Morbidity and Mortality Weekly Report

Supplement April 28, 2006 / Vol. 55

Global Epidemiology

Proceedings of the Third TEPHINET Conference —

Beijing, China, November 8–12, 2004

depar tment of health and human ser

department vices
services
Centers for Disease Control and Prevention
 MMWR

CONTENTS
The MMWR series of publications is published by the
Coordinating Center for Health Information and Service, Preface ................................................................................ 1
Centers for Disease Control and Prevention (CDC), U.S.
Department of Health and Human Services, Atlanta, GA 30333. Investigation of Avian Influenza (H5N1) Outbreak
in Humans — Thailand, 2004/Darin Areechokchai,
SUGGESTED CITATION C. Jiraphongsa, Y. Laosiritaworn, W. Hanshaoworakul,
General: Centers for Disease Control and Prevention. Global M. O’ Reilly ........................................................................ 3
epidemiology: proceedings of the third TEPHINET Occupational Injuries Among Workers in the Cleansing
Conference — Beijing, China, November 8–12, Section of the City Council’s Health Services
2004. MMWR 2006;55(Suppl). Department — Bulawayo, Zimbabwe, 2001–2002/
Specific: [Author(s)]. [Title of particular article]. In: Global Elizabeth Gonese, R. Matchaba-Hove, G. Chirimumba,
epidemiology: proceedings of the third TEPHINET Z. Hwalima, J. Chirenda, M. Tshimanga .............................. 7
Conference — Beijing, China, November 8–12, Progress Toward Tuberculosis Control and Determinants
2004. MMWR 2006;55(Suppl):[inclusive page of Treatment Outcomes — Kazakhstan, 2000–2002/
numbers]. Ekaterina Bumburidi, S. Ajeilat, A. Dadu,
I. Aitmagambetova, J. Ershova, R. Fagan,
Centers for Disease Control and Prevention M.O. Favorov ................................................................... 11
Julie L. Gerberding, MD, MPH Estimation of Measles Vaccination Coverage Using
Director the Lot Quality Assurance Sampling (LQAS) Method —
Tamilnadu, India, 2002–2003/Saravanan Sivasankaran,
Dixie E. Snider, MD, MPH P. Manickam, R. Ramakrishnan, Y. Hutin, M.D. Gupte ........ 17
Chief Science Officer
Bacterial Meningitis Among Cochlear Implant Recipients —
Tanja Popovic, MD, PhD Canada, 2002/Samantha D. Wilson-Clark, S. Squires,
Associate Director for Science S. Deeks ........................................................................... 21
Coordinating Center for Health Information Risk Factors for Neonatal Tetanus — Busoga Region,
and Service Uganda, 2002–2003/Sheba N. Gitta,
Steven L. Solomon, MD F. Wabwire-Mangen, D. Kitimbo, G. Pariyo ........................ 26
Director Risk Factors for Brucellosis — Leylek and Kadamjay
Districts, Batken Oblast, Kyrgyzstan, January–November,
National Center for Health Marketing
2003/Turatbek B. Kozukeev, S. Ajeilat, E. Maes,
Jay M. Bernhardt, PhD, MPH M. Favorov ....................................................................... 32
Director
Salmonellosis Outbreak Among Factory Workers —
Division of Scientific Communications Huizhou, Guangdong Province, China, July 2004/
Judith R. Aguilar Lunguang Liu, H.F. He, C.F. Dai, L.H. Liang, T.Li,
(Acting) Director L.H. Li, H.M. Luo, R. Fontaine ........................................... 36

Mary Lou Lindegren, MD Varicella Outbreak Among Primary School Students —

Beijing, China, 2004/Huilai Ma, R. Fontaine .................... 40
Editor, MMWR Series
Suzanne M. Hewitt, MPA
Managing Editor, MMWR Series
Teresa F. Rutledge
Lead Technical Writer-Editor
Jeffrey D. Sokolow, MA
Project Editor Notice

Beverly J. Holland This supplement is being published as a courtesy to the MMWR

Lead Visual Information Specialist readership. The findings and conclusions in these papers represent the
Lynda G. Cupell views of the authors and do not necessarily represent the views of CDC.
Malbea A. LaPete
Visual Information Specialists The material in this supplement originated at a meeting of the
Training in Epidemiology and Public Health Interventions Net-
Quang M. Doan, MBA work (TEPHINET) held in Beijing, China, during November
Erica R. Shaver 8–12, 2004. It was prepared for publication by the Coordinating
Information Technology Specialists Office for Public Health, Stephen C. Blount, MD, Director.
Vol. 55 / Supplement MMWR 1

Preface
Accurate epidemiologic information is essential for making This enables TEPHINET to offer global and regional organiza-
good decisions about developing, implementing, monitoring, tions access to public health professionals who conduct surveil-
and evaluating health policies. To ensure the quality of infor- lance and respond to health threats. Network members serve more
mation gathered, CDC created the Epidemic Intelligence Ser- than half of the world’s population, including the United States
vice (EIS) in 1951. Since then, approximately 32 countries (Figure). Just as EIS provides critical personnel for the U.S. pub-
have developed field epidemiology and allied training pro- lic health system, FETPs in TEPHINET provide essential staff
grams (FETPs) based on the EIS model, all sharing a com- to their own countries and regions.
mon principle of training through service. As founding members of TEPHINET, WHO and CDC
In the 1990s, work began to create a unified global network, maintain close partnerships with the organization. With
and in 1999, the Training Programs in Epidemiology and Pub- assistance from WHO and CDC, TEPHINET is piloting a
lic Health Interventions Network (TEPHINET) was formed. continuing quality improvement process that is expected to
TEPHINET is dedicated to strengthening international public lead to steady improvement in the quality of science and level
health capacity by enhancing competencies in applied epide- of service that FETPs provide. In addition to its training and
miology and public health practice. Nearly all of the 32 FETPs capacity building role, TEPHINET is a member of WHO’s
in the network began as partnerships between national minis- Global Outbreak Alert and Response Network (1) and pro-
tries of health, the Division of Epidemiology and Surveillance vides experts for WHO investigative teams.
Capacity Development (formerly the Division of International The new International Health Regulations require each
Health) of CDC’s Coordinating Office for Global Health, and nation to have competent public health personnel for epi-
the World Health Organization (WHO). demic surveillance and response (2). TEPHINET and its mem-
This supplement to the MMWR highlights the work of epi- ber programs will be major participants in achieving this goal.
demiologists who have graduated from TEPHINET member TEPHINET, WHO, and CDC are currently working together
programs. The articles were developed from abstracts presented to create new FETPs in seven countries or regions worldwide.
in Beijing, China, at the Third Global Scientific Conference Much of the work of FETP staff, trainees, and graduates is
of TEPHINET during November 8–12, 2004. Major fund- accomplished as they fulfill duties in their national public health
ing for the conference was provided by the Bill and Melinda systems and has not been published. Because of the value of
Gates Foundation. Approximately 230 field epidemiologists this applied science, we are pleased at this opportunity to share
from 40 countries participated to share best practices and new a selection of the many excellent studies produced by
ideas in field epidemiology, surveillance, response, and train- TEPHINET member programs and their graduates.
ing. The reports selected for publication in this supplement
Stephen B. Blount, MD Roberto Flores, MD
focus on the contributions TEPHINET members and their
Associate Director for Global Health and Chairman, TEPHINET
trainees and graduates have made in surveillance of and Director, Coordinating Office
response to emerging infectious diseases. They reflect the broad for Global Health, CDC
impact of the training programs.
Field epidemiology trainees studied major emerging prob-
FIGURE. Field epidemiology training and allied programs
lems of global concern (e.g., avian influenza in Thailand) and (FETPs), 2005
common outbreaks (e.g., salmonellosis and varicella in China).
Trainees applied their skills to investigate occupational inju-
ries in Zimbabwe, evaluate a tuberculosis control program in
Kazakhstan, and adapt statistical methodologies used for
industrial quality control to estimate vaccination coverage in
India. The Canadian contribution highlights the application
of epidemiologic methods to improve product safety for
cochlear implants, and two papers discuss how to use case-
control methodology to identify risk factors for brucellosis in
Kyrgyzstan and neonatal tetanus in Uganda.
A key characteristic of TEPHINET members’ training-through-
service programs is their location in national ministries of health. FETPs
2 MMWR April 28, 2006

References Acknowledgments
1. World Health Organization. Global Outbreak Alert and Response Net- The following persons assisted in the publication of this
work. Available at http://www.who.int/csr/outbreakwork/en.index.html. supplement: Edmond F. Maes, PhD, Angeli Abrol, MHSc, Linda
2. World Health Organization. Revision of the International Health Carnes, DrPA, Georgina Castro, MPH, Ra Shel Cromwell, MPH,
Regulations. Geneva, Switzerland: World Health Organization; 2005. Kathy Harben, Valerie Kokor, MBA, Brenda Lawver, Sharel Mitchell,
Available at http://www.who.int/gb/ebwha/pdf_files/WHA58/WHA
Patricia Simone, MD, Coordinating Office for Global Health, CDC.
58_3-en.pdf.

Peer Reviewers
The following persons served as peer reviewers for these proceedings: Frederick J. Angulo, DVM, PhD, Thomas Clark, MD, Nancy
Rosenstein, MD, Division of Bacterial and Mycotic Diseases, David Shay, MD, Division of Viral and Rickettsial Diseases, National Center
for Infectious Diseases; Janet Blair, PhD, Career Development Division, Office of Workforce and Career Development; Maryam B. Haddad,
MPH, MSN, Division of TB Elimination, National Center for HIV, STD, and TB Prevention; Rafael Harpaz, MD, John S. Moran, MD,
Epidemiology and Surveillance Division, John Stevenson, MS, Immunization Services Division, National Immunization Program; Mark
Papania, MD, Office of the Chief Science Officer, Bettylou Sherry, PhD, Division of Nutrition and Physical Activity, National Center for
Chronic Disease Prevention and Health Promotion; Victor M. Caceres, MD, Robert Fontaine, MD, Karen Gieseker, PhD, Rubina Imtiaz,
MBBS, Donna Jones, MD, Edmond Maes, PhD, James M. Mendlein, PhD, Henry Walke, MD, Mark White, MD, FACPM, Coordinating
Office for Global Health, CDC.
Vol. 55 / Supplement MMWR 3

Investigation of Avian Influenza (H5N1) Outbreak

in Humans — Thailand, 2004
Darin Areechokchai,1 C. Jiraphongsa,1 Y. Laosiritaworn,1 W. Hanshaoworakul,1 M. O’Reilly2
1Bureau of Epidemiology, Department of Disease Control, Ministry of Public Health, Nonthaburi, Thailand
2Thailand Ministry of Public Health–U.S. CDC Collaboration, Nonthaburi, Thailand

Corresponding author: Darin Areechokchai, Bureau of Epidemiology, Ministry of Public Health, Nonthaburi 11000, Thailand. Telephone: 662-5901734;
Fax: 662-5918581; E-mail: doggyrin@health3.moph.go.th.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: Beginning in late 2003, a substantial outbreak of influenza A (H5N1) virus spread among poultry in Thailand. On
January 23, 2004, the Ministry of Public Health (MPH) detected the first confirmed human case of H5N1 infection in humans.
Methods: During February–November 2004, the MPH’s Bureau of Epidemiology and provincial health offices worked
together to investigate the H5N1 outbreak in humans. Two studies were conducted: a descriptive study to describe clinical
manifestations and epidemiologic characteristic of the cases and a matched case-control study to determine risk factors for
persons who might subsequently become ill with H5N1.
Results: A total of 16 patients with confirmed H5N1 were identified for the case-control study. Fever and respiratory symp-
toms predominated. Leucopenia and thrombocytopenia were present respectively in nine (100%) and four (44%) persons
aged <15 years. Direct touching of unexpectedly dead poultry was the most significant risk factor (odds ratio = 29.0; 95%
confidence interval = 2.7–308.2). Overall mortality was 75%; mortality for persons aged <15 years was 90%, compared
with 57% for persons aged >15 years.
Conclusion: Avian influenza was more severe in children, who should avoid handling dead poultry during epizootics. Early
avian influenza in children resembled the more common dengue fever, but presence of cough and absence of hemoconcentration
distinguished avian influenza, which often progressed rapidly to acute respiratory distress syndrome, requiring intensive care.

Introduction provincial health offices worked together to investigate the

H5N1 outbreak in humans. Two studies were conducted: a
In 1997, the first outbreak of avian influenza A (H5N1) in
descriptive study to describe clinical manifestations and epi-
humans was reported in Hong Kong, and exposure to infected
demiologic characteristic of the cases and a matched case-
poultry was identified as the probable route of transmission
control study to determine risk factors for persons who might
(1–3). Since late 2003, this highly pathogenic virus has caused
subsequently become ill with H5N1.
disease outbreaks in poultry in at least eight East Asian coun-
tries, and fatalities among humans have been attributed to the
same virus in Cambodia, China, Indonesia, Thailand, and Descriptive Study
Vietnam (4,5). During December 2003–October 2004, a Case Definition
total of 17 cases of respiratory disease caused by H5N1
viruses were reported in Thailand; these cases coincided in Under Thailand’s nationwide avian influenza surveillance
time and place with the epizootic disease in poultry. Millions system, which BoE initiated in December 2003, any patient
of persons in Thailand raise backyard chickens near their admitted to a health-care facility with pneumonia or influenza-
houses, which has increased concern about the potential for a like illness who has a history of exposure to poultry is reported
substantial H5N1 outbreak in humans. to BoE. For this study, a case was defined as H5N1 illness
occurring in a person who had received a diagnosis of pneu-
monia or influenza-like illness and who had either a positive
Methods viral culture for H5N1 virus or confirmation of H5 strain by
real-time reverse transcription-polymerase chain reaction
During February–November 2004, the Bureau of Epide- (RT-PCR) (6,7).
miology (BoE) of the Ministry of Public Health (MPH) and
4 MMWR April 28, 2006

Laboratory Testing Info 2002 version 2. The Mantel-Haenszel summary chi square
Each patient was evaluated for H5N1 infection status at test was used for the matched data.
the time of hospital admission. Nasopharyngeal aspirates or
nasopharyngeal swabs were collected and submitted to the
Thai National Institute of Health or to Siriraj Hospital at
Results
Mahidol University for laboratory testing. The outbreak in humans occurred in two episodes:
January–March and August–October 2004 (Figure 1). Of 17
Data Collection and Analysis cases reported from the avian influenza surveillance system,
A standard form was used to collect information concern- one was excluded because of incomplete laboratory results;
ing clinical manifestations as recorded on medical records. 16 patients with confirmed H5N1 infection were included in
Clinical manifestations and clinical progressions were analyzed the study.
using Epi Info 2002 version 2 (CDC, Atlanta, Georgia) and Patients lived in 11 provinces, primarily in central Thai-
Excel (Microsoft, Redmond, Washington). land. Nine were male and seven female. Patients varied by age
(range: 2–58 years; median: 13 years). Of the 16 patients
Analytic Study included in the study, nine (56%) were aged <15 years. Over-
all mortality was 75%; mortality for persons aged <15 years
Study Design and Sampling was 90%, compared with 57% for persons aged >15 years
A matched case-control study was conducted during the (OR = 6.4; CI = 0.4–204.2).
outbreak using the same case definition as the descriptive study. The incubation period varied (range: 2–8 days; median:
Population registers in the primary health-care unit of each 3 days). All patients had fever. Respiratory symptoms pre-
village were used as the control sampling frame. Controls were dominated, including cough (12 patients [75%]), sputum
selected matching village and age ±1 year to each patient and (12 [75%]), dyspnea (11 [69%]), and rhinorrhea (seven
then randomly selecting four controls for each patient. For [44%]); four (25%) patients had gastrointestinal symptoms
one patient who was thought to have been infected through (Figure 2). All nine patients aged <15 years had leucopenia
human-to-human transmission, four controls were selected (white blood cell count: <5,000 cells/mm3; normal range:
from among health-care workers in the hospital to which the 5,000–10,000 cells/mm3), and four (44%) also had thromb-
patient was admitted. A control was defined as any person ocytopenia (platelet count: <100,000 cells/mm3; normal range:
who had no fever (i.e., temperature >100.4°F [>38.0°C]) 100,000–400,000 cells/mm3). The median hematocrit in per-
<7 days before or after the onset of illness of the matched sons aged <15 years was 39% (range: 30%–41%). Clinical
case-patient. With respect to exposure, unexpectedly dead progressions demonstrated rapid progressive pneumonia and
poultry was defined as death of >10% of
all poultry in a farm or house within FIGURE 1. Number* of persons with confirmed cases of avian influenza, by date
1 day or >40% within 3 days. of onset — Thailand, December 2003–October 2004
6
Data Collection and Analysis The first The second
round of Culling round of Culling
Interviews were conducted by physi- 5 outbreak poultry outbreak poultry
in poultry started in poultry started
cians in BoE’s Field Epidemiology Train- started started
4 Culling
ing Program. For persons who had died poultry
Culling
Number

poultry
or were aged <12 years, adult proxies were ended ended
3
interviewed. A standard questionnaire
was used to collect information regard- 2
ing demographic characteristics, under-
lying diseases, exposure to other persons 1
with suspected avian influenza, and
activities related to poultry in the week 0
1 15 29 12 26 9 23 8 22 5 19 3 17 31 14 28 12 26 9 23 9 20 4 18
before onset of illness in the matched
Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct
case-patient. In the univariate matched
2003 2004
analyses, the matched odds ratio (OR) Date
and 95% confidence interval (CI) were
* N = 16.
calculated for each exposure using Epi
Vol. 55 / Supplement MMWR 5

