1081

Persistent Diarrhea in a Cohort of Israeli Bedouin Infants: Role of Enteric

Pathogens and Family and Environmental Factors
Drora Fraser, Ron Dagan, Nurith Porat, Joseph El-On,
Soliman Alkrinawi, Richard J. Deckelbaum,
and Lechaim Naggan
University of the Negev and Primary Health Care Clinic (Kupat Cholim
Epidemiology and Health Services Evaluation Department, (Faculty of
Health Sciences) and Daniel Abraham International Center for Health
and Nutrition, Ben-Gurion University of the Negev, Beer-Sheva, and
Pediatric Infectious Disease Unit (Faculty of Health Sciences) and
Parasitology Laboratory (Department of Microbiology and
Immunology), Soroka University Medical Center and Ben-Gurion
Klalit), Rahat, Israel; Department of Pediatrics and Institute of Human
Nutrition, College of Physicians and Surgeons, Columbia University,
New York, New York

Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020

This study examined the role of enteric pathogens and infant, family, and household characteris-
tics in persistent diarrhea. Bedouin infants from southern Israel were followed from birth to age
18 – 23 months. During monthly home visits, stool samples were obtained, and feeding practices and
history of diarrhea were determined, and at age 3 months, an environmental assessment was done.
Diarrhea surveillance was either via a network covering all community health care facilities or via
weekly interviews with the mother. None of the enteric pathogens examined, including Cryptosporid-
ium parvum and enteroaggregative Escherichia coli, were associated with persistent diarrhea. In
multivariate analyses, age at first diarrheal illness and maternal age and maternal education were
independently and significantly associated with the risk of persistent diarrhea. These data suggest
that persistent diarrhea is a clinical entity that may be related less to a specific enteric pathogen
and more to the health experiences of children and their home environment.

Diarrhea, a major cause of morbidity in children in devel-
oping countries [1, 2], is also associated with high health care
costs in developed areas [3]. Persistent diarrhea, defined by
the World Health Organization as illness beginning as acute
diarrhea lasting §14 days [4], is associated with retardation of
growth, morbidity, and mortality [5, 6] and is a frequent finding
in malnourished children [7]. Persistent diarrhea is responsible
for 3% – 23% of diarrheal illnesses in various locations [8] and
has been estimated to be responsible for Ç50% of all child
diarrhea deaths in northeast Brazil [9, 10].
Different enteric pathogens are associated with persistent
diarrhea. These include rotavirus, Shigella and Campylobacter
species, Cryptosporidium parvum, and enteroaggregative Esch-
erichia coli (EAggEC) [11 – 16]. The results, however, are in-
Received 20 January 1998; revised 27 May 1998.
Informed consent was obtained from the parents or guardians of the children
in the study, and human experimentation guidelines of the US Department of
Health and Human Services were followed. The study was approved by the
Soroka University Medical Center – Ben Gurion University and Columbia Uni-
versity review boards.
Financial support: National Institute of Allergy and Infectious Diseases (AI-
26497), US-Israel Binational Science Foundation (BSF 90-00257), National
Academy of Sciences/Institute of Medicine (AID/ANE 0158-G-SS-9035-00).
Reprints or correspondence: Dr. Drora Fraser, Epidemiology and Health
Services Evaluation Dept., Faculty of Health Sciences, Ben-Gurion University
of the Negev, P.O. Box 653, Beer-Sheva 84105, Israel (fdrora@bgumail
.bgu.ac.il).
The Journal of Infectious Diseases 1998;178:1081–8
q 1998 by the Infectious Diseases Society of America. All rights reserved.
0022–1899/98/7804–0022$02.00
consistent between studies and between populations. For exam-
ple, while EAggEC were identified more frequently in
persistent than in acute diarrhea in India [13], Bangladesh [14],
and Mexico [17], in other studies from Bangladesh [15], Peru
[16], and in Cambodian refugees [18] these differences were
not found. Similarly C. parvum was detected more frequently
in persistent diarrhea in Bangladesh [15] and Guinea Bissau
[19] but not in Peru [16]. The lack of consistent results regard-
ing the cause of persistent diarrhea by specific enteric agents
raises the possibility that other factors associated with the chil-
dren, their family, or their immediate environment may contrib-
ute to the risk of developing persistent diarrhea [20].
The Bedouin people who were the subjects of this study live
in the Negev region in southern Israel. This population is in
transition from a seminomadic to a settled, semiurban lifestyle.
Since the establishment of the State of Israel and the closure
of international borders in the area, the Bedouins in Israel
who were desert nomads have become more settled and are
undergoing a process of urbanization [21]. However, despite
recent changes in lifestyle, the rates of hospitalizations for
infectious diseases of infants and young children in this popula-
tion are several fold higher than for the largely urban Jewish
population living in the same geographic region and served by
the same health care services [22]. Thus, Bedouin infants õ1
year old have nearly a 6-fold risk for hospitalizations (odds
ratio [OR], 5.6; 95% confidence interval [CI], 4.9 – 6.4 P õ
.0001) when compared with Jewish infants; while in the second
year of life, the risk is 2-fold (OR, 2.0; 95% CI, 1.6 – 2.4;
P õ .0001).

