American Journal of Epidemiology
Copyright C 1997 by The Johns Hopkins University School of Hygiene and Public Health
AHrightsreserved
Vol. 146, No. 3
Printed in U.SA
Risk Factors for Diarrheal Disease Incidence in Early Childhood: A
Community Cohort Study from Guinea-Bissau
Kare Molbak,1-2 Henrik Jensen, 12 Liselotte Ingholt,2 and Peter Aaby1-2
To determine risk factors for diarrhea, the authors followed an open cohort of 1,314 children from
Guinea-Bissau by weekly diarrhea recall interviews between April 1987 and March 1990. Data on feeding
practices and measles infection were available for all children and, for 531 children, comprehensive data on
explanatory variables were recorded. Of 57 variables, seven were independently associated with an increased
incidence of diarrhea. These were a recent (in the past 14 days) diarrheal episode, male sex, being weaned
from breast milk, not being looked after by the mother, head of the household being <30 years old, eating cold
leftovers, and drinking water from an unprotected public water supply. In breastfed children, only three
variables were associated with diarrhea, including prior diarrhea, male sex, and not being looked after by the
mother. Among weaned children, six variables delineated increased rates of diarrhea, including unprotected
public water supply, eating of cold leftovers, and lack of maternal education. Major determinants of persistent
diarrhea included weaning, lack of maternal education, and having pigs in the home. It is concluded that, in
addition to the promotion of breastfeeding, important interventions against diarrhea include improvements in
water supply, hygiene, and food handling. However, because of effect modification by breastfeeding, the
largest effects of these interventions will probably be among weaned children. Am J Epidemiol 1997;146:
273-82.
breast feeding; cohort
services
studies; Cryptosporidium;
Several studies have investigated risk factors for
childhood diarrhea, a major cause of mortality among
children in the less developed countries (1). These
studies have, in particular, addressed domestic and
environmental factors (2-5), including water and sanitation (6), or factors related to the mother's hygienic
practices and knowledge (7-9). The impact of feeding
practices, in particular, weaning practices and breastfeeding, has also been addressed (10), whereas other
studies have investigated anthropometric nutritional
status as a risk factor for diarrhea (11, 12). Crosssectional surveys have identified risk factors for prevalent cases (5, 13, 14), but incidence data better reflect
the risk of acquiring diarrhea. Although well-designed
case-control studies may reliably assess the risk for
incident cases (3, 4, 8, 15, 16), there are inherent
methodological problems in the collection of exposure
diarrhea, infantile; incidence; preventive
health
data and the selection of controls. While cohort studies
have the potential of yielding the most reliable data
assessing risk factors for diarrhea, only a few longitudinal studies from Africa have identified determinants
for diarrhea based on the collection of a large range of
exposure variables (17, 18).
This 3-year open cohort study of children <4 years
of age from periurban Guinea-Bissau examined a large
range of possible risk factors for diarrheal diseases.
MATERIALS AND METHODS
Fieldwork
The community-based study, which has been described elsewhere in detail, was conducted in a periurban district, Bandim II, in the capital of GuineaBissau (19-21). All children born after June 1984,
residing in or moving to 301 randomly sampled
houses, were included in the study that started in April
1987 and ended in March 1990. Children who moved
within the study area were followed from their new
homes. The children were excluded from the study in
their fourth year of life. Information on child morbidity and feeding patterns was obtained by weekly
household interviews. The study included 1,314 chil-
Received for publication November 21, 1996, and accepted for
publication March 26, 1997.
Abbreviations: Cl, confidence interval; RR, rate ratio.
1
Department of Epidemiology Research, Danish Epidemiology
Science Center, Statens Serum Institut, Copenhagen, Denmark.
2
Projecto de Saude de Bandlm, Bissau, Guinea-Bissau.
Reprint requests to Dr. KAre Malbak, Department of Epidemiology Research, Statens Serum Institut, Artillerivej 5, DK-2300
Copenhagen S, Denmark.
273
274
Molbak et al.
dren followed for 1,031 child-years at risk. The median follow-up time per child was 242 days (interquartile range, 102-438 days). The birth date, sex of child,
and twinning status were available for all children, and
information on feeding mode (exclusively breastfed,
partially breastfed, or weaned) and measles infection
was updated throughout the follow-up.
Weight and length were obtained at intervals of
approximately 3 months. While nude or dressed in
underwear, each child was weighed with a portable
Salter-type spring scale to the nearest 0.1 kg, and the
child's recumbent length (or stature for children above
2 years) was measured using a portable measuring
board. Information on birth weight was available for
411 children born at the central hospital in Bissau; 51
had birth weight below 2,500 g.
In a series of background interviews conducted in
the first year of the study, we obtained data on the
characteristics of the household (including water supply, sanitation, and domestic animals) and the mother.
These data were available for 707 and 688 children,
respectively. Comprehensive background information,
including information on the child's relation to the
caretaker, was available for 531 children. This part of
the data set (51.3 percent of the person-time), containing information about 57 exposure variables, was used
for the main multivariate model. A list of these variables may be obtained from the authors.
