Am. J. Trop. Med. Hyg., 87(6), 2012, pp. 1116–1118
Copyright © 2012 by The American Society of Tropical Medicine and Hygiene
Short Report: Higher Risk of Infection with Dengue at the Weekend among Male Singaporeans
Alex R. Cook,* Luis R. Carrasco, Vernon J. Lee, Eng Eong Ooi, Mark I-C Chen, David C. Lye, and Yee Sin Leo
Department of Statistics and Applied Probability, Faculty of Science, National University of Singapore, Singapore;
Program in Health Services and Systems Research, Duke-NUS Graduate Medical School Singapore, Singapore;
Saw Swee Hock School of Public Health, National University Health System, National University of Singapore, Singapore;
Department of Biological Sciences, Faculty of Science, National University of Singapore, Singapore; Communicable
Disease Centre, Tan Tock Seng Hospital, Singapore; Program in Emerging Infectious Diseases, Duke-NUS Graduate
Medical School Singapore, Singapore; Department of Medicine, Yong Loo Lin School of Medicine,
National University Health System, National University of Singapore, Singapore
the home, but when and where dengue transmission occurs is unclear, confounding control efforts. The authors estimate
days of the week in which dengue inpatients in Singapore were infected during the period 2006–2008, based on the day
they became febrile and historical data on the incubation period, using Bayesian statistical methods. Among male
inpatients, the relative risk of infection is an estimated 57% higher at the weekend, suggesting infections associated with
the home or leisure activities. There was no evidence of elevated risk of infection at the weekend for female inpatients.
The study motivates further research identifying locales frequented in the week leading up to onset to improve the
effective targeting of vector control efforts.
A growing body of evidence suggests that dengue infection in Singapore predominantly occurs away from
Dengue is endemic to Singapore, a tropical city state in
South East Asia, with year-round infection and circulation of
all four serotypes.1An intensive and well-regarded vector
control program was instigated in the 1960s and 70s2and led
to a period of low incidence of symptomatic disease; since the
1990s, however, dengue has resurged and was responsible for
around 50,000 reported cases and 100 deaths from 2000 to
2009.3This resurgence accompanied a shift in both infection
and clinical disease from children to adults.2,4
It is unclear in what locations infections occur, confound-
ing control efforts. As most people have different activities
and frequent different locales on weekdays and on the week-
end, being able to identify different risks of infection in these
two time periods would suggest locations or behaviors that
may be associated with a higher risk of dengue transmission
and that could be investigated further to devise enhanced
vector control measures. We therefore sought to estimate the
proportion of infections occurring in these two time periods,
using inpatient records of fever onset to infer the infection
day distribution using Bayesian statistical methods.
We reviewed data on all patients admitted to Tan Tock
Seng Hospital, the main hospital managing dengue patients
in Singapore, during 2006–2008 with 1) dengue infection con-
firmed by reverse transcription-polymerase chain reaction
(RT-PCR) or 2) who tested positive using immunoglobulin
M (IgM) and fulfilled the 1997 criteria for dengue fever or
the 2009 criteria for probable dengue or both.5,6For each
patient we determined a date of onset of fever (N = 2,126),
excluding those who could not recall the date of onset or
duration of fever, or with missing data (N = 11). The median
time from onset to hospitalization is 5 days; other demo-
graphic and clinical data are presented in Table 1. Of partic-
ular note, there was a paucity of children (routinely referred
instead to nearby KK Women’s and Children’s Hospital) and
twice as many men as women were admitted; in addition, as
in other settings,7women were more likely to present more
severe manifestations of dengue. Dates were converted to
days of the week for subsequent analysis. Because there are
differences in incidence of cases1and in dengue seropositivity8
between the genders, we consider men and women separately.
This study was approved by the Institutional Review Board,
National Healthcare Group, Singapore (DSRB E/08/567).
