Published by Oxford University Press on behalf of the International Epidemiological Association
Ó The Author 2005; all rights reserved. Advance Access publication 2 December 2005
International Journal of Epidemiology 2006;35:438–446
doi:10.1093/ije/dyi246
Mortality patterns during a war in Guinea-Bissau
1998–99: changes in risk factors?
Jens Nielsen,1,2* Henrik Jensen,2,3 Per Kragh Andersen2,3 and Peter Aaby1,2
Accepted
13 October 2005
Background The crude mortality rate of the whole population and the mortality of children
,5 years of age are the common indicators of the severity of a complex
emergency situation. However, these indicators rarely take account of differences
in socioeconomic conditions and vulnerability.
Methods
We followed a population in Guinea-Bissau, which fled when fighting took place
in the capital during the war in 1998–99. The population stayed close to the area
of conflict and returned as soon as a cease-fire was negotiated. A peace treaty was
signed after half-a-year. The following 6 months was a period of returning and
re-settlement, even though two outbreaks of fighting occurred.
Results
In the first half-year the mortality rate was 78% [mortality ratio (MR) 5 1.78;
95% CI 1.61–1.97] increased and mortality for children ,5 years of age doubled
(MR 5 2.07; 95% CI 1.79–2.38). In the last 6 months of the war, mortality was
slightly increased for children and not at all for the total population. In the first
half-year, households living in better houses and having members with schooling
were less affected. In the ‘re-settlement’ period two inequalities emerged; the
largest ethnic group, Pepel, continued to have high mortality when the mortality of
other groups declined; likewise girls continued to have an elevated mortality
whereas mortality of boys declined.
Conclusion
Whereas specific ‘free’ interventions reduced social inequalities for the groups
affected, for the total population health-inequalities were slightly amplified during
the war. Once the population returned to their urban homes, mortality fell to prewar levels even though some fighting continued, limited humanitarian aid was
available and the pre-war infra-structure had not been re-established.
Keywords
Introduction
During complex emergency situations (CESs), the vulnerability
of the affected populations is increased and the populations are
exposed to a higher risk of diseases and malnutrition. Most of the
evidence about the impact of CESs has come from refugees or
displaced persons in camps, which were distant from the centre
of conflict. During the war in Guinea-Bissau 1998–99, we
followed a population that was internally displaced during
periods of fighting but returned to their urban residence shortly
1
2
3
Bandim Health Project, Guinea-Bissau, West Africa.
Danish Epidemiology Science Centre, Statens Serum Institut, Denmark.
Department of Biostatistics, University of Copenhagen, Denmark.
* Corresponding author. Bandim Health Project, Danish Epidemiology Science
Centre, Statens Serum Institut, Artillerivej 5, DK-2300 Copenhagen S,
Denmark. E-mail: [email protected]
after the fighting had stopped. Hence, throughout the war they
stayed close to the centre of fighting and conflict.
The crude mortality rate (CMR) and the CMR of children
under five years of age (CMRU5) are key health indicators1,2 and
increases in these mortality levels have been used to characterize
CESs. In non-emergency Sub-Sahara countries, the CMR of a
population is expected to be ~0.4 per 10 000 per day and CMRU5
to be ~1.0 per 10 000 per day.3 When mortality doubles the
situation is normally classified as a CES.3–7 It has, however, been
observed that children .5 years and adults may be disproportionately more affected during a CES than younger children.3
In a CES, a wide range of potentially vulnerable social groups
is recognized including infants, children ,5, pregnant and lactating women, and older people. Many agencies target their
activities to such groups to maximize the impact of the humanitarian aid. Individuals may have different vulnerability during
a CES.2,3,8–10
438
MORTALITY PATTERNS DURING A WAR IN GUINEA-BISSAU
In peacetime, a population will have health inequalities related
to cultural and socioeconomic conditions.1,11–13 During a CES,
cultural and socioeconomic groups may be affected differently
and the increased risk of death may not affect all age groups
equally2,3 and inequalities may change during a CES.3,10 Health
inequalities might be reduced if the CES creates equally bad
conditions for everybody, for example, if all have to ‘run for their
life’. Conversely, health inequalities might be further strengthened if the public health sector breaks down and access to other
resources becomes even more important.
Humanitarian interventions might influence inequalities if
they target specific groups; since interventions are free-of-charge
they might be expected to diminish inequalities between groups.
Little is known about how non-targeted groups are affected; the
whole household might be expected to benefit from aid to
children ,5 years of age.
Inequality in infant and under five mortality in developing
countries favour the better-off households13 and households
with resources, like parental education. In a CES, schooling9,10
as well as presence of a male adult has been found to be
favourable.8,9,14 Inequalities in child mortality associated with
ethnic group, district, education of the mother, and economic
status had been identified before the war in Guinea-Bissau.15
During the war in Guinea-Bissau 1998–99, we compared
the key health indicators of CES (CMR and CMRU5) with the
observed mortality rates during the war and we investigated
the impact on cultural and socioeconomic health inequalities.
