Contents lists available at SciVerse ScienceDirect
Vaccine
jou rn al h om epa ge: www.elsev ier .com/ locate /vacc ine
mpact of Vi vaccination on spatial patterns of typhoid fever in the slums ofolkata, India
ohammad Alia,∗, Dipika Surb, Deok Ryun Kima, Suman Kanungob, Sujit K. Bhattacharyab,yomkesh Mannab, R. Leon Ochiaia, John Clemensa
International Vaccine Institute, Seoul, Republic of KoreaNational Institute of Cholera and Enteric Diseases, Kolkata, India
r t i c l e i n f o
rticle history:eceived 12 May 2011eceived in revised form 6 September 2011ccepted 8 September 2011vailable online 20 September 2011
eywords:accine trialeographic analysis
a b s t r a c t
A mass typhoid Vi vaccination campaign was carried out among approximately 60,000 slum residentsof Kolkata, India. This study evaluated the impact of the campaign on spatial patterns of typhoid fever.Eighty contiguous residential groups of households in the study area were randomized to receive eithera single dose of the Vi polysaccharide vaccine or a single dose of the inactivated hepatitis A vaccineas the control agent. Persons aged two years and older were eligible to receive the vaccine. Vaccineprotection against typhoid fever was monitored for two years after vaccination at both outpatient andinpatient facilities serving the study population. Geographic analytic and mapping tools were used inthe analysis. Spatial randomness of the disease was observed during the pre-vaccination period, which
isk areas turned into a significant pattern after vaccination. The high-risk areas for typhoid were observed in thearea dominated by the control clusters, and the low-risk areas were in the area dominated by the Viclusters. Furthermore, the control clusters surrounded by the Vi clusters were low risk for typhoid fever.The results demonstrated the ability of mass vaccination to change the spatial patterns of disease throughthe creation of spatial barriers to transmission of the disease. Understanding and mapping the disease
signi
risk could be useful for de
. Background
Typhoid fever is a severe multi-systemic illness caused byalmonella enterica serovar Typhi (Salmonella Typhi) and is char-cterized by classic prolonged fever. The reservoir of S. Typhi isumans. Typhoid fever is usually contracted by the ingestion of
ecally contaminated food or water [1]. In endemic areas, identifiedisk factors for the disease include: eating food prepared outside ofhe home, such as ice cream or flavored iced drinks from street ven-ors [2,3]; drinking contaminated water [4]; having a close contactr relative with a recent case of typhoid fever [2,5]; poor housingith inadequate facilities for personal hygiene [6]; and the use of
ntimicrobial drugs [5]. Paratyphoid fever is an illness similar toyphoid fever, but is usually much milder and is caused by a differ-nt Salmonella serotype, S. enterica serovar Paratyphi (Salmonella
aratyphi) [7].
Typhoid fever results in an estimated 216,000–600,000 deathsnnually, with almost all cases occurring in developing countries
∗ Corresponding author at: International Vaccine Institute, SNU Research Park,an 4-8 Nakseongdae-dong, Gwanak-gu, Seoul 151-919, Republic of Korea.el.: +82 2 881 1127; fax: +82 2 872 2803.
[8,9]. In developing countries, including India, it poses a major pub-lic health problem [10], and the disease is considered to be endemicin different parts of India. It has been estimated that more than12 million cases occur annually in developing countries (exclud-ing China), out of which 7.7 million cases occur in India alone [11].Though typhoid is recognized as a major public health problem inthe poor urban areas of India, no population-based data on typhoidfever exist for Kolkata, a large city with many urban slums. How-ever, surveillance conducted in two urban slums of Kolkata fromNovember 2003 to October 2004 found the typhoid incidence to be1.6 cases per 1000 population per year [12].