Data analysis
A sequence of days with diarrhea was regarded as
one episode, provided it was separated from previous
episodes by at least 2 diarrhea-free days. The persontime at risk was calculated as the observed days at risk
between episodes. The number of initiated episodes
and the person-time at risk were then classified by the
age of the child (intervals of 60 days for infants;
intervals of 180 days for children 2:360 days of age).
An episode of persistent diarrhea was an episode with
a duration of 14 days or more (22).
Bivariate analyses, adjusting for age, were carried
out for all exploratory variables. A level of significance of 0.10 was selected for including variables in
multivariate modeling, which was done separately in
the three major clusters of exposure variables (characteristics of the child, the household, and the mother).
The three models were reduced by backward stepwise
regression analyses. Finally, significant variables from
each of the three groups were fitted into a full model,
which was reduced to obtain the main model.
The effects of measles infection and nutritional status (birth weight < 2,500 g, weight-for-age, heightfor-age, or weight-for-height < — 2 Z scores of the
National Center for Health Statistics reference values
(23)) were evaluated in models with the major determinants from the analysis above as confounders. Specifically, the nutritional indices were included as timedependent variables. Observations on morbidity prior
to anthropometric measurements or more than 90 days
following a measurement were excluded.
The analysis of risk factors for persistent diarrhea
had a considerably lower number of outcomes and was
conducted in a multivariate model with age intervals
of 180 days. Because of the lower number of persistent
episodes, a p value of <0.20 was chosen as the cutoff
for the model. We also adjusted for low height-for-age
because of the importance of chronic malnutrition in
the pathogenesis of persistent diarrhea (11, 24, 25).
All the analyses were conducted by marginal Poisson
regression of the diarrhea rates. The generalized estimating equations method (26) was used. This method
is essentially a Poisson regression that accounts for
the intrachild correlation treated as a nuisance. This
was a sensible approach as the primary interest was
in the influence of risk factors for diarrheal diseases.
The estimated regression coefficients in this method
share the same interpretation as the estimated regression coefficient from the usual Poisson regression for
independent observations, namely, as log rate ratios.
Hypothesis testing was carried out by the Wald test.
The p value for exclusion of multivariate models was
0.10.
RESULTS
Incidence and duration of diarrheal diseases
A total of 10,485 episodes of diarrhea initiated on
376,475 days at risk yielded an incidence of 10.2
episodes per child-year at risk. Figure 1 shows the ageand sex-specific incidence. The highest incidence,
13.0 episodes per child-year at risk, occurred in the 6to 11-month age group. The median duration of diarrheal episodes was 3 days (interquartile range, 2-6
days). Of all episodes, 1,739 (16.6 percent) had a
duration of between 7 and 13 days, and 536 (5.1
percent) had a duration of 2:14 days. Persistent episodes accounted for 22.3 percent of all days with
diarrhea. The total number of days with diarrhea
divided by the total number of days of observation (the "longitudinal prevalence" (27)) was 12.2
percent.
Figure 2 shows the incidence of diarrheal diseases
and monthly rainfall during the 3 years. In all 3 years,
the incidence of diarrhea was highest in the months
before the maximum rainfall, with the exception of
1988 when an epidemic of diarrhea occurred in November and December. The age-adjusted rate ratio
was 1.22 (95 percent confidence interval (CI) 1.16Am J Epidemiol
Vol. 146, No. 3, 1997
Risk Factors for Diarrhea
0
1
2
3
275
4
Age (years)
FIGURE 1. Age- and sex-specific Incidence of diarrhea in 1,314 children followed for 1,031 child-years at risk, Bissau, Guinea-Bissau,
1987-1990.
1.28, generalized estimating equations method estimates) in the rainy season (from June to November).
The seasonality of persistent diarrhea was more pronounced, with an incidence in the rainy seasons of
337/184,422 (0.67 per child-year) compared with 199/
192,053 (0.38 per child-year) in the dry seasons (rate
ratio (RR) = 1.73, 95 percent CI 1.40-2.15).
Risk factors for diarrheal diseases
Of the 57 exposure variables included, 15 had a p
value of ^0.10 in the initial bivariate analyses. Table
1 shows age-adjusted and multivariate analyses of
child-related determinants. Diarrhea in the past 14
days (RR = 1.90) was a major risk factor. Higher rates
of diarrhea were also observed in partially breastfed or
weaned children, among boys, in children who were
not looked after by their mother, and among individuals with low height-for-age (i.e., stunted children).
Multivariate model A (table 1) investigated the association between low height-for-age and diarrhea, adjusting for age, sex, previous diarrhea, and feeding
mode. Stunted children had a rate ratio of 1.11 (95
percent CI 1.00-1.23, p = 0.045). Finally, model B
included information on the caretaker. Inclusion of
this variable resulted in a substantial loss of data.
Nevertheless, the effect of feeding practices remained
highly significant (p = 0.0001). In this subset of the
data, the main contrast was between breastfed and
Am J Epidemiol
Vol. 146, No. 3, 1997
weaned children, not between exclusively and partially breastfed children.