The Bandim Health Project (BHP) maintains a longitudinal,
demographic surveillance system in four districts of Bissau, the
capital. CMR, CMRU5, and inequalities in mortality associated
with cultural and socioeconomic differentials before the war
were used to forecast expected mortality rates and inequalities in
mortality during the war period, as they would have occurred
had the war not occurred. Observed mortality rates and
inequalities were compared with the expected inequalities and
mortality levels during the war.
Materials and methods
Complex emergency situation
From June 1998 to May 1999, Guinea-Bissau experienced an
armed conflict between a military Junta, backed by most of the
army and the population, and the president, who was supported
by troops from the two neighbouring countries: Senegal and
Guinée.16–19 There were several periods of fighting: June 7 to
July 26 and October 9 to December 1 in 1998, and January 30 to
February 15 and May 6–7 in 1999. In November 1998, a peace
agreement was signed in Abuja, Nigeria, including withdrawal
of foreign troops and return to the barracks of soldiers, a
deployment of peace-keeping troops and establishment of a
government of national unity. The peace-keeping troops arrived
late December 1998. From January 1999 the conflict changed
character from an open-war to a peace-settling process. In May
1999, government troops surrendered to the Junta.
Fighting was most intense in the capital; the residential areas
of Bissau were exposed to heavy shelling during periods of
fighting. The majority of the population fled, many taking
refuge on the peninsula just outside Bissau.20 By the end of 1999,
265 000 internally displaced persons (IDPs) had returned and out
439
of 7100 refugees, who left the country, 5300 had returned.21
Camps never became a prominent feature of the war in
Bissau; displaced people moved in with relatives, friends, or
strangers in the rural areas. As soon as people believed the
cease-fire would hold they started returning to the capital to
defend their property and because living conditions were
better than in the rural areas.
During the war, the public health systems did not function,22–24 but the central hospital supported by humanitarian aid
continued to receive patients.25 Most other public services did
not function. Less money was available as salaries were not paid
and the market was limited making access to food difficult.
Humanitarian aid provided a limited proportion of the missing
food26 and, although there was no real hunger, the situation was
difficult.
The Bandim Health Project and
humanitarian assistance
The BHP maintains a longitudinal, demographic surveillance
system in four districts of Bissau: Bandim I, Bandim II, Belem,
and Mindara. The project covers ~16% of the population in the
capital, Bissau. The districts of the BHP have a higher proportion
of the Pepel ethnic group and since this group has higher
mortality, one could expect a higher mortality in the BHP
population. On the other hand the health surveillance system
in these districts has facilitated access to healthcare, e.g. the
vaccination coverage is higher in the BHP areas. Hence, there is
no difference in the hospital case fatality of BHP and non-BHP
children hospitalized at the paediatric ward of the national
hospital (unpublished data). As part of the routine demographic
surveillance system, all houses are visited monthly to identify
new pregnancies, births, and deaths. Reported deaths are subsequently confirmed in an interview and the perceived causes
of deaths registered. Children ,3 years of age are visited at
home every 3 months to monitor nutritional status, breastfeeding, hospitalizations and immunizations. The older population was followed in connection with censuses every second or
third year.
Owing to the 3 monthly visits, information on date of death or
date of movement would be fairly accurate. However, some of
the deaths or migrations registered through the censuses were
not reported to the exact date but only up to month, season, or
year; these deaths or movements have been registered at the
middle of the reported period, implying excessive monthly
mortality rates in February (dry season), August (rainy season),
and June (year) for persons >3 years of age.
During the war, BHP took responsibility for humanitarian
aid activities, organizing consultations, providing medical drugs,
and distributing food to IDPs. When the population returned to
Bissau, BHP followed and continued to organize consultations at
the local health centres in the project area and organized food
distribution, when stocks were available. From September 1998
supplementary feeding was established for malnourished
children,27 from October 1998 the project organized vitamin A
supplementation for children ,5 years of age,28 and from
January 1999 impregnated bednets were distributed to pregnant
women and children ,2 years of age. For humanitarian
assistance, the routine home visits were extended to include
children ,5 years of age.
440
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Study populations
We used data from all persons resident in the four districts
Bandim I, Bandim II, Belem, and Mindara (study area) sometime
from January 1, 1995 to May 31, 1999. The war period covered
June 1998 to May 1999. In the war period, 49 731 persons were
resident for some time in the study area; of these 8933 were
,5 years of age, at least part of the period. Deaths due to actsof-war or accidents, as reported in a cause-of-death interview,
were censored. In the war period 1051 died, 916 deaths were
included in the analysis, and 135 deaths were censored; 439
children ,5 years of age died during the war, of these 430 were
included in the analysis and 9 were censored. We could not
clearly distinguish abortions, stillborn, and children who died
during the first day of life, therefore only children who survived
the first day of life were included. Vaccination coverage was
adequate in the study population.29
differential were calculated for each month January 1995 to May
1999. Based on RRs from January 1995 to May 1998 we
forecasted expected inequalities for the war period from June
1998 to May 1999.30 Changes in inequalities during wartime
were estimated as the relative rate ratio (RRR) of the observed
inequality relative to the expected inequality. RRR . 1 signifies a
strengthening of the inequality and RRR , 1 a reduction in the
inequality.
