Fever After a Stay in the Tropics
Diagnostic Predictors of the Leading Tropical Conditions
Emmanuel Bottieau, MD, Jan Clerinx, MD, Erwin Van den Enden, MD, Marjan Van Esbroeck, MD,
Robert Colebunders, MD, PhD, Alfons Van Gompel, MD, and Jef Van den Ende, MD, PhD
Abstract: Differential diagnosis of fever in travelers returning from
the tropics is extremely diverse. Apart from the travel destination,
other diagnostic predictors of tropical infections are poorly
documented in returning travelers. From April 2000 to December
2005, we prospectively enrolled all patients presenting at our
referral centers with fever within 1 year after visiting a tropical or
subtropical area. For clinical relevance, the diagnostic predictors of
the leading tropical conditions were particularly investigated in the
febrile episodes occurring during travel or within 1 month after
return (defined as early-onset fever).
In total, 2071 fever episodes were included, occurring in 1962
patients. Most patients were western travelers (60%) or expatriates
(15%). Regions of exposure were mainly sub-Saharan Africa (68%)
and southern Asia/Pacific (14%). Early-onset fever accounted for
1619 episodes (78%). Most tropical infections were related to
specific travel destinations. Malaria (mainly Plasmodium falcipa-
rum) was strongly predicted by the following features: enlarged
spleen, thrombocytopenia (platelet count <150
103/mL), fever
without localizing symptoms, and hyperbilirubinemia (total biliru-
bin level 1.3 mg/dL). When malaria had been ruled out, main
predictors were skin rash and skin ulcer for rickettsial infection
(mainly African tick bite fever); skin rash, thrombocytopenia, and
leukopenia (leukocyte count <4
103/mL) for dengue; eosinophil
count 0.5
103/mL for acute schistosomiasis; and enlarged spleen
and elevated alanine aminotransferase level (70 IU/L) for enteric
fever.
The initial clinical and laboratory assessment can help in
selecting appropriate investigations and empiric treatments for
patients with imported fever.
(Medicine 2007;86:18–25)
Abbreviations: LR = likelihood ratio; OR = odds ratio.
From Department of Clinical Sciences (EB, JC, EVdE, MVE, RC, AVG,
JVdE), Institute of Tropical Medicine, Antwerp; Unit of Tropical
Medicine (RC, JVdE), University Hospital Antwerp; and Faculty of
Medicine (RC), University of Antwerp; Antwerp, Belgium.
Address reprint requests to: Dr. Emmanuel Bottieau, Institute of Tropical
Medicine, Nationalestraat 155, 2000 Antwerp, Belgium. Fax: 32(0)3/
247.64.52; e-mail: EBottieau@itg.be.
Copyright n 2007 by Lippincott Williams & Wilkins
ISSN: 0025-7974/07/8601-0018
DOI: 10.1097/MD.0b013e3180305c48
18
Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
INTRODUCTION
The differential diagnosis of fever in travelers return-
ing from the tropics remains a challenge, particularly for
physicians less familiar with exotic pathology. Clinicians
face patients possibly exposed to various tropical pathogens,
and must consider many cosmopolitan infections as well
(infections with worldwide distribution)10. Most tropical
infections have a nonspecific presentation, and some may
evolve rapidly into severe disease if not adequately treated16.
Therefore, optimal management should aim at recognizing
promptly the most serious conditions, while minimizing
unnecessary workup and treatment for milder diseases7.
Morbidity profiles in international travelers vary
significantly according to the region of exposure10. In a
recent 5-year study2 investigating the etiology of fever after
a stay in the tropics in 1842 returning travelers or migrants,
we demonstrated that the clinical spectrum depended not
only on the geographic risk, but also on other factors such as
the type of traveler and the lapse between exposure and onset
of fever. Since the epidemiology of imported fever
considerably overlaps, additional elements of the initial
assessment are necessary to steer the diagnostic and
therapeutic strategy. Diagnostic predictors of several tropical
diseases have been investigated in endemic settings5, but
only incompletely in returning travelers, except for ma-
laria3,8,9,21. We conducted the current study to identify
in febrile travelers and migrants the clinical and lab-
oratory features that could help in diagnosing the most
frequent tropical conditions, and to quantify their predictive
contribution.
PATIENTS AND METHODS
Study Setting and Patients
From April 2000 to December 2005 we prospectively
recruited all consecutive patients attending the Institute of
Tropical Medicine (outpatient travel clinic, national refer-
ence center) and the University Hospital (emergency ward
and inpatient Unit of Tropical Diseases, tertiary-care
hospital) of Antwerp, Belgium, if they presented with fever
within 1 year after visiting a tropical or subtropical area. The
epidemiology and the etiologic spectrum of the first 1842
febrile episodes (from April 2000 to March 2005) have been
recently reported2.
Medicine  Volume 86, Number 1, January 2007
Medicine  Volume 86, Number 1, January 2007 Diagnostic Predictors of Imported Tropical Conditions