FIGURE 2. Clinical manifestations of influenza A (H5N1) TABLE 2. Number and percentage of influenza A (H5N1)
patients from onset of illness to hospital admission, by case-control study participants, by selected characteristics —
symptom and age — Thailand, December 2003–October 2004 Thailand, 2004
Cases Controls
Fever* (n = 16) (n = 64)
Cough Characteristic No. (%) No. (%)
Sputum Median no. of family members 4 5
Rhinorrhea Median age (yrs) 14 14
Dyspnea Sex
Vomiting Adult (age >15 yrs)
Male 9 (56) 31 (48)
Diarrhea Pediatric (age Female 7 (44) 33 (52)
† <15 yrs) Smoking
WBC <5,000 cells/cu.mm.
Yes 2 (13) 9 (14)
Platelet <100,000 cells/cu.mm.
No 14 (87) 55 (86)
0 10 20 30 40 50 60 70 80 90 100 Chronic medical condition
Percentage Yes 2 (13) 10 (16)
No 14 (87) 54 (84)
* Body temperature of >100.4°F (>38°C).
† White blood cell.
in 1997 or mutation of the virus might explain this differ-
ence. During December 16, 2004–December 8, 2005, global
acute respiratory distress syndrome (ARDS) in all persons who mortality among persons with human cases was 29% (5).
died (Table 1). In the matched case-control study, case- Basic laboratory profiles in children indicated leucopenia
patients and controls were similar with respect to the majority and thrombocytopenia. These laboratory results in children
of variables (Table 2). led attending physicians initially to suspect dengue fever in
From the univariate matched analyses, direct touching of the differential diagnosis. However, certain differences helped
unexpectedly dead poultry yielded the highest odds ratio clinicians differentiate between H5N1 infection and dengue
(OR = 29.0; 95% CI = 2.7–308.2). Having poultry die in or fever. In H5N1 cases, respiratory symptoms typically predomi-
near the house or performing activities related to poultry was nate, and case-patients did not exhibit hemoconcentration (7).
significantly associated with H5N1 disease (Table 3). No sta- The rapid progressive pneumonia and ARDS required physi-
tistically significant difference was identified among patients cians in the epizootic areas to be aware of the possibility of
reporting contact with persons with a suspected case of H5N1 infection before patients’ illness progressed to severe
human avian influenza (Table 3). pneumonia.
In the matched analyses, H5N1 disease was associated with
recent exposure to sick or unexpectedly dead poultry, particu-
Discussion larly direct touching of unexpectedly dead poultry. Only one
The H5N1 outbreak in humans in Thailand during patient was reported to have been exposed in a live poultry
January–March and August–October 2004 affected prima- market. In contrast, the 1997 Hong Kong outbreak was
rily children. Overall mortality was high (75%). During the related to exposure to live poultry market (2,9). A family cluster
same period, mortality among persons with human cases in was recognized in Thailand in September 2004, probably as a
Vietnam was also high (80%) (8). These data suggest that the result of person-to-person transmission during unprotected
disease has become more severe than that reported in Hong exposure to a critically ill index patient (10). Vietnam also
Kong in 1997, in which mortality was only 33% (2). How- reported probable human-to-human transmission in Febru-
ever, availability of early aggressive treatment in Hong Kong ary 2004 (11). Moreover, in Hong Kong in 1997, a study
demonstrated H5N1 infection among health-care workers
TABLE 1. Clinical progressions of influenza A (H5N1) patients,
exposed to a patient with H5N1 infection (12). The risk for
by symptoms and days after exposure to birds — Thailand, 2004 person-to-person transmission of H5N1 virus might increase,
No. of days either through viral mutation or reassortment. This empha-
Symptoms Median Range sizes the need for strong surveillance, early detection, and
Onset of illness with fever,* cough, intensive measures to protect unexpected person-to-person
or rhinorrhea (n = 16) 4 2–8 transmission.
Pneumonia (n = 13) 9 6–12
ARDS† (n = 12) 10 8–18
The findings in this report are subject to at least five limita-
* Defined as body temperature >100.4°F (>38°C). tions. First, because a substantial number of patients had died,
† Acute respiratory distress syndrome.
case interviews were conducted more frequently by proxy than
6 MMWR April 28, 2006

TABLE 3. Number and percentage of reported exposures associated with influenza A (H5N1) infection, by type of exposure —
Thailand, 2004
No. (%) of exposures
Case* Control†
Exposure (n = 16) (n = 64) OR§ (95% CI¶)
Direct touching of unexpectedly dead** poultry 10 (63) 12 (19) 29.0 (2.7–308.2)
Dressing poultry 5 (31) 4 (6) 17.0 (1.6–177.0)
Having unexpectedly dead poultry around the house 8 (50) 9 (14) 5.6 (1.5–20.7)
Plucking poultry 4 (25) 3 (5) 14.0 (1.3–152.5)
Being <1 m away from dead poultry 10 (63) 16 (25) 13.0 (1.8–96.3)
Storing products of sick or dead poultry in house 7 (44) 3 (5) 9.3 (2.1–41.3)
Direct touching of sick poultry 8 (50) (14) 5.6 (1.5–20.7)
Being <1 m away from sick poultry 9 (56) 14 (22) 3.8 (1.2–11.7)
Having contact with person with suspected or confirmed H5N1 illness 3 (19) 13 (20) 0.9 (0.2–4.4)
Visiting live poultry market 1 (6) 0
* H5N1 illness occurring in a person who received a diagnosis of pneumonia or influenza-like illness and who had either a positive viral culture for H5N1 virus
or confirmation of H5 strain by real-time reverse transcription-polymerase chain reaction (RT-PCR).
† Selected by matching village and age ±1 year to a person with a case and then randomly selecting four controls for each case-patient.
§ Odds ratio.
¶ Confidence interval.
** Death of >10% of all poultry in a farm or house within 1 day or of >40% of poultry within 3 days.

were control interviews. Second, recall bias might have and Timothy M. Uyeki, MD, National Center for Infectious
occurred as a result of the public’s high level of alarm. Diseases, CDC.
Third, exposure to poultry was part of the surveillance References
requirement, and results might have been biased toward 1. Buxton Bridges C, Lim W, Hu-Primmer J, et al. Risk of influenza A
these exposures. Fourth, because dengue serology was not (H5N1) infection among poultry workers, Hong Kong, 1997–1998.
documented for patients, combined dengue and H5N1 in- J Infect Dis 2002;185:1005–10.
fections cannot be ruled out. Finally, matching one case 2. Mounts AW, Kwong H, Izurieta HS, et al. Case-control study of risk
factors for avian influenza A (H5N1) disease, Hong Kong, 1997.
patient with health-care workers rather than community con-
J Infect Dis 1999;180:505–8.
trols might have introduced bias. 3. Katz JM, Lim W, Bridges CB, et al. Antibody response in individuals
During and after performance of this study, multiple rec- infected with avian influenza A (H5N1) viruses and detection of anti-
ommendations were provided to health-care workers. In areas H5 antibody among household and social contacts. J Infect Dis 1999;
that have unexpectedly dead poultry, clinicians should have a 180:1763–70.
high index of suspicion for avian influenza in patients with acute 4. Li KS, Guan Y, Wang J, et al. Genesis of a highly pathogenic and
potentially pandemic H5N1 influenza virus in eastern Asia. Nature
respiratory illness and quickly distinguish between H5N1 and 2004;430:209–13.
other viral infections (e.g., dengue fever) that have similar labo- 5. World Health Organization. Cumulative number of confirmed
ratory profiles in children. Persons (especially children) should human cases of avian influenza A (H5N1) reported to WHO. Avail-
avoid direct contact with unexpectedly sick or dead poultry able at http://www.who.int/csr/disease/avian_influenza/country/cases_
and not store products from unexpectedly dead poultry in the table_2005_12_07/en/index.html.
6. Apisarnthanarak A, Kijphati R, Thongphubeth K, et al. Atypical avian
home. When contact is unavoidable, persons should use per-
influenza. Emerg Infect Dis 2004;10:1321–4.
sonal protective equipment and wash their hands frequently. 7. Chotpitayasunondh T, Ungchusak K, Hanshaoworakul W, et al.
Long-term public health surveillance and control measures are Human disease from influenza A (H5N1), Thailand, 2004. Emerg
needed to monitor for person-to-person transmission and the Infect Dis 2005;11:201–9.
emergence of a potentially pandemic H5N1 influenza virus. 8. Tran TH, Nguyen TL, Nguyen TD, et al. Avian influenza A (H5N1)
in 10 patients in Vietnam. N Engl J Med 2004;350:1179–88.
9. Chan PK. Outbreak of avian influenza A (H5N1) virus infection in
Acknowledgments
Hong Kong in 1997. Clin Infect Dis 2002;34:S58–64.
Assistance in field investigations, advice, and technical support 10. Ungchusak K, Auewarakul P, Dowell SF, et al. Probable person-to-
was provided by the Bureau of Epidemiology, Kumnuan Ungchusak, person transmission of avian influenza A (H5N1). N Engl J Med
MD, Director; provincial health offices for Kanchanaburi, 2005;352:333–40.
Kamphaeng Phet, Khon Kaen, Nakhon Ratchasima, Petchabun, 11. Jane P. WHO investigates possible human-to-human transmission of
Lop Buri, Sukhothai, Suphan Buri, Chaiyaphum, Pathum Thani avian flu. Biomedical Journal 2004;328:308.
and Phrachin Buri provinces; staff members and fellows of the Thai 12. Buxton Bridges C, Katz JM, Seto WH, et al. Risk of influenza A
Field Epidemiology Training Program; the Thai National Institute (H5N1) infection among health care workers exposed to patients with
of Health; the Thai Ministry of Public Health–CDC Collaboration; influenza A (H5N1), Hong Kong. J Infect Dis 2000;181:344–8.
Vol. 55 / Supplement MMWR 7

Occupational Injuries Among Workers in the Cleansing Section

of the City Council’s Health Services Department —
Bulawayo, Zimbabwe, 2001–2002
Elizabeth Gonese,1 R. Matchaba-Hove,1 G. Chirimumba,1 Z. Hwalima,2 J. Chirenda,3 M. Tshimanga3
1Department of Community Medicine, University of Zimbabwe, Avondale, Zimbabwe
2Health Services Department, Bulawayo City Council, Bulawayo, Zimbabwe
3Health Studies Office, University of Zimbabwe, Harare, Zimbabwe

Corresponding author: Elizabeth Gonese, Department of Community Medicine, University of Zimbabwe, Box A 178, Avondale, Zimbabwe. Telephone: 263-91-285965;
Fax: 263-4-726803; E-mail: egonese@yahoo.co.uk.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: During 2001–2002, a total of 97 occupational injuries occurred among workers in the cleansing section of the
Bulawayo, Zimbabwe, City Council’s Health Services Department. This report describes a study that was conducted to
describe the nature of these injuries and determine the associated risk factors.
Methods: A retrospective, descriptive cross-sectional survey was conducted concerning occupational injuries incurred by workers in
the cleansing section during 2001–2002. A total of 153 workers who had been in the section as of January 1, 2001, and 23
senior managers and section supervisors were interviewed, the occupational injury register was reviewed, and a walk-through
survey was conducted to estimate risk factors.
Results: The occupational injury register indicated that during the study period, 62 workers sustained 67 injuries, including
one that was fatal. Of these 67 injuries, 27 (40%) involved workers who sustained cuts inside a box-type refuse removal truck,
and 11 (16%) involved workers who had sprained ankles and wrists as a result of improper lifting. Workers aged 18–25 years
were more likely to incur an injury than workers aged >25 years. Working as a bin loader and not having received preemployment
training were associated with injuries. None of the bin loaders had received preemployment training. Hazards identified
during the walk-through survey included use of small jacks in workshops, contact with biologic and chemical materials on
trucks and landfill sites, and poor use of protective clothing. Supervisors cited worker negligence as the main cause of injury,
whereas 72 (84%) workers cited lack of adequate protective clothing as a source of injury, and eleven (7%) workers cited use
of inappropriate equipment.
Conclusion: On the basis of the modifiable risk factors for injury identified in this study, the Bulawayo City Council drafted
a new health and safety training manual. New recruits now receive training before starting work on refuse collection trucks.

Introduction The Bulawayo City Council Health Services Department

(HSD) monitors the city’s public health standards, including
Occupational injuries continue to be a serious but poorly
refuse disposal services provided by the cleansing section.
documented problem in both developed and developing coun-
During 2001–2002, a total of 97 injuries occurred among
tries. The International Labour Organisation estimates that
HSD workers, one of which was fatal (4). Of these 97 inju-
worldwide, workers sustain 250 million unintentional occu-
ries, 67 (73%) occurred in the cleansing section, including
pational injuries and an additional 330,000 fatalities annu-
the one fatality. After the City Council and employees
ally (1). Reducing the number of injuries is one of Zimbabwe’s
expressed concern regarding the high incidence of occupa-
top 10 health priorities (2). The highest numbers of occupa-
tional injuries in the section, a study was conducted to
tional injuries in Zimbabwe occur in the construction, min-
describe the nature of injuries and estimate the associated risk
ing, and manufacturing industries. Service industries,
factors.
including local government authorities, accounted for 1,707
(12%) of the 13,854 occupational injuries reported during
2000 (3).
8 MMWR April 28, 2006

Methods Of the 51 injuries for which interviews could be conducted,

13 (25%) injuries were classified as major because they
Records of injuries that occurred during January 2001–
resulted in permanent scars and in loss of >3 working days, and
December 2002 were reviewed, and a cross-sectional survey
38 (74%) were minor. Only three of the 13 workers with major
was conducted in the HSD cleansing section during June 2003
injuries received any compensation, and the amount they
to determine demographic and personal risk factors associ-
received was negligible (average: <$1 U.S. per month). Data
ated with injuries. Of 188 eligible general workers, 153 (84%)
regarding whether the family of the deceased worker was
were interviewed along with 23 senior managers and section
receiving compensation were not available. HSD management
supervisors. Of 62 workers who had an injury during the study
is responsible for the initial assessment and forwarding of nec-
period, 51 (82%) were interviewed. All participants were
essary documentation and referral of workers to NSSA for pay-
informed that the purpose of the study was to find risk factors
ment of compensation. Neither workers nor management
and make recommendations to improve the work environ-
appeared to understand the compensation system.
ment. Persons interviewed gave informed written consent. A
walk-through survey of public restrooms, mechanical work-
shops, a landfill site, and the suburban refuse collection pro- Personal Risk Factors
cess also was conducted. A checklist adapted from CDC’s The section workforce included 147 (96%) men and six
National Institute of Occupational Health (5) and a form (4%) women. Risk factors associated with suffering an injury
(Occupational Health and Safety Form OHS IAF 2003) cre- were age 25 years (odds ratio [OR] = 3.2; CI = 1.6–9.2), work-
ated by the National Social Security Authority (NSSA) of Zim- ing as a bin loader (OR = 3.6; CI = 1.1–4.8), not having
babwe were used to identify environmental risk factors. Data received preemployment training (OR = 3.1; CI = 1.3–7.5),
were analyzed using Epi Info 2002 (CDC, Atlanta, Georgia). and subsequently rating the job as difficult to perform (POR =
Univariate analysis, frequencies, prevalence odds ratios, and 2.9; CI = 1.1–7.7) (Table 2). Potential environmental hazards
p values at 95% confidence intervals (CIs) were reported. A cited by workers included inappropriate equipment (cited by
result with a value of p<0.05 was considered statistically sig- 30 [22%] workers), dust (22 [14%]), biologic material (20
nificant. [13%]), sharp objects (17 [11%]), and chemicals (15 [10%]).
A total of 72 (84%) workers said provision of adequate per-
sonal protective equipment (PPE) would help in the preven-
Results tion of injury. Another 11 (7%) workers said that provision of
appropriate equipment could help in injury prevention. A total
Results from Register of Occupational of 29 (20%) workers could not name any potential hazards.
Injuries
The register of occupational injuries indicated that 62 work- Environmental Hazards
ers were injured at least once during the study period; of 67
A walk-through inspection revealed human waste on
injuries that occurred, one was fatal (Table 1). The death
restroom floors. Workers were at risk for splashes into exposed
occurred when a bin loader slipped and was run over by the
eye and mouthparts because they did not wear goggles or face
back wheels of a refuse truck. Injuries occurred throughout
masks during cleaning of restrooms. Workshops were over-
the week with no association to day of week or hour of day.
crowded, with dirty vehicles under repair. All workers wore

TABLE 1. Number and percentage of occupational injuries among workers in the cleansing section of the Bulawayo City Council
Health Services Department, by injury type and cause — Bulawayo, Zimbabwe, 2001–2002
Type Risk factor No. (%*)
Cut on hand, finger, thumb, or foot Broken glass or sharp objects 27 (40)
Sprained ankle or wrist Improper lifting or throwing technique, or running and disembarking from vehicle 11 (16)
Eye injury Dust, liquid, chemicals, or smoke 8 (12)
Shoulder injury Contact collision 8 (12)
Knee injury Contact collision, slip, or fall 5 (8)
Laceration of leg or finger Dog or scorpion bite 4 (6)
Sharp back pain Excessive effort in lifting 3 (5)
Trunk injury Run over by truck 1 (2)
Total 67
* Total might exceed 100% because of rounding.
Vol. 55 / Supplement MMWR 9