/ 9d52$$oc17 08-10-98 13:16:42 jinfa UC: J Infect

1082 Fraser et al.
This study examined the role of enteric pathogens as well
as infant, family, and household characteristics in persistent
diarrhea detected in the community and in infants and young
children from a newly urbanized population.
Methods
Study location. This study was conducted between November
1989 and July 1994 in a town 35 km north of Beer-Sheva, the
capital of the Negev. The community was chosen because it was
the first urban settlement of Bedouin tribes and it has well-devel-
oped community preventive and curative health care facilities. The
town was first settled in the early 1970s and has grown from 6000
inhabitants in 1976 to 28,000 in 1995 [23]. The population growth
is largely due to the birth rate, which is the highest for any popula-
tion group in Israel [23]. The Negev region in the southern area
accounts for more than half the landmass of the country and is a
sparsely populated semiarid desert, housing only 7.6% of the
Israeli population. The Bedouins constitute about a quarter of the
Negev population. The climate is desert-like with a mean annual
daily temperature of 207C and average annual rainfall of Ç140 mm.
Intake procedures. From November 1989 to December 1993,
infants of families from the town born at the Soroka University
Medical Center in Beer-Sheva were enrolled in the study. The
medical center is the only tertiary care institution in the area and
provides both delivery and postdelivery care. All Jewish and
ú99% of Bedouin women deliver at this hospital. The study intake
was designed to ensure an even distribution of births during the
different seasons of the year. A total of 263 healthy infants, birth
weight ú2500 g, were recruited; however, 12 (4.6%) of the fami-
lies who signed informed consent could not be located or the home
could not be reached on a regular basis, leaving 251 infants in the
study.
Routine follow-ups. Throughout the study, monthly home vis-
its were made and questionnaires were completed relating to feed-
ing habits of the infant, his or her health in the preceding month,
and health of family members. At age 3 months, an environmental
assessment was done that included observations of environmental
conditions of the home and yard and the presence of animals. In
addition, the physical conditions of the home and availability of
water, toilet facilities, and electricity were noted. Other observa-
tions included the hygienic condition of the infant, kitchen, home,
and yard.
Each month, mothers were asked to place 2 consecutive infant
stool specimens in tubes containing phenol-alcohol-formaldehyde
(PAF) [24] and another stool sample into a tube without PAF. All
tubes were refrigerated until collection within 48 h. Of the samples
for collection during routine surveillance, 93.2% were obtained.
Mothers who did not have refrigerators (23%) kept the samples
in a refrigerator at a relative’s or neighbor’s home. Stool samples
were delivered in an icebox to the laboratories at Soroka University
Medical Center for testing.
Parasitology studies were done on all routine monthly samples
and on diarrhea samples. Bacteriology and rotavirus testing were
done on all diarrhea samples and on a random 1 in 4 sample of
the monthly routine specimens. Adenovirus and astrovirus were
tested only for the first 164 children in the study and were not
sought in later samples due to the low yield obtained. Wet smears
/ 9d52$$oc17 08-10-98 13:16:42 jinfa
JID 1998;178 (October)
were used for detection of Giardia lamblia cysts. Modified acid-
fast staining was used for detection of C. parvum oocysts; positive
samples were confirmed by immunofluorescent assays (Merifluor
cryptosporidium indirect immunofluorescent detection procedure;
Meridian Diagnostics, Cincinnati). Because parasitology samples
were collected in pairs, detection of Giardia or Cryptosporidium
species in one of the samples was considered a positive result.
Stool cultures for Campylobacter, Shigella, and Salmonella species
were done by routine laboratory methods.
For E. coli, specimens were inoculated onto MacConkey’s agar
and incubated for 18–24 h at 377C. Five lactose-fermenting colo-
nies were picked, confirmed as E. coli, and frozen at 0727C in
sheep erythrocytes until DNA extraction was performed. E. coli
Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020
categories were determined by DNA hybridization with radioactive
probes prepared as described in Levine et al. [25]. The following
E. coli categories were examined: EAggEC, diffuse adherent, lo-
calized adherent enteropathogenic, enterotoxigenic expressing
heat-stable toxin (ETEC-ST), enterotoxigenic expressing heat-la-
bile toxin (ETEC-LT), enterotoxigenic expressing both toxins
(ETEC-ST/LT), enterohemorrhagic, and enteroinvasive (EIEC).
Rotavirus, enteric adenoviruses, and astrovirus were examined
by an ELISA developed at the Centers for Disease Control (At-
lanta) according to methods previously described [26–28]. Only
results of stool samples for which parasitology and bacteriology
testing were done are included in the present analysis.
Diarrhea morbidity surveillance. From November 1989 to De-
cember 1992, diarrhea morbidity was identified via a surveillance
network that was established at all health care locations in the
study area. Study nurses were stationed at the local primary health
care clinics, the local rehydration unit, and at the well infant and
mother clinics (Ministry of Health) responsible for the immuniza-
tion program and for monitoring of growth and development.
Mothers of infants in the study were issued special cards with
which they identified themselves to study personnel and that en-
abled faster medical attention. These cards were presented to com-
munity clinics associated with the study and to the hospital emer-
gency room. This morbidity surveillance system was supplemented
by maternal reports at monthly home visits (method 1). From
January 1993 to the end of the study (July 1994), after the primary
health care clinic and rehydration unit nurse left the project, mor-
bidity was ascertained weekly by interviewers trained to inquire
about diarrhea by a home visit (85% of homes) or by telephone
interview (15% of homes; method 2) and by information obtained
at the monthly home visits.
All sick infants were examined by health care personnel and,
when diarrhea was diagnosed, an episode form was completed.
In addition to routine monthly samples, stool samples were also
obtained for all diarrhea episodes as soon as they were identified
and at 7- to 10-day intervals thereafter, up to 7–10 days after
each episode. During diarrhea episodes, 73.0% of the samples that
should have been obtained were collected.
Definitions and data analysis. An episode of diarrhea was de-
fined as the passing of unformed stools (watery, mucoid, or bloody)
§4 times in a 24-h period for infants õ1 month old and §3 times
in a 24-h period for older children. A new episode of diarrhea was
preceded by 3 diarrhea-free days. Persistent diarrhea was defined
as an episode beginning as acute illness and lasting §14 days. A
stool sample was defined as ‘‘onset’’ if obtained on any of the
first 3 days of symptoms. Routine monthly (non-diarrhea) samples
UC: J Infect
JID 1998;178 (October) Persistent Diarrhea in Bedouin Infants 1083