Disease maps are useful tools to identify the geographical dis-tribution of disease incidence. The development of geographicinformation systems (GIS) over the last three decades has provideda powerful tool to examine spatial patterns, and is commonly usedin public health and epidemiologic research [13]. GIS may also helpinvestigators improve their study design, management, analysis,and interpretation of data to enhance the scientific quality of vac-cine trials [14]. GIS and spatial analyses have also been used tounderstand the impact of vaccination on the heterogeneous distri-
bution of disease in space [15].
In industrialized countries, typhoid fever was largely con-trolled through the improvement of water and sanitation systems.However, the development of such infrastructure requires huge
nvestment and is unlikely to reach slums and other high-risk areasor typhoid in developing countries for many years to come. Vac-ination can provide a near-term solution [16]. To demonstrateeduction of the burden of typhoid using vaccines, a mass vacci-ation using the Vi polysaccharide vaccine was carried out in thelums of Kolkata, India. The vaccine conferred 61% total protectiongainst typhoid fever in that population, and was associated withinimal side effects. Furthermore, the vaccine conferred signifi-
ant herd protection to non-vaccinees [17]. In this paper, we useIS methodologies to evaluate the impact of the mass vaccinationampaign on the spatial patterns of typhoid fever.
. Methods
.1. Study site and design
The study was conducted in Kolkata (formerly Calcutta), the cap-tal city of West Bengal in India. Kolkata is one of the most populousities of the world with a population of 13 million. Kolkata has threeeasons: the cool dry months from November to February, the hotry months from March to May, and the hot rainy months from
une to October.It is estimated that one-third of Kolkata residents live in nearly
500 overcrowded slums (locally called bustees). The slums have anrea of approximately 185 km2 and a total population of 4,572,876ccording to the census of India 2001. There are 141 wards inolkata. A contiguous area typical of the other city slums encom-asses most of Ward 29 and all of Ward 30 in eastern Kolkata isnown as Narkeldanga, and has a total population of about 60,000.his area was selected as the study site. In the study area, the streetsre narrow and congested, and upper-floor apartments look downn the tile roofs of one-story houses. Water is available throughunicipal taps but is supplied intermittently, requiring storage.
ommon municipal latrines are shared by many households, andewage collects in open drains, which tend to overflow duringhe rainy season, flooding adjacent homes. The water and sewageipelines lie close to each other and are prone to leakage [18].
The outline of the study area and many of its structures haveemained unchanged since the country’s independence from theritish Empire in 1947 [19]. A brief period of intense in-migrationccurred during the East Bengal (now Bangladesh) partition fromakistan when houses were divided up to accommodate theigrants. During the last two and half decades, infrastructure,
ncluding sanitary latrines, water points, and bathing platformsave been built, and nutrition centers and prenatal care haveecome available. Living conditions in the slum areas have
mproved for a large number of people but the demand for urbanand has far outpaced supply. Throughout the years, extensive sub-etting has resulted in overcrowding, as more and more people arequeezed into available housing [20]. Households in the study areaave an average of 1.5 rooms, a median of five members (range–30), and a median monthly family income of USD 67 [18].
Details on the study designed and study procedures were pre-iously reported [17]. In brief, 80 contiguous residential groups ofouseholds (called clusters) in the study area were randomized toeceive either a single dose of the Vi polysaccharide vaccine or a sin-le dose of the inactivated hepatitis A vaccine as the control agent.ersons aged two years and older were eligible to receive a vaccine.accine protection against typhoid fever was monitored for twoears after vaccination at both outpatient and inpatient facilitieserving the study population.
A census of the de jure population was conducted prior to vac-
ination to enumerate all households and individuals in the fieldrial area. Each household and individual in the census was assigned
unique study identification number. Each household was refer-nced geographically in a GIS database to allow spatial analyses of
(2011) 9051– 9056
the study population. The census and geographical mapping wereused to divide the area into 80 geographic clusters that served asthe units of randomization.