We used early 20th century volunteer challenge study data
to quantify uncertainty in the incubation period (i.e., time
from infection to symptom onset). These data were originally
published by Siler and others9and Simmons and others10and
recently reanalyzed by Nishiura and Halstead11; the strains
used in the two experiments were previously identified to be
dengue viruses (DENV-4 and DENV-1), respectively.12
Nishiura and Halstead11could find no statistically significant
differences between the incubation periods of the two sero-
types, and in this analysis we assume that this also holds for
DENV-2 and DENV-3, and for primary and subsequent
infections. We obtained the challenge data from the authors11
to which we fitted log-normal, Weibull, and gamma distribu-
tions. The log-normal fitted substantially better (the differ-
ence in posterior mean deviances being 10 and 55) and so all
subsequent analyses assumed this form for the incubation
period. The log-normal model was parameterized using
Bayesian methods,13with uniform priors for the log-mean
and log-standard deviation on the real line and positive part
of the real line, respectively. The posterior distribution for
these two parameters was then sampled using Markov chain
Monte Carlo integration with 100,000 iterations after burn-in
of 10,000 iterations. Visual inspection confirmed near multi-
variate normality of the posterior and so the mean and
covariance of the Markov chain Monte Carlo sample were
used to define a multivariate normal prior distribution for
the Tan Tock Seng data; these are tabulated in the Supple-
We then “back fitted” the infection day distribution by
combining the information from the onset data with that from
the volunteer challenge studies. The model used to do so
assumed a probability that a randomly selected dengue
patient was infected on a Saturday or Sunday of pWE(the
same for both days) with the probability it happened on a
*Address correspondence to Alex R. Cook, Department of Statistics
and Applied Probability, Block S16 level 7, 6 Science Drive 2, National
University of Singapore, Singapore 117546. E-mail: alex.richard.cook@
weekday being pWD= (1−2 pWE)/5. We estimated pWEas
before using Markov chain Monte Carlo within a Bayesian
analysis, combining a uniform prior distribution on (0, 0.5)
for pWEwith the aforementioned multivariate normal prior
for the incubation parameters, and 100,000 iterations follow-
ing 10,000 as burn-in. We derive an approximate two-sided
P value, denoted P e, for the hypothesis of equal risk of infec-
intervals in lieu of confidence intervals and appealing to the
relationship between the coverage of and endpoints of the
latter and the null hypothesis tested. All analysis was per-
formed in the R statistical environment.14
The estimated incubation period is presented in Figure 1
(top) and is consistent with the oft mentioned typical incu-
bation period of 4 to 7 days and a range of 3–14 days15; the
distribution of onset by day of the week is presented in
Figure 1 (middle). Among males, a rise toward the end of
the week is discernible and consistent with a higher risk of
infection at the weekend displaced by the incubation period;
for females, onset is almost uniform across the week. The
Bayesian analysis (Figure 1, bottom) yields estimates of the
proportion infected on a weekend day of 19.2% for males
(95% credible interval [CI]: 16.3%, 22.3%) versus on a
weekday of 12.3% (95% CI: 11.1%, 13.5%), i.e., a relative
risk of 1.57 (95% CI: 1.21, 2.01); for female inpatients, the
weekend estimate (15.2%, 95% CI: 11.2%, 19.3%) is similar
to that during the week (13.9%, 95% CI: 12.3%, 15.5%), and
the relative risk is 1.11 (95% CI: 0.72, 1.58). Approximate
two-sided P values suggest strong evidence (Pe=0:0007) that
is enigmatic and has been variously ascribed to declining herd
immunity, climate change, virus changes, less effective vector
control, and changes in patterns of infection.2,16,17A key unan-
swered question is where and when infection is acquired.
Answering this question would allow high-risk areas to be
identified and targeted for responsive and structured control
measures. In finding a significantly higher infection rate for
tion in the two time periods (i.e., of pWD= pWE) using credible
the male infection rate is not constant over the week; for
The resurgence of dengue in Singapore over the last 20 years
men at weekends than weekdays, this study suggests some
hypotheses for subsequent investigation. The working pattern
in Singapore is similar to that in other developed countries,
with most non-service workers working Monday to Friday,
suggesting that the increase in infections at weekends is asso-
ciated with either the home, or with leisure and other week-
end activities. In Singapore, a small proportion (~30%) of
cases can be linked to a cluster around the household,2and
there are few infections among children,4which have led to
speculation of a switch toward infection away from the
home, around which the current emphasis on vector control
focuses. Research in Taiwan also suggests infection not be
associated with time spent at home or the workplace.18Com-
bining the current study’s findings, that infection among
Singaporean men is higher on the weekend, with the impli-
cation of previous studies, that infection increasingly occurs
away from the home, suggests weekend leisure activities
may be a risk factor.