TABLE 2. Number and percentage of occupational injuries among bin loaders observed during the bin-collection pro-
workers in the cleansing section of the Bulawayo City Council Health
Services Department, by selected characteristics — Bulawayo,
cess, 27 (44%) had worn-out footwear, and 22 (32%)
Zimbabwe, 2001–2002 had overalls in bad condition. Only 17 (28%) wore
Injured Not injured respirator masks, and none wore head protection,
Characteristic No. (%) No. (%) OR* (95% CI†) although both are necessary to prevent injury. Work-
Age group (yrs)§ ers are issued two pairs of overalls and a pair of boots
twice annually. Heavy-duty gloves are issued every 2
18–25 13 (25) 9 (9) 3.2 (1.2–9.2)
26–64 38 (75) 93 (91) weeks, and face masks are issued on demand.
Marital status Poor disposal habits of medical practitioners
Married 41 (80) 85 (83) 0.8 (0.3–1.5) resulted in needles and bandages being present in
Not married 10 (20) 16 (17) refuse collection trucks, and uncovered biologic and
Medical condition chemical materials were observed at the landfill. Scav-
Yes 6 (12) 8 (8) 1.8 (0.6–5.7)
No 45 (88) 94 (92)
engers were observed at the dump site, which could
Education level
cause unintentional injuries. Another hazard was that
Secondary or above 33 (65) 46 (45) 2.2 (1.1–4.8) dust and noise were generated during the compact-
Primary or none 18 (35) 56 (55) ing process.
Work position The presence of the hazards identified in the walk-
Bin loader 43 (84) 54 (53) 3.6 (1.6–8.1)
Other 8 (16) 48 (47)
through survey was confirmed by 23 senior managers
Length in service (yrs)
and supervisors, who attributed injuries to the need
<3 14 (27) 20 (20) 1.6 (0.7–3.4) for manual labor and to workers being careless and
4–35 37 (73) 82 (80) failing to follow the departmental safety protocol
Preemployment training However, no written documented safety protocol was
No 41 (80) 58 (57) 3.1 (1.3–7.5) identified. Supervisors also said that workers sold
Yes 10 (20) 44 (43)
their allocation of work clothes.
Health and safety training
No 46 (90) 91 (91) 1.1 (0.3–3.9) Section supervisors who were directly
Yes 5 (10) 11 (11) responsible for the workers said that not much could
Job rating¶ be done in the section because of poor communica-
Difficult (3–5) 45 (88) 73 (72) 2.9 (1.1–7.7) tion in the department and that directors did not
Easy (1–2) 6 (12) 29 (28)
welcome requests, especially those involving capital
Knowledge of hazards
No 2 (4) 27 (26) 0.1 (0.02–0.5)
expenditure. Supervisors said risk could be reduced
Yes 49 (96) 75 (74) by providing modern equipment and adequate PPE
Total 51 102 and by reducing the workload.
* Odds ratio.
† Confidence interval.
§ Age groups were classified on the basis of surveys, which indicated statistically
significant patterns in the occurrence of injuries in three age groups (18–25, 26–55,
Discussion
Workers aged 18–25 years were more likely to suf-
fer an injury than workers aged >25 years. Among
overalls, but none wore gloves to avoid contamination from
workers in all industries, frequency of injuries has been dem-
materials left in refuse trucks. Mechanics used small jacks to
onstrated to decline with age (6). Younger workers might be
lift trucks and were at risk for injury as a result of heavy equip-
more prone to injury because of their lack of experience and
ment falling.
an inclination to take unnecessary risk.
Two types of trucks were used for refuse collection: an old
Bin loaders were more likely to be injured than toilet or
box-type truck and a modern hydraulic-compactor refuse-
street cleaners. The absence of training at the beginning of
collection truck (Figure). The compactor has side steps on
employment was also identified as a risk factor. Bin loading
which bin loaders stand, and workers do not come into con-
involves handling of heavy loads and working with moving
tact with rubbish. However, workers loading the box-type truck
trucks. Workers should be trained in good lifting techniques
must lift bins above their shoulder level, and workers inside
and how to avoid vehicular accidents. Incidence of back
the truck come into direct contact with hazardous materials
injuries has been reduced substantially after training in proper
and are at risk for sustaining cuts to their legs, feet, or fingers
lifting techniques (7).
while spreading out rubbish. Use of PPE was limited. Of 62
10 MMWR April 28, 2006

FIGURE. Box-type (top) and hydraulic compactor (bottom) can be made to prevent future injuries. Injuries that occurred
refuse-collection trucks — Bulawayo, Zimbabwe, 2003
in the cleansing section during 2001–2002 could have been
reduced or, in some instances, avoided by implementing ap-
propriate interventions. As a result of these findings, the
Bulawayo City Council has compiled a health and safety
manual for use in the department. New recruits in the bin
loading section are now trained before they start work. HSD
was advised to provide proper jacks in its mechanical work-
shops and to devise a plan to gradually phase out the old refuse
removal trucks. To reduce the risk for injury, small jacks and
old box trucks should be replaced with modern
hydraulic trucks, and proper use should be made of PPE.
References
1. Anan K. Editorial. African Newsletter on Occupational Health and
Safety. 1997;7(3):3.
2. Ministry of Health and Child Welfare. National strategic framework
for 1997 to 2007. Harare, Zimbabwe: Ministry of Health and Child
Welfare; 1999.
3. National Social Security Authority on Occupational Health and Safety.
Occupational injuries in sectors of Zimbabwe industries, 2000. On
SOURCE: Bulawayo City Council, Bulawayo, Zimbabwe, 2003 Guard 2000;7:23.
4. Bulawayo Health Services Department. Annual report 2003. Bulawayo,
Zimbabwe: Bulawayo Health Services Department; 2003.
The findings in this report are subject to at least two limita-
5. Bernard B. Musculoskeletal disorders (MSDs) and workplace factors, a
tions. First, a walk-through inspection might not have identi- critical review of epidemiologic evidence for work-related musculoskel-
fied all hazards. Second, interview data were missing for 18% etal disorders of the neck, upper extremity and low back. Cincinnati,
of persons with injuries and for 19% of persons without inju- OH: US Department of Health and Human Services, CDC, National
ries. If these persons represent different distribution of risk Institute for Occupational Safety and Health; 1997. Available at
factors than persons included in the study, ORs might be http://www.cdc.gov/niosh/ergoscil.html.
6. Alberta Association of Optometrists. Occupational injuries. Work Sight
underestimated or overestimated. 2000;9(3):1–5.
7. Kazutaka K, Wai-On P, Thurman JE. Low cost ways of improving work-
ing conditions: 100 examples from Asia. Geneva, Switzerland: Interna-
Conclusion tional Labour Organisation; 1989:1–5, 45–7, 156–7.
This report indicates the importance of keeping accurate
records of the causes of injury so appropriate interventions
Vol. 55 / Supplement MMWR 11

Progress Toward Tuberculosis Control and Determinants

of Treatment Outcomes — Kazakhstan, 2000–2002
Ekaterina Bumburidi, S. Ajeilat, A. Dadu, I. Aitmagambetova, J. Ershova, R. Fagan, M.O. Favorov
Applied Epidemiology Training Program, CDC/Central Asia Office, Division of International Health, Office of Global Health, CDC
Corresponding author: Michael O. Favorov, CDC, 1600 Clifton Rd, MS E-93, Atlanta, GA 30333. Telephone: 404-498-6070; Fax: 404-498-6065;
E-mail: MOF0@cdc.gov.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: In Kazakhstan, during 1995–2002, the annual notification rate per 100,000 population for new cases of
tuberculosis (TB) increased from 67.1 to 165.1. Beginning in 1998, public health authorities have used the national case
management strategy (DOTS) promulgated by the World Health Organization (WHO) to control TB. Intended goals of
DOTS include achieving a cure rate of >85% for persons with newly detected pulmonary TB sputum-smear–positive (PTB+)
cases and having PTB+ represent >65% of all PTB cases among adults. Surveillance data collected during 2000–2002 were
analyzed to evaluate progress toward achieving these goals and identify factors associated with specific treatment outcomes.
Methods: Surveillance data included the following nonidentifiable information on persons with newly reported cases of PTB:
dates of disease onset and treatment initiation; methods of diagnosis; treatment outcomes; HIV status; and selected demo-
graphic, socioeconomic, and behavioral characteristics. Cure rates and proportions of PTB+ cases were calculated on the basis
of the TB case definition and treatment outcome classification format outlined in DOTS guidelines issued by WHO.
Denominator data to calculate rates were obtained from the National Census Office of Kazakhstan. Logistic regression was
employed to investigate factors associated with treatment outcomes using Epi Info version 3.2.
Results: During 2000–2002, a total of 65,011 new cases of PTB were detected in Kazakhstan. The average annual country-
wide notification rate per 100,000 population was 146.0; provincial notification rates varied (range: 65.1–274.0). The
countrywide cure rate for newly detected PTB+ was 72.2%; provincial rates varied (range: 65%–81%). Of 59,905 cases of
PTB among adults during 2000–2002, a total of 26,804 (44.7%) were PTB+. Unfavorable treatment outcome of new
PTB+ cases was associated with alcohol abuse, homelessness, and previous incarceration.
Conclusion: The cure rate for new PTB+ cases and the proportion of cases of PTB+ among all adults with PTB were below
targeted goals. This might, in part, be explained by the 1998 adoption of DOTS. Improving program indicators requires
evaluation of detection efforts, laboratory diagnostic capabilities, and adherence to treatment regimens, especially in provinces
in which rates are high and among persons at high risk for unfavorable treatment outcomes.

Introduction mated incidence of sputum-smear–positive pulmonary TB

(PTB+) and >85% cure rate for newly detected PTB+ cases.
In 1993, the World Health Organization (WHO) adopted
The >85% cure rate was adopted on the basis of accumulated
a national case management strategy (DOTS) to reduce the
experience in Africa and certain districts of China (5,6).
increasing global burden of tuberculosis (TB), especially in
Performance indicators in the DOTS program use the pro-
developing countries (1). The five elements of the DOTS strat-
portion detected of PTB+, which is the most infectious form
egy are sustainable government commitment, quality assur-
of TB. PTB+ is associated with high mortality (7) and is the
ance of sputum microscopy, standardized short-course
most effective form of TB to use for bacteriologic monitoring
treatment (including direct observation of therapy), regular
of treatment progress (8). The proportion of newly detected
supply of drugs, and establishment of reporting and
PTB+ cases among the total number of adults with PTB
recording systems (2,3). The goal of DOTS is to reduce TB
reflects the proper application of diagnostic criteria. In coun-
morbidity and mortality and the chance of Mycobacterium
tries with a medium or high TB burden, when necessary labo-
tuberculosis developing resistance to primary treatment drugs
ratory resources are available and sputum smears for
(4). Target goals of TB control adopted in 1991 by the World
microscopy are taken from TB patients, PTB+ accounts for
Health Assembly (3) include >70% detection rate of the esti-
12 MMWR April 28, 2006

>50% of all TB cases and >65% of new PTB cases in adults new PTB+ cases for which records for DOTS treatment regi-
(3,9). Achieving a high (i.e., >85%) cure rate for PTB+ is a men were completed by the time of the study were treated as
critical priority for TB-control programs (3,9). Failure to members of a cohort that was used to study associations
achieve this rate results in continued infectiousness and pos- between treatment outcome and risk factors. The study was
sible development of multidrug-resistant (MDR) TB, which given institutional review board approval by the Kazakh
is resistant to at least isoniazid and rifampicin (10,11). State Medial University ethics committee.
During 1980–1994, the countrywide TB notification rate
per 100,000 population decreased from 88.2 to 59.7. How- Study Definitions
ever, since 1995, notification rates have increased constantly,
Diagnostic categories for newly reported cases of PTB (both
reaching 165.1 in 2002. TB mortality rates per 100,000 popu-
PTB+ and sputum-smear–negative PTB) and treatment out-
lation decreased from 17.2 in 1980 to 10.6 in 1991. During
comes* for newly diagnosed PTB+ were defined in accordance
1992–1998, mortality rates increased, reaching 38.4 in 1998;
with WHO guidelines (3). To study the associations among
however, rates decreased to 24.2 in 2002 (Kazakhstan Minis-
treatment outcome and risk factors, a new bi-level treatment
try of Health [MoH], unpublished data, 1980–2002).
outcome variable was defined on the basis of the DOTS out-
Increased TB mortality and morbidity have been attributed
come definitions: favorable (i.e., cured) and unfavorable (i.e.,
to socioeconomic changes and the deterioration of the health-
failed, died, or defaulted). Patients classified as having com-
care system in Kazakhstan after the country gained its inde-
pleted treatment or transferred out were excluded from the
pendence from the former Soviet Union in 1991 (12,13).
risk factor analysis because either information was lacking
To manage the increasing burden of TB in the country, in
regarding alteration of sputum-smear status from positive to
1998, the Kazakhstan MoH adopted and implemented a new
negative or they had been classified according to a new diag-
National Tuberculosis Program (NTP), whose objectives and
nostic category. Sociodemographic and behavioral variables
target goals are in accord with the DOTS strategy (14). To
in the surveillance database that were available for risk factor
implement the DOTS strategy in Kazakhstan, primary health-
analysis included age, sex, place of residence, homelessness,
care physicians and TB specialists received training in case-
employment status, employment in the medical profession,
detection policy, and laboratories were equipped with binocular
previous incarceration, and alcohol abuse. These variables were
microscopes. During 1997–2000, the number of sputum-
complete for >95% of records. Data were recorded by treat-
smear examinations increased from 661,000 to 1,170,000
ing physicians at the time of TB diagnosis (5–10 days after
(Kazakhstan MoH, unpublished data, 2001).
hospital admission) on the basis of information provided by
Since 1998, a uniform TB surveillance system has operated
patients.
in Kazakhstan. Beginning in 2000, all administrative territo-
ries collected surveillance data. In 2003, CDC conducted a
study to evaluate NTP performance. Annual data for 2000– Data Sources
2002 were analyzed to describe patterns of PTB notification By law, TB is a reportable disease in Kazakhstan. A stan-
by person and place, assess progress toward achieving NTP dard notification form is completed by the treating physician
target goals, and identify factors associated with unfavorable and then entered into an electronic database when a patient
treatment outcomes. receives a TB diagnosis in any TB treatment facility. The
national TB surveillance database contains nonidentified
information on cases among civilians from all 16 administra-
Methods tive regions (14 provinces and two large urban areas [Astana
and Almaty]); data on prisoners and military personnel are
Study Design
not included in this database. During 2000–2002, a total of
The MoH surveillance database in Kazakhstan was analyzed 99,111 extrapulmonary and pulmonary cases were registered
to estimate the PTB notification rate and describe NTP per-
formance indicators that could be used to measure progress. * Cure: person who became sputum-smear–negative in the last month of
Selection of appropriate progress indicators was guided by treatment and on at least one previous occasion; treatment failure: person
who continued to have sputum-smear–positive status at >5 months during
WHO recommendations (15). Two performance indicators treatment; treatment completed: person with PTB+ who completed treatment
were selected: the proportion of new PTB+ cases among but whose condition was not consistent with the criteria for either cure or
reported cases of PTB among adults and the proportion of failure; died: person who died of any cause during treatment; defaulter: person
who interrupted the treatment regimen for >2 consecutive months; and transfer
persons with newly detected PTB+ cases who were cured (3). out: person who moved to another health-care facility and was entered in a
Persons in the database, registered during 2000–2002, with new diagnostic category (i.e., transfer in).
Vol. 55 / Supplement MMWR 13

in the database; 65,011 (65.6%) were newly diagnosed PTB FIGURE. Average annual notification rate* of new cases of
pulmonary tuberculosis (N = 65,011), by age group and sex —
cases that were used for the descriptive analysis. Of these, 803 Kazakhstan, 2000–2002
(1.2%) persons were excluded from the analysis because of
300
lack of data on bacterioscopic investigation. Of 27,171 per- Male
sons in the database with newly diagnosed PTB+, complete Female
outcome data were available for 20,461 (75.3%) persons and
were used for the risk factor analysis. Of these, 285 (1.4%) 200
were children aged <15 years; because of these limited num-

Rate
bers, children were not included in the risk factor analysis.
Population denominator data were obtained from the National
Census Office of Kazakhstan. 100

Statistical Analyses
Data analyses were performed using Epi Info, version 3.2 0
(CDC, Atlanta, Georgia). Population estimates at midpoint <1 1–14 15–24 25–34 35–44 45–54 55–64 >65

of the 2000–2002 study period were used to calculate the Age group (yrs)
average annual notification rates for new PTB by sex, age, * Per 100,000 population.
and region. Chi-square tests were used to evaluate differences
in notification rates and estimated target indicators by region.
Logistic regression analysis was used to study associations TABLE 1. Number of reported new pulmonary tuberculosis
(PTB) and PTB smear-positive (PTB+) cases among adults,
between the study-defined treatment outcome variable and by region — Kazakhstan, 2000–2002
sociodemographic and behavioral risk factors. All variables that No. of new PTB+ cases
were significant at the two-sided α = 0.05 level in univariate analy- Region PTB cases No. (%)
ses were included in the final multivariate logistic model; North Kazakhstan 2,582 1,538 (59.6)
adjusted odds ratios (AORs) were used as measures of association. Kostanay 4,061 2,309 (56.9)
Almaty 4,337 2,147 (49.5)
Acmolensk 3,962 1,931 (48.7)
Almaty City 2,154 1,038 (48.2)
Results Jambilsk 3,716 1,715 (46.2)
Karaganda 5,494 2,455 (44.7)
Descriptive Statistics Kzyl-Orda 4,525 1,979 (43.7)
Atirau 2,747 1,199 (43.6)
During 2000–2002, the countrywide average annual PTB Mangistau 1,988 855 (43.0)
notification rate was 146 per 100,000 population. Rates in West Kazakhstan 3,247 1,361 (41.9)
Actobe 4,071 1,694 (41.6)
urban and rural areas were similar (146.1 and 144.9, respec- East Kazakhstan 5,398 2,196 (40.7)
tively). However, rates differed significantly (p<0.001) by Pavlodar 3,729 1,498 (40.2)
region (range: 65.0 [Almaty City]–274.0 [Kzyl-Orda]). In gen- South Kazakhstan 6,122 2,277 (37.2)
Astana City 1,772 612 (34.5)
eral, the highest rates were observed in the country’s western
Total 59,905 26,804 (44.7)
regions (Actobe, Atirau, Mangistau, and West Kazakhstan).
Age-specific rate patterns were similar for men and women.
The highest rates occurred among persons aged 25–34 years human immunodeficiency virus (HIV)–infected patients were
(men: 277; women: 241). Among persons aged >25 years, identified during the study period.
rates were consistently higher for men than for women (Figure).
By region, significant differences were reported in the pro- Risk Factor Analysis
portion of sputum-smear–positive cases among adults with Among 20,176 persons aged >15 years, 547 (2.7%) patients
newly reported PTB cases (range: 34.5% [Astana]–59.6% had finished the treatment course but their sputum-smear
[Northern Kazakhstan]) (p<0.001) (Table 1); no region status was unknown, and 650 (3.2%) patients were transferred
attained the goal of >65%. The cure rate for persons with new out; both groups were excluded from the risk factor analysis.
PTB+ cases varied (range: 65% [Eastern Kazakhstan]–81% Of 18,979 patients included in the analysis, 4,422 (23.3%)
[Almaty City]) (p<0.001) (Table 2). No region attained the had unfavorable treatment outcomes (i.e., died, failed, or
goal of >85%. Among 99,111 persons with TB, 100 (0.1%) defaulted). The proportion recorded as having an unfavor-
14 MMWR April 28, 2006