Patients were informed of the objectives of this
observational study. Data were rendered anonymous accord-
ing to the Belgian legislation. The protocol was approved by
the ethical committee of our institution.
Definitions and Diagnostic Procedure
The definitions and diagnostic procedure have been
extensively described in a previous publication2. Briefly,
fever was defined by an axillary temperature 38 8C, or by a
combination of febrile sensation and chills and sweats,
within 3 days before consultation. A new febrile episode was
defined in the same patient if fever and associated symptoms
had completely abated during at least 1 week. Tropics and
subtropics corresponded to any nonindustrialized country at
least partly situated between 358 North and 358 South
latitude.
Patients were divided in 4 categories: western travelers
(natives of western countries who stayed less than 6 months
in the tropics); expatriates (western patients living abroad for
more than 6 months); visiting friends and relatives (VFR)
travelers (natives of the tropics who had been residing for >1
year in Europe and returned for a visit of <6 months to their
country of origin to visit friends and relatives); and foreign
visitors/migrants (natives of the tropics arriving for the first
time in Europe).
Clinical data were collected during the consultation
using electronic preformatted case-record forms. All patients
were subjected to a standardized first-line laboratory
screening set including a total blood cell count with
differential, liver and kidney function tests, and inflamma-
tory parameters. Diagnostic investigations were performed
according to the clinician’s decision.
Final diagnoses were either etiologic or clinical, and
they were classified as tropical conditions, cosmopolitan
infections (infections with worldwide distribution), and non-
infectious diseases according to strict case definitions2.
Febrile episodes that did not fulfill the criteria for etiologic
or clinical diagnosis were considered as ‘‘unknown cause.’’
Diagnosis of the Leading Tropical Conditions
The leading tropical diseases, on which this study was
focused (Table 1), were diagnosed according the following
definitions:
1. Malaria: demonstration of Plasmodium trophozoites in
blood smears and/or positive rapid diagnostic (2- or 3-
band) test13;
2. Rickettsial infection: fourfold or greater rise in reciprocal
antibody titer against Rickettsia conorii, Rickettsia typhi,
or Orientia tsutsugamushi by immunofluorescent anti-
body test, or a clinical presentation suggestive for African
tick bite fever with a single antibody titer 1/40 against
R. conorii; or an indisputable clinical presentation of
African tick bite fever (primary eschar and secondary
rash) with a dramatic response to doxycycline;
3. Dengue: appearance of IgM (by enzyme-linked immuno-
sorbent assay) and/or fourfold or greater rise of IgG (by

TABLE 1. Prevalence of the Tropical Infections According to the Delay Between Date of Return and Onset of Fever (n = 2071)
Before or Within Total Within
1 Month After Return During 2nd and 3rd From 4th to 12th 12 Months
(1619 cases) Month (228 cases) Month (224 cases) (2071 cases)

No. (%) No. (%) No. (%) No. (%)

P. falciparum malaria 401 (24.8) 29 (12.7) 10 (4.5) 440 (21.2)
Non-falciparum malaria* 34 (2.1) 41 (18) 38 (17) 113 (5.5)
Rickettsial infectiony 70 (4.3)   70 (3.4)
Dengue 64 (4)   64 (3.1)
Acute schistosomiasis 28 (1.7) 9 (3.9) 1 (0.4) 38 (1.8)
Enteric feverz 15 (0.9) 1 (0.4) 16 (0.8)
Protozoan enteritisx 12 (0.7)  2 (0.9) 14 (0.7)
Amebic liver abscess 8 (0.5) 1 (0.4) 1 (0.4) 10 (0.5)
Histoplasmosis 6 (0.4)   6 (0.3)
Helminthic enteritis 3 (0.2) 2 (0.9) 1 (0.4) 6 (0.3)
Hepatitis E 4 (0.2) 1 (0.4)  5 (0.2)
Löffler syndrome 3 (0.2) 1 (0.4)  4 (0.2)
Other tropical diseasesk 11 (0.6)   11 (0.6)
Total tropical diseases 659 (40.7) 85 (36.8) 53 (23.7) 797 (38.4)
*Including P. vivax (n = 53), P. ovale (n = 43), and P. malariae (n = 17) malaria.
y
Including R. africae (n = 58), R. conorii (n = 5), R. typhi (n = 4), and O. tsutsugamushi (n = 3) infection.
z
Including S. typhi (n = 9) and S. paratyphi A (n = 7).
x
Including Cyclospora species (n = 7), Cryptosporidium species (n = 4), E. histolytica (n = 2), and Isospora belli (n = 1).
k
Including human African trypanosomiasis (n = 3), sarcocystosis (n = 3), enteritis Shigella dysenteriae (n = 3), relapsing fever (n = 1), and
angiostrongyloidiasis (n = 1).

