.8, 92.0)
20.85 (21.12, 20.58)
0.28 (0.13, 0.44)
20.53 (20.75, 20.32)
HAZ – height-for-age Z-score; WHZ – weight-for-height Z-score; WAZ – weight-for-age Z-score.
Values in parentheses are 95% confidence intervals.
Bold font indicates a statistically significant regional difference, P , 0.05.
AMN Renzaho 802
lower proportion of households classified as ‘food
uncertain’ (difference: 223.5%, 95% CI: 228.9, 218.2%)
but a significantly higher proportion of households
classified as ‘food insecure without hunger’ (difference:
22.9%, 95% CI: 17.1, 28.9%) compared with Maewo. The
proportion of households classified as ‘food insecure with
hunger in children’ did not vary by location.
The overall prevalence of malnutrition was found to be
19.7% (95% CI: 15.8, 23.6%) for underweight, 28.2% (95%
CI: 23.8, 32.6%) for stunting and 8.8% (95% CI: 5.9, 11.8%)
for wasting. The prevalence of underweight was
significantly lower in Maewo than Ambae (difference:
27.6%, 95% CI: 213.6, 21.7%) but no significant
difference was detected for wasting (difference: 22.8%,
95% CI: 211.9, 6.1%) and stunting (difference: 20.8%,
95% CI: 26.5, 4.8%) between the two locations. After
adjusting for the child’s age and gender, multiple
regression coefficients (Table 4) also depicted relationship
differences in children’s nutritional status by location. The
adjusted model incorporating age, gender and location
explained 21.7% (P , 0.0001) of the variance for WAZ,
17.9% (P , 0.01) of the variance for HAZ and 17.5%
(P , 0.01) of the variance for WHZ. The model suggests
that children in Maewo Island had higher WAZ and HAZ,
and thus were less likely to be underweight and stunted
than children of the same age living in Ambae Island.
The overall crude mortality rate (CMR) and under-five
mortality rate (U5MR) were found to be 0.53/10000 per
day (95% CI: 0.48, 0.59/10000 per day) and 0.78/10000
per day (95% CI: 0.62, 0.97/10000 per day), respectively.
CMR was significantly lower in Maewo than Ambae
(difference: 20.12/10000 per day, 95% CI: 20.23,
20.01/10000 per day), while the U5MR was comparable
in both islands (difference: 20.05/10000 per day, 95% CI:
20.37, 0.05/10000 per day) (see Table 5). Malaria, acute
respiratory infection and diarrhoeal diseases were the
frequently reported causes of death (Fig. 1). While in
absolute terms the proportion of deaths due to malaria,
diarrhoeal diseases and violence was lower in Maewo than
Ambae, these differences were negligible.
This study is the first to explore health and nutrition
outcomes in Maewo and Ambae islands. Consistent with
the study hypothesis, the evaluation found Maewo to
have better health and nutritional outcomes than Ambae.
With international data estimating the prevalence of
underweight and stunting at 19.7 and 19.1%, respect-
ively, for Vanuatu36(Fig. 2), the current data suggest that
Maewo had a prevalence of underweight that was lower
than the national average but a higher prevalence of
Table 3 Prevalence of food insecurity by location
Food security statusMaewo (n ¼ 406)
18.5 (15.0, 22.6)
32.5 (28.1, 37.2)
15.3 (12.1, 19.2)
33.7 (29.3, 38.5)
Ambae (n ¼ 411)
20.0 (16.4, 24.1)
9.0 (6.6, 12.2)
38.2 (33.6, 43.0)
32.9 (28.5, 37.6)
All (n ¼ 817)
19.2 (16.7, 22.1)
20.7 (18.0, 23.6)
26.8 (23.9, 30.0)
33.3 (30.1, 36.6)
Food security (%)
Food uncertainty (%)
Food insecurity without hunger (%)
Food insecurity with hunger in children (%)
Values in parentheses are 95% confidence intervals.
Bold font indicates a statistically significant regional difference, P , 0.05.