Expressing inequalities as monthly mortality RRs allows
the inequality to be different over calendar time; in a ‘standard’
relative-risk analysis the inequality will be assumed to be
constant over calendar time even though it is controlled for
seasonal variations and trend. Mortality and inequalities were
calculated on a monthly basis but are presented for 3 month
intervals.
Statistical methods
Study designs
Observed CMRs and CMRU5s during the war were compared
with standard mortality rates and local expected mortality based
on mortality from January 1995 to May 1998; accidents and
deaths due to acts-of-war were censored. We expected an
increased mortality throughout the whole war period; a doubled
mortality was expected in the first 6 months of the war and an
increased but abating mortality in the last part of the war, as the
war changed to a peace-settling process. Further, CMRU5 was
compared with crude mortality over five (CMRO5) to investigate
an age differential in mortality.
Health inequalities may decline owing to free-of-charge
interventions; on the other hand existing inequalities might
be exacerbated because those managing well before the war
cope better with difficult living conditions. Hence, we initially
expected the inequalities to be strengthened, but as interventions
were implemented the inequalities might disappear. In the last
part of the war, when people returned inequalities were expected
to return to pre-war levels. We investigated changes in
inequalities in CMRU5 during the war associated with differentials in socioeconomic status and household resources:
Ethnicity: Pepel vs other ethnic groups; Pepel is the largest
group (36%), indigenous to the study area, and normally has
higher mortality than other ethnic groups.
District: Bandim vs the more urban districts, Belem and
Mindara; Bandim usually has higher mortality than Belem
and Mindara.
Economic status expressed by type of roof: straw vs solid roof.
Mother’s education: 0–4 years of schooling vs .4 years of
schooling.
Highest education in the household: 0–4 years of schooling
vs .4 years of schooling.
Gender composition of household: only female adults vs
mixed households.
Gender: female vs male.
Inequality in mortality (CMRU5) associated with a differential
factor was expressed as a rate ratio (RR) between the mortality
rate of the disfavoured and the privileged groups; for example,
inequality associated with maternal education for children
,5 years of age: RRmother’s education 5 CMRU5<4 years schooling/
CMRU5.4 years schooling. Inequalities (RR) associated with a
Some of the deaths or migrations registered through the censuses
(persons >3 years of age) were not reported to the exact date
but only up to month, season, or year; these deaths or
movements were registered at the middle of the reported period,
implying excessive monthly mortality rates in February (dry
season), August (rainy season), and June (year). To circumvent
this, we used methods adjusting for interval censoring31 when
estimating monthly mortality rates.
It had not been registered which deaths had been reported to
the exact date and which had not, i.e. interval censored. Hence,
we cannot know exactly the dates of death that were accurate
and that were interval censored. Based on the fraction of
excessive deaths in the months where interval censored death
had been registered (February, June, and August) we simulated
100 assignments of interval censoring and used the average mean
and variance of these32 as estimates.
Since monthly mortality rates for children ,5 adjusted for
interval censoring and without such adjustment showed no
systematic difference, estimates of CMRU5 and the corresponding inequalities in mortality under-5 were not adjusted for
interval censoring. However, CMR and CMRO5 were adjusted
for interval censoring.
Expected monthly mortality rates during wartime were
achieved by forecasting data from January 1995 to May 1998 controlling for trend and seasonal variation into the war period.30
Inequalities, expressed as mortality RRs, were forecasted in a
log-transformed Gaussian regression with estimation error.
RRRs were estimated monthly as the ratio between the
observed and the expected mortality ratios. The variance of log
(RRR) was obtained as the sum of the variances of the two
independent log (RRobserved) and log (RRexpected).
Monthly RRRs were averaged, weighted by the inverse of the
variance, to quarter and half year, adjusted to dry and rainy
season, and for the whole period; test for homogeneity showed
how appropriate the summarizations were.
Analyses were done in SAS Release 8.2 (SAS Institute,
Cary, NC).