were defined as samples obtained up to 1 week before a diarrhea Table 1. Number and rates of diarrhea episodes per 100 months of
episode or those obtained §1 week from the last day of diarrhea. follow-up recorded in the study population by two different surveil-
Stool samples collected at other times were excluded from the lance methods (age of the child at onset and length of diarrheal
present analysis. illness).
Means were compared using Student’s t test or analysis of vari-
Method 1 Method 2 Both
ance (for a 3-group comparison) with Scheffe test for a posteriori
testing between groups. For contingency table analysis, x2 or Fish- Characteristic Rate (no. of episodes)
er’s exact test was used as appropriate. Rates were compared ac-
cording to the method outlined by Smith and Morrow [29]. ORs Age group (months)
and 95% CIs were computed, and multiple logistic regression was 0–5 13.4 (123) 9.2 (38) 12.1 (161)
used to determine associations between persistent diarrhea and 6 – 11 24.6 (235) 20.1 (85) 23.2 (320)
suspected risk factors. Pearson’s correlation coefficients were com- 12 – 17 18.6 (177) 21.3 (90) 19.4 (267)
puted to determine the association between maternal age and edu- 18 – 23 9.7 (91) 12.5 (40) 10.5 (131)

Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020

cation and environmental factors, feeding practices, and illness
parameters.
Results
A total of 234 (93%) of the 251 children starting follow-up
are included in this analysis. Most of the children, 192 (82.1%),
completed §2 years of follow-up; 27 (11.5%) completed 20 –
22 months and 15 (6.4%) completed 18 or 19 months of follow-
up. Total follow-up was 5339 child-months (445 child-years):
3761 (70%) under method 1 of morbidity surveillance and 1578
(30%) child-months of observation under method 2. Seventeen
families dropped out at various stages due to unwillingness to
continue or because of frequent and prolonged absences from
the home address. Most of the families (91.5%) lived in brick
houses, but 20 (8.5%) resided in a hut or the traditional Bedouin
tent. In 56 (23.9%) of the homes, the floor was beaten earth,
36 (15.4%) had outdoor toilets, 23 (9.8%) had outdoor water
supplies, and 18 (7.7%) had no electricity. In 104 (44.4%)
homes, domestic animals were found; in 116 (49.6%), grazing
animals were observed around the dwelling; and in the yards
of 28 homes (12.0%), camels or donkeys were present at the
time of the environmental assessment. All homes had access
to water supplies via Israel’s main water carrier but in 23
(9.8%) homes, the taps were located outdoors.
In the Bedouin population studied, breast-feeding is highly
prevalent and continues into the child’s second year of life.
Only 4 mothers (1.7%) never breast-fed their children. The
mean age at weaning was 11.3 months (SD, 6.7; median, 10);
however, mothers introduced milk supplements at an average
age of 3.5 months (SD, 2.6; median age, 2.0 months).
In total, 879 diarrhea episodes were detected during the sur-
veillance (table 1). Under method 1, 626 episodes were re-
corded for a rate of 16.6 per 100 child months of follow-up.
By method 2, the detection rate was 16.0 per 100 child months
of follow-up (253 episodes). Persistent diarrhea episodes, last-
ing §14 days, constituted 19.6% of the diarrhea episodes under
method 1 and 14.6% under method 2. There were no statisti-
cally significant differences in detection by age at onset or by
length of episode between the methods. Therefore, all data
were pooled for the following analyses. Ninety-four children
had persistent diarrhea. Of those children, 53 had 1 episode,
All ages 16.6 (626) 16.0 (253) 16.5 (879)
Length of illness (days)
1–7 9.2 (347) 9.8 (154) 9.4 (501)
8 – 13 4.2 (156) 3.9 (62) 4.1 (218)
§14 3.3 (123) 2.3 (37) 3.0 (160)
NOTE. Rates were compared to Smith and Morrow [29]. No statistically
significant differences were found between the 2 methods for rates of reporting
of diarrheal illness either by age at onset or by length of illness.
29 had 2, 6 had 3, and 6 children had 4 – 7 persistent diarrhea
episodes.
Although, on average, persistent diarrhea constituted only
18.2% of all diarrhea episodes, 45.2% of all days of illness
were from persistent episodes (figure 1), with the highest rates
recorded at age £3 months. At this age, however, we found
no enteric parasites or viruses, ETEC-ST, EIEC, or shigellae.
Other enteric agents were detected at lower rates at £3 months
than at older ages.
The rates at which enteric agents were detected were com-
pared between those for non-diarrhea routine stool samples and
stools samples obtained at diarrhea onset (table 2). C. parvum,
Campylobacter species, and rotavirus were found more fre-
quently in samples obtained during diarrheal illness. When
detection rates of all ETEC-expressing heat-stable toxin colo-
nies were considered together, they were found significantly
more frequently in diarrhea-onset stools (4.5%) than in routine
non-diarrhea samples (1.7%; OR, 2.76; 95% CI, 1.03 – 7.86).
In contrast, G. lamblia was more often detected in routine than
in diarrhea specimens.
Only 40.3% (509/1262) of routine stools and 35.8% (68/190)
of diarrhea-onset stool samples were free of any of the enteric
agents tested for in this study. Thus, even in routine, non-diarrhea
samples, in 6 of any 10 samples at least 1 enteric agent was
detected. In addition, in 33.8% (254/752) of all routine samples
positive for an enteric pathogen, §2 pathogens were detected. In
onset samples, §2 pathogens were detected in 30.5% (36/118)
of all samples in which an enteric agent was detected.
No significant changes in the direction or magnitude of the
difference between routine and onset samples was found when
the analysis shown in table 2 was repeated for samples con-
taining only a single pathogen. The exception was for Shigella
species where the difference became statistically significant