Table 2 shows age-adjusted bivariate and multivariate analyses of environmental, domestic, and maternal
risk factors. There was an increased rate of diarrhea in
children from households headed by a person under 30
years of age, households with domestic animals, and
families who had the practice of eating cold leftovers.
Diarrheal rates were reduced among children from
families who prepared at least two meals per day.
There was, in addition, a strong and independent association with the type and ownership of water supply,
with a 1.4 rate ratio among children from families with
a public, unprotected water supply compared with
those who used their own, protected supply.
Among the maternal factors there were independent
associations between diarrheal rates and the mother's
ethnic group and education. Specifically, children of
mothers who had S:7 years in school had a 25 percent
reduced diarrheal rate. None of the variables related to
the father were associated with diarrhea.
All variables from the three groups (i.e., child,
household, and maternal factors) were fitted into a
single multivariate model, which was reduced by stepwise exclusion of nonsignificant exposure variables.
Table 3 shows the main model that included seven
variables. Diarrhea in the past 14 days, male sex, being
a child weaned from breast milk, not looked after by
the child's mother, having a young head of household,
eating of cold leftovers, and drinking from an unpro-
276
Molbak et al.
1,000
30
re
RAINFALL
25
800
ALL DIARRHEA
PERSISTENT DIARRHEA
600
3
3
400
200
Apr
Jul
Oct Jan Apr Jul Oct Jan Apr Jul
1989
1988
1987
Oct Jan
1990
FIGURE 2. Monthly incidence of all diarrhea, persistent diarrrhea, and rainfall In 1,314 children followed for 1,031 child-years at risk, Bissau,
Guinea-Bissau, 1987-1990. * , persistent diarrhea: episodes per 10 child-years.
tected water supply were all associated with increased
rates of diarrhea.
Risk factors for diarrheal diseases in breastfed
and weaned children
Several of the factors examined in the initial exploratory analyses exhibited significant interactions with
age (data not shown). Age as an effect modifier may
indicate that different sets of risk factors might be
present in breastfed and weaned children. Table 4
shows multivariate models stratified by breast feeding.
Previous diarrhea was a major risk factor in both
groups. Among breastfed children, boys and children
who were not taken care of by their mother had higher
TABLE 1. Child-relatsd risk factors for diarrheal diseases, estimated by the generalized estimating equations method, Bissau,
Guinea-Bissau, 1987-1990
Rate ratio of diarrhea
No.
of
children
Blvarlate analyses
MuMvartate analyses:):
Model A
Model B
Population
time
(days)
No.
of
1,314*
376,375
1,079
590
197,179
142,838
Feeding mode
Male
sex
Dtarttwa
fripast
14 days
Exclusively
breastfed
Partially
breastfed
Weaned
10,485
1.15(1.04-127)t
1.90(1.77-2.03)
1.00
1.23(1.08-1.40)
1.62(1.37-1.91)
5,360
3,793
1.15(1.04-128)
1.13(0.98-1.31)
1.63 (150-1.77)
1.71 (134-150)
1.00
1.00
1.05(0.88-1.25)
0.91 (0.71-1.18)
1.52(124-1.88)
1.34(1.00-1.79)
Rate ratio of diarrhea
Mother
String
Other
1.00
134(1.06-1.71)
122(059-151)
Welght-for-age
S-2
Z score
1.09(059-121)
—§
1.00
135(1.16-2.07)
1.43(1.11-1.85)
Nti
Ml
Relation to primary caretaker
Blvarlate analyses
MuUvartate analyses
Model A
Model B
Hetgrrt-tor-age
S-2
Z sco re
1.15(1.05-127)
1.11 (1.00-123)
1.10(055-127)
* Information on sex and feeding mode was availabletoraO 1,314 children,
t Numbers In parentheses, 95% confidence Interval.
t MutUvartate model A hctuded heJghl-for-age Z score, feeding mode, sex, and prior diarrhea, whereas model B tnduded relation to caretaker as addlional
explanatory variable.
§ —, variable omitted from model.