n 2007 Lippincott Williams & Wilkins 19

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Bottieau et al
immunofluorescent antibody test) on paired sera, or a
suggestive illness with single positive IgM and/or single
IgG titer 1/1280) and no alternative diagnosis;
4. Acute schistosomiasis: exposure to surface water in an
endemic area and a positive serology by enzyme-linked
immunosorbent assay and/or indirect hemagglutination
(titer 160) or egg detection in stool and/or urine in non-
immune patients with eosinophil count 0.5
103/mL;
5. Enteric fever: demonstration of Salmonella typhi or
Salmonella paratyphi in blood and/or stool cultures.
Similar internationally recognized case definitions
were used for the other tropical conditions and the
cosmopolitan infections4,27.
Diagnostic Predictors of the Leading
Tropical Conditions
Based on clinical relevance, patients were divided in 2
groups: those with fever that started during the travel or within
1 month upon return (defined as early-onset fever) and those
with fever that developed at a later stage (late-onset fever).
Among patients with early-onset fever, the diagnostic
predictors of malaria were first investigated by comparing
the malaria cases with all the other febrile episodes. Then,
fever due to malaria was excluded from analysis, and each of
the remaining tropical diseases was successively compared
with all other (non-malaria) causes of fever. We opted for
this 2-step procedure to avoid bias for shared predictors, and
because malaria should be systematically and promptly ruled
out in any febrile traveler returning from an endemic area. In
late-onset fever, only malaria cases were compared with the
other causes because the remaining tropical etiologies were
too few (see Table 1).
Statistical Analysis
Statistical analyses were performed with SPSS soft-
ware, version 14 (SPSS Inc. Chicago, IL). Categorical data
were compared by using the chi-square test, and continuous
data by using appropriate nonparametric tests. The labora-
tory variables were dichotomized using accepted thresholds
for normality. Characteristics associated with the leading
tropical diagnoses in bivariate analysis were included in a
stepwise backward multivariate analysis. Crude likelihood
ratios (LRs) of the associated features were calculated using
standard formulas. Adjusted LRs were obtained by substi-
tuting the natural logarithm of the crude LRs to the
dichotomous data (positive LR if present feature and
negative LR if absent) in the multivariate analysis, and by
correcting them with the logistic coefficient, according to the
Spiegelhalter-Knill-Jones procedure19.
RESULTS
Epidemiologic Characteristics
of the Study Population
During the study period, 2071 febrile episodes
occurred in 1962 patients, including 1245 (60%) in western
20
Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Medicine  Volume 86, Number 1, January 2007
travelers, 300 (14.5%) in expatriates, 286 (14%) in travelers
visiting friends and relatives, and 240 (11.5%) in foreign
visitors/migrants. Mean age was 36 years (range, 5 mo–82
yr). The most visited region was sub-Saharan Africa (68%)
followed by southern Asia/Pacific (14%), Latin America
(7%), and North Africa/Middle East (5%). Forty patients
(2%) had visited more than 1 tropical region within the
month before consultation (Table 2). Most patients (1328,
64%) consulted our institutions directly without any previous
contact with another practitioner. In total, 565 patients
(27.5%) were hospitalized immediately or later during the
course of the disease.
Clinical Spectrum According to Travel
Destination and Latency Period
Tropical diseases were diagnosed in 797 (38%) of all
2071 fever episodes. Global prevalence of cosmopolitan
infection (697, 34%), fever of unknown cause (529, 25%)
and non-infectious disease (48, 2%) was similar to that we
previously reported2. Prevalence of the main diseases per
travel destination is reported in Table 2; P. falciparum
malaria, rickettsial infections, and acute schistosomiasis
were almost exclusively diagnosed after a stay in sub-
Saharan Africa, while most cases of dengue and enteric fever
were found in travelers returning from southern Asia/Pacific.
Dengue was the most frequent tropical cause of fever after a
stay in Latin America.
In early-onset fever (n = 1619), tropical infections
were found in 41% of the cases, and the differential
diagnosis was large (see Table 1). In contrast, malaria
(mainly non-falciparum cases) was the most predominant
tropical etiology (118/138, 86%) of late-onset fever.
Presenting Features of the Leading
Tropical Diseases
Table 3 shows the main clinical and laboratory
characteristics of the leading tropical conditions. Rickettsial
infection, dengue, and acute schistosomiasis were almost
exclusively diagnosed in western travelers or expatriates.
Abdominal symptoms (vomiting and/or diarrhea and/or
abdominal pain) were often mentioned in all diseases except
in rickettsial infection. In contrast, respiratory complaints
(cough and/or dyspnea and/or thoracic pain) were frequent
only in acute schistosomiasis and enteric fever. Abnormal
physical signs were rather infrequent except enlarged lymph
nodes and skin ulcers in rickettsial infection, a skin rash in
dengue and rickettsial infection, and an enlarged spleen
(assessed clinically) in malaria and enteric fever. Median
leukocyte count was lower in dengue than in other diseases.
Malaria cases were associated with a lower platelet count
and a higher total bilirubinemia. Eosinophil count was often
very high at presentation in acute schistosomiasis. Median
level of alanine aminotransferase was higher in cases of
dengue and enteric fever. The level of lactate dehydroge-
nase was frequently elevated (700 IU/L) in all 6 leading
tropical diseases. Median C-reactive protein level was rather
low (<5 mg/dL) in dengue, rickettsial infection, and acute
schistosomiasis.
n 2007 Lippincott Williams & Wilkins
Medicine  Volume 86, Number 1, January 2007 Diagnostic Predictors of Imported Tropical Conditions

TABLE 2. Prevalence of the Main Etiologies of Fever According to the Last Visited Region
Sub-Saharan Africa Southern Asia/Pacific Latin America North Africa/Middle East
(1401 cases) (381 cases) (146 cases) (103 cases)

No. (%) No. (%) No. (%) No. (%)