2.2. Administration of agents
The vaccines were administered between November 27 andDecember 31, 2004. One vaccination center was set up in each clus-ter, and 20 of the 80 clusters achieved vaccination coverage in aweek. Individuals were eligible for receipt of the vaccine assignedto the cluster if they were two years old or older at the time of vac-cination, had no fever, were not pregnant or lactating, and providedwritten informed consent to participate in the study. All subjectsor their guardians provided written informed consent. In the caseof children, principle of observing assent was also followed. A totalof 37,673 subjects received a dose of a study vaccine. All vaccineeswere observed for 30 min following dosing to detect and manageimmediate adverse events. Vaccination was undertaken at specialvaccination centers, one for each sub-region, and the schedule ofvaccination was arranged so that vaccinations with the Vi vaccineand the control agent were temporally balanced.
2.3. Surveillance for typhoid fever
To detect enteric fever in the study population, five study clinicswere established. These clinics gave routine care for all outpa-tient visits, and referred patients with severe disease for hospitalcare. Private medical care providers in the two study wards wereencouraged to refer patients to these study clinics for diagnosticevaluation. In addition, the emergency rooms, outpatient clinics,and inpatient wards of the Infectious Disease Hospital and B.C. RoyChildren’s Hospital monitored all patients from the study area pre-senting with febrile illness. In each of these treatment settings,all subjects from the study area presenting with a history of atleast three days of fever were examined by a study physician. Dataon the subject’s history and physical findings were systematicallydocumented on a structured clinical case report form, after verbalinformed consent was obtained and a blood specimen was collectedfor culture. For the blood culture, approximately 5–8 milliliters (ml)of blood was collected to inoculate Bactec Plus Aerobic® culturebottles (Becton Dickinson, New Jersey, USA); for younger children(less than 13 years of age), about 3–5 ml of blood was inoculatedinto Bactec Peds Plus® culture bottles. Standard biochemical andserologic methods were used to identify S. Typhi and S. Paratyphi[21]. When a blood culture yielded either S. Typhi or S. Paratyphi, astudy team was dispatched to the home of the patient within sevendays to verify that the subject whose name was given at the treat-ment site had indeed visited the treatment site for care on the datenoted in the surveillance.
2.4. Spatial clusters
To detect potential geographic areas that are high- and low-riskfor disease, the spatial scan test has been widely used in recenttimes [22–25]. The spatial scan test is also suitable for unevengeographic distribution of cases and population density [26]. Weused the spatial scan test implemented through SaTScan® [23] toidentify unique non-random spatial clusters that are higher riskand lower risk for typhoid and paratyphoid illnesses. We assumedthat the incidences of typhoid and paratyphoid followed a Poissondistribution. Under the null hypothesis, the incidence of disease ina particular location is proportional to the number of residents in
that location [26]. Using SaTScan®, we estimated the probabilitythat the frequency of disease at each peak surpasses that expectedby chance. We set the space limitations to 50% population at risk,which allowed us to scan for both large and small clusters of disease
M. Ali et al. / Vaccine 29 (2011) 9051– 9056 9053
Table 1Summary of statistics of the study variables.
Pre-vaccination period First year of post-vaccinationperiod
Second year ofpost-vaccination period
Period (one year) November 2003 to October2004
January 2005 to December2005
January 2006 to December2006
Number of address locations in GIS 8727 8713 8531Vi or control agent recipients 37,673 37,578 36,376Typhoid cases (incidence/100,000, 95% CI) 73 (194, 154–244) 61 (162, 126–208) 69 (190, 150–240)Paratyphoid cases (incidence/100,000, 95% CI) 39 (104, 76–141) 39 (104, 76–142) 62 (170, 133–218)
raetsCuicawiiS
2
Dtpt2
2
ibaRt
2
ys
3
cpmtaoydswd
p-Value for the difference in incidence ratesbetween typhoid and paratyphoid
0.001
isk. We took into account the observed number of cases insidend outside the circle when calculating the highest likelihood forach circle. This circle was the most probable cluster and had a ratehat was the least likely to happen by chance alone. The statisticalignificance of possible clusters was calculated using 999 Montearlo simulations [27]. Purely spatial analysis was performedsing circular windows. If the frequency of the disease at a location
s more frequent than that is expected, then the location wasonsidered as a high-risk area; and if the frequency of the diseaset a location is less frequent than that is expected, then the locationas considered as a low-risk area. The output from SaTScan® was
mported into ArcGIS (Version 9.2, California, USA) to map signif-cant (p < .01) clusters at higher risk and lower risk for S. Typhi and. Paratyphi.