Results
Crude mortality of the study population during the war is shown
in Figure 1 (left). The CMR was not constant through the
MORTALITY PATTERNS DURING A WAR IN GUINEA-BISSAU
441
Figure 1 Crude mortality rate (CMR) during the war in Guinea-Bissau 1998–99. Accidents and acts-of-war censored. Mortality rates (death per day
per 10 000) Mortality rate ratios 0.4 line: standard for developing countries observed/expected Full line: estimate. Dotted lines: 95% CI
Table 1 Relative changes in mortality and pre-war levels during the war in Guinea-Bissau 1998–99
Relative change in mortality rate during the war
Before signed peace treaty
a
Pre-war mortality rate
January 1995–May 1998
Crude mortality
0.43 (0.41–0.44)
Signed peace treaty and returning
June–
August 1998
September–
November 1998
December 1998–
February 1999
March–
May 1999
1.83 (1.58–2.13)
1.74 (1.52–2.00)
1.11 (0.94–1.31)
0.99 (0.84–1.18)
1.78 (1.61–1.97)
1.05 (0.93–1.19)
1.43 (1.32–1.54)
Crude mortality under 5
1.02 (0.96–1.08)
1.78 (1.45–2.19)
2.37 (1.94–2.90)
1.40 (1.10–1.77)
2.07 (1.79–2.38)
1.20 (0.91–1.59)
1.31 (1.09–1.57)
1.73 (1.55–1.94)
Note: Relative changes in mortality (observed mortality rate/expected mortality rate) were calculated monthly and have then been averaged to quarter, half-year,
and the whole period; accidents and acts-of-war were censored in the analysis.
a
Death per day per 10 000.
different months of the war (Phomogeneity , 0.01). An expected
local time-constant mortality would have been 0.43 (95% CI
0.41–0.44), a little higher than the expected standard mortality in
developing countries: 0.4 death per day per 10 000. Observed
CMR was compared with the local expected mortality (Table 1;
Figure 1, right). A crude mortality elevated with ~78% (MR 5
1.78; 95% CI 1.61–1.97) was observed from June to November
1998. From December 1998 to May 1999 crude mortality was
‘normal’.
Mortality of children under-5 (CMRU5) during the war is
shown in Figure 2 (left) and, as for CMR, it varied over time
(Phomogeneity , 0.01). The expected mortality of children ,5 (1.0
death per day per 10 000) fits well with a local time-constant
expected mortality rate of 1.02 (95% CI 0.96–1.08). Mortality
during the war was compared with the local expected mortality
(Figure 2, right and Table 1). Mortality for children ,5 was 2.07
(95% CI 1.79–2.38) times higher from June to November 1998
compared with the expected mortality. From December 1998 to
February 1999 CMRU5 was significantly increased with an MR of
1.40 (MR 5 1.40; 95% CI 1.10–1.77) and in the last 3 months
mortality under-5 was slightly increased.
When comparing the expected CMRU5/CMRO5 ratio with
the observed ratio during the war, there was no evidence of a
disproportional effect between children ,5 and, children .5 and
adults (Figure 3).
The impact of cultural and socioeconomic differentials on
mortality inequalities during the war is shown in Table 2.
Though the Pepel ethnic group usually has higher mortality
than other groups (RR 5 1.14; 95% CI 1.07–1.21), this differential was further deepened (RRR 5 1.60; 95% CI 1.14–2.25)
in the last 6 months of the war. The pre-war increased mortalitylevel in Bandim compared with Belem and Mindara (RR 5 1.43;
95% CI 1.32–1.55) remained unchanged throughout the war. In
the pre-war years, living in houses with straw-roofs was
associated with higher mortality (RR 5 1.10; 95% CI 0.99–
1.21); this inequality was significantly higher with a RRR of
1.81 (95% CI 1.18–2.77) in the first 3 months of the war and it
was slightly increased in the second half-year of the war. Living
in a house with straw roof was the only inequality that differed
significantly for the whole war period.
The better childhood survival associated with .4 years of
schooling of the mother (RR 5 1.52; 95% CI: 1.38–1.67) in the
pre-war period was not affected during the war. However, having
no one with .4 years of schooling in the household, which in
the pre-war period was associated with a slight inequality in
mortality (RR 5 1.10; 95% CI 1.00–1.23), was associated to a
significantly higher mortality from September to November 1998
(RRR 5 2.07; 95% CI 1.31–3.28); afterwards the inequality in
mortality was as before the war (RRR 5 1.13; 95% CI 0.74–1.74).
In the pre-war period, households with female adults only had
442
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Figure 2 Crude mortality of children under five (CMRU5) during the war in Guinea-Bissau 1998–99. Accidents and acts-of-war censored. Mortality
rates (death per day per 10 000) Mortality rate ratios 1.0 line: standard for developing countries observed/expected Full line: estimate. Dotted
lines: 95% CI
Figure 3 Relative changes in mortality between observed and expected under five to over-five mortality during the war in Guinea-Bissau 1998–99.
Accidents and acts-of-war censored. Note: Relative mortality rate ratio (RRR) is the relative ratio between the observed mortality rate ratio
(CMRU5/CMRO5) and the expected mortality rate ratio
slightly increased mortality compared with mixed households
(RR 5 1.07; 95% CI 0.99–1.16); this did not change during
the war.