Tropical condition 628 (44.8) 137 (36) 24 (16.4) 6 (5.8)
P. falciparum malaria 426 (30.4) 8 (2.1) 0 1 (1)
Non-falciparum malaria 70 (5) 33 (8.7) 6 (4.1) 4 (3.9)
Rickettsial infection 63 (4.5) 6 (1.6) 0 1 (1)
Dengue 2 (0.1) 50 (13.1) 12 (8.2) 0
Acute schistosomiasis 38 (2.7) 0 0 0
Enteric fever 3 (0.2) 13 (3.4) 0 0
Other tropical condition 26 (1.9) 27 (7.1) 6 (4.1) 0
Cosmopolitan infection* 406 (30) 158 (41.5) 66 (45.2) 43 (41.7)
Respiratory tract infectiony 115 (8.2) 32 (8.4) 19 (13) 4 (3.9)
Bacterial enteritisz 55 (3.9) 36 (9.4) 14 (9.6) 15 (14.6)
Mononucleosis-like syndromex 41 (2.9) 25 (6.6) 7 (4.8) 6 (5.8)
Skin/soft tissue infection 46 (3.3) 17 (4.5) 5 (3.4) 7 (6.8)
Genitourinary infection 41 (2.9) 11 (2.9) 8 (5.5) 3 (2.9)
Other cosmopolitan infection 108 (7.7) 37 (9.7) 13 (8.9) 8 (7.8)
Unknown etiology 343 (24.5) 81 (21.3) 49 (33.6) 41 (39.8)
Non-infectious disease 24 (1.7) 5 (1.3) 7 (4.8) 12 (11.7)
Note: 40 additional patients had visited more than 1 region within the month before consultation.
*Infections with worldwide distribution.
y
Including pharyngitis/tonsillitis, otitis, sinusitis, bronchitis (all diagnosed clinically), pneumonia (diagnosed by chest X-rays), Chlamydia/Mycoplasma
pneumoniae infection, and Influenza A/B.
z
Including infections with Shigella species, Campylobacter species, Salmonella species, Yersinia species, and presence of white/red blood cells in stool
examination without isolation of pathogenic bacteria.
x
Including primary infection with cytomegalovirus, Toxoplasma gondii, Epstein-Barr virus, and human immunodeficiency virus (HIV).

Early-Onset Fever: Predictors of the Leading
Tropical Diseases
Malaria cases (n = 435, 92% of P. falciparum) were first
compared with all other patients with early-onset fever.
Adjusted odds ratios (OR) and LRs of the variables
independently associated with malaria are shown in Table 4.
Hematologic values were missing in 30/1619 (2%) fever
episodes, and biochemistry data were incomplete in 93
(5.5%). The highest positive LRs were found for thrombocy-
topenia (platelet count <150
103/mL) (5.8), hyperbilirubi-
nemia (total bilirubin level 1.3 mg/dL) (5.2), enlarged
spleen (assessed clinically) (5.1), and fever without localizing
symptom (4.5). In our setting, post-test probabilities for
malaria were 68% [(435/1184)
5.8 = 2.13; 2.13/(1 + 2.13) =
68%] for thrombocytopenia, 66% for hyperbilirubinemia, and
reached 80% when both predictors were present.
In a second step, we compared successively the cases
of dengue, rickettsial infection, acute schistosomiasis, and
enteric fever with all other fevers, excluding malaria
(Table 5). For the diagnosis of dengue, the highest positive
LRs were found for a stay in Latin America (29), a stay in
southern Asia/Pacific (7.6), leukopenia (leukocytes < 4

103/mL) (3.3), skin rash (2.8), and thrombocytopenia (2). In
case of leukopenia, the post-test probability of dengue
reached respectively 30% in a febrile traveler returning from
Latin America, and 38% when returning from southern Asia/
Pacific. Rickettsial infection was strongly predicted by the
presence of a skin rash and particularly of a skin ulcer
(respective positive LR of 3.8 and 11). These clinical
features increased the posttest probability to roughly 20%
and 50%, respectively, in a febrile traveler returning from
sub-Saharan Africa. Eosinophilia had an extremely strong
confirming power for acute schistosomiasis (positive LR =
31.8). Finally, an enlarged spleen (assessed clinically), a stay
in southern Asia/Pacific, and an elevated alanine amino-
transferase level were the strongest predictors of enteric
fever (positive LRs = 10, 4.1, and 2.5, respectively) in non-
malaria febrile patients.
Late-Onset Fever: Predictors of Malaria
In another analysis, malaria cases (n = 118, 67% non-
falciparum) were compared with all other causes of late-
onset fever. Features strongly associated (p < 0.001) with
malaria were similar to those seen in early-onset fever:
thrombocytopenia (OR, 13.0; 95% CI, 6.7–25.1), enlarged
spleen (OR, 7.1; 95% CI, 3.3–15.8), hyperbilirubinemia (OR,
5.8; 95% CI, 2.6–12.9), and fever without localizing
symptom (OR, 5.6; 95% CI, 2.8–11.2). In addition, their

n 2007 Lippincott Williams & Wilkins 21

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Bottieau et al Medicine  Volume 86, Number 1, January 2007

TABLE 3. Main Clinical and Laboratory Features of the Leading Imported Tropical Diseases
Non-Falciparum Rickettsial Acute
Falciparum Malaria Malaria Infection Dengue Schistosomiasis Enteric Fever
(n = 440) (n = 113) (n = 70) (n = 64) (n = 38) (n = 16)
Median age, yr (IQR) 36.5 (28.5–48.5) 36 (26–50) 50 (31–57) 35.5 (27–42.5) 30.5 (24.5–45.5) 28 (20–40.5)
Western traveler 171 (39) 63 (56) 67 (96) 55 (86) 34 (89) 11 (69)
Expatriate 106 (24) 22 (19) 3 (4) 6 (9) 3 (8) 1 (6)
VFR traveler 101 (23) 10 (9) 0 3 (5) 1 (3) 2 (12)
Foreign visitor/migrant 62 (14) 18 (16) 0 0 0 2 (12)
Fever >7 d 71 (16) 27 (24) 9 (13) 12 (19) 15 (39) 8 (50)
Fever 398C 241 (55) 56 (50) 21 (30) 37 (58) 11 (29) 5 (31)
Headache 374 (85) 84 (74) 40 (57) 54 (84) 20 (53) 11 (69)
Myalgia/arthralgia 297 (67) 86 (76) 37 (53) 53 (83) 27 (71) 8 (50)
Abdominal symptom* 248 (56) 33 (29) 13 (19) 27 (42) 18 (47) 14 (87)
Respiratory symptomy 77 (17) 16 (14) 15 (21) 18 (28) 19 (50) 8 (50)
Enlarged lymph node 11 (2) 6 (5) 36 (51) 8 (12) 2 (5) 0
Enlarged spleen 96 (22) 31 (27) 4 (6) 3 (5) 2 (5) 7 (44)
Skin rash 13 (3) 4 (3) 44 (63) 30 (47) 1 (3) 0 (0)
Skin ulcer(s) 3 (1) 1 (1) 57 (81) 0 (0) 2 (5) 0 (0)
No localizing symptom 148 (34) 51 (45) 0 8 (12) 8 (21) 2 (12)
Median leukocyte count, 5 (4–6.3) 5.6 (4.6–7.1) 5.9 (5–7.6) 4.4 (2.7–6.5) 9.6 (8–11.2) 5.8 (4.3–7.2)