.5. Study period
Since the vaccination was carried out during November andecember 2004, we considered the one year pre-vaccination period
o be from November 2003 to October 2004, the first year of theost-vaccination period from January 2005 to December 2005, andhe second year of the post-vaccination period to be from January006 to December 2006.
.6. Ethics
The underlying main project “Effectiveness of the Vi vaccinen Kolkata, India” that this study was a part of was approvedy the ethics committees of the National Institute of Cholerand Enteric Diseases (NICED) and the Indian Council of Medicalesearch (ICMR), as well as by the Institutional Review Board ofhe International Vaccine Institute.
.7. Role of the funding sources
The sponsors of this research had no role in the design and anal-sis of this study, the writing of this paper, or in the decision toubmit this paper for publication.
. Results
There were 37,763 recipients of either the Vi or hepatitis A vac-ine residing at 8727 addresses in the GIS database during there-vaccination period (Table 1). Out of these recipients, 95 subjectsigrated out or died after vaccination and before the beginning of
he post-vaccination period (January 2005), leaving 37,578 subjectst 8713 locations in the GIS database to be analyzed for the first yearf the post-vaccination period. Before the beginning of the secondear of the post-vaccination period, 1202 subjects migrated out or
ied, leaving 36,376 persons at 8531 locations to be analyzed in theecond year of the post-vaccination period. The incidence of S. Typhias 194/100,000 and the incidence of S. Paratyphi was 104/100,000uring the pre-vaccination period (Table 1). During the first year
0.03 0.54
of post-vaccination, there was a lower rate of S. Typhi incidencethan that of the pre-vaccination period. The incidence rate for S.Paratyphi in the first year of post-vaccination remained the sameas that for the pre-vaccination period. In the second year of the post-vaccination period, the incidence rate of S. Typhi was 190/100,000and the incidence rate of S. Paratyphi was 170/100,000. A signif-icant difference was seen in the incidence rates between typhoidand paratyphoid fever during the pre-vaccination period and thefirst year of the post vaccination period, but was not seen in thesecond year of the post-vaccination period (Table 1).
Through the spatial scan test using SaTScan® and by setting50% of the study population at risk, we observed a small area(0.02 km2) that was significantly high risk (p = .01) for S. Typhiinfection during the pre-vaccination period in the southwest ofthe study area. No areas that were significantly low risk for thedisease were observed as a result of the spatial analysis (Fig. 1).The risk for infection by S. Typhi inside the high-risk area was4.79 times the risk than that outside the high-risk area. For paraty-phoid, there was no significantly high- or low-risk area during thepre-vaccination period. In the first year of the post-vaccinationperiod, there were one significantly high-risk area (0.08 km2) for S.Typhi infection (p < .01) and one significantly low-risk (p = .01) area(0.43 km2) for the disease (Fig. 2). The high-risk area had a relativerisk of 6.12 and the low-risk area had a relative risk of 0.23, in com-parison to the risk outside the area. Similar to the pre-vaccinationperiod, there was no high- or low-risk area for S. Paratyphi dur-ing the first year of the post-vaccination period. In the secondyear of post-vaccination period, there were two small (0.001 km2
and 0.003 km2) significantly high-risk areas (p < .01) and one large(though not as large as that in the first year of the post-vaccinationperiod) area (0.18 km2) of significant low-risk (p = .01) for S. Typhi(Fig. 3). The high-risk area for S. Typhi infection had a relativerisk of 16.89, and the low-risk area had no risk (relative risk was0.0), in comparison to the risk outside the area during the secondyear of the post-vaccination period. A significantly high-risk area(0.18 km2) for S. Paratyphi infection was noticed in the second yearof the post-vaccination period (p < .01), with a relative risk of 5.14inside the high-risk area compared to that outside the high-riskarea.