Table 4 Prevalence of stunting, underweight and wasting and adjusted* coefficients (b) of HAZ, WAZ and WHZ as a function of
n %† Adjusted b %‡Adjusted b%§ Adjusted b
29.5 (23.1, 35.8)
26.7 (20.4, 33.1)
Ref23.6 (19.4, 27.7)
16.0 (11.8, 20.3)
Ref 9.2 (5.1, 13.2)
8.4 (4.5, 12.4)
0.9 (0.4, 1.5) 0.9 (0.5, 1.3) 0.2 (20.1, 0.5)
34.0 (28.3, 39.6)
25.6 (21.4, 29.8)
Ref29.9 (24.4, 35.3)
14.5 (11.0, 17.7)
Ref12.3 (8.4, 16.3)
0.0 (20.6, 0.5) 0.1 (20.3, 0.5)2 0.3 (20.7, 0.0)
25.6 (19.3, 31.9)
30.5 (24.4, 36.6)
Ref14.8 (9.6, 19.9)
23.8 (18.1, 29.5)
Ref 5.7 (2.3, 9.1)
11.4 (7.1, 15.7)
2 0.6 (21.1, 20.1)
r2¼ 0.179 (P , 0.01)
2 0.7 (21.1, 20.2)
r2¼ 0.217 (P , 0.0001)
2 0.4 (20.7, 20.1)
r2¼ 0.175 (P , 0.01)
HAZ – height-for-age Z-score; WAZ – weight-for-age Z-score; WHZ – weight-for-height Z-score.
Values in parentheses are 95% confidence intervals.
Bold font indicates a statistically significant difference from the reference (Ref), P , 0.05.
*Model adjusted for factors in the table.
†Prevalence of stunting defined as HAZ ,22.
‡Prevalence of underweight defined as WAZ ,22.
§Prevalence of wasting defined as WHZ ,22.
Food insecurity, malnutrition and mortality in Vanuatu803
stunting. However, the above-reported national preva-
lences for underweight and stunting date back to 198336.
It is thus possible that the current national malnutrition
prevalence could be higher than the above figures.
Nevertheless, the prevalence of malnutrition (under-
weight, stunting and wasting) found in Maewo was far
lower than the prevalence of malnutrition in Melanesia,
Asia and Oceania and the average for all developing
countries (see Fig. 2).
The current data also suggest that the CMR was lower in
Maewo than Ambae. However, compared with national
mortality averages as at 200337(CMR: 5/1000 inhabitants
per year – equivalent to 0.14/10000 per day, U5MR: 42/
1000 live births per year – equivalent to 1.15/10000 per
day), the CMR was 3.5 and 4.4 times higher in Maewo and
Ambae, respectively. In stark contrast, both islands
recorded lower U5MR than the national average, with
U5MR found to be 64.3% lower in Maewo and 34.1% lower
in Ambae. Nevertheless, mortality and nutritional out-
comes found in Maewo indicate that using existing
community livelihood and resources yields desired
sustained outcomes. We recently replicated this trend in
other countries across Africa such as in Mozambique and
Lesotho11,38. This is in contrast to the current approach to
international aid, which is overwhelmingly dominated by
the reliance on external inputs when dealing with
dependent on international assistance39–41.
and thus makingthem
Although Maewo displayed more favourable outcomes
on food security indicators than Ambae, food insecurity
remains a big problem in both islands and the percentage
of households classified as food-insecure was comparable
to that reported in countries undergoing economic
crisis28,29. Such a trend could be due to the fact that food
insecurity results from many factors, ranging from socio-
economic, political and cultural to environmental, many of
which were beyond the scope of the project as indeed is
the case in many developing countries42. Another reason
could be that both islands experienced a series of natural
disasters, mainly cyclones43. Both islands were hit by
cyclone Ivy on 25–27 February 2004 which destroyed
many coconut palms, fruit and breadfruit trees and
knocked down a number of homes and health posts. A
number of families were left without shelter, food and
access to medication43. Thus, the destruction of the
livelihood system in Maewo and Ambae islands by the
cyclone could have led to the high level of food insecurity
observed in both islands, as the evaluation was
undertaken just 4 months after cyclone Ivy hit.