1 Nl, not Included (p > 0.10).
Am J Epidemiol
Vol. 146, No. 3, 1997
Risk Factors for Diarrhea
277
TABLE 2. Environmental, domestic, and maternal risk factors for diarrhea! diseases, estimated by the
generalized estimating equations method, Bissau, Guinea-Bissau, 1987-1990
Rate ratio of diarrhea
Btvartale analyses
MulUvartate model*
Environmental and domestic factors
Head of household aged <30 years
No. of children <5 years old/room
<2
Possess domestic animals
Food habits
No. of meals/day
1
^2
Eating of cold leftovers
Water supply
Private
Protected§
Unprotected
PuWic
Protected
Unprotected
1.24(1.04-1.47)f
1.26(1.06-1.50)
1.00
Nit
1.13(0.98-1.30)
1.13(0.99-1.28)
1.12 (0.98-1.27)
0.80(0.71-0.91)
1.31 (1.16-1.48)
1.00
0.88 (0.78-1.00)
1.27(1.12-1.44)
1.00
1.17(0.94-1.46)
1.00
1.25(1.00-1.55)
1.18(1.03-1.36)
1.45(1.19-1.76)
1.21 (1.06-1.39)
1.41 (1.15-1.71)
1.00
Maternal factors
Mother's ethnic group
Papel
Balante
Manjaco
Muslim
Others
Duration of stay in Bissau
<5 years
5-9 years
£10 years
Mother's education (years in school)
None
1-4 years
5-6 years
£7 years
1.00
0.96(0.91-1.03)
1.37(1.30-1.45)
0.83 (0.77-0.89)
0.90 (0.85-0.96)
1.00
0.95(0.78-1.15)
1.37(0.16-1.63)
0.80(0.64-1.00)
0.91 (0.76-1.08)
0.84 (0.73-0.98)
0.91 (0.77-1.09)
1.00
Nl
1.00
1.12(0.98-1.29)
0.97 (0.80-1.17)
0.76 (0.63-0.90)
1.00
1.10(0.96-1.27)
0.98(0.81-1.19)
0.75 (0.61-0.91)
* The multvarlate modelforenvtronmerdal and domestic (actors Included 704 children foDowsd for 263,010 days (7,296
events; median, ejght per chid), whereas the mutUvartate model tor maternal (actors Included 688 children foBowed for 262,222
days (7,323 everts; median, eight per chfld).
t Numbers In parentheses, 95% confidence interval.
t Nl, not Included (p > 0.10).
§ A protected supply was defined as a tap, a hand pump, or a shallow well wtth Intact apron.
rates of diarrhea. Among weaned children, the eating
of cold leftovers, lack of maternal education, age of
family head being less than 30 years, and drinking
from an unprotected water supply independently demonstrated an increased risk for diarrhea.
persistent diarrhea, and children of educated mothers
had a lower incidence of persistent diarrhea. There
was also a tendency of an increased rate in children
from families possessing pigs.
DISCUSSION
Risk factors for persistent diarrhea
Table 5 summarizes risk factors for persistent diarrhea. The analysis included 648 children. Most (75.2
percent) of the children did not experience persistent
diarrhea, while 251 episodes were recorded among
161 of the children. Children with a recent episode of
diarrhea and weaned children had increased rates of
Am J Epidemiol
Vol. 146, No. 3, 1997
Diarrhea] diseases remain a major cause of child
morbidity and mortality in low-income societies, and
the aim of the present study was to identify risk factors
to improve health care and aid impact assessment of
interventions. Of the 57 exposure variables included in
the analyses, only a few exhibited strong associations
with diarrheal rates. A previous recent diarrhea! illness
278
Molbak et al.
TABLE 3. Main reduced multivariate model* of risk factors for diarrhea) diseases, estimated by the
generalized estimating equations method, Bissau, Guinea-Bissaui, 1987-1990
Deter minart
Diarrhea in past 14 days
No
\fes
Sex
Girl
Boy
Breastfeeding mode
Exclusively
Partially
Weaned
Relation to primary caretaker
Mother
Sibling
Other
Age of head of family
<30 years
£30 years
Eating of cold leftovers
No
Water supply
Private
Protected^
Unprotected
Public
Protected
Unprotected
No. of
children
Proportion of
person-time
f%)
Rate ratio
n
r
value
<0.0O1
73.7
26.3
1.00
49.7
50.3
1.00
1.19(1.03-1.37)
0.017
6.4
1.00
0.98 (0.77-1.24)
1.22(0.93-1.62)
0.031
38.3
55.3
89
85
357
18.9
15.6
65.5
1.00
1.66(1.25-2.21)
1.51 (1.15-1.97)
0.002
82
449
12.8
87.2
1.00
0.80(0.64-1.00)
0.046
353
178
65.6
34.4
1.00
1.25(1.08-1.45)
0.003
152
70
28.1
13.2
1.00
1.51 (1.18-1.94)
0.003
256
53
48.1
10.6
1.28(1.07-1.54)
1.42(1.09-1.85)
261
270
1.95(1.78-2.13)t
* The analysis included data from 531 children followed for 192,970 days at risk (median, 335), with 5,481
events (median, seven per child; interquartile range, 3-15).
t Numbers in parentheses, 95% confidence interval,
i A protected supply was defined as a tap, a hand pump, or a shallow well with intact apron.
seemed to be the major factor, even after adjusting for
intrachild correlation by the generalized estimating
equations method. This has been demonstrated in a
few other studies (12, 28, 29) and is open to alternative
explanations as stressed by Black (30). It is possible
that a diarrheal episode may alter the host defenses
directly by damage to the intestinal epithelium or
indirectly by a decrease in immune responsiveness to
subsequent infection. Diarrheal losses of micronutrients such as zinc may impair recovery of the intestinal
mucosa, compromise the immune response, and increase the risk of another diarrhea] episode (31).