The resulting maps show that the high-risk area for typhoidfever in the first year of the post-vaccination period was an expan-sion of the risk area from the pre-vaccination period, and theexpansion took place mostly in the area where the control agentswere administered. In contrast to the pre-vaccination period, alarge significantly low-risk area (p = .01) for typhoid fever wasnoticed in the northern part of the study area where the Vi vac-cines were mostly administered. The high-risk areas for typhoidfever in the second year of the post-vaccination period were fullysuperimposed on the high-risk area from the first year of the post-
vaccination period, and were entirely located in the control clusters.However, the significantly low-risk area in the second year of thepost-vaccination period was partially superimposed on the low-risk area from the first year of the post-vaccination period. The
9054 M. Ali et al. / Vaccine 29 (2011) 9051– 9056
Fig. 1. Spatial distribution of typhoid and paratyphoid cases and the high- and low-risk areas of typhoid and paratyphoid fever during the pre-vaccination period, Kolkata,India.
Fig. 2. Spatial distribution of typhoid and paratyphoid cases and the high- and low-risk areas of the diseases during the first year of the post-vaccination period, Kolkata,India.
M. Ali et al. / Vaccine 29 (2011) 9051– 9056 9055
F w-riskI
hvc
4
iaiowsrcataTaptvt
sspvarian
ig. 3. Spatial distribution of typhoid and paratyphoid cases and the high- and londia.
igh-risk area for paratyphoid fever in the second year of the post-accination period appeared to be in the middle of the study area,omprising both Vi and hepatitis A clusters.
. Discussion
This study identified one small high-risk area for typhoid fevern the southern part of the study area and no significantly low-riskrea for the disease, indicating spatial randomness of typhoid fevern the study area during the pre-vaccination period. In the first yearf the post-vaccination period, a large area of high risk for typhoidas observed in the southwest part of the study area; and in the
econd year of the post vaccination period, two small areas of highisk were observed. These high risk areas were dominated by theontrol (hepatitis A vaccine) clusters. In contrast, a large low-riskrea was observed in the northern part of the study areas in bothhe first and second years of the post-vaccination period, and thesereas of low risk for typhoid were dominated by the Vi clusters.he elevated disease risk in the southwest part of the study area,nd the low disease risk in the northern part of the area during theost-vaccination period indicate difference in spatial patterns ofhe disease risk after vaccination than that observed during the pre-accination period. This illustrates the ability of mass vaccinationo change spatial patterns of disease risk.
The size of the high-risk area in the pre-vaccination period ismaller than that in the post-vaccination period, and is mostlyuperimposed on the high-risk area from the post-vaccinationeriod. The expanded part of the high-risk area in the post-accination period is located further south in the study area where
number of control clusters are located. Since the north in the high-
isk area was mostly bordered by Vi clusters, any risk of the diseasen that part of the study area was not observed. Due to chance vari-tions in randomization of the clusters, some control clusters in theorthern part of the study area were surrounded by Vi clusters, and
areas of diseases during the second year of the post-vaccination period, Kolkata,
these clusters experienced decreased disease burden during thepost-vaccination period. This suggests that by vaccinating a groupof people in an area, spatial barriers to transmission of the dis-ease can be created. Therefore, non-vaccinees near the communityreceiving the vaccines may benefit, possibly due to a herd protec-tive effect of the vaccine, as observed in our previous study [17].The results also raise the possibility that vaccination with Vi aroundareas at high risk of typhoid may assist in disease control.