Prior to the war, there was no difference in the female/male
under 5 mortality ratio (RR 5 1.01; 95% CI 0.91–1.13). During
the first 6 months of the war, when mortality was increased, the
sexes were equally affected (RRR 5 0.99; 95% CI 0.76–1.29)
(Table 3). In the last 6 months of the war, mortality for boys
levelled to normal (RR 5 0.98; 95% CI 0.75–1.27). Even though
mortality for girls decreased in this period, it remained
significantly elevated compared with normal (RR 5 1.41; 95%
CI 1.10–1.79). Hence, in the last 6 months of the war, girls
had 47% (RRR 5 1.47; 95% CI 1.03–2.09) higher mortality
than boys.
Discussion
We followed the population in four districts of the capital
through wartime, from June 1998 to May 1999. The population
was followed both when present in Bissau and when they fled
the capital, i.e. were IDPs. Most reports of mortality from
war-emergencies have been from refugee camps or long-time
displaced persons separated from where the underlying conflict
is occurring. We followed a population that was displaced within
a short distance while active fighting was going on, but returned
a few months after fighting had stopped. Hence, the population stayed in the area of the conflict. Though deaths due to actsof-war and accidents were censored, mortality was increased in
the first 6 months of the war, which was the period with the
heaviest fighting and when people spent most time away from
MORTALITY PATTERNS DURING A WAR IN GUINEA-BISSAU
443
Table 2 Inequalities in mortality (rate ratios) for children ,5 years of age associated with socioeconomic differentials and household resources
before the war and changes in these during the war in Guinea-Bissau 1998–99
Relative change in mortality ratio during the war
Pre-war mortality
ratio for social
inequality indicators
January 1995–May 1998
Ethnicity (Pepel)
1.14 (1.07–1.21)
Before signed peace treaty
Signed peace treaty and returning
June–
August 1998
September–
November 1998
December 1998–
February 1999
March–
May 1999
0.87 (0.61–1.23)
0.88 (0.61–1.27)
1.80 (1.15–2.82)
1.36 (0.81–2.29)
0.87 (0.68–1.12)
1.60 (1.14–2.25)
1.08 (0.89–1.33)
District (Bandim)
1.43 (1.32–1.55)
0.96 (0.64–1.43)
0.75 (0.50–1.12)
1.30 (0.74–2.29)
0.85 (0.64–1.13)
0.81 (0.45–1.45)
1.03 (0.69–1.55)
0.90 (0.72–1.14)
Economic status (Straw roof)
1.10 (0.99–1.21)
1.81 (1.18–2.77)
0.83 (0.47–1.48)
1.58 (0.86–2.89)
1.38 (0.98–1.94)
1.20 (0.61–2.38)
1.40 (0.89–2.20)
1.38 (1.05–1.82)
Education of mother
(<4 years of schooling)
1.52 (1.38–1.67)
0.74–(0.49–1.12)
1.11 (0.69–1.79)
0.90 (0.52–1.56)
0.88 (0.65–1.20)
1.12 (0.57–2.19)
0.98 (0.64–1.50)
0.92 (0.71–1.18)
Education in household
(<4 years of schooling)
1.10 (1.00–1.23)
0.82 (0.52–1.31)
2.07 (1.31–3.28)
1.10 (0.63–1.92)
1.31 (0.95–1.81)
1.18 (0.60–2.31)
1.13 (0.74–1.74)
1.24 (0.96–1.61)
Household composition
(Only female adults)
1.07 (0.99–1.16)
1.17 (0.78–1.76)
1.16 (0.75–1.79)
1.36 (0.82–2.26)
1.16 (0.86–1.57)
0.87 (0.46–1.65)
1.15 (0.77–1.71)
1.16 (0.91–1.47)
An inequality in mortality associated to a differential is estimated as the mortality rate ratio (RR) between mortality rates in the two differential groups. The relative
change in inequality (RRR) is the ratio: observed mortality rate ratio/expected mortality rate ratio (RRR 5 RRobs/RRexp).
Note: Relative changes in inequalities (RRR) were calculated monthly and have then been averaged to quarter, half-year, and the whole period; accidents and
acts-of-war were censored in the analysis.
Table 3 Inequalities in mortality (rate ratios) for children ,5 years of age according to gender and the change in mortality during the war in
Guinea-Bissau 1998–99
Relative change in mortality rate during the war
Pre-war
a
mortality rate
January 1995–May 1998
Relative female mortality under five
1.03 (0.91–1.02)
Before signed peace treaty
Signed peace treaty and returning
June–
August 1998
September–
November 1998
December 1998–
February 1999
March–
May 1999
1.62 (1.24–2.11)
1.72 (1.31–2.27)
1.65 (1.20–2.28)
1.14 (0.79–1.65)
1.67 (1.38–2.02)
1.41 (1.10–1.79)
1.56 (1.34–1.81)
Relative male mortality under five
1.01 (0.90–1.07)
1.59 (1.22–2.08)
1.78 (1.35–2.35)
1.15 (0.82–1.63)
1.68 (1.39–2.04)
0.79 (0.52–1.18)
0.98 (0.75–1.27)
1.39 (1.19–1.62)
Female male mortality ratio
Relative inequality female/male
1.01 (0.91–1.13)
Change in mortality ratio during the war
1.01 (0.70–1.46)
0.97 (0.66–1.42)
1.41 (0.88–2.24)
0.99 (0.76–1.29)
1.56 (0.91–2.69)
1.47 (1.03–2.09)
1.14 (0.92–1.41)
A relative change in mortality is: observed mortality rate/expected mortality rate. The inequality in mortality associated to gender is estimated as the mortality rate
ratio (RR) between mortality rates in the two gender groups. The relative change in inequality (RRR) is the ratio: observed mortality rate ratio/expected mortality
rate ratio (RRR 5 RRobs/RRexp).