Limitations of the study
this field. Our organisation has undertaken to conduct ex-
assess the extent to which projects have contributed to
poverty reduction. For the Maewo Capacity Building
Suchan uncommonly longrecall period may haveaffected
the accuracy of the estimates. Nevertheless, the use of a
calendar of events and of immunisation cards (given the
adequate maternal and child health (MCH) observed in the
region), to check the last date the child attended the MCH
centre prior to death and to validate the reported date of
studies using a relatively shorter recall period in this region
oedema (indicating severe malnutrition irrespective of
WHZ) was not collected due to the lack of skill of data
collectors to detect oedema given their low educational
level and the high likelihood of bias and misclassification.
Thus the prevalence of wasting may have been
Cause of deaths
Fig. 1 Major causes (%) of death by location. *Includes ulcers
(stomach), liver diseases and cardiac diseases
Table 5 Crude mortality rate (CMR) and under-5 mortality rate (U5MR) by location
Maewo (A)Ambae (B) A–B
No. of deaths during recall period
No. of newborns during recall period
Deaths/10000 per day
No. of deaths during recall period
Total ,5 years of age
Deaths/10000 per day
0.47 (0.39, 0.55) 0.59 (0.51, 0.67)2 0.12 (20.23, 20.01)
0.76 (0.54, 0.99)0.81 (0.59, 1.03)20.05 (20.37, 0.05)
Bold font indicates a statistically significant regional difference, P , 0.05.
AMN Renzaho 804
underestimated in this study. Furthermore, in the absence
on verbal reports on morbidity and cause of death.
This paper examines an ex-post evaluation of the Maewo
Capacity Building project, examining the impact of the
project five years after completion. Project initiatives in
reducing the risks of mortality and malnutrition. This was
achieved by investing in human capital in terms of transfer
of skills, mobilising the wider community in the form of
the opportunity to take advantage of the newly created
possibilities. Using a cross-sectional ‘external control
group’ design, this paper demonstrates that it is possible
to draw conclusions about project effectiveness where
baseline data are incomplete or absent. Shifting from
donor-driven evaluations to impact evaluations (whether
periodic, on project conclusion or ex-post) has greater
reporting back to the beneficiaries about project impact
and transformational development in their community. An
impact evaluation can speak more to people than
evaluations based on accountancy data and, put more
simply, can the answer the question ‘While we met all the
requirements of the project, what was the real impact for
the people?’ Public health nutritionists working in the field
are well versed in the collection and interpretation of
anthropometric data for evaluation of nutritional interven-
tions such as emergency feeding programmes. These same
skills can be used to conduct impact evaluations, even
some time after project completion, to elucidate lessons to
be learned and shared. These skills can also be applied
more widely to projects which impact on the longer-term
nutritional status of communities and their food security.
The author is a Program Quality Advisor at World Vision
Australia. Any opinions, findings and conclusions
expressed in this paper are those of the author and
represent a summary of consultations with the commu-
nities. These views do not necessarily reflect the views of
World Vision Australia as an entity and its partners.
The author declares that this is an original manuscript
which has not been submitted to or published in another
journal. The author also declares that there has not been
any financial or other relationship that might lead to a
conflict of interest.
Special thanks go to the staff and community of World
1 Centre for Public Health Nutrition (CPHN). Food Security
Options Paper: A Planning Framework and Menu of Options
for Policy and Practice Interventions. Improving Nutrition in
NSW Series. New South Wales, Australia: CPHN, 2003.
Pelletier DL, Frongillo EA, Schroeder DG, Habicht JP. The
effect of malnutrition on child mortality in developing
countries. Bulletin of the World Health Organization 1995;
Food and Agriculture Organization of the United Nations
(FAO). The State of Food Insecurity in the World: Monitoring
Progress Towards the World Food Summit and Millennium
Development Goals. Rome: FAO, 2003.
Burgermeister J. Number of chronically hungry people is
rising by 5m a year. British Medical Journal 2005; 327: 1303.
Welegtabit SR. Food Security Strategies for Vanuatu. Work-
ing Paper Series No. 58. Bogor, Indonesia: Coarse Grains,
Pulses, Roots and Tuber Crops Centre, 2001.
United Nations Programme on HIV/AIDS (UNAIDS). 2004
Report on the Global HIV/AIDS Epidemic: 4th Global Report.
Geneva: UNAIDS, 2004.