The importance of breastfeeding in the prevention
of diarrheal diseases and, in particular, of persistent
diarrhea is emphasized by the present study. Breastfeeding patterns in Guinea-Bissau are characterized by
nearly universal initiation and high rates of prolonged
breastfeeding (32, 33). In addition to maternal milk,
most children receive water from their first month of
life. As early as 3 months of age, half of the children
receive supplements of solids or gruels, which may
have been stored for some time at ambient tempera-
ture. The beneficial effect of prolonged breastfeeding
has been discussed previously (21). However, breastfeeding may also be an effect modifier of other risk
factors for diarrhea, which is one of the main conclusions from the present study (table 5). Only a few risk
factors were found for breastfed children. The lower
incidence among children who were looked after by
their mothers may be related to the quantity of breast
milk consumed; that is, a mother who spends most of
her day together with her child may breastfeed more
frequently than a mother who is away from her child
for several hours during the daytime. The fact that
there was no difference in the effects of being cared
for by an older sibling or another person supports this
interpretation. Bukenya and Nwokolo (15) from Papua
New Guinea found an increased risk of diarrhea when
the mother had been temporarily absent, and Taylor et
al. (14) from Grenada found an increased diarrhea
prevalence when a child was caretaker. It was striking
that there were no associations with socioeconomic or
environmental variables nor was there any apparent
relation between maternal education and diarrhea as
Am J Epidemiol
Vol. 146, No. 3, 1997
Risk Factors for Diarrhea
279
TABLE 4. Multtvariate analysis of risk factors for diarrheal diseases in breastfed and weaned children,
estimated by the generalized estimating equations method, Bissau, Guinea-Bissau, 1987-1990
Rate ratio byfeedingmode
Diarrhea in past 14 days
Male sex
Relation to primary caretaker
Mother
Sibling
Other
Eating of cold leftovers
No
Yes
Ethnic group
Papel
Balante
Manjaco
Muslim
Others
Mother's education
None
1-4 years
5-6 years
£7 years
Age of head of family
<30 years
£30 years
Water supply
Private
Protected!]
Unprotected
Public
Protected
Unprotected
Breastfed*
Weanedt
1.89(1.70-2.10)*
1.20(1.03-1.39)
1.92(1.73-2.14)
Nl§
1.00
1.35(1.05-1.73)
1.31 (1.07-1.61)
Nl
Nl
1.00
1.31 (1.13-1.53)
Nl
1.00
1.16 (0.91-1.48)
1.16(0.94-1.42)
0.74(0.53-1.01)
0.96(0.78-1.18)
Nl
1.00
0.97(0.82-1.15)
0.83 (0.65-1.06)
0.75 (0.58-0.96)
Nl
1.00
0.83(0.67-1.02)
Nl
1.00
1.25(0.94-1.67)
1.23(1.02-1.50)
1.45(1.13-1.86)
* A total of 388 children; person-time, 89,649 days; 2,601 events (median, five/child).
t A total of 461 children; person-time, 107,515 days; 2,962 events (median, four/child).
i Numbers in parentheses, 95% confidence interval.
§ Nl, not included (p > 0.10).
H A protected supply was defined as a tap, a hand pump, or a shallow well with intact apron.
long as the children were breastfed children. By contrast, in fully weaned children, there were strong and
independent associations with several socioeconomic,
domestic, and environmental variables. These included the type and ownership of water supply, eating
of cold leftovers, and the mother's ethnic group and
education.
Effect modification by breastfeeding (or age, as a
proxy for feeding practices) has not been sufficiently
addressed in other observational or intervention studies of diarrheal diseases. It is obviously of interest to
be aware of such interactions, in particular, if the aim
of a health program is to target vulnerable groups such
as infants. The large variation in the magnitude of the
reported effects of water and sanitation programs may
to some degree be explained by effect modification
(6). While an interaction between different environmental interventions has been discussed in a paper by
Am J Epidemiol
Vol. 146, No. 3, 1997
VanDerslice and Briscoe (34), there has been little attention to effect modification by breastfeeding. For example, an evaluation of a water and sanitation project with
a health education component from Bangladesh found an
impact on the incidence of diarrhea in children over 6
months but not in younger, breastfed infants (35). Regarding ownership of water as a proxy for quantity and
accessibility, the finding of an effect of ownership and
type of water supply is in line with the experience from
water and sanitation programs (6).
Of the other variables from the household cluster,
we found that eating of cold leftovers increased diarrheal rates. Previous studies have suggested that storage of food for later consumption is a risk factor for
diarrheal diseases (2, 15, 36). This issue has received
little attention in intervention programs. According to
our data, improved food hygiene may decrease diarrheal rates primarily in older, fully weaned children.
280
Melbak et al.