/ 9d52$$oc17 08-10-98 13:16:42 jinfa UC: J Infect

1084 Fraser et al.
(OR, 15.4; 95% CI, 1.09 – 430.7; P Å .037). The statistically
significant associations shown in table 2 were not found be-
cause the lower number of observations in the analysis was
restricted to samples with single organisms.
In order to define the enteric pathogens associated with per-
sistent diarrhea, the detection rate of enteric agents was deter-
mined in samples obtained at the onset of diarrhea according
to length of illness (table 3). None of the enteric pathogens were
identified more frequently in persistent diarrhea. All ETEC-ST
colonies combined were detected in 9.5% of persistent samples
versus in 1.7% in non-diarrhea stools (P Å .057). C. parvum
was found significantly more often in episodes that lasted £14
days and was detected only in 1 of the 52 sets of samples
obtained during persistent illness. Rotavirus and Campylo-
bacter species were associated with episodes lasting £1 week,
while Campylobacter species and astrovirus were associated
with illness lasting 8 – 13 days. When the analysis was repeated
only for samples in which a single pathogen was detected
(66.2% of routine and 69.5% of onset samples), no significant
change in the magnitude or direction of the differences between
the diarrhea groups and routine samples was found, but the
statistical significance of the differences was not retained due
to the decreased number of observations. Shigellae were an
exception and were significantly associated with diarrhea last-
ing 8 – 13 days (P Å .003).
To investigate the possibility that sequential infections may
contribute to the occurrence of persistent diarrhea, we exam-
/ 9d52$$oc17 08-10-98 13:16:42 jinfa
JID 1998;178 (October)
Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020
Figure 1. Distribution of days ill with per-
sistent diarrhea and persistent diarrhea by age
in months.
ined persistent episodes for which stool samples were obtained
during the first week of illness and at least once during one of
the following 2 weeks. In 22.2% (14/63) of those episodes,
different enteric pathogens were detected between the first
week and subsequent weeks.
Other factors that may contribute to the incidence of persis-
tent diarrhea, such as infant, family, and environmental factors,
were also sought. In order to do so, the status of each child
was defined according to the longest episode of diarrhea experi-
enced. Ninety-four children experienced §1 persistent diarrhea
episode. Sixty-six children had §1 diarrhea episode lasting
8 – 13 days and no episodes of persistent illness; 55 children
had diarrhea episodes lasting £7 days. For 19 children, no
diarrheal illness of any length was recorded. All results pre-
sented from here on refer to analyses that included each child
only once, according to status defined by the longest diarrhea
episode experienced by the child.
The mean number of diarrhea episodes experienced, the
total days of illness, and the age at the first recorded episode
were compared between the groups by the longest episode
of diarrhea experienced (table 4). All comparisons were sta-
tistically significant and showed clear trends. Thus, children
who during follow-up had §1 persistent episode experienced
the most diarrhea episodes (mean, 5.8), had the most days
of illness (nearly 2 months in total), and had the first episode
of diarrhea at the youngest age (mean, 4.7 months) compared
with other children.
UC: J Infect
JID 1998;178 (October) Persistent Diarrhea in Bedouin Infants 1085

Table 2. Detection of enteric pathogens in nondiarrhea routine sam- Table 3. Detection of enteric pathogens at onset of diarrhea in
ples and during onset of diarrheal illness. diarrhea illness by length of episode and in nondiarrhea stools.

Routine Onset Routine Length of episode (days)

stool stool stool
samples samples Enteric pathogens samples 1–7 8 – 13 §14

Enteric pathogen % (no. positive) OR (95% CI) P* Giardia lamblia 24.6 (310) 17.2 (24) 19.0 (11) 13.5 (7)
Cryptosporidium parvum 1.7 (22) 5.0* (7) 6.8* (4) 1.9 (1)
Giardia lamblia 24.6 (310) 17.0 (42) 0.62 (0.43 – 0.90) .011 EAggEC† 26.0 (215) 20.3 (16) 28.1 (9) 33.3 (7)
Cryptosporidium DAEC† 24.2 (200) 20.3 (16) 15.6 (5) 9.5 (2)
parvum 1.7 (22) 4.8 (12) 2.85 (1.31 – 6.11) .006 LA-EPEC† 6.7 (55) 7.6 (6) 12.5 (4) 4.8 (1)
EAggEC 26.0 (215) 24.2 (32) 0.91 (0.58 – 1.42) NS ETEC-ST† 1.2 (10) 2.6 (2) 0 4.8 (1)
DAEC 24.2 (200) 17.4 (23) 0.66 (0.40 – 1.09) NS ETEC-LT† 8.0 (66) 3.8 (3) 9.4 (3) 4.8 (1)

Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020

LA-EPEC 6.7 (55) 8.3 (11) 1.27 (0.67 – 2.60) NS ETEC-ST/LT† 0.5 (4) 1.3 (1) 3.1 (1) 4.8 (1)
ETEC-ST 1.2 (10) 2.3 (3) 1.90 (0.41 – 7.59) NS EHEC† 0.4 (3) 0 3.1 (1) 0
ETEC-LT 8.0 (66) 5.3 (7) 0.64 (0.26 – 1.50) NS EIEC† 0.1 (1) 0 0 0
ETEC-ST/LT 0.5 (4) 2.3 (3) 4.78 (0.84 – 25.5) NS Campylobacter species 9.0 (114) 15.2* (15) 20.0* (8) 3.7 (1)
EHEC 0.4 (3) 0.8 (1) 2.09 (0.04 – 26.3) NS Shigella 0.6 (7) 0 5.0* (2) 0
EIEC 0.1 (1) 0 — — NS Salmonella 0.4 (5) 1.0 (1) 0 0
Campylobacter Rotavirus 2.0 (19) 7.0* (5) 6.3 (2) 5.3 (1)
species 9.0 (114) 14.5 (24) 1.70 (1.03 – 2.79) .038 Astrovirus 2.3 (11) 2.1 (1) 12.5* (3) 0
Shigella 0.6 (7) 1.2 (2) 2.20 (0.22 – 11.7) NS Enteric adenovirus 0.4 (2) 2.1 (1) 0 0
Salmonella 0.4 (5) 0.6 (1) 1.53 (0.03 – 13.8) NS
Rotavirus 2.0 (19) 6.6 (8) 3.51 (1.38 – 8.71) .007 NOTE. Data are % (no. positive).
Astrovirus 2.3 (11) 4.5 (4) 2.04 (0.53 – 7.14) NS * P õ .05 compared with routine stool samples.
†
Enteric adenovirus 0.4 (2) 1.1 (1) 2.73 (0.05 – 52.8) NS E. coli categories were as follows: enteroaggregative, EAggEC; diffuse
adherent, DAEC; localized adherent enteropathogenic, LA-EPEC; enterotoxi-
genic expressing heat-stable toxin, ETEC-ST; enterotoxigenic expressing heat-
NOTE. Data are rates (%) and odds ratios (OR) comparing onset to routine
labile toxin, ETEC-LT; enterotoxigenic expressing both toxins, ETEC-ST/LT;
detection and 95% confidence intervals (95% CI).
enterohemorrhagic, EHEC; and enteroinvasive, EIEC.
* P ú .05 are noted as NS (not significant).
†
E. coli categories were as follows: enteroaggregative, EAggEC; diffuse
adherent, DAEC; localized adherent enteropathogenic, LA-EPEC; enterotoxi-
genic expressing heat-stable toxin, ETEC-ST; enterotoxigenic expressing heat-
labile toxin, ETEC-LT; enterotoxigenic expressing both toxins, ETEC-ST/LT; final model. Increased maternal age or years of maternal educa-
enterohemorrhagic, EHEC; and enteroinvasive, EIEC. tion resulted in a lower risk of persistent diarrhea for the child
as did a delay in the first diarrhea episode. Thus, for every
We assessed the relationship between infant, family, and 10-year increase in maternal age, the risk fell by 46%, and if
environmental factors and persistent diarrhea (table 5). While a mother completed primary school, the risk for persistent diar-
young age at first diarrhea episode and low maternal age and rhea in her child fell by 44%.
education at enrollment were associated with increased risk for Correlation coefficients were used to examine the association
persistent diarrhea, children living in a hut or tent, the presence between maternal age and education and other factors. Low
of an outdoor toilet, and breast-feeding in the second year of maternal education was associated with higher parity (P õ
life lowered the risk for persistent diarrhea. The associations
between dwelling type and maternal education and persistent
diarrhea were close to statistical significance (P õ .10). None Table 4. Number of diarrhea episodes, total days of diarrheal illness,
of the other variables were associated with §1 persistent ill- and age of child at first diarrhea episode in three groups of children.
ness: exposure to animals; physical and hygienic conditions of
Longest diarrhea episode
the dwelling, yard, and kitchen; food and kitchen utensil storage experienced (days)
conditions; infant hygiene; weaning before age 3 months; and
other family and infant variables, such as infant gender, family 1–7 8 – 13 §14 P
size, and birth order.
No. of children 55 66 94
All of the variables in table 5 were available for stepwise
No. of diarrhea
analysis, but as shown in table 6, only age at first diarrheal episodes 2.0 (1.3) 3.4 (2.2) 5.8 (3.1) õ.001*
illness and maternal age and maternal education were indepen- Total days with
dently and significantly associated with the risk of §1 persis- diarrhea 7.8 (5.9) 21.5 (11.8) 57.5 (35.3) õ.001*
tent diarrhea episode. Maternal education was close to statisti- Age at first diarrhea
episode, months 8.8 (4.4) 7.0 (4.0) 4.7 (3.4) õ.001*
cal significance (P Å .064). Since maternal age and education
were inversely related (i.e., young mothers had more years of NOTE. Data are mean (SD).
education), the presence of both variables was required in the * All means significantly different.

/ 9d52$$oc17 08-10-98 13:16:42 jinfa UC: J Infect

1086 Fraser et al. JID 1998;178 (October)

.001), older age at enrollment (P õ .001), more days of diarrhea Table 6. Multivariate logistic regression defining the risk factors
(P Å .036), more diarrhea episodes (P Å .031), having an associated with persistent diarrhea in Bedouin children.
outdoor water supply (P Å .010), having an outdoor toilet (P
Characteristic b OR 95% CI P
Å .009), animal excrement in the yard (P Å .001), infant ap-
pearing very or moderately dirty (P Å .001), and breast-feeding Age at first diarrhea
beyond age 1 year (P Å .001). Young maternal age at enroll- episode (months) 00.2155 0.81 0.74 – 0.88 õ.001
ment was also associated with more days of illness with diar- Maternal age at
enrollment (years) 00.0627 0.94 0.90 – 0.98 .006
rhea (P õ .001) and more diarrhea episodes (P Å .002) in the
Maternal education (years) 00.0699 0.93 0.87 – 1.00 .064
follow-up period.
NOTE. OR, odds ratio; CI, confidence interval.