TABLE 2. Number of new pulmonary tuberculosis sputum- notification rates were higher among males. Similar results
smear–positive (PTB+) cases* and cure rates, by region —
Kazakhstan, 2000–2002
have been reported in other countries (18–20). The reasons
Cured for this difference in TB notification rates are not known. No
No. of new
Region PTB+ cases No. (%) evidence exists that health-seeking behaviors are substantially
Almaty City 834 676 (80.8) different in countries of the former Soviet Union and could
South Kazakhstan 1,649 1,321 (80.1) not have accounted for male-female differences. Genetic fac-
Kzyl-Orda 1,540 1,201 (78.0)
Astana City 359 272 (75.8)
tors or transmission dynamics might explain this difference.
Atirau 895 669 (74.9) During 2000–2002, the proportion of PTB+ cases among
North Kazakhstan 1,288 962 (74.7) the total number of new PTB cases among adults in Kazakhstan
West Kazakhstan 1,036 769 (74.2)
was 44.7% (program target: >65%). In certain areas, the pro-
Jambilsk 1,285 937 (73.0)
Kostanay 1,867 1,353 (72.5) portion was lower (e.g., Astana-City: 34.5%; South
Almaty 1,742 1,252 (71.9) Kazakhstan: 37.2%). The DOTS strategy was adopted recently
Acmolensk 1,500 1,070 (71.3) (i.e., in 1998), and the quality of laboratory services and the
Actobe 1,088 762 (70.1)
Mangistau 662 453 (68.4) application of the diagnostic criteria might not be up to the
Karaganda 1,755 1,160 (66.1) required standard, especially in certain areas.
Pavlodar 1,236 804 (65.0) Since the DOTS strategy was implemented, the supply of
East Kazakhstan 1,725 1,119 (64.9)
primary TB drugs in TB-treatment facilities has been adequate;
Total 20,461 14,780 (72.2)
* Occurring among persons for whom complete treatment outcome data
consequently, this factor cannot explain the low cure rate for
were available. new PTB+ cases, especially in certain regions (e.g., East

able outcome varied by age group. In univariate TABLE 3. Number and percentage of persons with new sputum-smear–
analyses, all studied variables were statistically sig- positive pulmonary tuberculosis (PTB+) cases, by selected demographic
nificant at the 0.05 level and were included in the characteristics and type of analysis — Kazakhstan, 2000–2002
final multivariate logistic model. Unfavorable treat- New PTB+ cases Univariate Multivariate
(N = 18,979) (crude) analysis analysis
ment outcomes for new PTB+ cases were associ-
Characteristic No. (%) COR* (95% CI†) AOR§ (95% CI)
ated with alcohol abuse, homelessness, previous Residence
incarceration, unemployment, being male, and ur- Urban 10,309 (54.4) 1.5 (1.4–1.6) 1.4 (1.3–1.5)
ban residence. Being medical personnel was pro- Rural¶ 8,649 (45.6) 1.0 1.0
tective. AORs for unfavorable treatment outcomes Sex
Male 11,938 (62.6) 1.5 (1.4–1.6) 1.2 (1.1–1.3)
increased with age (Table 3). Female¶ 7,041 (37.4) 1.0 1.0
Alcohol abuse
Yes 762 (4.0) 3.2 (2.8–3.7) 2.1 (1.8–2.5)
Discussion No¶ 18,203 (96.0) 1.0 1.0
Homeless
The high countrywide notification rate of new Yes 313 (1.7) 3.4 (2.7–4.3) 2.2 (1.7–2.7)
PTB cases (146 per 100,000 population for the No¶ 18,653 (98.3) 1.0 1.0
Previous incarceration
3-year study period) underscores the importance Yes 1,043 (5.5) 1.8 (1.6–2.1) 1.5 (1.3–1.7)
of controlling TB in Kazakhstan. Substantial No¶ 17,923 (94.5) 1.0 1.0
regional differences were recorded; in certain Unemployed
Yes 11,614 (61.2) 1.4 (1.3–1.5) 1.5 (1.4–1.6)
areas (e.g., Kzyl-Orda, 274 per 100,000 popula- No¶ 7,364 (38.8) 1.0 1.0
tion), the notification rate was approximately twice Medical personnel
the national average. As in other countries, regional Yes 229 (1.2) 0.5 (0.3–0.7) 0.6 (0.4–0.9)
No¶ 18,736 (98.8) 1.0 1.0
differences might result from different underlying Age group (yrs)
socioeconomic conditions and the quality of the 15–24¶ 4,340 (23.0) 1.0 1.0
performance of local TB-control programs (16,17). 25–34 5,316 (28.0) 1.4 (1.2–1.5) 1.3 (1.1–1.4)
Further research is needed to identify the reasons 35–44 4,324 (23.0) 1.9 (1.7–2.1) 1.7 (1.5–1.9)
45–54 2,723 (14.0) 2.4 (2.1–2.7) 2.1 (1.9–2.4)
for regional differences, so corrective measures can 55–64 1,376 (7.0) 2.2 (1.9–2.6) 2.3 (2.0–2.7)
be taken as needed. >65 900 (5.0) 2.3 (1.9–2.6) 2.8 (2.3–3.3)
The age-specific notification rate of new PTB * Crude odds ratio.
† Confidence interval.
cases was similar for children aged <15 years for § Adjusted odds ratio.
both sexes. However, for persons aged >15 years, ¶ Referent group.
Vol. 55 / Supplement MMWR 15

Kazakhstan and Pavlodar, both 65%). Data were not available to 4. Pio A, Chaulet P. Tuberculosis handbook. Geneva, Switzerland: World
explore reasons for the low cure rate (e.g., the prevalence of Health Organization; 1998 (WHO/TB/98.253).
5. Murray CJL, Dejonhe E, Chum HJ, Nyangulu DS, Salomao A, Styblo
MDR-TB). The single study available indicated that high
K. Cost effectiveness of chemotherapy for pulmonary tuberculosis in
(14.2%) MDR-TB rates among new TB patients in Kazakhstan three sub-Saharan African countries. Lancet 1991;338:1305–8.
might contribute to a low cure rate in the country (21). 6. China Tuberculosis Control Collaboration. Results of directly observed
Unemployment and urban place of residence were highly short-course chemotherapy in 112,842 Chinese patients with smear-
prevalent (>54%) in the study group and associated with positive tuberculosis. China Tuberculosis Control Collaboration. Lancet
unfavorable treatment outcomes; this could have contributed 1996;347:358–62.
7. Rouillon A, Perdrizet S, Parrot R. Transmission of tubercle bacilli: the
to the low cure rate. Other factors (i.e., homelessness, history effects of chemotherapy. Tubercle 1976;57:275–99.
of incarceration, alcohol abuse, and age >55 years) had more 8. Harries AD, Nyirenda TE, Banerjee A, Boeree MJ. Treatment out-
effect on the unfavorable treatment outcome; however, their come of patients with smear-negative and smear-positive pulmonary
overall effect on the cure rate might have been limited tuberculosis in the National Tuberculosis Control Programme, Malawi.
because of low prevalence. Trans R Soc Trop Med Hyg 1999;93:443–6.
9. World Health Organization. Compendium of indicators for monitor-
Among medical personnel, the protective odds ratio for
ing and evaluating national tuberculosis programs. Geneva, Switzer-
unfavorable treatment outcome (AOR = 0.6) might be land: World Health Organization; 2004 (WHO/HTM/TB/2004.344).
explained by their stricter adherence to treatment regimen. The 10. Kimerling ME, Slavuckij A, Chavers S, et al. The risk of MDR-TB
high percentage of unfavorable treatment outcomes among per- and polyresistant tuberculosis among the civilian population of Tomsk
sons aged >55 years was attributable to deaths from any cause; city, Siberia, 1999. Int J Tuberc Lung Dis 2003;7:866–72.
however, among persons in other age groups, the majority of unfa- 11. Frieden TR, Sterling T, Pablos-Mendez A, Kilburn JO, Cauthen GM,
Dooley SW. The emergence of drug-resistant tuberculosis in New York
vorable treatment outcomes were attributable to treatment failure. City. N Engl J Med 1993;328:521–6. Erratum in: N Engl J Med
Kazakhstan is in the beginning stage of an HIV epidemic. 1993;329:148.
During 2000–2002, a total of 100 (0.1%) cases of HIV were 12. Dzhunusbekov AD, Khazhibaeva ZI, Dametov US. Epidemiologic situ-
identified among 99,111 persons with TB. The contribution of ation of tuberculosis in the Republic of Kazakhstan. Probl Tuberk
HIV infection to the TB burden is believed to be insubstantial. 1997;1:25–7.
13. Raviglione MC, Rieder HL, Styblo K, Khomenko AG, Esteves K, Kochi
The findings in this report are subject to at least one limita-
A. Tuberculosis trends in eastern Europe and the former USSR.
tion. All study factors were measured by the treating physi- Tubercle Lung Dis 1994;75:400–16.
cians on the basis of patient self-reports. Patients might not 14. World Health Organization. Forty-fourth World Health Assembly,
have reported certain behaviors, particularly those regarded resolution and decisions. Geneva, Switzerland: World Health Organi-
as socially unacceptable (e.g., alcohol abuse). As a result, the zation; 1991 (WHA 44/1991/REC/1).
low prevalence of these behaviors among TB patients and their 15. Broekmans JF, Migliori GB, Rieder HL, et al. Standardized tuberculosis
treatment outcome monitoring in Europe. Recommendations of a Work-
consequent low overall impact on the cure rate might reflect ing Group of the World Health Organization (WHO) and the Euro-
inaccurate self-reporting. pean Region of the International Union Against Tuberculosis and Lung
Continued evaluation is needed to improve performance of Disease (IUATLD) for uniform reporting by cohort analysis of treat-
the TB program in Kazakhstan. Improving program indicators ment outcome in tuberculosis patients. Eur Respir J 2002;19:765–75.
requires evaluation of detection efforts, laboratory diagnostic 16. Mangtani P, Jolley DJ, Watson JM, Rodrigues LC. Socioeconomic
deprivation and notification rates for tuberculosis in London during
capabilities, and adherence to treatment regimens, especially in
1982–91. BMJ 1995;310:963–6.
provinces where rates are high and among persons at high risk for 17. Bhatti N, Law MR, Morris JK, Halliday R, Moore-Gillon J. Increas-
unfavorable treatment outcomes, so recommendations for im- ing incidence of tuberculosis in England and Wales: a study of the
provement can be offered as needed. In addition, laboratory quality likely causes. BMJ 1995;310:967–9.
control and quality assurance for TB culture should be further 18. Rose AMC, Watson JM, Graham C, et al. Tuberculosis at the end of
implemented in the country to assist in assessing the impact of the 20th century in England and Wales: results of a national survey in
1998. Thorax 2001;56:173–9.
MDR TB on the cure rate of PTB patients. 19. Martinez AN, Rhee JT, Small PM, Behr MA. Sex differences in the
References epidemiology of tuberculosis in San Francisco. Int J Tuberc Lung Dis
1. World Health Organization. Treatment of tuberculosis: guidelines for 2000;4:26–31.
national programmes. Geneva, Switzerland: World Health Organization; 1997. 20. Borgdorff MW, Nagelkerke NJ, Dye C, Nunn P. Gender and tubercu-
2. Raviglione MC, Snider DE, Kochi A. Global epidemiology of tuber- losis: a comparison of prevalence surveys with notification data to
culosis: morbidity and mortality of a worldwide epidemic. JAMA explore sex differences in case detection. Int J Tuberc Lung Dis
1995;273:220–6. 2000;4:123–32.
3. World Health Organization. Treatment of tuberculosis: guidelines for 21. World Health Organization. Anti-tuberculosis drug resistance in the
national programs. Geneva, Switzerland: World Health Organization; world: third global report. Geneva, Switzerland: World Health Orga-
2003 (WHO/CDS/TB/2003.313). nization; 2004.
16 MMWR April 28, 2006

Estimation of Measles Vaccination Coverage Using

the Lot Quality Assurance Sampling (LQAS) Method —
Tamilnadu, India, 2002–2003
Saravanan Sivasankaran,1,2 P. Manickam,1 R. Ramakrishnan,1 Y. Hutin,1,3 M.D. Gupte1
1National Institute of Epidemiology, Chennai, India
2National Polio Surveillance Project, Pondicherry, India
3World Health Organization, Field Epidemiology Training Programme, Chennai, India

Corresponding author: Saravanan Sivasankaran, Surveillance Medical Officer, National Polio Surveillance Project, 30 Chellan Nagar, Pondicherry, India 605011.
Telephone: 91-413-2211113; Fax: 91-413-2354925; E-mail: drsrvnndph@yahoo.com.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: As part of the global strategic plan to reduce the number of measles deaths in India, the state of Tamilnadu
aims at >95% measles vaccination coverage. A study was conducted to measure overall coverage levels for the Poondi Primary
Health Center (PPHC), a rural health-care facility in Tiruvallur District, and to determine whether any of the PPHC’s six
health subcenters had coverage levels <95%.
Methods: The Lot Quality Assurance Sampling (LQAS) method was used to identify health subcenters in the PPHC area
with measles vaccination coverage levels <95% among children aged 12–23 months. Lemeshow and Taber sampling plans
were used to determine that the measles vaccination status of 73 children aged 12–23 months had to be assessed in each health
subcenter coverage area, with a 5% level of significance and a decision value of two. If more than two children were unvac-
cinated, the null hypothesis (i.e., that coverage in the health subcenter was low [<95%]) was not rejected. If the number of
unvaccinated children was two or fewer, the null hypothesis was rejected, and coverage in the subcenter was considered to be
good (i.e., >95%). All data were pooled in a stratified sample to estimate overall total coverage in the PPHC area.
Results: For two (33.3%) of the six health subcenters, more than two children were unvaccinated (i.e., coverage was <95%).
Combining results from all six health subcenters generated a coverage estimate of 97.7% (95% confidence interval = 95.7–98.8)
on the basis of 428 (97.7%) of 438 children identified as vaccinated.
Conclusion: LQAS techniques proved useful in identifying small health areas with lower vaccination coverage, which helps
to target interventions. Monthly review of vaccination coverage by subcenter and village is recommended to identify pockets of
unvaccinated children and to maintain uniform high coverage in the PPHC area.