Demographic and clinical data on dengue inpatients, Tan Tock Seng
Hospital, Singapore, 2006–8*
Characteristic Proportion (N = 2115)
Aged < 20
Satisfying criteria for
Dengue fever 1997
Severe dengue 2009
*Eleven other patients who did not have data on onset of fever were excluded from
analysis. Except where otherwise noted, in parenthesis, demographic data for the resident
population (i.e., Singapore citizens and permanent residents) are presented for comparison;
these are derived from the 2010 census of population published by the Singapore Depart-
ment of Statistics. Detailed data on non-residents, that is, foreigners’ resident in Singapore,
but without permanent residence status, are not published by the Department of Statistics.
†Proportion of citizens to the entire population, which consists of citizens, permanent
residents, and non-residents.
volunteer challenge studies, empirical distribution of onset days
among dengue inpatients, Tan Tock Seng Hospital, Singapore,
2006–8, and inferred distribution of infection days for these inpa-
tients. Top: incubation period data of dengue viruses 1 (DENV-1)
and 4 (DENV-4) extracted from Nishiura and Halstead11(bars); the
inferred distribution is overlaid (solid line, posterior mean, dashed
line 95% credible interval [CI]). Middle: distribution of day of onset
of fever in dengue patients at Tan Tock Seng Hospital by sex. Bot-
tom: posterior mean (dot) and 95% CI (lines) for the per day prob-
ability of infection, for weekdays and weekends, by sex; the dashed
line corresponds to what would be expected were infection rates
homogeneous across the week.
Estimated incubation period from early 20th century
DENGUE INFECTION AT THE WEEKEND IN SINGAPORE
There are several limitations to the study: 1) in the absence Download full-text
of data to the contrary, we assume the same distribution for all
four serotypes and for primary and subsequent infections.
Although this may not be correct, we do not have data to
inform alternative incubation period distributions for different
combinations of serotypes and clinical phenotypes, or to quan-
tify differences between these combinations. 2) It is likely that
a small proportion of patients testing positive on IgM but
negative on RT-PCR are not currently infected with dengue.
3) Almost everyone in our cohort is an adult, so we are unable
to assess the timing of pediatric infections, which in Singapore
constitute about 15% of notified cases,2lower than in many
neighboring countries. 4) To generalize from the day of the
week of infection of adult inpatients to that of all adults
requires making the assumption that the probability of even-
tually requiring hospitalization is the same regardless of the
day the infection occurred. Most importantly, 5) the study
design is unable to answer the question of where transmission
occurs—we think that a similar approach, combining elicited
information on locales frequented in the week leading up to
onset with the onset distribution, would allow enhanced
targeting of vector control to areas in which transmission is
more likely to have occurred.
Anecdotally, Singaporeans, especially males, engage more in
leisure activities outside the house on the weekend and thus, in
light of our findings, future investigation of weekend leisure
activities, such as visits to parks, shopping and food centers,
and other recreation areas, as a source of putative dengue
infection would be warranted and may prove fruitful.
Received October 17, 2011. Accepted for publication July 1, 2012.
Published online November 5, 2012.
Note: Supplemental data appears at www.ajtmh.org.
Acknowledgments: We thank Hiroshi Nishiura for providing the
onset data from Reference 11.
Financial support: This work was supported by the National University
of Singapore to ARC and LRC; and the National Medical Research
Council (grants NMRC/H1N1R/005/2009 to ARC, LRC, and MICC,
and NMRC/TCR/005/2008 to DCL, YSL, and EEO).