103/mL (IQR)
Median platelet count, 113 (69–165) 122 (89–179) 219 (169–272) 166 (108–240) 255 (202–311) 204 (153–299)

103/mL (IQR)
Median eosinophil count, 0.03 (0–0.07) 0.04 (0.01–0.1) 0.04 (0.02–0.1) 0.03 (0.01–0.08) 1.61 (1.03–2.35) 0.01 (0–0.02)

103/mL (IQR)
Median LDH level, 763 (620–1071) 713 (579–904) 674 (590–814) 741 (584–916) 673 (557–831) 1025 (736–1486)
IU/L (IQR)
Median ALT level, 43 (29–71) 36 (26–54) 49 (34–67) 59 (34–129) 42 (33–73) 75 (39–141)
IU/L (IQR)
Median total bilirubinemia, 1.1 (0.7–1.7) 1 (0.7–1.5) 0.6 (0.4–0.8) 0.5 (0.3–0.7) 0.5 (0.4–0.7) 0.7 (0.5–0.9)
mg/dL (IQR)
Median CRP level, 7 (2.5–12.5) 5.5 (2–10.5) 2 (0.5–5) 0.5 (0–1) 3 (1.5–4) 7 (3.5–9)
mg/dL (IQR)
Abbreviations: IQR = interquartile range; VFR = visiting friends and relatives; LDH = lactate dehydrogenase; ALT = alanine aminotransferase;
CRP = C-reactive protein.
*Including vomiting and/or diarrhea and/or abdominal pain.
y
Including cough and/or dyspnea and/or thoracic pain.
Note 1: All results are expressed in no. (%), except otherwise specified.
Note 2: P value was < 0.05 between diseases for each characteristic (by chi-square for categorical parameters or Kruskal-Wallis test for continuous
parameters).

predictive weight was rather similar (positive LRs = 6.9, 3.9,
4.7, and 3.5, respectively).
DISCUSSION
The diagnostic approach to imported fever consists
mainly of a predictive assessment based on destination-
specific risk and disease presentation. We conducted the
current study to identify diagnostic predictors for the leading
tropical diseases in travelers and to quantify their confirming
power in this population. Our findings may be useful for
physicians less familiar with travel medicine, who might be
less aware of the clinical and laboratory pattern of the main
imported tropical conditions. Key qualities of the current
study are the sample size, the large proportion of unreferred
patients as well as outpatients, reducing the selection bias
toward most serious conditions, and the standardized
prospective data gathering, with a low rate of missing
laboratory values.
Several limitations must be mentioned, however. First,
due to the strict case definitions, some less typical cases
might have been misdiagnosed as ‘‘unknown cause,’’
especially in case of early workup and absence of repeated
serology in a self-limiting disease. Second, the strong bias
toward exposure to sub-Saharan Africa led to an overrepre-
sentation of malaria cases, distorting the weight of predictors
shared with other diseases. We tried to circumvent that

22 n 2007 Lippincott Williams & Wilkins

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Medicine  Volume 86, Number 1, January 2007 Diagnostic Predictors of Imported Tropical Conditions

TABLE 4. Adjusted Odd Ratios and Adjusted Likelihood Ratios of the Variables Independently Associated With Malaria (n = 435) in
Patients With Early-Onset Fever (n = 1619)
Frequency in Frequency in
Malaria Cases Non-Malaria Cases Adjusted OR Adjusted Adjusted
Variable (435 patients) (1184 patients) (95 % CI) P LR + LR 

No. (%) No. (%)