Introduction plications. To evaluate progress toward achieving the first

The goal of the World Health Organization’s global measles objective, the Field Epidemiology Training Programme and
strategic plan is to reduce by half the estimated number of local public health officials estimated coverage levels by using
measles deaths by 2005 compared with 1999 estimates (1). a Lot Quality Assurance Sampling (LQAS) survey (3). This
The target goal set by the Indian Ministry of Health and Family evaluation was conducted for fiscal year 2002–2003 (April 1,
Welfare is to reduce measles mortality 66% by 2010 com- 2002–March 31, 2003) to identify health subcenters with
pared with 2000 estimates (2). Increasing, sustaining, and measles vaccination coverage of <95% among children aged
documenting high routine coverage is essential to achieve sus- 12–23 months and to estimate the overall measles vaccina-
tainable reduction of measles mortality. tion coverage in the same age group in the primary health
The southern Indian state of Tamilnadu has set three objec- center area.
tives for its measles control program: 1) measles vaccination LQAS is a stratified random sampling method in which small
coverage of >95% through routine vaccination of children samples randomly selected from each stratum are used to
aged 9–12 months, 2) ring vaccination after occurrence of determine whether coverage in a stratum exceeds a specific
measles cases, and 3) treatment of ill persons to prevent com- threshold. Compared with the commonly used cluster survey
Vol. 55 / Supplement MMWR 17

technique, LQAS can be used to identify areas with low vacci- Only those children who received measles vaccination at age
nation coverage (3). The operational feasibility of LQAS to 9–12 months were considered vaccinated. The measles vacci-
evaluate vaccination coverage among limited population units nation status of the child was obtained by reviewing vaccina-
in India has been demonstrated (4,5). tion cards if available or by interviewing the mother or guardian
when cards were not available. Children were excluded from
the survey if a parent or guardian was not available.
Methods
In April 2003, the LQAS survey was conducted in the Poondi Analysis of the Survey
Primary Health Center (PPHC) area, Tiruvallur District, Two analyses were conducted, an LQAS analysis in each
Tamilnadu, India. This primary health center serves approxi- stratum and a pooled analysis for the total sample. Card
mately 27,000 persons distributed in six health subcenter availability in subcenters varied (range: 7%–37%). The LQAS
areas (approximately 4,500 persons per health subcenter).* analysis in each stratum was based on testing a hypothesis.
The null hypothesis (Ho) was that the coverage in the health
Sampling Methods subcenter was <95% (Ho: p<95%, low performance). The
A simple random sample of the population served by PPHC alternate hypothesis (Ha) was that the coverage in the health
was used, divided into six strata representing the six health subcenter was >95% (Ha: p>95%, high performance). In each
subcenters. District health authorities expected measles cov- stratum of 73 persons, two (2.7%) persons were considered
erage in the PPHC area to be >95%. A decision value (d) of as the decision value (i.e., the threshold for analysis). In prac-
two nonimmunized children and an acceptable alpha value of tice, when more than two of 73 children were identified as
0.05 were set. On the basis of this information, sample size unvaccinated, coverage in the stratum was considered to be
(n) was estimated by using Lemeshow and Taber LQAS tables <95%. If no more than two unvaccinated children were iden-
(6). A plan with n = 73 was used; this single-stage sampling tified, coverage in the stratum was considered to be >95%. All
plan accepted an alpha error of 5%. In each health subcenter, data were pooled to estimate overall total PPHC coverage.
the total number of eligible children (i.e., those aged 12–23 Overall measles vaccination coverage was calculated from the
months) available was assumed to be 100. Within each stra- total number of eligible children vaccinated compared with
tum, 73 children were assessed for measles vaccination status, the total number of eligible children surveyed. Because this
for an overall sample of 438 children in the PPHC area. was a simple random sample, with no design effect, best esti-
Each health subcenter included multiple villages. To deter- mates and 95% confidence intervals (CIs) were calculated using
mine the number of children to be selected in each village in standard methods.
each stratum, a list of villages was constructed for each stra-
tum with the number of households and the cumulative num-
ber of households; 73 random numbers were selected using Results
random number tables. Once the number of children had been LQAS Analysis
determined for each village, researchers randomly selected as
many households as children were needed. For each selected In two health subcenters (Neyveli and Vellathukkottai), more
household, any eligible child was included. When no eligible than two children were unvaccinated, and vaccination coverage
child was identified in the house, the next houses to the right was considered to be <95% (Table). In the other four subcenters
were surveyed until an eligible child was identified. When more (Chitampakkam, Meyyur, Nambakkam, and Poondi), the num-
than one eligible child was present in a house, only the youngest ber of unvaccinated children did not exceed two, and vaccina-
one was included. Children’s ages were estimated using birth tion coverage was considered to be >95% (Table).
certificates or any other records showing the date of birth.
When no written documentation was available, the age given Pooled Analysis
by the mother or the guardian was used. The pooled analysis was made on the basis of an overall
* A primary health center is a basic health unit staffed by a medical officer and
sample of 438 children (73 from each subcenter). Of these,
health team that provides integrated curative and preventive health-care 428 (97.7%) had been vaccinated against measles, and overall
services to a rural population of approximately 20,000–30,000 persons. A measles vaccination coverage in Poondi was considered to be
health subcenter is a peripheral outpost staffed by an auxiliary nurse midwife 97.7% (CI = 95.7%–98.8%) (Table).
who provides primary health-care services (e.g., mother-and-child care, family
planning, and vaccination) for a population of approximately 3,000–5,000
persons.
18 MMWR April 28, 2006

TABLE. Measles vaccination coverage, by health subcenter — Poondi Primary On the basis of the results of the study, an analy-
Health Center, Tiruvallur District, Tamilnadu, India, 2002–2003*
sis of vaccination coverage by health subcenter
Coverage
and village during regular monthly review meet-
No. of children Estimate
ings was recommended to identify missed pock-
Health subcenter Surveyed Vaccinated Unvaccinated LQAS† (95% CI§)
ets of unvaccinated children and to continue
Neyveli 73 69 4 <95% NA¶
Vellathukkottai 73 70 3 <95% NA
efforts to maintain uniform high vaccination cov-
Meyyur 73 71 2 >95% NA erage in the PPHC area. The recommendation
Chitampakkam 73 72 1 >95% NA was followed by all the medical officers and helped
Poondi 73 73 0 >95% NA them identify groups of unvaccinated children
Nambakkam 73 73 0 >95% NA
Total 438 428 10 NA 97.7%
and improve the measles vaccination coverage.
(95.7%–98.8%) No outbreaks of measles were reported subse-
* Fiscal year 2002–2003 (April 1, 2002–March 31, 2003). quently in the study area. However, measles sur-
† Lot quality assurance sampling.
§ Confidence interval.
veillance was limited by the unavailability of
¶ Not applicable. long-term trend data, and the validity of the
surveillance system has not been estimated.
The findings of this study are subject to at least
Discussion one limitation. Vaccination cards were available for only 91 (21%)
of 438 children surveyed, which could have resulted in overesti-
This study determined that the overall vaccination cover- mating measles vaccination coverage. The supply of vaccination
age level in the PPHC area was 97.7%, which is consistent cards was limited, and those parents who did receive cards often
with the state target level of >95%. Two (33.3%) subcenters did not preserve them. To aid in future surveys, a sufficient quan-
had coverage levels below the target level of >95%; these tity of vaccination cards should be made available to health workers
results were used to target interventions to these low-coverage in each health subcenter. This will facilitate monitoring and evalu-
areas. ation efforts. Health workers should be trained to understand the
LQAS techniques provide a rapid and simple determina- importance of the cards and how to use them properly, and fami-
tion of output quality and are used in industry for quality- lies should be educated to understand the need to keep the cards in
assurance purposes. The strategy and goals of LQAS in the a safe place.
health field are similar to those in the manufacturing field In December 2004, an outbreak of measles was reported in
(7). LQAS analysis is based on testing a hypothesis rather than the Cuddalore district of Tamilnadu. Estimated measles vacci-
on estimating a proportion. Because LQAS is based on strati- nation coverage was approximately 96% (9), indicating that a
fied random sampling, results from lot samples can be com- measles outbreak can occur among a well-vaccinated popula-
bined to obtain a point estimate for the entire population, tion when a single-dose measles vaccination strategy is employed.
allowing for a small sample size. LQAS procedures were use- A substantial measles outbreak also was reported during 1999–
ful for identifying small health areas with lower measles vacci- 2000 in Sri Lanka, where single-dose measles vaccination
nation coverage. This information, combined with further coverage since 1996 was >90% (10). For future measles out-
assessment of performance problems and timely corrective breaks in an area with a single-dose measles schedule to be pre-
action, has been used to improve vaccination coverage in the vented, a possible strategy to provide a second dose might be
district. LQAS techniques also could be used to assess perfor- considered.
mance as part of routine monitoring or supervisory activities
of routine vaccination. References
For a vaccination program to achieve its goal, a sufficient 1. World Health Organization. Measles mortality reduction and regional
elimination, strategic plan 2001–2005. Geneva, Switzerland: World
number of doses must be administered at the appropriate ages. Health Organization; 2001.
Coverage levels are therefore a key process indicator of perfor- 2. Department of Family Welfare, Ministry of Health and Family Wel-
mance. Monitoring this indicator at the population level pro- fare. Multi year strategic plan 2005–2010, Universal Immunization
vides an overall assessment of program performance. Programme. New Delhi, India: Government of India; 2005:24–6.
Operational units with poorer coverage should be identified 3. World Health Organization. Description and comparison of the methods
of cluster sampling and lot quality assurance sampling to assess immuni-
so performance can be improved (8). LQAS techniques are a
zation coverage. Geneva, Switzerland: World Health Organization; 2001.
particularly useful way of monitoring indicators of coverage,
as these techniques provide a rapid and simple determination
of output quality.
Vol. 55 / Supplement MMWR 19

4. Singh J, Jain DC, Sharma RS, Verghese T. Evaluation of immuniza- 8. Lanata CF, Stroh G Jr, Black RE, Gonales H. An evaluation of lot
tion coverage by lot quality assurance sampling compared with quality assurance sampling to monitor and improve immunization
30-cluster sampling in a primary health center in India. Bull World coverage. Int J Epidemiol 1990;19:1086–90.
Health Organ 1996;74:269–74. 9. Mohan A, Murhekar MV, Porkaipandiyan RT, Hutin Y, Wairagkar NS,
5. Murthy BN, Radahakrishna S, Venkatasubramanian S, et al. Lot qual- Gupte MD. An outbreak of measles in Cuddalore, Tamilnadu, India.
ity assurance sampling for monitoring immunization coverage in In: Gupte MD, ed. Tsunami: Indian Council of Medical Research
Madras City. Indian Pediatr 1999;36:555–9. response. Chennai, India: National Institute of Epidemiology;
6. Lemeshow S, Hoshmer DW Jr, Klar J, Lwanga SK. Adequacy of sample 2005:18–24.
size in health studies. New York, NY: John Wiley & Sons; 1990:198. 10. Puvimanasinghe JP, Arambepola CK, Abeysinghe NM, et al. Measles out-
7. Lemeshow S, Taber S. Lot quality assurance sampling: single and double break in Sri Lanka, 1999–2000. J Infect Dis 2003;187(Suppl 1):S241–5.
sampling plans. World Health Statistics Quarterly 1991;44:115–32.
20 MMWR April 28, 2006

Bacterial Meningitis Among Cochlear Implant Recipients —

Canada, 2002
Samantha D. Wilson-Clark,1,2 S. Squires,1 S. Deeks1
1Immunization and Respiratory Infections Division, Public Health Agency of Canada, Ottawa, Canada
2Canadian Field Epidemiology Program, Public Health Agency of Canada, Ottawa, Canada

Corresponding author: Samantha D. Wilson-Clark, Region of Waterloo Public Health, 99 Regina St S, 3rd floor, Waterloo, ON N2J 4V3. Telephone: 519-883-2004
ext. 5413; Fax: 519-883-2241; E-mail: wsamanth@region.waterloo.on.ca.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: In July 2002, a cluster of bacterial meningitis (BM) cases was identified among European cochlear implant
recipients (CIRs), prompting Health Canada to conduct a retrospective cohort study to determine the rate of BM infection
among Canadian CIRs and to identify risk factors for acquiring BM.
Methods: A survey was mailed to 1,432 Canadian CIRs who had received implants during January 1995–July 2002 to
assess occurrence of postimplant BM infection. Data collection included demographics, episodes of meningitis, and vaccina-
tion status.
Results: A total of 1,024 (72%) surveys were completed. Median age of CIRs at implantation was 16 years (range: 7 months–
81 years). Five (0.5%) cases of BM infection were reported (two pneumococcal, one meningococcal, and two of unknown
etiology); one CIR died. Four cases occurred among children aged <18 years. Time between implantation and BM infection
varied (range: 7 months–7.7 years; median: 11 months). The rate of BM infection per 1,000 person-years was 0.7 among
CIRs aged >18 years and 2.9 among those aged <18 years. The proportion of CIRs vaccinated against pneumococcal and
meningococcal disease was low (46% and 41%, respectively). Preimplant meningitis was identified as a risk factor for
postimplant BM (p = 0.002). No other risk factors evaluated were associated with an increased risk for BM infection.
Conclusion: CIRs have a high rate of postimplant BM infection. Preimplant BM infection was identified as a risk factor.
Cases of BM infection might have been prevented through vaccination.

Introduction facturer A (2). This device differed from other similar manu-
factured devices in that it had two components, an electrode
Meningitis is an inflammation of the lining of the brain’s
array and a positioner, rather than a single electrode array.
surface, often as a result of a bacterial or viral infection. Chil-
The positioner increased electrical signal transmission to the
dren aged <2 years are most at risk for meningitis. Among
auditory nerve, particularly among patients with malforma-
infants, early symptoms of meningitis include fever, irritabil-
tions of the cochlea. On July 23, manufacturer A issued a
ity, lethargy, and loss of appetite. Among children and adults,
voluntary recall of the device in France (3). After discussions
other symptoms might include headache, stiff neck, photo-
with the European regulatory authorities, manufacturer A sub-
phobia, nausea and vomiting, and confusion or alteration in
sequently issued a voluntary worldwide withdrawal of the
consciousness (1).
device. On July 26, the two-piece device, which had been
Cochelar implants are medical devices that electronically
licensed for use in Canada in November 1998, was withdrawn
stimulate the auditory nerves in the cochlea (inner ear),
from the Canadian market.
allowing persons with severe hearing loss to perceive sound.
On July 29, Health Canada issued an alert warning that
In early July 2002, a cluster of cases of bacterial meningitis
CIRs might be at greater risk for meningitis (4). Subsequently,
(BM) infection was identified among European cochlear
the Immunization and Respiratory Infections Division of
implant recipients (CIRs) who had received an implant pro-
Health Canada (now the Public Health Agency of Canada)
duced by manufacturer A (2). European implant surgeons had
investigated the extent of BM infection among CIRs in
hypothesized that the cluster of BM infections they had
Canada. To understand the magnitude of this problem in
observed among their patients might be related to a recently
Canada, a retrospective cohort study was conducted. The three-
licensed cochlear implant device (device A) produced by manu-
Vol. 55 / Supplement MMWR 21

fold purpose of the investigation was to determine the rate of TABLE 1. Number and percentage of cochlear implant recipients
(CIRs), by selected characteristics — Canada, 2002
BM among CIRs in Canada, identify risk factors for
Characteristic No. (%)
postimplant BM, and recommend public health action based
Age at time of implant
on investigation findings. Pediatric (<18 yrs) 482 (47.5)
Adult (>18 yrs) 532 (52.5)
Median: 16 yrs
Methods Range: 7 mos–81 yrs
Sex
The study cohort included all 1,432 recipients of cochlear Female 523 (51.1)
implant devices in Canada with implant dates during January Male 501 (48.9)
1995–July 2002. Cohort members were identified using manu- Implant manufacturer
A 391 (38.4)
facturer implant registries. At the time of the study, two manu- B 628 (61.6)
facturers were licensed in Canada. A case was defined as one Date of implant
occurring in a CIR who reported having BM infection since Before January 1, 1999 294 (35.2)
receiving an implant. On or after January 1, 1999 541 (64.8)
A self-administered questionnaire in both French and Bacterial meningitis before implant
Yes 71 (7.3)
English was mailed to CIRs or, if deceased, their next of kin. No 905 (92.7)
The questionnaire was designed by modifying tools developed Bacterial meningitis after implant
for a similar investigation in the United States (5). The survey Yes 5 (0.5)
was completed by the recipient or by a parent or guardian if No 984 (99.5)
Received vaccination
the recipient was aged <16 years or was incapable of complet-
Pneumococcal 432 (45.7)
ing the survey. Questions addressed included cause of hearing Meningococcal serogroup C 381 (41.1)
loss, history of meningitis infection, vaccination status, and Hib* 369 (40.0)
risk factors (e.g., household smoking, other children in the Received vaccination before implant†
Pneumococcal 40 (11.6)
household, and otitis media infections). Questions were
Meningococcal serogroup C 32 (10.9)
divided into two periods: before and after receiving a cochlear Hib 155 (55.4)
implant. Only nonnominal data were collected. Data were Additional implanted devices§ 33 (6.9)
collected during November 25, 2002–March 31, 2003. To Cause of deafness
obtain the highest response rate possible to the mailed ques- Acquired 408 (40.1)
Congenital 177 (17.4)
tionnaire, Dillman’s Total Design Method for mailed surveys
Other 200 (19.7)
was followed, with certain modifications (6). Ethics approval Unknown 233 (22.9)
for the study was obtained from Health Canada’s Research Household smoker 279 (22.9)
Ethics Board. Other household children 275 (26.9)
Data were entered in EpiData 2.1a (The EpiData Associa- Child care attendance 320 (31.5)
tion, Odense, Denmark, 2001–2002). Univariate and * Haemophilus influenzae type b.
† Certain CIRs were not old enough to be eligible for vaccination before
bivariate analysis, including relative risks and chi-square tests, implantation.
were conducted using Epi Info 6.04d (CDC, Atlanta, Geor- § Includes middle-ear tubes and ventribuloperitoneal and endolymphatic

gia). To determine if the introduction of the two-piece shunts.

implant affected incidence of BM infection, data were strati-
fied by implant manufacturer and date of implant (i.e., before sented in proportion to their market share (approximately 33%
and after January 1, 1999, when the two-piece implant device for manufacturer A and 67% for a second manufacturer
was licensed in Canada). [manufacturer B]).
Among 98 CIRs who had an episode of meningitis infec-
tion before receiving a cochlear implant, 71 (72%) had infec-
Results tions that were bacterial. Of seven cases of meningitis
A total of 1,024 CIRs completed and returned the survey reported after a cochlear implant, five (71%) were bacterial;
(overall response rate: 72%). Median follow-up time (from the type of bacteria was reported for three cases. Two CIRs
implant to BM infection or end of the study period) for CIRs reported Streptococcus pneumoniae as the causative agent, and
was 42 months (Table 1). Among respondents, males and one reported Neisseria meningitidis (Table 2). None of the BM
females were equally represented, and approximately half episodes occurred during the perioperative period (0–30 days
(n = 482) were aged <18 years. Manufacturers were repre- postimplantation).
22 MMWR April 28, 2006

TABLE 2. Cases of postimplant bacterial meningitis (BM) among cochlear implant recipients — Canada, 2002
Postimplant BM
Interval between
Preimplant Previous implantation
Age group (yrs)* Sex BM Type vaccination and infection Outcome
1–4 Male No Pneumococcal No 20 mos Recovered
5–9 Female No Unknown —† 7 mos Recovered
10–14 Female Yes Meningococcal No 11 mos Died
15–17 Male Yes Pneumococcal No 7 yrs, 9 mos Recovered
65–70 Female Yes — — 12 mos Recovered
* No cases occurred among persons aged 18–64 years.
† No data available.