TABLE 5. Risk factors for persistent diarrhea, derived by generalized estimating equations, Bissau,
Guinea-Bissau, 1987-1990*
Determinant
No of
cfiBdren
Proportion o)
person-time
(%)
n
Rale ratio
value
Age of child (months)
0-5
6-11
12-17
18-23
24-29
30-35
£42
Sex of child
Girl
Boy
Season
Dry
Rainy$
Diarrhea in previous 14 days
No
Yes
Breastfeeding mode
Exclusively
Partially
Weaned
Height-for-age
>-2 Z sco re
<,-2 Z score
Maternal education
None
1-4 years
5-6 years
z.7 years
Pigs in household
No
Yes
9.8
14.1
14.6
15.0
16.9
16.6
13.1
0.40(0.14-1.13)
0.32 (0.12-0.86)
0.22 (0.08-0.62)
0.23 (0.08-0.66)
0.05 (0.01-0.29)
49.7
50.3
1.00
1.33(0.86-2.06)
0.196
50.3
49.7
1.00
1.57(1.08-2.30)
0.019
74.8
25.2
1.00
2.44 (1.72-3.47)
8.7
1.00
1.68(0.69-4.07)
4.40(1.59-12.12)
0.002
41.0
50.3
75.5
24.5
1.00
1.17(0.68-2.01)
0.560
322
176
90
60
44.9
28.7
15.0
11.3
1.00
1.18(0.72-1.92)
0.85(0.47-1.52)
0.35 (0.17-0.71)
0.010
381
267
58.2
41.8
1.00
1.44(0.93-2.22)
0.098
317
331
1.00
0.030
0.75 (0.34-1.65)t
<0.001
* The data consisted of 648 children followed for 150,798 days at risk; 251 episodes of persistent diarrhea were
recorded in 161 of the children (range, 1-7 episodes/child),
f Numbers in parentheses, 95% confidence interval.
t June to November.
Nutritional status
Persistent diarrhea
Malnutrition as a risk factor for diarrhea has been
investigated in many studies. Whereas some have reported a higher incidence of diarrhea in malnourished
children (11, 12, 37-39), others have been unable to
demonstrate such an association (40) or have found a
more striking relation with episode duration (28, 41,
42). We found a tendency of an increased incidence of
diarrhea in stunted children. The effects of malnutrition were not strong enough to be included in the final
multivariate model. This does not preclude that there
is a real, albeit modest increase in the overall rates of
diarrhea in malnourished children. As mentioned,
there were relatively sparse data in the full multivariate model that included information on only 531 children.
In Guinea-Bissau, persistent diarrhea accounts for
only 5 percent of diarrheal episodes. Nevertheless,
these episodes contribute almost 1 of 4 prevalent days
and more than 50 percent of deaths from diarrhea (19).
These findings are in line with other recent studies of
persistent diarrhea (43, 44). Prevention and treatment
of persistent diarrhea are now recognized as a major
challenge for the health care system in low income
societies (30). The strong association between persistent diarrhea and prior diarrheal disease and young age
corroborates studies from Peru, Bangladesh, and India
(24, 29, 30, 45). Furthermore, we found increased
rates of persistent diarrhea in children from households keeping pigs, although the confidence interval
was wide (RR = 1.44, 95 percent Cl 0.93-2.22). We
Aw J Epidemiol
Vol. 146, No. 3, 1997
Risk Factors for Diarrhea
have previously demonstrated an increased risk of
cryptosporidiosis in households with pigs (46) and
shown that Cryptosporidium is a major cause of persistent diarrhea and mortality in early childhood (20).
In Guinea-Bissau, pigs are not usually kept in pigsties;
they roam in the close surroundings of the houses and
often sleep in the same room as the humans. Although
the keeping of animals may be a proxy for overall poor
hygiene, it is more likely that the association between
pigs and persistent diarrhea is a reflection of the significance of cryptosporidiosis and perhaps other zoonotic infections. Animal keeping has previously been
identified as a risk factor for child mortality (21, 47)
and morbidity (2). We suggest that control of domestic
animals may be a potential measure in the prevention
of childhood diarrhea that should be further investigated.
Except for having pigs in the home, none of the
domestic or environmental variables were risk factors
for persistent diarrhea. This may suggest that host- and
care-related factors are the most important determinants of persistent diarrhea. Improved case management of acute diarrhea, improved nutrition including
breastfeeding, and prevention of malnutrition are
likely to be preventive measures against persistent
diarrhea.
Conclusions
Promotion of breastfeeding remains the major preventive measure against diarrhea in developing countries. In Guinea-Bissau, where breastfeeding was
nearly universal and prolonged for most (but not all)
children, this may be achieved by discouraging the
cessation of breastfeeding associated with disease of
the child or the mother. This accounts for a large part
of premature terminations of lactation (33). Furthermore, breastfeeding modifies the effect of other important determinants. Because of this effect modification, the largest effect of improvements in water
supply, food hygiene, and other environmental interventions will probably be among weaned children.
However, breastfeeding patterns vary among countries, and programs for the reduction of morbidity
from diarrhea may need to adapt their strategy accordingly. Thus, in areas with a shorter duration of breastfeeding, domestic and environmental interventions
may be more important for infants than in areas with
prolonged breastfeeding. The different epidemiology
of diarrhea between weaned and breastfed children
may have significant implications for future studies of
diarrheal diseases in developing countries.
Am J Epidemiol
Vol. 146, No. 3, 1997
281
ACKNOWLEDGMENTS
This work was supported by the Science and Technology
for Development Programme of the Commission of the
European Community (contracts TS2-0179 and TS30060); by the University of Copenhagen, Denmark; by the
Danish Council for Development Research (grant
104.Dan.8/1114); and by the Ministry of Public Health in
Guinea-Bissau.