Discussion

Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020

Bedouins residing in southern Israel provide an opportunity half of the children by supplemental milk-feeding from age 2
to examine factors associated with pediatric persistent diarrhea months. We examined the prevalence of enteric agents before
in a community undergoing urbanization. This study, which age 3 months and found that most enteric agents were either
followed infants from birth to age 2 years, investigated the nondetectable or were detected at much lower levels than in
associations between specific enteric agents, infant characteris- older children. The low prevalence of enteric agents at this age
tics, family and environmental factors, and persistent diarrhea. is probably due to the combined effects of the passive protec-
While we did not find any association between persistent diar- tion young infants have from their mothers and from lower
rhea and specific enteric pathogens, we found that the age at exposure rates. This finding together with the high rates of
which a child experienced the first diarrhea episode and mater- milk supplementation given at an early age suggest that it is
nal age and education were strongly associated with the risk unlikely that a high proportion of persistent diarrhea episodes
of persistent diarrhea during the first 2 years of life. at this age is due to specific enteric pathogens, and other causes
Before age 3 months, persistent diarrhea was responsible for need to be sought.
76% of days of diarrhea illness; persistent diarrhea accounted Our surveillance methods may have underestimated the over-
for 43% of all diarrhea episodes detected. Similar findings of all number of episodes of diarrhea in the community, although
higher rates of detection of persistent diarrhea in the very young both (one passive — relying on use of health care services and
have come from India, Bangladesh, and Guatemala [13, 14, one active — weekly surveillance) yielded similar rates of diar-
30]. In the very young, much persistent diarrhea (sometimes rhea both by age at onset and by length of episode.
termed intractable diarrhea of infancy) is believed to result Several enteric agents were detected more frequently in stool
from milk protein intolerance either as a result of exposure to samples obtained during diarrheal illness: C. parvum, Campylo-
cow’s milk or soy protein or following an infection [31]. In bacter species, rotavirus, and ETEC expressing heat-stable tox-
this population, breast-feeding is highly prevalent and contin- ins. These enteric agents have been associated with diarrheal
ues into the second year of life accompanied, however, in illness in pediatric and adult populations worldwide [15, 19,
32, 33]. In contrast G. lamblia in this study was found more
frequently in non-diarrhea than in diarrhea specimens. Our
findings, which are similar to those from Guinea Bissau and
Table 5. Association between selected maternal, infant, and house-
hold factors and persistent diarrhea in Bedouin children by univariate Bangladesh [15, 19], question the role of Giardia as a pathogen
logistic regression analysis. in areas in which diarrhea is endemic [34].
In our search for enteric agents associated with persistent
Characteristic b OR 95% CI P diarrhea, we did not detect higher rates for any of the enteric
agents examined. EAggEC, which have been implicated in
Age at first diarrhea
episode (months) 00.2177 0.80 0.74 – 0.87 õ.001 persistent diarrhea in India [13], Bangladesh [14], and Mexico
Maternal age at [17], were detected frequently both in non-diarrhea (26%) and
enrollment (years) 00.0576 0.94 0.91 – 0.98 .004 in diarrhea-related samples (20.3% – 33.3%). This E. coli cate-
Child breast-fed beyond gory was the enteric agent most frequently detected in the
1 year of age (1 Å
Bedouin pediatric population but was unrelated to diarrheal
yes, 0 Å no) 00.6427 0.53 0.31 – 0.90 .019
Outdoor toilet (yes Å 1, illness. This is consistent with studies in Bangladeshi, Peruvian,
indoor Å 0) 00.8139 0.44 0.20 – 0.99 .048 and Cambodian refugee children that did not find an association
Living in hut or tent between EAggEC and persistent diarrhea [15, 16, 18]. In our
(yes Å 1, house Å 0) 01.0658 0.34 0.11 – 1.06 .064 study, DNA probes were used to define the different E. coli
Maternal education
categories and may contribute to the difference between our
(years) 00.0531 0.95 0.89 – 1.01 .088
findings and those of other studies. In most published studies
NOTE. OR, odds ratio; CI, confidence interval. in which EAggEC were more frequently detected in persistent