The overall incidence of BM infection for this cohort was TABLE 3. Number of cases and relative risk for postimplant
1.8 per 1,000 person-years of observation (95% confidence bacterial meningitis (BM) among cochlear implant recipients,
by selected characteristics — Canada, 2002
interval [CI] = 0.6–4.2). Among CIRs aged <18 years, inci-
Postimplant BM
dence was 2.9 (CI = 0.8–7.3); among CIRs aged <6 years, Yes No
incidence was slightly lower (2.0; CI = 0.2–7.1). Incidence (No. (person- RR†
among adults aged >18 years was 0.7 per 1,000 person-years Characteristic cases) months*) (95% CI§)
of observation (CI = 0.0–4.1). BM before implant
Yes 3 2,003 23.1
Incidence of BM infection among adults aged >18 years No 2 30,920 (3.2–197.3)
before and after 1999 did not vary (2.2 and 2.0 per 1,000 person- Other implanted devices
years observation, respectively). Although not statistically Yes 2 3,380 6.0
significant, incidence of BM infection among persons aged No 3 30,573 (1.1–36.1)
<18 years was higher on or after January 1, 1999, than before Otitis media after implant
(4.0 and 2.2 per 1,000 person-years, respectively). Among chil- Yes 2 5,336 3.2
No 3 25,750 (0.5–19.2)
dren aged <6 years, incidence was 1.5 (CI = 0.0–8.1) before
Child care attendance
1999 and 3.1 on or after January 1, 1999 (CI = 0.1–17.2). Yes 1 11,402 0.4
All five persons with BM infection had received meningo- No 4 17,593 (0.04–3.5)
coccal vaccine, and four had received pneumococcal vaccine. Implant manufacturer
However, of the three persons for whom the causative agent A 2 11,685 1.3
B 3 22,383 (0.2–7.6)
was known, none had received vaccination against the
Child in household
implicated agent before the postimplant episode of meningitis. Yes 1 9,139 0.7
Potential risk factors assessed for postimplant BM infection No 4 25,086 (0.1–6.1)
included a history of otitis media, household smoking, and Household smoker
children living in the household. None was statistically sig- Yes 1 10,820 0.5
No 4 23,358 (0.1–4.8)
nificant (Table 3). A previous episode of BM infection was
* Number of months each study participant contributed to the study cohort.
identified as a risk factor for postimplant BM infection (rela- † Relative risk, calculated by using the number of cases as the numerator
tive risk [RR] = 23.1; CI = 3.2–197.3; p = 0.002). Having and the number of person-months of observation for each person in the
had other implanted devices was associated with an increased cohort as the denominator.
§ Confidence interval.
risk for BM infection; however, this association was not sta-
tistically significant (p = 0.081). Discussion
No difference in risk for BM infection by implant manu-
facturer was noted (p = 1.0). The cohort included recipients The rate of BM infection per 1,000 person-years among
who had received the device with a positioner; however, the CIRs aged >18 years was 0.7, compared with 2.9 among CIRs
type of implanted device (with or without a positioner) was aged <18 years. Results of a similar study of CIRs in the United
not well reported, and whether the positioner was an inde- States during the same period have been published (5). Inci-
pendent risk factor could not be determined. dence of BM infection among CIRs was 3.9 among U.S. chil-
dren aged <6 years, compared with 2.0 among Canadian
children in the same age group during the period when device
A was on the market. Although different methodologies were
used, certain key study questions were similar, allowing com-
Vol. 55 / Supplement MMWR 23

parisons between the two studies. In Canada, no cases of and that did not contain the most recent mailing address of
perioperative BM infection were identified among CIRs, every CIR in Canada. CIRs who did not register their
whereas in the U.S. study, the rate of perioperative BM infec- implant devices were not included in the cohort. However,
tion was 2.1 cases per 1,000 procedures (5). Why incidence because incomplete mailing addresses were evenly distributed
of perioperative infections is higher among CIRs in the United among the two implant manufacturers, no selection bias was
States is not known. likely introduced as a result. Second, as with all self-
These findings of increased incidence of BM infection administered questionnaires, reported medical histories might
among CIRs aged <18 years since the device with the positioner not be accurate, particularly for adults who might have lost
was introduced in 1998 are similar to findings published pre- their hearing many years before receiving a cochlear implant.
viously (5). No increased risk for BM infection was identified Third, medical details (e.g., type of meningitis and whether
among adult CIRs after the device with the positioner was treatment was received for episodes of otitis media) were poorly
introduced. reported. Response rates for medical details varied widely
In Canada, surveillance of invasive diseases (e.g., meningi- (range: 40%–80%) compared with responses to other ques-
tis) is organism specific and includes all forms of invasive dis- tions (e.g., ever having otitis media, ever having meningitis,
ease. Incidence of invasive pneumococcal disease ranges from and cause of deafness [range: 90%–99%]). The cause of BM
11.6 to 17.3 per 100,000 population, whereas incidence of was based on self-report and was not verified with medical
invasive meningococcal disease ranges from 0.6 to 1.6 per records. In addition, subtyping of bacteria that caused
100,000 population (7,8), compared with an observed inci- postimplant BM was not reported. The type might not have
dence of 1.8 per 1,000 person-years among CIRs in this study. been vaccine preventable. Certain questions pertaining to vac-
CIRs have multiple potential underlying conditions that cination history were misinterpreted. For example, Hib vac-
might increase their risk for BM infection above that of the cine was often indicated as having been received yearly,
general Canadian population. The ideal comparison group indicating confusion between Hib vaccine and annual influ-
for CIRs would be a population of severe-to-profoundly deaf enza vaccine. Finally, the study was conducted in English and
persons who do not have cochlear implants; however, such French. CIRs who could not read either of these languages
data are not available. During 1994–2001, the overall annual were systematically excluded from the cohort and might have
incidence of BM infection in the general Canadian popula- BM at a different rate from CIRs who read English or French.
tion ranged from 3.2 to 3.7 per 100,000 population (9).
Preliminary study results were presented to Canada’s
National Advisory Committee on Immunization (NACI). In Conclusion
February 2003, NACI recommended that CIRs be consid- In this study, CIRs had postimplant BM infection at a rate
ered at high risk for both Haemophilus influenzae type b (Hib) of approximately 1.8 cases per 1,000 person-years of observa-
and invasive pneumococcal disease and should receive vacci- tion. In addition, children with cochlear implants had BM
nation according to the high-risk schedule (10). In addition, infection at a higher rate than adults (2.9 and 0.7 per 1,000
CIRs, like all Canadians, should be up-to-date on all routine person-years of observation, respectively). Because of the
vaccinations, including meningococcal C conjugate vaccines. increased risk for BM infection among CIRs, health-care pro-
These vaccines are recommended for all Canadian children fessionals should ensure that CIRs and persons considering
aged <5 years, adolescents, and young adults. cochlear implants are vaccinated against bacteria that com-
CIRs, their parents, and caretakers should be aware of the monly cause meningitis. The only risk factor identified for
signs and symptoms of meningitis and seek medical attention having postimplant BM was preimplant BM infection. CIRs
if they occur. Medical professionals should be aware of the and their families should be aware of the signs and symptoms
potential for BM infection among CIRs, be vigilant for the of meningitis and seek prompt attention if they occur.
signs and symptoms of meningitis in this population, and
educate their patients accordingly. Primary health-care pro- Acknowledgments
viders of CIRs or persons considering cochlear implants should Six other members of the Cochlear Implants and Bacterial Meningitis
ensure that these persons are fully vaccinated according to Collaborative Working Group contributed to this report: Attar Chawla,
NACI guidelines (10). MD, Barbara Harrison, Irwin Hinberg, MD, Damian Kakwaya, Fred
The findings in this report are subject to at least four limi- Lapner, MD, and Tara Tucker, MD. Lee Lior, MD, Eleni Galanis,
tations. First, the retrospective cohort was established by MD, and Arlene King, MD, also contributed to this report.
using manufacturer registry data that were likely incomplete
24 MMWR April 28, 2006

References 6. Dillman DA. Mail and telephone surveys: the total design method.
1. Tunkel AR, Scheld WM. Acute meningitis. In: Mandell GL, Bennett New York, NY: Wiley; 1978.
JE, Dolin R, eds. Principles and practice of infectious diseases. 5th ed. 7. Squires SG, Deeks SL, Tsang RSW. Enhanced surveillance of invasive
New York, NY: Churchill Livingstone; 2000. meningococcal disease in Canada, 1 January 1999 through 31
2. Hannoverschen Cochlear Implant-Gesellschaft e.V. Meeting on post December 2001. Can Commun Dis Rep 2004;30:17–28.
cochlear implantation meningitis. Schiphol Airport, Amsterdam, the Neth- 8. National Advisory Committee on Immunization. Statement on rec-
erlands July 5, 2002 [Minutes; German]. Available at http://www.hcig.de. ommended use of pneumococcal conjugate vaccine. Can Commun
3. French Public Health Agency for Health Product Safety. Recall of Dis Rep 2002;28:ACS–2.
Clarion cochlear implants with a positioner made by Advanced Bion- 9. MacDonald D, Deeks SL, Squires SG, Medaglia A, Tam T. Hospital-
ics Corporation following cases of meningitis, July 23, 2002 [French]. ization for bacterial meningitis in Canada, 1994–2001 [Poster presen-
Available at http://agmed.sante.gouv.fr/htm/alertes/filalert/dm02 tation]. Canadian Immunization Conference, Montréal, Québec,
0706.htm. Canada, December 5–8, 2004.
4. Health Canada. Cochlear implant recipients may be at greater risk for 10. National Advisory Committee on Immunization. Immunization rec-
meningitis. Available at http://www.hc-sc.gc.ca/dhp-mps/medeff/ ommendations for cochlear implant recipients. Can Commun Dis Rep
advisories-avis/prof/2002/implant_cochle_nth_ah_e.html. 2003;29:ACS–2.
5. Reefhuis J, Honein MA, Whitney CG, et al. Risk of bacterial meningitis
in children with cochlear implants. N Engl J Med 2003;349:435–45.
Vol. 55 / Supplement MMWR 25

Risk Factors for Neonatal Tetanus —

Busoga Region, Uganda, 2002–2003
Sheba N. Gitta,1 F. Wabwire-Mangen,1 D. Kitimbo,2 G. Pariyo1
1Makerere University Institute of Public Health, Kampala, Uganda
2District Directorate of Health Services, Jinja, Uganda

Corresponding author: Sheba N. Gitta, Makerere University Institute of Public Health, P.O. Box 7072, Kampala, Uganda. Telephone: 256-77-2479403;
Fax: 256-41-531807; E-mail: sgitta@iph.ac.ug or sgitta@med.mak.ac.ug.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Background: Uganda has not achieved the 2005 neonatal tetanus (NNT) global elimination target set by the World Health
Organization (WHO). The Busoga region has the highest recorded level of NNT incidence in Uganda. To understand the
reasons for this high incidence, a study was conducted to identify NNT risk factors.
Methods: During March–May 2004, a matched case-control study was conducted in the Busoga region for a 2-year study
period (2002–2003). Matching variables were sex, residence, and date of birth. A total of 24 cases of NNT (according to the
WHO case definition) were identified from hospital records, and 96 community controls (children who survived the neonatal
period) were selected.
Results: Bivariate analysis indicated that neonates with NNT were more likely to have been delivered outside a health
facility, on an unclean surface, without use of gloves, or by unskilled attendants. Mothers of these neonates were less likely to
report vaccination during previous pregnancies, administration of 2 doses of tetanus toxoid (TT) during the study pregnancy,
or use of certain intravaginal substances (most commonly, herbs) at onset of labor. Multivariate analysis indicated that
unclean delivery surfaces (odds ratio [OR] = 38.8; 95% confidence interval [CI] = 2.9–518.1) and primigravidae mothers
(OR = 79.5; CI = 1.8–3,472.2) were associated with NNT. Administration of 2 doses of TT during pregnancy, vaccination
during previous pregnancies, and intravaginal application of certain substances were protective against NNT.
Conclusion: These findings underscore the importance of having clean delivery surfaces and of mothers receiving 2 doses of
TT during pregnancy. Implementation of these measures might help eliminate NNT from the Busoga region of Uganda.

Introduction A previous study attributed NNT in Busoga to unhygienic

delivery practices, application of harmful substances on the
Neonatal tetanus (NNT) is a fatal yet preventable disease
umbilical cord, and lack of TT immunization among moth-
that accounts for 14% of annual neonatal deaths worldwide
ers but did not document the magnitude of risk associated
(1). In 1997, a total of 248,000 NNT deaths occurred glo-
with these factors (6). Worldwide, NNT has been associated
bally; 95,000 NNT deaths occurred in Africa, 4,600 of which
with prenatal, perinatal, and neonatal factors, including
occurred in Uganda (2). In 2000, the World Health Organi-
unhygienic delivery and cord-care practices and lack of
zation (WHO) targeted NNT for global elimination by 2005
maternal TT immunization (7). Risk for contamination and
(3). Uganda is one of the 57 countries that have not yet elimi-
subsequent occurrence of NNT remains high for several days
nated NNT. In Uganda, with 195 cases, the Busoga region
after delivery until the baby’s cord wound heals (8).
has the highest NNT incidence (4). In 2002, the Uganda Min-
NNT cases continue to be reported in Busoga. For this rea-
istry of Health (MoH) launched an NNT elimination cam-
son, district health teams in the region recommended that a
paign in Busoga. After the second round of mass tetanus toxoid
study be conducted to identify NNT risk factors as a basis for
(TT) immunization during the NNT elimination campaign,
planning and implementing appropriate NNT elimination
approximately 86% of women of reproductive age in all dis-
strategies.
tricts in the region had received 2 doses of TT (5).
26 MMWR April 28, 2006

Methods Data were double-entered and validated by using Epi Info

6.04 (CDC, Atlanta, Georgia). Epi Info also was used for
A community-based, matched case-control study was con-
bivariate analysis; stratified Mantel-Haenszel analysis produced
ducted to identify delivery and umbilical cord-care practices
matched odds ratios (ORs) and 95% confidence intervals
that predispose neonates to NNT and to compare maternal
(CIs). For multivariate analysis, data were exported to SPSS
TT immunization status for both groups. The study was con-
10.0 (Statistical Package for the Social Sciences for Windows,
ducted during March–May 2004 in five districts that make
version 10.0; SPSS; Chicago, Illinois); conditional
up the Busoga region (Bugiri, Iganga, Jinja, Kamuli, and
logistic regression was performed to identify the predictors
Mayuge). Respondents were the mothers of the neonates with
for NNT. All variables entered in the model were dichoto-
NNT and matched controls.
mized, with neonatal tetanus case (Yes/No) as the outcome
Mothers of neonates with NNT were identified by screen-
variable. The initial logistic regression model included all vari-
ing NNT investigation forms from all hospitals in the Busoga
ables that were statistically significant during bivariate analy-
region for January 2002–December 2003. A case was defined
sis (e.g., delivery place, birth attendant, delivery surface, use
as confirmed NNT according to the WHO case definition
of intravaginal substances, use of gloves, maternal TT status
(i.e., history of normal sucking and crying for the first 2 days
in study, and previous pregnancies). Other variables that were
of life, onset of illness at age 3–28 days, and inability to suck
not significant on bivariate analysis but that are known risk
followed by generalized stiffness and/or spasms) occurring in
factors for NNT (e.g., handwashing, type of cord-cutting tool,
a child born in the Busoga region during the study period (5).
cord dressings, and cord tie used) and potential confounders
A control was defined as a child who had survived the neona-
(e.g., maternal age) also were included in the model. Mea-
tal period, who was matched for sex and date of birth (range:
sures of interest were adjusted ORs and CIs for NNT risk
±6 months), and who lived in the same village as a neonate
factors. Goodness of fit of the final model was tested by using
with NNT. A control was selected from each of the first four
Nagelkerke R2 value.
households closest to the home of a child with NNT that had
a child consistent with the control definition. If a household
had more than one eligible control, the child whose date of Results
birth was closest to that of the neonate with NNT was selected.
Multiple controls were selected to increase the power of the A total of 40 NNT cases were identified from the records,
study (9). Children whose mothers did not live in the Busoga but only 24 mothers could be traced in the community. Full
region at the time of delivery were excluded from the study. names were not available for three mothers, and addresses of
The questionnaire used was adapted from the WHO NNT nine mothers were not recorded in the hospital records. In four
case investigation form (5). It was modified before the study cases, parents had moved and could not be located. Four
was conducted on the basis of qualitative data about the most matched controls were selected for each of the 24 cases, for a
probable NNT risk factors collected from 80 mothers who total of 96 controls and a study sample of 120.
participated in eight focus group discussions and interviews
with eight traditional birth attendants. Questionnaires were Bivariate Analysis
pretested among 20 Jinja district mothers who were not part The majority of neonates with NNT and controls had moth-
of the final study sample. ers who were aged >20 years, were currently married, and had
Trained interviewers used the pretested, interviewer- attained at least a primary education (Table 1). The percent-
administered questionnaire to collect data on maternal socio- age of mothers who were primigravidae (i.e., this was their
demographic characteristics, delivery practices, umbilical first pregnancy) was 25% for neonates with NNT and 12.5%
cord-care practices, and maternal TT immunization status. for controls. The majority of mothers (17 [70.8%] of 24 neo-
The delivery surface was categorized as clean or unclean; any nates with NNT and 62 [64.6%] of 96 controls) had had one
surface other than a new plastic sheet or operating theater to four previous deliveries. No statistically significant associa-
table was classified as unclean. The mother of a neonate with tions were observed between maternal sociodemographic char-
NNT was always interviewed before that of the controls to acteristics and NNT.
ensure accurate matching. The Uganda National Council of Maternal TT immunization status of the two study groups
Science and Technology and Makerere University Institute of was compared (Table 2). Maternal TT immunization status
Public Health Higher Degrees Research and Ethics Commit- was determined by history because only 12.5% of mothers
tee approved this study, and all participating mothers signed whose neonates had NNT and 26% of controls had TT
consent forms.
Vol. 55 / Supplement MMWR 27