Stay in sub-Saharan Africa 406 (93) 677 (57) 20.2 (11.0–37.1) < 0.001 1.89 0.09
Expatriate 105 (24) 151 (13) 1.8 (1.1–3.0) 0.01 1.63 0.90
VFR traveler 95 (22) 112 (9) 1.7 (1.0–2.9) 0.05 1.56 0.93
No or incomplete chemoprophylaxis* 379 (87) 828 (70) 3.5 (2.1–5.6) < 0.001 1.30 0.37
Headache 369 (85) 638 (54) 4.1 (2.6–6.4) < 0.001 1.50 0.36
Vomiting 150 (34) 164 (14) 2.6 (1.6–4.2) < 0.001 2.03 0.81
Enlarged spleen 108 (23) 56 (5) 4.1 (2.3–7.1) < 0.001 5.10 0.81
No localizing symptom 152 (35) 139 (12) 7.3 (4.6–11.5) < 0.001 4.48 0.65
Platelet count <150
103/mLy 300/433 (70) 120/1156 (10) 15.8 (10.3–24.3) < 0.001 5.76 0.37
LDH 700 IU/Ly 254/416 (61) 288/1110 (26) 2.4 (1.6–3.6) < 0.001 1.67 0.68
ALT 70 IU/Ly 108/416 (26) 192/1110 (17) 0.51 (0.31–0.83) 0.007 0.58 1.16
Total bilirubinemia 1.3 mg/dLy 158/416 (38) 58/1110 (5) 6.9 (4.0–12.0) < 0.001 5.17 0.70
Abbreviations: See previous table.
*Chemoprophylaxis included chloroquine/proguanil, mefloquine, atovaquone/proguanil, and doxycycline.
y
No./No. tested (%).
Note 1: In addition to the above-mentioned variables, the following parameters associated with malaria by bivariate analysis were also entered in the
multivariate model: male sex, foreign visitor/migrant, fever 398C, myalgia, and leukocyte count <4
103/mL. The goodness of fit (Hosmer and Lemeshow
test) chi-square of this model remained non-significant during the 6 steps (p = 0.20 at the last step).

problem by excluding malaria in the second step of the
analysis, assuming that this diagnosis is a priority and is
rather straightforward. Third, the clinical and laboratory
features were recorded at presentation, but some may have
subsequently evolved during the course of the disease.
Fever occurring during or within the first month after
travel is the most challenging situation for clinicians in terms
of potential disease spectrum and associated morbidity. We
focused our investigations, therefore, on this critical post-
travel period. As shown elsewhere, an enlarged spleen8,21,
thrombocytopenia3,8,21, and hyperbilirubinemia3 were all
strong predictors of malaria. LRs we calculated were
generally lower than those reported in other studies, but this
may be due mainly to the adjustment procedure used here,
which provided the real (independent) diagnostic weight of
each feature. Also the relative burden of P. falciparum cases,
the selection of controls, and the proportion of missing
laboratory data may explain some differences in the
predictive values. The advantage of determining LRs is that
they are independent of the disease prevalence. Post-test
probabilities can therefore be calculated for any setting or
destination, as illustrated. It is noteworthy that, in patients
with late-onset fever, malaria (although mainly due to non-
falciparum species) was predicted by the same criteria and
with similar LRs as in early-onset fever. This implies that
when 1 of these predicting criteria is met, the travel history,
even for the distant past, should be scrutinized and malaria
should be intensively looked for, including by repeated
testing if the initial investigation is negative.
Dengue is highly endemic in most popular destinations
of southern Asia/Pacific and Latin America11. Diagnostic
predictors have been investigated almost exclusively in
endemic areas5,24,25. Case-control series in travelers were
small and retrospective17,20. Among the usual dengue clin-
ical features22,26, only skin rash had a substantial predictive
weight in our population. Both leukopenia and throm-
bocytopenia were good laboratory predictors of dengue in
febrile travelers in whom malaria had been excluded, similar
to observations in endemic areas24,25.
In this study, disease presentation of rickettsial
infection mirrored that of African tick bite fever, by far the
most common rickettsiosis encountered. Although epidemi-
ology of rickettsiosis varies widely, all cases were pooled for
analysis because of their similar clinical features and
treatment12, and because attributing a clinical rickettsiosis
to a specific species is not always clear-cut. In contrast with
the only other case-control series18 that we know has been
published, skin rash and skin ulcers were the only 2
independent predictors for this diagnosis, but enlarged lymph
nodes were not. Apart from rickettsial disease, the differential
diagnosis of an ulcerative skin lesion in a feverish patient
includes mainly a bacterial superinfection of a wound or
insect bite, and only exceptionally East African trypanoso-
miasis, cutaneous anthrax, and bubonic plague7. As
leukocytosis is extremely infrequent in rickettsial infection,
and is the rule in bacterial superinfection, it may help to
select the appropriate antibiotic in febrile patients with
atypical skin ulcers when an empirical treatment is needed.

n 2007 Lippincott Williams & Wilkins 23

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Bottieau et al Medicine  Volume 86, Number 1, January 2007