We did not find any changes in spatial patterns of paratyphoidfever during the first year of the post-vaccination period. However,we observed a high-risk area for paratyphoid in the second year ofthe post-vaccination period. Unlike the high-risk area for typhoid,the high-risk area for paratyphoid was not dominated by the controlclusters. This suggests that the change in spatial patterns of typhoidfever in the post-vaccination period was due to the effect of themass vaccination campaign and not due to ecological change of thearea over time. And the stability of spatial patterns of paratyphoidfever between pre- and post-vaccination periods illustrates thatthe use of Vi vaccine may not change the spatial epidemiology ofparatyphoid fever. Investigating the causes of the higher incidenceof paratyphoid fever and the formation of a high-risk area in thesecond year of the post-vaccination period were beyond the scopeof this work, but may be considered in a future study.
The limitation of this study is that the comparison of high-and low-risk areas before and after vaccination was made throughvisual interpretation. However, there is a clear distinguishable pat-tern of the risk areas before and after vaccination, and one can easilyidentify this variation.
The findings of this paper have important public health impli-cations for vaccine distribution policy. It is desirable that vaccines
should be allocated in such a way in order to avert as many infec-tions as possible. Our results suggest that for a vaccine such as Vi,with known indirect protective capability [17], a community-basedvaccine distribution policy may help reduce the disease burden in
9 ne 29
aonasottmt
A
eaKatd
dtw
R
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
056 M. Ali et al. / Vacci
n effective way. This of course does not guarantee the protectionf non-vaccinees after the waning of vaccine protection in vacci-ees. A vaccine trial is important for evaluating vaccine efficacy,s well as its effectiveness [28]. The spatial analysis in this studyhows the ability of mass vaccination to change spatial patternsf the disease, possibly through the creation of spatial barriers toransmission of the disease, and illustrates in a spatial fashion howhe indirect benefit of vaccination can be achieved. These findings
ay be useful for formulating a cost-effective vaccination strategyo control disease.
cknowledgements
This study was supported by the Diseases of the Most Impov-rished Program, funded by the Bill & Melinda Gates Foundationnd by the governments of the Republic of Korea, Sweden, anduwait. GlaxoSmithKline donated the vaccines used in the studynd performed serological assays, but they provided no funding forhe study. The Japanese International Cooperation Agency suppliediagnostic equipment and reagents for the blood culture.
Contributors: MA, DS, DRK, SK, RLO, and JC participated in theesign, conduct, and analysis of the study, and in the writing ofhe manuscript. SKB and BM participated in the analysis and in theriting of the manuscript. Conflict of interest: None declared.
eferences
[1] Parry CM, Hien TT, Dougan G, White NJ, Farrar JJ. Typhoid fever. N Engl J Med2002;347(22):1770–82.
[2] Black RE, Cisneros L, Levine MM, Banfi A, Lobos H, Rodriguez H. Case–controlstudy to identify risk factors for paediatric endemic typhoid fever in Santiago,Chile. Bull World Health Organ 1985;985(63):899–904.
[3] Luby SP, Faizan MK, Fisher-Hoch SP, Syed A, Mintz ED, Bhutta ZA, et al. Riskfactors for typhoid fever in an endemic setting, Karachi, Pakistan. EpidemiolInfect 1998;120:129–38.
[4] Mermin JH, Villar R, Carpenter J, Roberts L, Samaridden A, Gasanova L, et al. Amassive epidemic of multidrug-resistant typhoid fever in Tajikistan associatedwith consumption of municipal water. J Infect Dis 1999;179:1416–22.
[5] Luxemburger C, Chau MC, Mai NL, John W, Tran TH, Julie AS, et al. Risk factors fortyphoid fever in the Mekong Delta, southern Viet Nam: a case–control study.Trans R Soc Trop Med Hyg 2001;95:19–23.
[8] Crump JA, Luby SP, Mintz ED. The global burden of typhoid fever. Bull WorldHealth Organ 2004;82:346–53.
[9] Ivanoff B, Levine MM, Lambert PH. Vaccination against typhoid fever: presentstatus. Bull World Health Organ 1994;72:957–71.