The authors thank Queba Djani, Domingos Sanca, Mario
Rui Lopes, Paulo Umbasse', Jean Gomes Si, Faustino
Umbunque, and Aquinaldo Djana\ who all participated in
the fieldwork, and Niels H0jlyng, Kim Knudsen, Astrid
Permin, Lene Brink, and Uffe Gansted for their contributions. In particular, they wish to acknowledge the late Anja
Vollmer and the late Henning Andersen.
REFERENCES
1. Bern C, Martines J, De Zoysa I, et al. The magnitude of the
global problem of diarrhoeal disease: a ten-year update. Bull
World Health Organ 1992;70:705-14.
2. Knight SM, Toodayan W, Caique WC, et al. Risk factors for
the transmission of diarrhoea in children: a case-control study
in rural Malaysia. Int J Epidemiol 1992;21:812-18.
3. Traore E, Cousens S, Curtis V, et al. Child defecation behaviour, stool disposal practices, and childhood diarrhoea in
Burkina Faso: results from a case-control study. J Epidemiol
Community Health 1994;48:270-5.
4. Mertens TE, Jaffar S, Fernando MA, et al. Excreta disposal
behaviour and latrine ownership in relation to the risk of
childhood diarrhoea in Sri Lanka. Int J Epidemiol 1992;21:
1157-64.
5. Huttly SR, Blum D, Kirkwood BR, et al. The epidemiology of
acute diarrhoea in a rural community in Imo State, Nigeria.
Trans R Soc Trap Med Hyg 1987;81:865-70.
6. Esrey SA, Potash JB, Roberts L, et al. Effects of improved
water supply and sanitation on ascariasis, diarrhoea, dracunculiasis, hookworm infection, schistosomiasis, and trachoma.
Bull World Health Organ 1991;69:609-21.
7. Alam N, Wojtyniak B, Henry FJ, et al. Mothers' personal and
domestic hygiene and diarrhoea incidence in young children in
rural Bangladesh. Int J Epidemiol 1989;18:242-7.
8. Dikassa L, Mock N, Magnani R, et al. Maternal behavioural
risk factors for severe childhood diarrhoeal disease in Kinshasa, Zaire. Int J Epidemiol 1993;22:327-33.
9. Stanton BF, Clemens JD. Socioeconomic variables and rates
of diarrhoea] disease in urban Bangladesh. Trans R Soc Trop
Med Hyg 1987;81:278-82.
10. Feachem RG, Koblinsky MA. Interventions for the control of
diarrhoeal diseases among young children: promotion of
breast-feeding. Bull World Health Organ 1984;62:271-91.
11. Schorling IB, McAuliffe JF, De Souza MA, et al. Malnutrition
is associated with increased diarrhoea incidence and duration
among children in an urban Brazilian slum. Int J Epidemiol
1990;19:728-35.
12. Baqui AH, Black RE, Sack RB, et al. Malnutrition, cellmediated immune deficiency, and diarrhea: a communitybased longitudinal study in rural Bangladeshi children. Am J
Epidemiol 1993;137:355-65.
13. Tsui AO, DeClerque J, Mangani N. Maternal and sociodemographic correlates of child morbidity in Bas Zaire: the effects
of maternal reporting. Soc Sri Med 1988;26:701-13.
14. Taylor SM, Frank J, White NF, et al. Modelling the incidence
of childhood diarrhea. Soc Sci Med 1986;23:995-1002.
15. Bukenya GB, Nwokolo N. Transient risk factors for acute
childhood diarrhoea in an urban community of Papua New
282
Molbak et al.
Guinea. Trans R Soc Trop Med Hyg 1990;84:857-60.
16. Gorter AC, Sandiford P, Smith GD, et al. Water supply,
sanitation and diarrhoeal disease in Nicaragua: results from a
case-control study. Int J Epidemiol 1991;20:527-33.
17. Pickering H, Hayes RJ, Tomkins AM, et al. Alternative measures of diarrhoeal morbidity and their association with social
and environmental factors in urban children in The Gambia.
Trans R Soc Trop Med Hyg 1987:81:853-9.
18. Freij L, Wall S. Quantity and variation in morbidity: THAIDanalysis of the occurrence of gastroenteritis among Ethiopian
children. Int J Epidemiol 1979;8:313-25.
19. M0lbak K, Aaby P, Ingholt L, et al. Persistent and acute
diarrhoea as the leading causes of child mortality in urban
Guinea Bissau. Trans R Soc Trop Med Hyg 1992;86:216-20.
20. M0lbak K, H0jlyng N, Gottschau A, et al. Cryptosporidiosis in
infancy and childhood mortality in Guinea Bissau, West Africa. BMJ 1993;307:417-20.
21. M0lbak K, Gottschau A, Aaby P, et al. Prolonged breast
feeding, diarrhoeal disease, and survival of children in GuineaBissau. BMJ 1994;308:1403-6.