Authors’ addresses: Alex R. Cook, Saw Swee Hock School of Public
Health, National University of Singapore, Singapore, and Depart-
ment of Statistics and Applied Probability, National University of
Singapore, Singapore, E-mail: firstname.lastname@example.org. Luis
R. Carrasco, Department of Biological Sciences, National University
of Singapore, Singapore. Vernon J. Lee, Communicable Disease Cen-
tre, Tan Tock Seng Hospital, Singapore, and Saw Swee Hock School of
Public Health, National University Health System, National University
of Singapore, Singapore. Eng Eong Ooi, Program in Emerging Infec-
tious Diseases, Duke-NUS Graduate Medical School Singapore,
Singapore. Mark I-C Chen, David C. Lye, and Yee Sin Leo, Commu-
nicable Disease Centre, Tan Tock Seng Hospital, Singapore.
1. Koh BK, Ng LC, Kita Y, Tang CS, Ang LW, Wong KY, James L,
Goh KT, 2008. The 2005 dengue epidemic in Singapore: epide-
miology, prevention and control. Ann Acad Med Singapore 37:
2. Ooi E-E, Goh K-T, Gubler DJ, 2006. Dengue prevention and
35 years of vector control in Singapore. Emerg Infect Dis 12:
3. Ministry of Health, Singapore, 2011. MOH Weekly Publication of
Statistics on Local Infectious Disease Situation. Available at:
Accessed August 1, 2011.
4. Ooi EE, Hart TJ, Tan HC, Chen SH, 2001. Dengue sero-
epidemiology in Singapore. Lancet 357: 685–686.
5. WHO, 1997. Dengue Hemorrhagic Fever: Diagnosis, Treatment,
Prevention and Control. Second edition. Geneva: World
6. WHO, 2009. Dengue: Guidelines for diagnosis, Treatment Prevention
and Control. New edition. Geneva: World Health Organization.
7. Guzman MG, Kouri GP, Bravo J, Soler M, Vazquez S, Santos M,
Villaescusa R, Basanta P, Indan G, Ballester JM, 1984. Dengue
hemorrhagic fever in Cuba. II. Clinical investigations. Trans R
Soc Trop Med Hyg 78: 239–241.
8. Yew YW, Ye T, Ang LW, Ng LC, Yap G, James L, Chew SK, Goh
KT, 2009. Seroepidemiology of dengue virus infection among
adults in Singapore. Ann Acad Med Singapore 38: 667–675.
9. Siler JF, Hall MW, Hitchens AP, 1926. Dengue: its history, epi-
demiology, mechanism of transmission, etiology, clinical mani-
festations, immunity, and prevention. Philipp J Sci 29: 1–302.
10. Simmons JS, St. John JH, Reynolds FHK, 1931. Experimental
studies of dengue. Philipp J Sci 44: 1–247.
11. Nishiura H, Halstead SB, 2007. Natural history of dengue virus
(DENV)-1 and DENV-4 infections: reanalysis of classical stud-
ies. J Infect Dis 195: 1007–1013.
12. Halstead SB, 1974. Etiologies of the experimental dengues of
Siler and Simmons. Am J Trop Med Hyg 23: 974–982.
13. Gelman A, Carlin JB, Stern HS, Rubin DB, 2003. Bayesian Data
Analysis. Second edition. Boca Raton, FL: Chapman & Hall/
CRC (Taylor & Francis Group).
14. R Development Core Team, 2011. R: a language and environment
for statistical computing. R Foundation for Statistical Comput-
ing, Vienna, Austria. Available at: http://www.R-project.org.
15. Centers for Disease Control and Prevention, 2012. CDC Health
Information for International Travel 2012. New York: Oxford
16. Ooi EE, Wilder-Smith A, Ng LC, Gubler DJ, 2010. The 2007
dengue outbreak in Singapore. Epidemiol Infect 138: 958–961.
17. Hii YL, Rocklo ¨v J, Ng N, Tang CS, Pang FY, Sauerborn R, 2009.
Climate variability and increase in intensity and magnitude
of dengue incidence in Singapore. Global Health Action 2.
18. Ko Y-C, ChenM-J,Yeh S-M,1992.The predisposing and protective
factors against dengue virus transmission by mosquito vector. Am
J Epidemiol 136: 214–220.
COOK AND OTHERS