Note: Relative changes in mortality and inequalities were calculated monthly and have then been averaged to quarter, half-year, and the whole period; accidents
and acts-of-war were censored in the analysis.
a
Death per day per 10 000.
their normal residence. Crude mortality of the whole study
population was 78% increased in the first half-year of the war
and mortality during the whole year was 43% increased.
Mortality of children ,5 was doubled in the first half-year, i.e.
a CES according to international definitions. The increased
mortality was only found in the first part of the war, there being
virtually no increase in the last 6 months of the war after a peace
treaty had been signed and people returned. We found no
444
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
evidence of a disproportional impact on mortality for children ,5
compared with persons .5 years, censoring for accidents and
deaths due to war.
Mortality inequalities associated with cultural and socioeconomic differentials, including schooling in the household
and type of roof, were affected in the first half-year of the war. If
type of roof can be interpreted as economic status, then children
in better-off households were less vulnerable in the first quarter
of the war, but the advantage decreased possibly as resources
were used; the disadvantage of people living in houses with straw
roofs remained throughout the whole war period. The pre-war
inequality in CMRU5 associated with schooling of the mother
was unchanged during the war. However, schooling in the
household had a beneficial effect on childhood mortality towards
the end of first half-year of the war. It was unexpected that the
inequality associated with mother’s schooling was unchanged,
but education in the household reduced vulnerability. This might
be because a child’s vulnerability is associated not merely with
the mother’s resources but is linked to household resources and
the ability to seek aid. Education in the household had the largest
effect in the phase when humanitarian aid became available and
the majority of people still were IDPs. In the second half-year of
the war when most people had returned, the inequality related to
schooling was as before the war.
The capital, Bissau, is located in the Pepel homeland of GuineaBissau and Pepeis constitute the largest group in the study area. It
might have been easier for the Pepel households in the beginning
of the war because their relatives and family-land were close by.
However, in the second part of the war, mortality declined
more slowly for the Pepel group than other ethnic groups. The
reason for this is unknown, but the Pepel homeland, supporting
the Pepeis returning to Bissau, may have been comparatively
more exhausted because all IDPs initially fled to the Pepel area
surrounding Bissau. Some from other ethnic groups continued
to their ethnic homeland, but all had a stop-over in the Pepel
homeland. Hence, the change in inequality that seems associated
to ethnicity may actually have been associated with being coresident with IDPs, which has been shown to have a deleterious
impact on childhood mortality.20
During the first 6 months of the war, when CMRU5 was
elevated, the mortality for boys and girls were equally affected. In
the last 6 months of the war, however, mortality for boys
normalized, but mortality for girls levelled more slowly. There is
no clear explanation for this occurrence. There is no indication of
marked sex-differential treatment in this society and mortality
levels were equal before the war. The pattern could have been
due to a different return-pattern, for example, girls staying longer
outside Bissau, while boys returned earlier. However, we found
no difference in migrations pattern of boys and girls during the
war (data not shown).
We have earlier shown an equalizing impact on health
inequalities of humanitarian aid interventions during the war
such as vitamin A supplementation28 and distribution of drugs
free of charge at the paediatric ward.25 Hence, it might be
surprising that most general inequalities maintained their
pre-war level. However, these interventions were probably too
limited to have an impact on the overall mortality levels.
The war in Guinea-Bissau was a CES even though its duration
and mortality level may have been below what has been
observed in other African CESs. However, even in short CESs the
affected individuals suffer both in terms of health and material
conditions as looting is common and resources are exhausted.
In the Bissau war, mortality was increased during the first
6 months, when the fighting was most intense. Though mortality
was increased, most inequalities in childhood mortality remained
unaffected, except an advantage for those with better housing
conditions and a short-term beneficial effect of schooling in
the household, which might be associated with a better ability
of attaining humanitarian aid.
Little is known about conditions for war-affected people who
stay close to the area of fighting because they maintain the hope
that fighting will soon stop and they may be able to return to
better living conditions and to save whatever material conditions
they had been able to accumulate. It is not surprising that
mortality was increased in this period, but it was surprising that
no stronger changes in inequalities in childhood mortality
associated with cultural and socioeconomic differentials occurred
during this period. We had expected that the miserable living
conditions that most people had as IDPs20 would have tended
to reduce inequalities between different groups.