/ 9d52$$oc17 08-10-98 13:16:42 jinfa UC: J Infect

JID 1998;178 (October) Persistent Diarrhea in Bedouin Infants
diarrhea than in subjects with acute diarrhea or in controls, the
HEp-2 assay was used. However, in one prospective study
from Peru in which the HEp-2 assay was used, no association
between EAggEC and persistent diarrhea was found [16].
While C. parvum was associated with diarrhea in general and
with diarrhea episodes lasting £13 days, it was only detected in
1 of the 52 sets of stool samples obtained at the onset of persistent
diarrhea. This finding contradicts results from other studies in
which this enteric agent was associated with persistent diarrhea
[15, 19] but is in agreement with a report that failed to find
association between C. parvum and persistent diarrhea [16].
Most available data on the role of enteric pathogens in persis-
tent diarrhea are from studies in developing populations. The
Bedouins studied differ in that they have access to piped water
and to modern preventive and curative medical services. The
Bedouin population should be considered as one that is interme-
diate between developing and developed populations as indi-
cated by their transition from seminomadism to urbanization.
Thus, our results may not be directly applicable to developing
populations. Nevertheless, our findings add to the knowledge
of geographic and intrapopulation differences regarding patho-
genicity of various enteric pathogens around the globe. Those
differences may be related to density of exposure, host charac-
teristics (e.g., extent of breast-feeding and malnutrition), or to
strain differences in different areas. Answers for these geo-
graphic differences may be obtained from multipopulation
studies with unified protocols and standardized laboratory
methods.
Regardless of enteric pathogens, factors associated with the
child’s immediate environment and diarrhea experience con-
tribute to the risk of persistent illness. Maternal education and
age were both associated with reduced risk for persistent illness
in our population and in others. In Zaire, for example, education
was strongly and inversely associated with risk of diarrhea
[35], while maternal age was inversely related to diarrhea in
Cali, Columbia [36]. In a study in Guinea Bissau, fewer moth-
ers of children with diarrhea had attended school than the moth-
ers of controls [19]. Low levels of education and young mater-
nal age were risk factors for persistent diarrhea in Burma and
Chile, respectively [37, 38]. Thus, the association between ma-
ternal education and age are consistent between different popu-
lations around the world.
When the correlations of maternal age and maternal educa-
tion with other factors were examined in our study, it became
apparent that maternal age and education were strongly associ-
ated with variables related to living conditions, the overall
diarrhea experience of the child during the observation period,
household hygiene practices, family size, and breast-feeding
practices. We have little doubt that these factors reflect the
overall level of knowledge and behavior in the home, including
habits that are related to the prevention and treatment of illness,
in particular diarrhea. Maternal practices during diarrheal ill-
ness were recently shown to affect the length of diarrhea epi-
sodes in a study in Kenya [39].
/ 9d52$$oc17 08-10-98 13:16:42 jinfa
1087
In addition to maternal age and education level, we found
that the child’s experience of diarrhea at a young age was the
strongest risk factor for persistent diarrhea. The child’s age at
first diarrhea episode may act as an indicator for the overall
diarrhea experience of the child during the first 2 years of life.
Thus, those with first illness at a younger age will accumulate
more episodes and more days of illness. Age at first diarrhea
episode may also be an expression for the density of exposure
and for diarrhea treatment provided. In Italy, early onset of
diarrhea was associated with increased incidence of persistent
diarrhea in hospitalized children [40], while in Bangladesh,
recent diarrhea episodes were associated with persistent diar-
Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020
rhea in a case-control study [41].
An important predictor of persistent diarrhea in developing
countries is malnutrition [18, 42, 43]. In our study, anthropometric
measures were obtained at the maternal and child health clinic,
usually during an immunization visit unrelated to illness. These
data therefore cannot be used to investigate the role of undernutri-
tion as a predictor for persistent diarrhea in the present study.
After reviewing the available data, we propose that persistent
diarrhea is a clinical entity that may be related less to a specific
enteric agent and be more reflective of the health experiences of
a given child and the child’s home environment. It is possible
that an infectious event initiates this process, resulting in malaise,
loss of appetite, and changes in the intestinal mucosa. This leads
to loss of fluids and difficulties in nutrient absorption. In the
weakened nutritional and immunologic state resulting, one of the
explanations for persistent diarrhea raised by Penny [44] may
come into play, namely a secondary event or sequential infections
may cause the prolonged illness. In populations such as ours,
where the background level of exposure to enteric agents is high
(60% of non-diarrhea routine samples contained one of the enteric
pathogens examined), the probability of secondary events or se-
quential infections resulting in persistent illness is high as shown
by our data. This would be especially true in a home environment
in which preventive practices are not practiced. This hypothesis
is consistent with data indicating lack of consistent association
between specific enteric agents and persistent diarrhea found in
different populations and different studies. It is also consistent
with data showing associations between caregiver characteristics
and persistent illness that have been described in studies world-
wide. The caregiver characteristics influence both the degree of
exposure the child has to various enteric agents and the modes
of treatment undertaken.
In conclusion, shifting attention from the search for a specific
enteric agent that causes persistent diarrhea to examining the
immediate environment of the children themselves will allow
us to gather appropriate data to plan community-based interven-
tions. The aim of these interventions would be to reduce the
heavy burden of persistent diarrhea and its consequences in
high-risk populations.
References
1. Claeson M, Merson MH. Global progress in the control of diarrheal dis-
eases. Pediatr Infect Dis J 1990; 9:345 – 55.
UC: J Infect
1088 Fraser et al.
2. Bern C, Martines J, de Zoysa I, Glass RI. The magnitude of the global
problem of diarrhoeal disease: a ten year update. Bull World Health
Organ 1992; 70:705 – 14.
3. Glass RI, Lew JF, Gangarosa RE, LeBaron CW, Ho MS. Estimates of