TABLE 1. Number and percentage of neonates with neonatal tetanus intravaginal substances (most frequently herbs) (OR
(NNT) and controls, by selected maternal characteristics — Busoga
region, Uganda, 2002–2003
= 0.2; CI = 0.1–0.8) (Table 3). Bivariate analysis
Children
indicated that the only umbilical cord-care practice
with NNT Controls Matched that was associated with NNT was the birth atten-
Characteristic No. (%) No. (%) OR* (95% CI†) dant not wearing gloves (OR = 3.8; CI = 1.1–13.1;
Maternal age (yrs) p = 0.06) (Table 2).
<20 3 (12.5) 16 (16.7) 0.7 (0.2–2.8)
>20 21 (87.5) 80 (83.3) 1.0
Marital status
Multivariate Analysis
Unmarried 4 (16.7) 10 (10.4) 1.9 (0.5–7.5) The best fitting model for NNT risk factors
Married 20 (83.3) 86 (89.6) 1.0
explained 71.4% (Nagelkerke R2) of the variation
Mother’s education level
None/Primary 22 (91.7) 72 (75.0) 3.7 (0.8–16.9)
observed in the outcome variable. Within this model,
Secondary 2 (8.3) 24 (25.0) 1.0 the mother receiving 2 doses of TT during the study
Mother’s occupation pregnancy and a history of receiving TT in previous
Agriculture related 8 (33.3) 25 (26.0) 2.0 (0.5–8.6) pregnancies were significantly protective (p = 0.007)
Not agriculture related 16 (66.7) 71 (74.0) 1.0 (Table 4). Receiving only 1 TT dose in study preg-
No. of previous deliveries nancy was not protective and was excluded from the
None 6 (25.0) 12 (12.5) 2.5 (0.8–8.4)
>1 18 (75.0) 84 (87.5) 1.0 final model. Delivery on an unclean surface and
* Odds ratio. primigravidae mothers were associated with increased
† Confidence interval.
risk for NNT. Use of intravaginal substances at
onset of labor was significantly protective, with
immunization cards. Significantly fewer mothers of neonates mothers of neonates with NNT less likely to have used them
with NNT (20.8%) than control mothers (66.7%) had (p = 0.017).
received the recommended 2 doses of TT by the time of
delivery (OR = 0.2; CI = 0.1–0.5; p = 0.0001). Mothers of
neonates with NNT also were less likely to have had TT Discussion
immunization for previous births (OR = 0.1; CI = 0.02–0.3; This study identified two risk factors for NNT and two
p<0.001). The majority of mothers in both groups had protective factors in the Busoga region of Uganda: unclean
received prenatal care during the study pregnancy, but this delivery surfaces and primigravidae mothers. Unclean deliv-
was not identified as being protective. ery surfaces were the most likely source of tetanus organisms,
All four investigated delivery practices were associated with underscoring the importance of using clean delivery surfaces
NNT (Table 2). Mothers of neonates with NNT were more to prevent NNT. This finding concurs with results of other
likely than control mothers to have delivered outside health studies that identified unclean delivery surfaces as a risk fac-
facilities and to have had unskilled birth attendants; 70.8% of tor for NNT (8,10). Children born to primigravidae mothers
these mothers had unskilled attendants, including traditional were at more risk for having NNT than those born to multi-
birth attendants (six of 24), friends or relatives (six of 24), parous mothers. The association might be confounded by
and nursing assistants (five of 24). In contrast, only 31.3% of receipt of TT before pregnancy, which was not captured in
controls were delivered by unskilled attendants, including tra- this study.
ditional birth attendants (12 of 96), friends or relatives (nine Intravaginal application of local medicines (most commonly,
of 96), nursing assistants (four of 96), and self (five of 96). herbs) at onset of labor is widely practiced as it is thought to
Doctors delivered no neonates with NNT and 3.1% (three of ease labor by widening the birth canal. This traditional prac-
96) of controls. The rest were delivered by midwives and nurses. tice was negatively associated with acquisition of NNT,
Mothers of neonates with NNT were more likely than con- implying that it might have had a protective effect. This find-
trol mothers to have delivered on unclean delivery surfaces ing has not been reported previously and requires further clari-
(e.g., mats, sacks, uncovered delivery beds, and uncovered fication. Association might result from reporting bias, as a
floors). Certain mothers (three [12.5%] of 24 mothers of chil- mother of a neonate with NNT conceivably might have with-
dren with NNT and 36 [37.5%] of 96 controls) reported hav- held information about use of these substances for fear of
ing applied substances into the vagina at onset of labor, a being blamed for her child’s illness. Alternatively, applying
traditional practice in this region; mothers of neonates with such substances immediately after bathing, as is the practice,
NNT were less likely than control mothers to have used might have ensured that clean hands were used, thus decreas-
28 MMWR April 28, 2006

TABLE 2. Number and percentage of neonates with neonatal tetanus (NNT) ing the risk for contaminating the substances and
and controls, by delivery practices, cord-care practices, and maternal
tetanus toxoid immunization status — Busoga, Uganda, 2002–2003
the birth canal. This finding is contrary to the
Neonates
increased risk for NNT observed in other studies of
with NNT Controls Matched pregnant women who had predelivery intravaginal
Characteristic No. (%) No. (%) OR* (95% CI†) exposure to ghee or coconut oil (8,11). Possibly the
Maternal immunization status herbs used as intravaginal medications had antibacterial
at delivery properties. Further research is required to determine if
1 dose of TT§ in study pregnancy
Yes 13 (54.2) 85 (88.5) 0.1 (0.04–0.5¶)
these herbs have such medicinal properties.
No 11 (45.8) 11 (11.5) 1.0 Absence of the mother receiving 2 doses of TT
2 doses of TT** in study pregnancy vaccination during pregnancy was identified as a risk
Yes 5 (20.8) 64 (66.7) 0.2 (0.1–0.5¶) factor for NNT, which concurs with findings of pre-
No 19 (79.2) 32 (33.3) 1.0
TT in previous pregnancies††
vious studies conducted in Uganda and Nigeria
Yes 9 (37.5) 76 (79.2) 0.1 (0.02–0.3¶) (6,11,12). The study indicated that <80% of all
No 15 (62.5) 20 (20.8) 1.0 mothers in Busoga region had received recom-
Delivery practices mended 2 doses of TT, which suggests that >20%
Delivery place
Outside health facility 13 (54.2) 26 (27.1) 4.5 (1.4–13.1¶) of children born in the region are still at risk for
Health facility 11 (45.8) 70 (72.9) 1.0 having NNT because of low 2-dose TT coverage.
Birth attendant§§ A discrepancy was noted between prenatal care
Unskilled 17 (70.8) 30 (31.3) 7.3 (2.1–16.9¶)
Skilled 7 (29.2) 66 (68.8) 1.0
attendance and maternal TT vaccination. Although
Delivery surface¶¶ approximately 75% of mothers of neonates with NNT
Unclean 12 (50.0) 14 (14.6) 6.7 (2.1–21.3¶) had prenatal care, only half received >1 dose of TT,
Clean 12 (50.0) 82 (85.4) 1.0 indicating that health-care workers are missing oppor-
Used intravaginal
substances***
tunities to vaccinate pregnant women. This finding has
Yes 3 (12.5) 36 (37.5) 0.2 (0.1–0.8¶) been documented previously (13–15). This discrep-
No 21 (87.5) 60 (62.5) 1.0 ancy might result from mothers receiving prenatal care
Cord-care practices only when pregnancy is so far advanced that they can
Attendant washed hands
Yes 16 (72.7) 74 (78.7) 1.0 receive only 1 dose of TT before delivery.
No 6 (27.3) 20 (21.3) 1.5 (0.5–5.0) The type of tool used to cut the umbilical cord
Attendant wore gloves was not identified as a risk factor for NNT. This
Yes 16 (69.6) 80 (85.1) 1.0
No 7 (30.4) 14 (14.9) 3.8 (1.1–13.1¶)
finding is consistent with a previous study conducted
Cutting tool in Senegal (16) and is probably attributable to the
New razor blade 14 (63.6) 51 (58.0) 1.0 use of new razor blades and scissors. In addition, no
Scissors 8 (36.4) 37 (42.0) 0.7 (0.2–2.7) mother reported having applied any harmful sub-
Thread cord tie
Yes 17 (70.8) 71 (74.0) 0.8 (0.3–2.4)
stance (e.g., cow dung or mud) to the newborn’s
No 7 (29.2) 25 (26.0) 1.0 cord wound. This suggests that a positive shift in
Cloth cord tie cord-care culture has occurred among mothers in
Yes 4 (16.7) 13 (13.5 1.3 (0.4–4.7) Busoga; certain unhygienic cord practices identified
No 20 (83.3) 83 (86.5) 1.0
Applied substance on cord
previously are no longer practiced (6). This change
Yes 6 (25.0) 33 (34.4) 0.7 (0.3–1.7) in cord-care practices might be attributable to wide-
No 18 (75.0) 63 (65.6) 1.0 spread health education activities in this region,
Prenatal-care attendance especially during recent mass NNT elimination cam-
None 6 (25.0) 14 (14.6) 1.8 (0.6–5.3)
At least one visit 18 (75.0) 82 (85.4) 1.0 paigns. It also might result from public awareness of
* Odds ratio. risk for acquisition of HIV associated with use of
† Confidence interval.
§ Received >1 tetanus toxoid dose during study pregnancy.
unsterilized instruments.
¶ Statistically significant at 0.05 level. The findings in this report are subject to at least
** Received >2 tetanus toxoid doses during study pregnancy. five limitations. First, use of unverified maternal TT
†† Answered “yes” to question, “Were you ever immunised with TT in previous preg-
nancies or during the recent mass TT immunisation campaign?” vaccination histories might have biased study find-
§§ Skilled = doctors, midwives, and nurses; any other = unskilled. ings. Second, health education provided during the
¶¶ Clean = only new plastic sheet or operating theatre; any other = unclean.
*** Answered “yes” to question, “Did you apply any substance into the vagina at onset
NNT elimination campaign might have introduced
of labor?” reporting bias; however, such bias would be
Vol. 55 / Supplement MMWR 29

TABLE 3. Percentage of mothers* who used intravaginal Conclusion

substances at onset of labor, by type of substance used —
Busoga region, Uganda, 2002–2003 This study identified two risk factors for NNT: unclean
Mothers of children Mothers of
Substance with neonatal tetanus controls
delivery surfaces and primigravidae mothers. Protective factors
Herbs 67% 86%
were the mother receiving 2 doses of TT in current pregnancy,
Burnt snake skin or snail shells 33% 8% history of TT in a previous pregnancy, and application of intrav-
Ghee 0 3% aginal substances at onset of labor. Cord-care practices were not
Vaseline 0 3%
NNT risk factors.
* Mothers of children with neonatal tetanus: n = three; controls: n = 36.
On the basis of these findings, the following recommenda-
tions are offered:
TABLE 4. Adjusted odds ratios (AORs) and 95% confidence • Health-care workers should routinely assess the TT sta-
intervals (CIs) for variables retained in the final multivariate tus of all female patients of reproductive age and vacci-
model of risk factors for neonatal tetanus — Busoga region,
Uganda, 2002–2003
nate them according to MOH vaccination guidelines.
Variable AOR (95% CI)
• Health-care workers should educate mothers, especially
Parity
primigravidae mothers, about the importance of receiv-
Primigravidae mothers 79.5 (1.8–3472.2*) ing 2 doses of TT during pregnancy and the need to pur-
Mothers with previous births 1.0 chase clean delivery kits.
Delivery surface† • MoH and district health teams (DHTs) should collabo-
Unclean 38.8 (2.9–518.1*)
Clean 1.0
rate with social marketing organizations to provide preg-
Birth attendant§
nant women with clean delivery kits (containing gloves,
Unskilled 9.0 (0.9–89.4) cord ligatures, blade, plastic delivery sheet, and cotton
Skilled 1.0 wool) at a subsidized fee.
Used intravaginal substances¶ • DHTs should budget for clean delivery kits, which they
Yes 0.04 (0.003–0.6*)
should provide at no cost to mothers who cannot afford
No 1.0
to buy.
Type of cord tie used
Thread 0.1 (0.014–1.1) • The possible prophylactic effects of application of intrav-
Other 1.0 aginal substances should be studied further.
2 doses of TT in study pregnancy**
Yes 0.05 (0.005–0.4*) References
No 1.0 1. World Health Organization. Field manual for neonatal tetanus elimi-
nation. Geneva, Switzerland: World Health Organization; 1999:14.
1 dose of TT in previous pregnancies††
Yes 0.04 (0.004–0.4*) 2. World Health Organization. Neonatal tetanus: progress towards the
No 1.0 global elimination of neonatal tetanus, 1990–1997. Available at http://
* Statistically significant at 0.05 level.
www.who.int/vaccines-diseases/diseases/NeonatalTetanus.shtml.
† Clean = only new plastic sheet or operating theatre; any other = unclean. 3. World Health Organization Regional Office for Africa. Neonatal teta-
§ Skilled = doctors, midwives, and nurses; any other = unskilled. nus cases and deaths drastically under-reported in West Africa. Vac-
¶ Answered “yes” to question, “Did you apply any substance into the cine Preventable Diseases Bulletin 2000;6:2.
vagina at onset of labor?” 4. United Nations Children’s Fund. Tetanus question and answer book-
** Received >2 tetanus toxoid doses during study pregnancy.
†† Answered “yes” to question, “Were you ever immunized with tetanus let for leaders. Kampala, Uganda: United Nations Children’s Fund;
toxoid in previous pregnancies?” 2003:12.
5. Ministry of Health and National Expanded Program for Immuniza-
tion. A field guide for maternal and neonatal tetanus elimination in
Uganda. Kampala, Uganda: Uganda National Expanded Program for
nondifferential. Third, recall bias might have occurred because
Immunization; 2002:1–45.
mothers of neonates with NNT are likely to have a better 6. Musoke GM. Epidemiology of tetanus in Busoga [Dissertation].
recall of circumstances surrounding delivery and cord care than Kampala, Uganda: Makerere University; 1975.
control mothers. Birth-date matching attempted to equalize 7. Stanfield JP, Galazka A. Neonatal tetanus in the world today. Bull World
recall difficulties for both mothers of cases and controls. Health Organ 1984;62:647–69.
Fourth, NNT status of participants could not be masked; how- 8. Bennett J, Macia J, Traverso H, Banoagha S, Malooly C, Boring J.
Protective effects of topical antimicrobials against neonatal tetanus.
ever, nonmedical data collectors were employed to minimize Int J Epidemiol 1997;26:897–903.
interviewer bias. Finally, only 60% of mothers of neonates 9. Gordis L. Epidemiology. 2nd ed. Philadelphia, PA: WB Saunders Com-
with NNT could be traced, which reduced study sample size. pany; 2000:148–51.
However, the matched case-control study design with four
controls per case minimized the effect of this limitation.
30 MMWR April 28, 2006

10. Quddus A, Luby S, Rahbar M, Pervaiz Y. Neonatal tetanus: mortality 14. Davies-Adetugbo AA, Davies-Adetugbo AA, Torimiro SEA, Ako-Nai
rate and risk factors in Loralai District, Pakistan. Int J Epidemiol KA. Prognostic factors in neonatal tetanus. Trop Med Int Health
2002;31:648–53. 1998;3:9.
11. Hlady WG, Bennette JV, Samadi AR, et al. Neonatal tetanus in rural 15. Idema CD, Harris BN, Ogunbanjo GA, Durrheim DN. Neonatal teta-
Bangladesh: risk factors and toxoid efficacy. Am J Public Health nus elimination in Mpumalanga Province, South Africa. Trop Med
1992;82:1365–9. Int Health 2002;7:622–4.
12. Babaniyi O, Parakoyi B. Cluster survey for poliomyelitis and neonatal 16. Leroy O, Garenne M. Risk factors of neonatal tetanus in Senegal. Int
tetanus in Ilorin, Nigeria. Int J Epidemiol 1991;20:515–20. J Epidemiol 1991;20:521–6.
13. Buekens P, Tsui A, Kotelchuck M, Degraft-Johnson M. Tetanus
immunization and prenatal care in developing countries. Int J Gynaecol
Obstet 1995;48:91–4.
Vol. 55 / Supplement MMWR 31

Risk Factors for Brucellosis — Leylek and Kadamjay Districts,

Batken Oblast, Kyrgyzstan, January–November, 2003
Turatbek B. Kozukeev,1,2 S. Ajeilat,1 E. Maes,3 M. Favorov4
1Applied Epidemiology Training Program, Central Asia Regional Program, CDC; 2State Department of Sanitary-Epidemiological Surveillance, Kyrgyzstan;
3Coordinating Office for Global Health, CDC; 4Central Asia Regional Program, CDC

Corresponding author: Michael O. Favorov, 1600 Clifton Rd, MS E-93, Atlanta, GA 30333. Telephone: 404-498-6070; Fax: 404-498-6065; E-mail: MOF0@cdc.gov.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: Brucellosis is a zoonotic disease that is associated with chronic serious sequelae in humans. During 1997–2002,
the reported incidence of human brucellosis in Kyrgyzstan increased nearly twofold, from 20 to 36 per 100,000 popula-
tion. In 2002, the highest incidence of brucellosis was reported in two rural districts of Batkan Oblast: Leylek (106 per
100,000 population) and Kadamjay (80 per 100,000 population). During January–November 2003, trainees from the
Applied Epidemiology Training program in Central Asia conducted a matched, hospital-based, case-control study to identify
risk factors for brucellosis and describe the epidemiology of disease in these two districts.
Methods: Brucellosis cases were defined on the basis of epidemiologic, clinical, and laboratory criteria. During January–
November 2003, a total of 100 persons with confirmed brucellosis were identified in the infectious disease wards of the two
district hospitals; these persons were matched by age and date of admission to 100 controls who were admitted to other hospital
wards for unrelated conditions. Data on socioeconomic and occupational factors and history of exposure to animals and
animal products were collected by using a structured questionnaire. Conditional logistic regression was used to study the
association between exposure variables and brucellosis.
Results: Among the 100 persons with confirmed brucellosis during the study period, 86 (86%) owned farm animals, and
45 (45%) became ill during April–May, the birthing season for farm animals. Multivariate analysis indicated that
brucellosis was associated with exposure to aborted farm animals in the household (odds ratio [OR] = 29.8; 95% confidence
interval [CI] = 4.4–203.4) and consumption of home-made milk products obtained from bazaars or neighbors (OR =
11.4; CI = 1.6–83.9). Knowledge of the mode of brucellosis transmission appeared to be protective against disease transmis-
sion (OR = 0.2; CI = 0.03–0.8).
Discussion: Exposure to aborted home-owned animals and consumption of home-made milk products obtained from bazaars
or neighbors were identified as probable sources of human brucellosis infections in the study districts. This finding suggests that
brucellosis spreads among farm animals in the area and that home-made milk products are not adequately pasteurized.
Conclusion: To reduce the burden of brucellosis in Batken Oblast, veterinary services should be improved, and health
education programs should be increased. Implementing these measures should minimize exposure to farm animals and reduce
the risk for infection from locally produced milk products.