22. Persistent diarrhoea in children in developing countries: memorandum from a WHO meeting. Bull World Health Organ
1988:66:709-17.
23. Hamill PV, Drizd TA, Johnson CL, et al. NCHS growth
curves for children birth-18 years. United States. Vital Health
Stat [11] 1977;165:i-iv, 1-74.
24. Baqui AH, Sack RB, Black RE, et al. Cell-mediated immune
deficiency and malnutrition are independent risk factors fof
persistent diarrhea in Bangladeshi children. Am J Clin Nutr
1993:58:543-8.
25. Bhandari N, Bhan MK, Sazawal S, et al. Association of
antecedent malnutrition with persistent diarrhoea: a case j
control study. BMJ 1989:298:1284-7.
26. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13-22.
27. Morris SS, Cousens SN, Kirkwood BR, et al. Is prevalence of
diarrhea a better predictor of subsequent mortality and weight
gain than diarrhea incidence? Am J Epidemiol 1996;144:
582-8.
28. Chowdhury MK, Gupta VM, Bairagi R, et al. Does malnutrition predispose to diarrhoea during childhood? Evidence from
a longitudinal study in Matlab, Bangladesh. Eur J Clin Nutr
1990:44:515-25.
29. Sazawal S, Bhan MK, Bhandari N, et al. Evidence for recent
diarrhoeal morbidity as a risk factor for persistent diarrhoea: a
case-control study. Int J Epidemiol 1991;20:540-5.
30. Black RE. Persistent diarrhea in children of developing countries. Pediatr Infect Dis J 1993;12:751-61.
31. Sazawal S, Black RE, Bhan MK, et al. Zinc supplementation
in young children with acute diarrhea in India. N Engl J Med
1995:333:839-44.
32. Gunnlaugsson G, da Silva MC, Smedman L. Determinants of
delayed initiation of breastfeeding: a community and hospital
study from Guinea-Bissau. Int J Epidemiol 1992;21:935-40.
33. Jakobsen MS, Sodemann M, M0lbak K, et al. Reason for
termination of breastfeeding and the length of breastfeeding.
Int J Epidemiol 1996;25:115-21.
34. VanDerslice J, Briscoe J. Environmental interventions in developing countries: interactions and dieir implications. Am J
Epidemiol 1995;141:135-44.
35. Aziz KM, Hoque BA, Hasan KZ, et al. Reduction in diarrhoeal diseases in children in rural Bangladesh by environmental and behavioural modifications. Trans R Soc Trop Med
Hyg 1990:84:433-8.
36. Black RE, Brown KH, Becker S, et al. Contamination of
weaning foods and transmission of enterotoxigenic Escherichia coli diarrhoea in children in rural Bangladesh. Trans R
Soc Trop Med Hyg 1982;76:259-64.
37. Tomkins A. Nutritional status and severity of diarrhoea among
pre-school children in rural Nigeria. Lancet 1981;l:860-2.
38. el Samani EF, Willett WC, Ware JH. Association of malnutrition and diarrhea in children aged under five years. A
prospective follow-up study in a rural Sudanese community.
Am J Epidemiol 1988;128:93-105.
39. Sepulveda J, Willett W, Munoz A. Malnutrition and diarrhea.
A longitudinal study among urban Mexican children. Am J
Epidemiol 1988;127:365-76.
40. Chen LC, Huq E, Huffman SL. A prospective study of the risk
of diarrheal diseases according to the nutritional status of
children. Am J Epidemiol 1981;114:284-92.
41. Black RE, Brown KH, Becker S. Malnutrition is a determining
factor in diarrheal duration, but not incidence, among young
children in a longitudinal study in rural Bangladesh. Am J Clin
Nutr 1984:39:87-94.
42. Bairagi R, Chowdhury MK, Kim YJ, et al. The association
between malnutrition and diarrhoea in rural Bangladesh. Int J
Epidemiol 1987;16:477-81.
43. Victora CG, Huttly SR, Fuchs SC, et al. International differences in clinical patterns of diarrhoeal deaths: a comparison of
children from Brazil, Senegal, Bangladesh, and India. J Diarrhoeal Dis Res 1993:11:25-9.
44. Bhandari N, Bhan MK, Sazawal S. Mortality associated with
acute watery diarrhea, dysentery and persistent diarrhea in
rural north India. Acta Paediatr 1992;81(suppl 381):3-6.
45. Bhan MK, Bhandari N, Sazawal S, et al. Descriptive epidemiology of persistent diarrhoea among young children in
rural northern India. Bull World Health Organ 1989;67:281-8.
46. M0lbak K, Aaby P, H0jlyng N, et al. Risk factors for Cryptosporidium diarrhea in early childhood: a case-control study
from Guinea-Bissau, West Africa. Am J Epidemiol 1994;139:
734-40.
47. Pickering H, Hayes RJ, Ng'andu N, et al. Social and environmental factors associated with the risk of child mortality in a
peri-urban community in The Gambia. Trans R Soc Trop Med
Hyg 1986:80:311-16.
Am J Epidemiol
Vol. 146, No. 3, 1997