morbidity and mortality rates for diarrheal disease in American children.
J Pediatr 1991; 118:527 – 33.
4. World Health Organization. Diarrhoeal diseases control. Persistent diar-
rhoea in children. Geneva, WHO, 1987; CDD/DDM/85.1.
5. Bhan MK, Arora NK, Ghai KR, Khoshoo V, Bhandari N. Major factors
in diarrhoea related mortality among rural children. Indian J Med Res
1986; 83:9 – 12.
6. Black RE, Brown KH, Becker S. Effects of diarrhea associated with spe-
cific enteropathogens on the growth of children in rural Bangladesh.
Pediatrics 1984; 73:799 – 805.
7. Faveau V, Henry FJ, Briend A, Yunus M, Chakraborty J. Persistent diar-
rhea as a cause of childhood mortality in rural Bangladesh. Acta Paediatr
1992; 81(suppl 381):12 – 14.
8. Black RE. Persistent diarrhea in children of developing countries. Pediatr
Infect Dis J. 1993; 12:751 – 61.
9. McAuliffe JR, Shields DS, De Sousa MA, Sakell J, Schorling J, Guerrant
RL. Prolonged and recurring diarrhea in the northeast of Brazil: exami-
nations of cases from a community-based study. J Pediatr Gastroenterol
Nutr 1986; 5:902 – 6.
10. Lima AAM, Fang G, Schorling JB, et al. Persistent diarrhea in northeast
Brazil: etiologies and interaction with malnutrition. Acta Paediatr Pedi-
atr 1992; 81(suppl 381):39 – 44.
11. Khoshoo V, Bhan MK, Jayashree S, Kumar R, Glass RI. Rotavirus infec-
tion and persistent diarrhoea in young children [letter]. Lancet 1990;
336:1314 – 5.
12. Huttly SRA, Hoquet BA, Aziz MA, et al. Persistent diarrhoea in a rural
area of Bangladesh: a community-based longitudinal study. Int J Epide-
miol 1989; 18:964 – 9.
13. 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.
14. Henry FJ, Udoy AS, Wanke CA, Aziz K. Epidemiology of persistent
diarrhea and etiologic agents in Mirzapur, Bangladesh. Acta Pediatr
1992; 81(suppl 381):27 – 31.
15. Baqui AH, Sack RB, Black RE, et al. Enteropathogens associated with
acute and persistent diarrhea in Bangladeshi children under 5 years of
age. J Infect Dis 1992; 166:792 – 6.
16. Lanata CF, Black RE, Maurtua D, et al. Etiologic agents in acute vs.
persistent diarrhea in children under three years of age in periurban
Lima, Peru. Acta Paediatr 1992; 81(suppl 381):32 – 8.
17. Cravioto A, Tello A, Navarro A, et al. Association of Escherichia coli
HEp-2 adherence patterns with type and duration of diarrhoea. Lancet
1991; 337:262 – 4.
18. Arthur JD, Bodhidatta L, Echeverria P, Phuphaisan S, Sakti P. Diarrheal
disease in Cambodian children at a camp in Thailand. Am J Epidemiol
1992; 135:541 – 51.
19. Molbak K, Wested N, Hojlyng N, et al. The etiology of early childhood
diarrhea: a community study from Guinea-Bissau. J Infect Dis 1994;
169:581 – 7.
20. Bentley ME. Household behaviors in the management of diarrhea and
their relevance for persistent diarrhea. Acta Paediatr 1992; 81(suppl
381)49 – 54.
21. Bar-Zvi S. Characteristics of the Bedouin’s life in the Negev — prior to
their sedenterization (in Hebrew) In: Gradus Y, Shmueli A, eds. Tel-
Aviv, Israel: Eretz Hanegev, 1979:621 – 30.
22. Levy A, Fraser D, Vardi H, Dagan R. Hospitalizations for infectious
diseases in Jewish and Bedouin children in southern Israel. Eur J Epide-
miol 1998; 14:179 – 86.
/ 9d52$$oc17 08-10-98 13:16:42 jinfa
JID 1998;178 (October)
23. Central Bureau of Statistics. Statistical abstracts of Israel, no. 47. Jerusa-
lem: Hemed Press, 1996.
24. El-On J, Dagan R, Fraser D, Deckelbaum RJ. Detection of Cryptosporid-
ium and Giardia intestinalis in Bedouin children from southern Israel.
Int J Parasitol 1994; 24:409 – 11.
25. Levine MM, Ferreccio C, Prado V, et al. Epidemiologic studies of E. coli
diarrheal infections in low socioeconomic level periurban community
in Santiago, Chile. Am J Epidemiol 1993; 138:849 – 69.
26. Grohmann G, Glass RI, Gold J, et al. Outbreak of human calicivirus
gastroenteritis in a day-care center in Sidney, Australia. J Clin Microbiol
1991; 29:544 – 50.
27. Moe CL, Allen JR, Monroe, SS et al. Detection of astrovirus in pediatric
stool; samples by immunoassay and RNA probe. J Clin Microbiol 1991;
29:2390 – 5.
Downloaded from https://academic.oup.com/jid/article/178/4/1081/837903 by guest on 29 September 2020
28. Grohmann GS, Glass RI, Pereira HG, et al. Enteric viruses and diarrhea
in HIV-infected patients. N Engl J Med 1993; 329:14 – 20.
29. Smith PG, Morrow RH. Methods for field trials of interventions against
tropical diseases. A toolbox. New York: Oxford University Press, 1991.
30. Baqui AH, Black RE, Sack RB, Yunus MD, Siddique AK, Chowdhury
HR. Epidemiological and clinical characteristics of acute and persistent
diarrhea in rural Bangladeshi children. Acta Paediatr 1992; 81(suppl
381):15 – 21.
31. Vanderhoof JA. Diarrhea, In: Wyllie R, Hyams JS eds. Pediatric Gastroin-
testinal Disease. Pathophysiology, Diagnosis, Management. Philadel-
phia: W. B. Saunders, 1993; 187 – 97.
32. MacKenzie WR, Hoxie NJ, Proctor ME, et al. A massive outbreak in
Milwaukee of Cryptosporidium infection transmitted through the public
water supply. N Engl J Med 1994; 331:161 – 7.
33. Wolf MK, Taylor DN, Boedeker EC, et al. Characterization of enterotoxi-
genic Escherichia coli isolated from US troops deployed in the Middle
East. J Clin Microbiol 1993; 31:851 – 6.
34. Flanagan PA. Giardia — diagnosis, clinical course and epidemiology. A
review. Epidemiol Infect 1992; 109:1 – 22.
35. Dikassa L, Mock N, Magnani R, et al. Maternal behavioural risk factors
for severe childhood diarrhoeal disease in Kinshasa, Zaire. Int J Epide-
miol 1993; 22:327 – 33.
36. Bertrand WE, Walmus BF. Maternal knowledge, attitudes and practice as
predictors of diarrhoeal disease in young children. Int J Epidemiol 1983;
12:205 – 10.
37. U KM, Khin M, Wai NN, Hman NW, Myint TT, Butler T. Risk factors
for the development of persistent diarrhoea and malnutrition in Burmese
children. Int J Epidemiol 1992; 21:1021 – 9.
38. Araya M, Baiocchi N, Espinoza J, Brusner O. Persistent diarrhoea in the
community. Characteristics and risk factors. Acta Paediatr Scand 1991;
80:181 – 9.
39. Mirza NM, Caulfield LE, Black RE, Macharia WM. Risk factors for
diarrheal duration. Am J Epidemiol 1997; 146:776 – 85.
40. Guarino A, Spagnuolo MI, Russo S, et al. Etiology and risk factors of
severe and protracted diarrhea. J Pediatr Gastroenterol 1995; 20:
173 – 8.
41. Sazawal S, Bhan MK, Bhandari N, Clemens J, Bhatnagar S. Evidence for
recent diarrhoeal morbidity as a risk factor for persistent diarrhoea: a
case-control study. Int J Epidemiol 1991; 20:540 – 5.
42. Baqui AH, Sack RB, Black RE, Chowdhury HR, Yunus M, Siddique AK.
Cell-mediated immune deficiency and malnutrition are independent risk
factors for persistent diarrhea in Bangladeshi children. Am J Clin Nutr
1993; 58:543 – 8.
43. Lima AA, Fang G, Schorling JB, et al. Persistent diarrhea in northeast
Brazil: etiologies and interactions with malnutrition. Acta Paediatr 1992;
81(suppl 381):39 – 44.
44. Penny ME. Commentary. Persistent diarrhea in children of developing
countries. Pediatr Infect Dis J 1993; 12:762 – 4.
UC: J Infect