TABLE 5. Adjusted Odd Ratios and Adjusted Likelihood Ratios of the Variables Independently Associated With Dengue (n = 64),
Rickettsial Infection (n = 70), Acute Schistosomiasis (n = 28) and Enteric Fever (n = 15) in Early-Onset Fevers, After Exclusion of
Malaria Cases (n = 1184)
Adjusted OR Adjusted Adjusted
Cases* Controls* (95 % CI) P LR + LR
Dengue (n = 64 patients) (n = 1120 patients)
Stay in southern Asia/Pacific 50 (78) 245 (22) 64.6 (14.7–284) < 0.001 7.58 0.13
Stay in Latin America 12 (19) 90 (8) 49.3 (9.9–246.3) < 0.001 29.18 0.61
Headache 54 (84) 584 (52) 4.2 (1.7–10.5) 0.002 1.54 0.37
Myalgia 53 (83) 526 (47) 2.8 (1.2–6.5) 0.01 1.41 0.51
Skin rash 30 (47) 146 (13) 3.9 (1.9–7.9) < 0.001 2.80 0.67
Leukocyte count <4
103/mL 25 (39) 78/1092 (7) 4.5 (1.8–11.5) 0.001 3.33 0.74
Platelet count <150
103/mL 27 (43) 93/1092 (8) 2.6 (1.1–6.1) 0.03 1.96 0.83
LDH 700 IU/L 37/61 (61) 251/1047 (24) 2.2 (1.1–4.5) 0.03 1.59 0.72
Rickettsial infection (n = 70 patients) (n = 1114 patients)
Stay in sub-Saharan Africa 63 (90) 614 (55) 7.7 (2.9–21.0) < 0.001 1.64 0.22
Western traveler 67 (96) 790 (71) 3.9 (1.1–14.2) 0.03 1.35 0.14
Skin rash 44 (63) 132 (12) 7.1 (3.2–15.6) < 0.001 3.77 0.50
Skin ulcer(s) 57 (81) 46 (4) 43.7 (20.3–100) < 0.001 11.1 0.27
Acute schistosomiasis (n = 28 patients) (n = 1156 patients)
Western traveler or expatriate 27 (96) 970 (84) 11.8 (1.3–105) 0.03 1.24 0.11
Myalgia 19 (68) 560 (48) 4.1 (1.1–15.0) 0.03 1.81 0.44
Eosinophil count 0.5
103/mL 26 (93) 47/1128 (4) 431 (84–2210) < 0.001 31.85 0.06
LDH 700 IU/L 14/26 (54) 274/1080 (25) 3.9 (1.0–14.5) 0.04 2.33 0.58
Enteric fever (n = 15 patients) (n = 1169 patients)
Stay in southern Asia/Pacific 13 (87) 282 (24) 25 (4.7–132.2) < 0.001 4.12 0.14
Any abdominal symptom 13 (87) 558 (48) 7.6 (1.5–38.5) 0.015 1.81 0.25
Enlarged spleen 6 (40) 48 (4) 17.7 (4.4–70.2) < 0.001 10.0 0.63
ALT level 70 IU/L 9 (60) 183/1096 (17) 4.9 (1.4–16.5) 0.011 2.53 0.56
Abbreviations: See previous tables.
*All figures represent no. or no./no. tested and (%).
Note 2: In addition to the above-mentioned variables per disease, the following parameters associated with each disease by bivariate analysis were also
successively entered in the multivariate model: western traveler, fever 39 8C, and ALT 70 IU/L for dengue; male sex, enlarged lymph node, and LDH 700
IU/L for rickettsial infection; male sex, stay in sub-Saharan Africa, respiratory symptom, and fever without localizing symptom for acute schistosomiasis; LDH
700 IU/L for enteric fever. The goodness of fit chi-square remained non-significant for each step of each model.

Acute schistosomiasis is frequently diagnosed in febrile
travelers returning from sub-Saharan Africa. Eosinophilia
was by far the strongest predictor of this condition, although
eosinophil levels were not always elevated at the initial
consultation. This laboratory feature should trigger a careful
questioning to reveal fresh water exposure as well as
appropriate parasitic and/or serologic investigations1. Al-
though a large variety of helminths can cause fever and
eosinophilia in travelers7, Löffler syndrome or strongyloido-
sis were seen only rarely in our experience.
Data of typhoid and paratyphoid fever cases were
pooled because they are clinically indistinguishable14,15,23.
Despite the small number of cases, we identified an enlarged
spleen as another strong diagnostic predictor besides travel to
southern Asia/Pacific. Abdominal symptoms and elevated
liver enzymes may also suggest enteric fever6. In practice,
distinguishing enteric fever from dengue is the most pressing
problem, because both conditions share the same geographic
risk and a similar presentation, and may evolve into severe
disease. The C-reactive protein level might be an interesting
discriminative criterion to initiate empiric antibiotherapy,
until blood culture results are available. However this finding
requires further prospective evaluation.
Because of the rarity of the conditions, predictive
factors for other serious tropical conditions, such as amebic
liver abscess, African human trypanosomiasis, histoplasmo-
sis, and hepatitis E, could not be determined in the study
population. This applies as well to travel-related cosmopolitan
infections such as hepatitis A/B, leptospirosis, or primary HIV
infection. Multicentric studies would probably be required to
further characterize these neglected conditions in travelers.
In conclusion, the optimal management of fever after a
stay in the tropics depends on early recognition of the main
tropical diseases. Until more reliable and rapid diagnostic
tests become available, the diagnostic and therapeutic
strategy has to be based on the geographic risk and on the
initial assessment. Most clinical and laboratory predictors
identified in endemic settings can be used in febrile travelers