We analysed mortality associated with the poor living and
health conditions for the civil population during the war. This, in
Africa, mainly means increased risk of infectious diseases. We
censored direct war-related deaths as these will have other risk
factors and patterns. To our surprise, the number of war-related
deaths among the civil population increased in the last 6 month
period. This might be owing to most people remaining in the
capital during the last outbreaks of fighting; many civilians
sought safety in mission stations in the periphery of the fighting
area. One of these stations was hit by bombs and 60–70
individuals were killed on the final day of the war.
The first 6 months of the war coincided with the rainy
season whereas the last 6 months were the dry season. Hence,
the marked differential effect during the first 6 months could
be related to people being more susceptible to the impact of a
CES in the rainy season with immune suppression and reduced
hygienic levels. However, it is also noteworthy that mortality
normalized already when a peace treaty had been signed and
people started to return more permanently, and not when the
total war-period was over. The reduced crowding and improved
hygiene when people returned to their normal homes might
have been very important for disease transmission and mortality
patterns.20 During the war, the rebel radio consistently advised
the population not to return to the capital owing to the risk of
fighting and bombing. Even though the general population
was in favour of the rebels, most people did return home feeling
living conditions to be better and more secure in spite of the
breakdown of public systems and risk of further fighting. Our
analysis would suggest that they probably took the right decision
even though bombing did kill some people during the final
assault on Bissau in May 1999. Humanitarian aid programme
may benefit from trying to re-settle IDPs as quickly as possible.
Acknowledgements
ECHO (European Commission, Brussels, Belgium), The Swedish
Embassy in Guinea-Bissau and Danida funded humanitarian
aid during the war in Guinea-Bissau. The Danish Council for
Development Research and ECHO supported studies of the war.
We acknowledge the field assistants of BHP, without whom
MORTALITY PATTERNS DURING A WAR IN GUINEA-BISSAU
data collection would have been impossible. Contributions: Peter
Aaby was responsible for the humanitarian work of the Bandim
Health Project during the war. Peter Aaby took part in planning
the design of the study and the collection of data. Henrik Jensen
and Per K. Andersen supervised statistical methods and analyses.
Jens Nielsen designed the study, implemented data control, did
all statistical analyses and wrote the first draft of the paper. All
authors contributed to the final version. Sponsorship: The Danish
445
Council for Development Research and ECHO, European Union,
supported studies of the war.
Conflict of Interest
The authors have declared no conflicts of interest.
KEY MESSAGES
Infectious disease mortality increased 2-fold during an armed conflict in Guinea-Bissau, when all inhabitants of the
capital had to flee their home
Peace settlement and returning home had an immediate beneficial impact on mortality levels
Overall the war maintained or deepened social inequalities in mortality, even though specific humanitarian aid
interventions reduced inequalities
References
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Wagstaff A. Socioeconomic inequalities in child mortality: comparisons
across nine developing countries. Bull World Health Organ 2000;
78:19–29.
Guha-Sapir D, van Panhuis WG. The importance of conflict-related
mortality in civilian populations. Lancet 2003;361:2126–28.
17
18
19
20
Davis AP. Targeting the vulnerable in emergency situations: who is
vulnerable? Lancet 1996;348:868–71.
Hakawill PA, Moren A. Monitoring and evaluation of relief
programmes. Trop Doct 1991;21:24–28.
21
CDC. Famine-affected, refugee, and displaced populations: recommendations for public health issues. MMWR Recomm Rep 1992;41:1–76.
22
Boss LP, Toole MJ, Yip R. Assessments of mortality, morbidity, and
nutritional status in Somalia during the 1991–1992 famine. JAMA
1994;272:371–76.
23
The Sphere Project. The Steering Committee for Humanitarian Response and
InterAction. Humanitarian Charter and Minimum Standards in Disaster
Response. Oxford, England: Oxford Publishing, 1998.
24
Goma Epidemiology Group. Public health impact of Rwandan refugee
crisis: what happened in Goma, Zaire, in July, 1994? Lancet 1995;345:
339–44.
25
Cutts FT, Santos CD, Novoa A, Dacid P, Macassa G, Soares AC. Child
and maternal mortality during a period of conflict in Beira City,
Mozambique. Int J Epidemiol 1996;25:349–56.
26
Kiros GE, Hogan D. War, famine and excess child mortality in Africa:
the role of parental education. Int J Epidemiol 2001;30:447–55.
27
Hobcraft J. Women’s education, child welfare and child survival:
a review of the evidence. Health Transit Rev 1993;3:159–75.
Rutstein SO. Factors associated with trends in infant and child
mortality in developing countries during the 1990s. Bull World Health
Organ 2000;78:1256–70.
Wagstaff A. Poverty and health sector inequalities. Bull World Health
Organ 2002;80:97–105.
Ager A, Ager W, Long L. The differential experience of Mozambican
refugee women and men. J Refugee Studies 1995;8:265–87.