10] Edelman R, Levine MM. Summary of an international workshop on typhoidfever. Rev Infect Dis 1986;8:329–49.
11] Patnaik KC, Kapoor PN. A note on incidence of typhoid in Delhi. Ind J Med Res1967;55:228–40.
12] Sur D, von Seidlein L, Manna B, Dutta S, Deb AK, Sarkar BL, et al. The malaria andtyphoid fever burden in the slums of Kolkata, India: data from a prospectivecommunity-based study. Trans R Soc Trop Med Hyg 2006;100(8):725–33.
13] Toprak D, Erdogan S. Spatial analysis of the distribution of typhoidfever in Turkey. Int Arch Photogram Remote Sens Spatial Inform Sci2008;XXXVII(B8):1367–72.
14] Ali M, Clemens J. Editorial: Ecological aspects in vaccine trials. Expert Rev Vac-cines 2008;7(3):279–81.
15] Ali M, Emch M, Yunus M, Clemens J. Modeling spatial heterogeneity of diseaserisk and evaluation of the impact of vaccination. Vaccine 2009;27:3724–9.
16] DeRoeck D, Jodar L, Clemens J. Putting typhoid vaccination on the global healthagenda. N Engl J Med 2007;357(11):1069–71.
17] Sur D, Ochiai RL, Bhattacharya SK, Nirmal K, Ganguly NK, Ali M, et al. A cluster-randomized effectiveness trial of Vi typhoid vaccine in India. N Engl J Med2009;361:335–44.
18] Sur D, Manna B, Deb AK, Niyogi SK, Kanungo S, Sarkar BL, et al. Correlates fordiarrhoea and treatment uptake in an impoverished slum area of Kolkata, India.J Health Popul Nutr 2004;22:130–8.
19] Kundu A, Nag P. Atlas of the city of Calcutta and its environs. Calcutta: NationalAtlas and Thematic Mapping Organisation, Ministry of Science and Technology,Government of India; 1996.
20] Thomas FC, Calcutta. The human face of poverty. London: Penguin Books Ltd.;1997. p. 3–166.
21] Bopp CA, Brenner FW, Wells JG, Strockbine NA. Escherichia, Shigella andSalmonella. In: Murray PR, editor. Manual of clinical microbiology. 7th ed.Washington, DC: ASM Press; 1999. p. 459–74.
22] Kulldorff M, Feuer EJ, Miller BA, Freedman LS. Breast cancer clustersin the northeast United States: a geographic analysis. Am J Epidemiol1997;146:161–70.
23] Kulldorff M, Athas WF, Feurer EJ, Miller BA, Key CR. Evaluating cluster alarms:a space–time scan statistic and brain cancer in Los Alamos, New Mexico. Am JPublic Health 1998;88:1377–80.
24] Hjalmars U, Kulldorff M, Gustafsson G, Nagarwalla N. Childhood leukaemia inSweden: using GIS and a spatial scan statistic for cluster detection. Stat Med1996;15:707–15.
25] Gosselin P, Lebel G, Rivest S, Douville-Fradet M. The Integrated System for Pub-lic Health Monitoring of West Nile Virus (ISPHM-WNV): a real-time GIS forsurveillance and decision-making. Int J Health Geogr 2005;4:21.
26] Sheehan TJ, DeChello LM, Kulldorff M, Gregorio DI, Gershman S, Mroszczyk M.The geographic distribution of breast cancer incidence in Massachusetts 1988to 1997, adjusted for covariates. Int J Health Geogr 2004;3:17.
27] Mostashari F, Kulldorff M, Hartman JJ, Miller JR, Kulasekera V. Dead bird clus-
ters as an early warning system for West Nile virus activity. Emerg Infect Dis2003;9:641–6.
28] Clemens J, Brenner R, Rao M, Tafari N, Lowe C. Evaluating new vac-cines for developing countries. Efficacy or effectiveness? JAMA 1996;275(5):390–7.