24 n 2007 Lippincott Williams & Wilkins

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Medicine  Volume 86, Number 1, January 2007
as well to speed up the diagnosis and treatment of the leading
tropical conditions.
REFERENCES
1. Bottieau E, Clerinx J, de Vega MR, Van den Enden E, Colebunders R,
Van Esbroeck M, Vervoort T, Van Gompel A, Van den Ende J. Imported
Katayama fever: Clinical and biological features at presentation and
during treatment. J Infect. 2006;52:339–345. [Epub 2005 Sep 19].
2. Bottieau E, Clerinx J, Schrooten W, Van den Enden E, Wouters R, Van
Esbroeck M, Vervoort T, Demey H, Colebunders R, Van Gompel A,
Van den Ende J. Etiology and outcome of fever after a stay in the
tropics. Arch Intern Med. 2006;166:1642–1648.
3. Casalino E, Le Bras J, Chaussin F, Fichelle A, Bouvet E. Predictive
factors of malaria in travelers to areas where malaria is endemic. Arch
Intern Med. 2002;162:1625–1630.
4. Centers for Disease Control and Prevention. Case definitions for
infectious conditions under public health surveillance. MMWR Morb
Mortal Wkly Rep. 1997;46:1–57.
5. Chadwick D, Arch B, Wilder-Smith A, Paton N. Distinguishing dengue
fever from other infections on the basis of simple clinical and laboratory
features: application of logistic regression analysis. J Clin Virol. 2006;
35:147–153.
6. Connor BA, Schwartz E. Typhoid and paratyphoid fever in travellers.
Lancet Infect Dis. 2005;5:623–628.
7. D’Acremont V, Ambresin AE, Burnand B, Genton B. Practice
guidelines for evaluation of fever in returning travelers and migrants.
J Travel Med. 2003;10 (Suppl 2):S25–S52.
8. D’Acremont V, Landry P, Mueller I, Pecoud A, Genton B. Clinical and
laboratory predictors of imported malaria in an outpatient setting: an aid
to medical decision making in returning travelers with fever. Am J Trop
Med Hyg. 2002;66:481–486.
9. Doherty JF, Grant AD, Bryceson AD. Fever as the presenting complaint
of travellers returning from the tropics. QJM. 1995;88:277–281.
10. Freedman DO, Weld LH, Kozarsky PE, Fisk T, Robins R, von
Sonnenburg F, Keystone JS, Pandey P, Cetron MS. Spectrum of disease
and relation to place of exposure among ill returned travelers. N Engl J
Med. 2006;354:119–130.
11. Jelinek T, Muhlberger N, Harms G, Corachan M, Grobusch MP,
Knobloch J, Bronner U, Laferl H, Kapaun A, Bisoffi Z, Clerinx J,
Puente S, Fry G, Schulze M, Hellgren U, Gjorup I, Chalupa P, Hatz C,
Matteelli A, Schmid M, Nielsen LN, da Cunha S, Atouguia J, Myrvang
B, Fleischer K. Epidemiology and clinical features of imported dengue
fever in Europe: sentinel surveillance data from TropNetEurop. Clin
Infect Dis. 2002;35:1047–1052.
n 2007 Lippincott Williams & Wilkins
Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Diagnostic Predictors of Imported Tropical Conditions
12. Jensenius M, Fournier PE, Raoult D. Rickettsioses and the international
traveler. Clin Infect Dis. 2004;39:1493–1499.
13. Marx A, Pewsner D, Egger M, Nuesch R, Bucher HC, Genton B, Hatz
C, Juni P. Meta-analysis: accuracy of rapid tests for malaria in travelers
returning from endemic areas. Ann Intern Med. 2005;142:836–846.
14. Maskey AP, Day JN, Phung QT, Thwaites GE, Campbell JI,
Zimmerman M, Farrar JJ, Basnyat B. Salmonella enterica serovar
Paratyphi A and S. enterica serovar Typhi cause indistinguishable
clinical syndromes in Kathmandu, Nepal. Clin Infect Dis. 2006;42:
1247–1253.
15. Meltzer E, Sadik C, Schwartz E. Enteric fever in Israeli travelers: a
nationwide study. J Travel Med. 2005;12:275–281.
16. Ryan ET, Wilson ME, Kain KC. Illness after international travel. N Engl
J Med. 2002;347:505–516.
17. Shirtcliffe P, Cameron E, Nicholson KG, Wiselka MJ. Don’t forget
dengue! Clinical features of dengue fever in returning travellers. J R
Coll Physicians Lond. 1998;32:235–237.
18. Smoak BL, McClain JB, Brundage JF, Broadhurst L, Kelly DJ, Dasch
GA, Miller RN. An outbreak of spotted fever rickettsiosis in U.S. Army
troops deployed to Botswana. Emerg Infect Dis. 1996;2:217–221.
19. Spiegelhalter DJ, Knill-Jones RP. Statistical and knowledge-based
approaches to clinical decision-support systems, with an application in
gastroenterology. J R Statist Soc A. 1984;147:35–77.
20. Sung V, O’Brien DP, Matchett E, Brown GV, Torresi J. Dengue fever
in travelers returning from southeast Asia. J Travel Med. 2003;10:
208–213.
21. Svenson JE, Gyorkos TW, Maclean JD. Diagnosis of malaria in the
febrile traveler. Am J Trop Med Hyg. 1995;53:518–521.
22. Teichmann D, Gobels K, Niedrig M, Grobusch MP. Dengue virus
infection in travellers returning to Berlin, Germany: clinical, laboratory,
and diagnostic aspects. Acta Trop. 2004;90:87–95.
23. Vollaard AM, Ali S, Widjaja S, Asten HA, Visser LG, Surjadi C, van
Dissel JT. Identification of typhoid fever and paratyphoid fever cases at
presentation in outpatient clinics in Jakarta, Indonesia. Trans R Soc Trop
Med Hyg. 2005;99:440–450.
24. Watt G, Jongsakul K, Chouriyagune C, Paris R. Differentiating dengue
virus infection from scrub typhus in Thai adults with fever. Am J Trop
Med Hyg. 2003;68:536–538.
25. Wilder-Smith A, Earnest A, Paton NI. Use of simple laboratory features
to distinguish the early stage of severe acute respiratory syndrome from
dengue fever. Clin Infect Dis. 2004;39:1818–1823.
26. Wilder-Smith A, Schwartz E. Dengue in travelers. N Engl J Med.
2005;353:924–932.
27. World Health Organization. WHO recommended surveillance stand-
ards. Second edition. Geneva: World Health Organization. 1999:1–116.
25