Aaby P, Jensen H. Crowding and Health in Low-Income Settlements. Case
Study Report, Bissau, UNCHS (Habitat). Copenhagen: COWIconsult,
1995.
Gomes JC. Polon di bra. Um documento de refflexao sobre uma
Guerra devastadora. Desnecessaria e injustamente imposta ao povo da
Guine-Bissa. INACEP 1998.
28
29
30
31
32
Soronda. Revista de Estudos Guineenses. Número Especial 7 de Junho.
INEP Guinea-Bissau 2000.
Rudebeck L. On Democracy’s Sustainability: Transition in Guinea-Bissau.
Stockholm: SIDA, 2001.
Co JV. As consequencias juridico-constitucionais do conflito politicomilitar da Guine-Bissau. Rete Guinea Bissau, 2001.
Aaby P, Gomes J, Fernandes M, Djana Q, Lisse I, Jensen H. Nutritional
status and mortality of refugee and resident children in a non-camp
setting during conflict: follow up study in Guinea-Bissau. BMJ
1999;319:878–81.
UNHCR. Population data summary sheet. Available at: http://www.
unhcr.ch/cgi-bin/texis/vtx/statistics (1999).
Gustafson P, Gomes VF, Vieira CS et al. Tuberculosis mortality during a
civil war in Guinea-Bissau. JAMA;286:599–603.
Aaby P, Jensen H, Garly ML, Balé C, Martins C, Lisse I. Routine
vaccinations and child survival in a war situation with high mortality:
effect of gender. Vaccine 2002;21:15–20.
Aaby P, Garly ML, Balé C et al. Survival of previously measlesvaccinated and measles-unvaccinated children in an emergency
situation: an unplanned study. Paediatr Infect Dis J 2003;22:798–805.
Sodemann M, Veirum J, Biai S et al. Reduced case fatality among
hospitalised children during a war in Guinea-Bissau: a lesson in equity.
Acta Paediatr 2004;93:959–64.
Tin H. O Benefı́cio do Fracasso: PAM, Ajuda Alimantar e Sobrevivência
Local na Guiné-Bissau. 1998–99. Soronda. Revista de Estudos
Guineenses. Novo Série no 3. INEP Guinea-Bissau 2002: 77–145.
Nielsen J, Martins C, Valentiner-Branth P, Aaby P. Malnourished
children and supplementary feeding during war-emergency in
Guinea-Bissau 1998–99. Am J Clin Nutr 2004;80:1036–42.
Nielsen J, Benn CS, Balé C, Martins C, Aaby P. Vitamin A supplementation during war-emergency in Guinea-Bissau 1998–99. Acta
Trop 2005 (in press).
Aaby P, Martins C, Bale C, Lisse I. Assessing measles coverage by
maternal recall in Guinea-Bissau. Lancet 1998;352:1229.
Nielsen J, Jensen H, Andersen PK. Creating a reference to a complex
emergency situation using Time Series methods: war in Guinea-Bissau
1998–99. J Appl Stat 2005;32:75–86.
Carstensen B. Regression models for interval censored survival data:
application to HIV infection in Danish homosexual men. Stat Med
1996;15:2177–89.
Harrell FE. Regression Modelling Strategies. New York: Springer-Verlag
Inc., 2001.
446
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Map of the Warzone in Guinea-Bissau. Reproduced from: Aaby P, Gomes J, Fernandes M, Djana Q, Lisse L, Jensen H. Nutritional status and mortality
of refugee and resident children in a non-camp setting during conflict: follow up study in Guinea-Bissau. BMJ 1999;319:878–81, with permission from
the BMJ Publishing Group.
Published by Oxford University Press 2006
Advance Access publication 15 March 2006
International Journal of Epidemiology 2006;35:446–447
doi:10.1093/ije/dyl018
Commentary: When should we monitor
mortality in humanitarian crises?
Richard Garfield
Accepted
24 January 2006
This paper on mortality patterns in Guinea-Bissau1 contributes to
the current extensive discussions on monitoring mortality rates
during humanitarian crises. Mortality is the most sincere
expression of vulnerability and, when well monitored, should
direct efforts for relief, protection, and humanitarian intervention. Unfortunately, areas with humanitarian crises in the
world today seldom enjoy an active case finding system like that
set up by the authors prior to conflict in Guinea-Bissau. One
Columbia University, 617 West 168th Street, New York, NY 10032, USA.
E-mail: [email protected]
lesson from this research is the importance of maintaining, and if
possible enhancing, any such system in areas of crisis. Failing
this, at best we get occasional special mortality studies that are
easily biased by rapid population movements and difficult to
interpret owing to their cross-sectional nature. This is just
the situation in most of the major humanitarian crises in
the last decade, including the Democratic Republic of
Congo, Afghanistan, and Iraq. Without a system of on-going
monitoring, even if special studies are accurate, the information
is frequently not available to act on in a timely basis. In GuineaBissau, monitoring helped minimize all these problems.