Southern Africa Development Community. Regional
Emergency Food Security Assessment Report. Harare: Food,
Agriculture and Natural Resources Vulnerability Assessment
Food and Agriculture Organization of the United Nations
(FAO). FAO and SIDS: Challenges and Emerging Issues in
Agriculture, Forestry and Fisheries. Rome: FAO, 2004.
Levine S, Chastre C. Missing the Point: An Analysis of Food
Security Interventions in the Great Lakes. London: Overseas
Development Institute, 2004.
Renzaho AMN, Renzaho CC. In the shadow of the volcanoes:
the impact of intervention on the nutrition and health status
of Rwandan refugee children in Zaire two years on from
exodus. Nutrition and Dietetics 2003; 60: 85–91.
Renzaho AMN, Mahony G, David J, Campbell D. World
Vision Food Aid Program in Tete Province – Mozambique:
MaewoMelanesiaOceaniaAsiaAll developing countries
Fig. 2 Prevalence (%) of malnutrition: Maewo versus international data. Source: adapted from reference 34
Food insecurity, malnutrition and mortality in Vanuatu805
Impact Evaluation and Documentation of Lessons Learned.
Melbourne: World Vision Australia, 2004.
Hofmann C. Measuring the Impact of Humanitarian Aid.
Humanitarian Policy Group Research Paper No. 15. London:
Overseas Development Institute, 2004.
Hansch S. Measuring program impact: lessons in evalu-
ation from humanitarian aid. Democracy at Large 2005;
1(3). Also available at http://www.democracyatlarge.org/-
Accessed 10 October 2005.
Adelski E, Hill R, Mason J, MeClelland D, Muscat R. Complex
Humanitarian Emergencies and USAID’s Humanitarian
Response. Washington, DC: Centre for Development
Information and Evaluation, US Agency for International
Petticrew M, Roberts H. Evidence, hierarchies and typolo-
gies: horses for courses. Journal of Epidemiology and
Community Health 2003; 57: 527–9.
Habicht JP, Victora CG, Vaughan JP. Evaluation designs for
adequacy, plausibility and probability of public health
programme performance and impact. International Journal
of Epidemiology 1998; 28: 10–8.
Boe S. Maewo Island Capacity Building Project: End of
Project Report. Port Vila: World Vision Vanuatu, 1999.
Penama Provincial Council. Penama Five-year Social
Development Plan: 2003–2007. Saratamata: Penama
Province, Ministry of Home Affairs, Republic of Vanuatu,
Wassertheil-Smoller S. Biostatistics and Epidemiology: A
Primer for Health Professionals. New York: Springer-Verlag,
Food and Agriculture Organization of the United Nations
(FAO). Conducting Small-scale Nutrition Surveys: A Field
Manual. Rome: FAO, 1990.
Campbell MJ, Julious SA, Altman DG. Sample sizes for
binary, ordered categorical, and continuous outcomes in
two group comparisons. British Medical Journal 1995; 311:
Henderson RH, Sundaresan T. Cluster sampling to assess
immunisation coverage: a review of experiences with a
simplified sampling method. Bulletin of the World Health
Organization 1982; 60: 253–60.
Salama P, Assefa F, Talley L, van der Veen A, Gotway C.
Malnutrition, measles, mortality and the humanitarian
response during a famine in Ethiopia. Journal of the
American Medical Association 2001; 286: 563–71.
Me ´decins Sans Frontie `res. Nutrition Guidelines to Facilitate
the Application of Fundamental Concepts and Principles
Necessary for the Assessment of Nutritional Problems and
The Implementation of Nutritional Programmes in Emer-
gency Situations, 1st ed. Paris: Me ´decins Sans Frontie `res,
Radimer KL, Olson CM, Campbell CC. Development of
indicators to assess hunger. Journal of Nutrition 1990; 120:
Hamilton WL, Cook JT, Thompson W, Buron L, Frongillo JE,
Olson CM, et al. Household Food Security in the United States
in 1995. Washington, DC: US Department of Agriculture
Food and Consumer Service, Office of Analysis and
Evaluation and Abt Associates, 1997.
27Kendall A, Olson CM, Frongillo E Jr. Validation of the
Radimer/Cornell measures of hunger and food insecurity.
Journal of Nutrition 1995; 125: 2793–801.
Welch K, Mock N, Netrebenko O. Measuring hunger in the
Russian Federation using the Radimer–Cornell hunger scale.
Bulletin of the World Health Organization 1998; 76: 143–6.
Studdert L, Frongillo E, Valois P. Household food insecurity
was prevalent in Java during Indonesia’s economic crisis.
Journal of Nutrition 2001; 131: 2685–91.
Frongillo EA Jr, Rauschenbach BS, Olson CM, Kendall A,
Colmenares AG. Questionnaire-based measures are valid for
the identification of households with hunger and food
insecurity. Journal of Nutrition 1997; 127(Suppl.):
Hatloy A, Torheim LE, Oshaug A. Food variety–a good
indicator of nutritional adequacy of the diet? A case study
from an urban area in Mali, West Africa. European Journal of
Clinical Nutrition 1998; 52: 891–8.
Hoddinott J. Choosing Outcome Indicators of Household
Food Security. Washington, DC: International Food Policy
Research Institute, 1999.
National Center for Health Statistics. A Growth chart for
International use in Maternal and Child Health Care.
Guidelines for Primary Health Care Personnel. Geneva:
World Health Organization, 1978.
Renzaho A, Burns C, Reidpath D. Measuring refugee
Malnutrition in the under-fives: pit falls in practice.
Grein T, Checchi F, Escriba JM, Tamrat A, Karunakara U,
Stokes C, et al. Mortality among displaced former UNITA
members and their families in Angola: a retrospective cluster
survey. British Medical Journal 2003; 327: 650.
de Onis M, Monteiro C, Akre J, Clugston G. The worldwide
magnitude of protein–energy malnutrition: an overview
from the WHO global database on child growth. Bulletin of
the World Health Organization 2003; 71: 703–12.
United Nations Children’s Fund. At a glance: Vanuatu
infobycountry/vanuatu_statistics.html. Accessed 10 March
Renzaho AMN. Evaluation of the Nthabiseng and Nazareth
ADP Food Security Project in Lesotho: Summary of Key
Achievements and Suggested Way Forward. Melbourne:
World Vision Australia, 2005.
Hoddinott J. Examining the Incentive Effects of Food Aid on
Household Behaviour in Rural Ethiopia. Washington, DC:
International Food Policy Research Institute, 2003.
Schmalbruch G. Linking relief and development starts with
addressing food insecurity. The ACP-EU Courier 2003; 197:
World Bank. World Bank Aid and Reform in Africa: Lessons
from Ten Case Studies. Washington, DC: World Bank, 2001.
Renzaho AMN. Human right to food security in refugee
settings: rhetoric versus reality. Australian Journal of
Human Rights 2002; 8: 43–56.
European Centre on Pacific Issues. Climate change: cyclones
batter Pacific region. Europe Pacific Solidarity Bulletin 2004;
Appendix – Survey items used in Maewo and Ambae to assess household food security and their respective
Over the last 12 months: MaewoAmbaeAll
1.Have there been occasions when you (head of the family)
worried that you could run out of food before you could
get money to buy more food before the next harvest?
Have there been occasions when you felt that you might
not be able to afford adequate food to feed all members
of the family during the planting–harvest season?
Have there been occasions when you felt like ‘I wish
I could buy more food if I had more money’ during
the planting–harvesting season?
Has your family eaten the same type of food for several
consecutive days because you did not have enough money
to buy different food all season around?
Has your family run out of food for more than a day on more
than two occasions because you did not harvest enough food
or did not have money to buy food?
Have there been occasions when your household relied on
only a few kinds of low-cost food to feed children
because you were running out of food or did not have
enough money to buy food?
Did you (or family members living in this household)
ever cut the size of your meals or skip meals for at least
once every month between the planting and harvest
periods because there wasn’t enough money for food?
Have all children living in this household not had enough
to eat because you did not have enough food or money to
Have you (head of the family) not had enough money to buy
different types of food for your children to diversify
Have any of the family members living in this household
ever eaten less than they felt you should
because there wasn’t enough food?
Did you or other adults in your household ever not eat for
a whole day because there wasn’t enough food or money to
Have any of the children in this household lost weight
because you didn’t have enough food or money
to buy food?
3. 77.9 84.381.2
Note: Participants answered either yes or no to each of the questions.
Food insecurity, malnutrition and mortality in Vanuatu 807