Introduction former Soviet Union. An estimated 3,180,000 (65%) persons

live in rural areas. Agriculture accounts for 44% of the country’s
Brucellosis is an infectious zoonotic disease that is associ-
gross domestic product (GDP), and half of the country’s agri-
ated with chronic debilitating infections in humans and
cultural GDP comes from the livestock subsector (2). Animal
reproductive failure in domestic animals (1). Although bru-
husbandry is the dominant form of agricultural production
cellosis occurs worldwide, it is more common in countries
in the mountainous regions of Kyrgyzstan.
that do not have effective public health and domestic animal
During the Soviet era, the collective-farm system that
health programs.
existed in the country accounted for the majority of livestock
Kyrgyzstan (estimated 2003 population: 4,892,808) is an
production; a high level of state veterinary and sanitary con-
almost entirely mountainous country in Central Asia that
trol was maintained over the livestock population and pro-
became independent in 1991 after the disintegration of the
32 MMWR April 28, 2006

duction. The collapse of the Soviet Union in 1991 led to Methods

social changes and economic hardship that affected all gov-
During January–November 2003, in response to a request
ernment sectors. The state farming system disappeared, forc-
from the Kyrgyz MoH, trainees from the Applied Epidemiol-
ing families, especially those in rural areas, to raise farm animals
ogy Training Program in Central Asia conducted a hospital-
themselves. These changes in the agricultural sector resulted
based, matched case-control study in Leylek and Kadamjay
in the loss of the state control over the livestock population
districts of Batken Oblast to identify risk factors for brucello-
and a drastic decline in the quality of veterinary services. In
sis and describe the epidemiology of disease in these districts.
addition, as a result of deteriorating economic conditions,
During the study period, 100 persons with confirmed cases
Kyrgyzstan was unable to maintain the health infrastructure
of brucellosis were identified in the infectious disease wards
inherited from the Soviet era, and the quality and availability
of the two district hospitals. A case was defined as illness
of health services declined, especially in rural areas, where
occurring in a permanent resident of one of the two study
neither veterinary nor sanitary inspection agencies provide
districts who was admitted to either of the two local hospitals
adequate health services (3,4).
and who received a clinical diagnosis of brucellosis for the
Brucellosis can be transmitted to humans through contact
first-time on the basis of the following symptoms: fever
with animals or their products; it is an occupational hazard to
(>99.5°F [>37.5°C]) for >5 days, headache and arthralgias,
persons engaged in certain professions (e.g., veterinarians,
and Wright agglutination test dilution >1:100. A control was
slaughterhouse workers, and farmers) (5–7). During 1997–
defined as a patient with no recent history of brucellosis who
2002, countrywide incidence of human brucellosis increased
was a permanent resident of one of the two study districts and
substantially, from 20 to 36 cases per 100,000 population
who was admitted to the same hospital with a noninfectious
(Kyrgyz Ministry of Health [MoH], unpublished data, 2003)
condition during the study period. One control per case was
(Figure 1); during the same period, the State Veterinary
selected and matched by age (±2 years) and date of admission
Department did not report an increase in incidence of brucel-
(<2 weeks of the brucellosis patient’s date of hospitalization).
losis in animals. The greatest increase in incidence during this
The Schlesselman formula for matched case-control design
5-year period (from six to 68 cases per 100,000 population)
was used to calculate sample size, with the following param-
was reported in Batken Oblast, a mountainous area in the
eters: the significant level was set at 0.05, the desired power
southwestern part of the country. In two districts of this oblast,
was 0.8, and the case-to-control ratio was 1:1. To adjust for
Leylek and Kadamjay, incidence of human brucellosis in 2002
confounding, 10% was added to the result. The total number
was 106 and 81 per 100,000 population, respectively.
of pairs needed was 100.
Informed consent was obtained from
all study participants, and person-to-
FIGURE 1. Reported rate* of brucellosis, by year — Leylek and Kadamjay districts, person interviews were conducted. Data
Batken Oblast, Kyrgyzstan, 1990–2002
on socioeconomic and occupational fac-
120 tors and on exposure to animals and ani-
mal products were collected by using a
100 Kyrgyzstan structured questionnaire. Clinical and
Leylek laboratory data were abstracted from
80 medical records. A period of 2 months
Kadamjay
Rate

before date of illness onset for cases and

60 the same period for corresponding con-
trols was used as the relevant time for
40 measuring these variables.
Data management and analysis were
20 performed using Epi Info, version 3.2
(CDC, Atlanta, Georgia). Conditional
0 logistic regression was used to study the
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
association between exposure variables
Year and brucellosis. Potential risk factors
* Per 100,000 population. were first assessed in the univariate
analysis. All variables with p values of
Vol. 55 / Supplement MMWR 33

<0.2 in univariate analysis were selected for multivariate analy- Multivariate analysis indicated that exposure to aborted home-
sis. The stepwise method of model building was used to arrive owned animals (OR = 29.8; CI = 4.4–203.4), exposure to home-
at the final statistical model. A p value of <0.05 was defined as made milk products purchased in bazaars or from neighbors
statistically significant. (OR = 11.4; CI = 1.6–83.9), and being of Kyrgyz nationality
(OR = 4.8; CI = 1.2–20.3) were independent risk factors for
having brucellosis. Knowledge of the mode of brucellosis trans-
Results mission was protective (OR = 0.2; CI = 0.03–0.8). Owning a
Of 100 persons with brucellosis who participated in the cow at home was borderline associated with brucellosis (OR = 4.5;
study, 69 (69%) became ill during February–May; with the CI = 0.9–23.5) (Table 2).
highest number of illnesses beginning during late April–early
May (Figure 2). The age of ill persons varied (range: 1–75
years; mean: 32 years); 66 were males and 87 were of Kyrgyz Discussion
nationality (Table 1). Twelve patients were collective farm Although brucellosis can occur any time, the majority of
workers; 86 patients reported owning farm animals at home, cases occurred during February–May 2003, which is the
including cattle (n = 80), goats (n = 70), dogs (n = 50), and birthing season for farm animals. In this season, Brucella can
sheep (n = 49). Among the 86 persons with brucellosis who be excreted in high numbers from infected animals, and the
had farm animals at home, 70 (81%) cleaned barns, 60 (70%) likelihood of exposure to infected animals increases. Similar
assisted in animal delivery, and 22 (26%) slaughtered animals trends have been observed in other countries (7,8).
(Table 1). Humans become infected with Brucellae by coming into
Univariate analysis indicated that exposure to aborted ani- contact with animals or animal products that are contami-
mals (OR = 16.0; CI = 4.5–99.2), assistance in animal deliv- nated with these bacteria. Inadequately heated milk and other
ery (OR = 5.6; CI = 2.3–16.3), keeping goats at home (OR = dairy products from infected animals are a primary source of
5.2; CI = 2.1–15.2), cleaning animal barns (OR = 3.0; infection (6,9,10). Person-to-person transmission of Brucella
CI = 1.1–9.2) and being of Kyrgyz nationality (OR = 2.4; is extremely rare (1,10). In the two districts studied, exposure
CI = 1.1–5.6) were significantly associated with risk for hav- to aborted home-owned animals and eating home-made milk
ing brucellosis; limited evidence in the univariate analysis sug-
products purchased in bazaars or from neighbors were inde-
gested that milk products purchased in bazaars or from pendent risk factors for having brucellosis. Being of Kyrgyz
neighbors (OR = 1.7; CI = 0.9–3.2) might be associated with nationality was associated with brucellosis; this might be
brucellosis. explained by residual confounding or to other unmeasured
risk. Knowledge of the mode of brucellosis transmission was
protective, which underscores the
FIGURE 2. Number of brucellosis cases,* by week of onset — Leilek and Kadamjay importance of health education in pre-
districts, Batken Oblast, Kyrgyzstan, January–October 2003 venting of brucellosis. Increased govern-
18 ment efforts are necessary to improve
education regarding risk factors for bru-
16
cellosis transmission, especially in rural
14 areas, where human contact with
12 domestic animals is widespread.
Brucellosis in Batken Oblast is prob-
Number

10
ably spread among home-owned ani-
8 mals, and home-made milk products
6 obtained from bazaars or neighbors
4
probably are not adequately pasteur-
ized. The spread of brucellosis in farm
2 animals might be attributable to the
0 privatization of collective farms as a
1 15 29 12 26 12 26 9 23 7 21 4 18 2 16 30 13 27 10 24 8 22
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
result of the changed political and eco-
nomic situation in the country. Collec-
Week beginning
tive farm animals were distributed
* N = 100. among small private farms; families who
34 MMWR April 28, 2006

TABLE 1. Number and percentage of persons with brucellosis, TABLE 2. Risk factors for human brucellosis — Leylek and
by selected characteristics — Leylek and Kadamjay districts, Kadamjay districts, Batken Oblast, Kyrgyzstan, January–
Batken Oblast, Kyrgyzstan, January–November 2003 November 2003
Characteristic (n = 100) No. (%) Adjusted
Sex Risk factors OR* (95% CI†)
Male 66 (66) Kyrgyz nationality 4.8 (1.2–20.3)
Female 34 (34) Kept cows 4.5 (0.9–23.5)
Nationality Kept goats 1.6 (0.4–6.5)
Kyrgyz 87 (87) Cleaned animal barns 3.3 (0.6–17.3)
Other 13 (13) Sheared animals 0.4 (0.1–2.0)
Education Slaughtered animals 0.7 (0.1–3.1)
No formal 4 (4) Was exposed to aborted animals
Primary 19 (19) in the household 29.8 (4.4–203.4)
Secondary 61 (61) Milked cows 0.5 (0.1–2.6)
Secondary special* 14 (14) Obtained milk products from bazaar
or neighbor 11.4 (1.6–83.9)
Postsecondary 2 (2)
Knew how brucellosis is acquired 0.2 (0.03–0.8)
Type of domestic animal owned (n = 100)
* Odds ratio.
Cattle 80 (80) † Confidence interval.
Goat(s) 70 (70)
Dog(s) 50 (50)
Sheep 49 (49) protective clothes, especially when assisting in delivery; not per-
Any 86 (86) mitting children to have contact with animals; and having sick
Relative distribution of cases by type animals checked by a veterinarian) during contact with animals
of exposure to animals (n = 86)
Cleaned barns 70 (81)
and adherence to adequate sanitary standards (e.g., boiling or
Assisted during delivery 60 (70) pasteurizing) when processing milk and milk products. Brucel-
Exposed to aborted animals in household 48 (56) losis health education brochures and flyers are being distrib-
Milked cows 29 (34) uted at infectious disease hospitals, local clinics, and health
Sheared sheep 27 (31)
Slaughtered animals 22 (26) outposts. In addition, MoH has applied for a grant from the
* Combines general education with specialized training in a professional field. World Bank to fund brucellosis prevention efforts.
References
own these animals might disregard or not be aware of sanitary 1. Corbell JM. Brucellosis: an overview. Emerg Infect Dis 1997;2:213–21.
and health requirements necessary to prevent transmission of 2. World Bank. Kyrgyz Republic–Sheep Development Project. Available
brucellosis to humans. A probable consequence of privitization at http://lnweb18.worldbank.org/ECA/ECSSD.nsf/ProjectProfiles/
749CEF035AFA61F885256B07000EC174.
of the animal sector is an increased volume of home-made
3. Swiss Agency for Development and Cooperation. Kyrgyz-Swiss Health
animal food products with inadequate sanitary control over Reform Support Project. Available at http://www.swisscoop.kg/index.
production. php?navID=22085.
Despite the increase in the number of registered human bru- 4. United Nations Office for the Coordination of Humanitarian Affairs.
cellosis cases, the State Department of Veterinary Services did Kyrgyzstan: focus on brucellosis in south. Available at http://www.
not report an increase in the number of brucellosis cases among irinnews.org/report.asp?ReportID=37604&SelectRegion=Central_Asia
&SelectCountry=KYRGYZSTAN%3CBR%3E.
animals during the same period. This might reflect the inad-
5. Martínez MC, Jiménez PA, Blanco CM, et al. Brucellosis outbreak
equacy of veterinary services. In addition, persons who own due to unpasteurized raw goat cheese in Andalucia (Spain), January–
their own animals might not seek veterinary service if needed March 2002. Euro Surveill 2003;8:164–8. Available at http://212.234.
because of fear that they might lose their source of income. 146.164/em/v08n07/0807-223.asp.
These findings were discussed by MoH staff and were pre- 6. Issa H, Jamal M. Brucellosis in children in south Jordan. World Health
sented at conferences attended by staff of international agencies Organization Eastern Mediterranean Health Journal 1999;5:895–902.
Available at http://www.emro.who.int/Publications/EMHJ/0505/05.htm.
(e.g., the World Bank and the Swiss Emergency Relief Agency).
7. Chomel BB, DeBess EE, Mangiamele DM, et al. Changing trends in
Public health and veterinary officials in Kyrgyzstan have deter- the epidemiology of human brucellosis in California from 1973 to 1992;
mined that the best approach to reduce the country’s brucello- a shift toward foodborne transmission. J Infect Dis 1994;170:1216–23.
sis burden is to focus their resources on improving health 8. Taylor PM, Perdue JN. The changing epidemiology of human brucel-
education. Together with the nongovernment organization Rural losis in Texas, 1977–1986. Am J Epidemiol 1989;130:160–5.
Activists for Health, MoH has developed a prevention cam- 9. Hartigan P. Human brucellosis: epidemiology and clinical manifesta-
tions. Irish Veterinary Journal 1997;50:179–80.
paign that targets persons on rural privately owned farms with
10. Corbell JM. Brucellosis: epidemiology and prevalence worldwide. In:
livestock. Prevention messages are delivered through the mass Young EJ, Corbell JM, eds. Brucellosis: clinical and laboratory aspects.
media and focus on use of protection methods (e.g., wearing Boca Raton, FL: CRC Press; 1989:26–37.
Vol. 55 / Supplement MMWR 35

Salmonellosis Outbreak Among Factory Workers —

Huizhou, Guangdong Province, China, July 2004
Lunguang Liu,1,2 H.F. He,3 C.F. Dai,2 L.H. Liang,3 T. Li,2 L.H. Li,2 H.M. Luo,2 R. Fontaine4
1Chinese Field Epidemiology Training Program, Beijing, China
2Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, China
3Huizhou City Center for Disease Control and Prevention, Huizhou, Guangdong Province, China
4Division of International Health, Coordinating Office for Global Health, CDC

Corresponding author: Lunguang Liu, Sichuan Center for Disease Prevention and Control, 40 Huaishu Street, Chengdu, Sichuan 610031, China. Telephone:
852-28-86770124; Fax: 852-28-86770121; E-mail: liulunguang@vip.sina.com.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: During July 9–14, 2004, an outbreak of gastroenteritis occurred among workers at an electronics factory in
Huizhou, Guangdong Province, China; 199 cases were reported. A case-control investigation was initiated to identify the
agent and the mode of transmission.
Methods: Stool samples were collected from 142 workers and food handlers and cultured for enteric pathogens. A question-
naire concerning meals and foods eaten in the factory cafeteria during July 11–13 was administered to 92 ill workers and 100
controls.
Results: Of approximately 2,000 workers who worked during the outbreak, 197 (10%) had illness consistent with the case
definition. Salmonella enteritidis was identified from 44 (31%) of 142 stool samples collected from ill workers. Ill workers
were more likely than controls to have eaten breakfast in the factory cafeteria during July 11–13. Of eight foods served at
breakfast in the factory cafeteria, three were associated with illness: cake, bread (on July 12 only), and congee (i.e., rice
porridge). Stratification of bread and congee exposure by cake consumption indicated that only bread eaten on July 12 was
associated with gastroenteritis. The cake was baked on July 11, and a mixture that included raw eggs was poured on top; the
cake was then stored at room temperature and served for breakfast on 3 consecutive days (July 11–13). The bread was stored
together with the cake on July 11 and 12. No leftover food was tested.
Discussion: The investigation indicated that an outbreak of S. enterica serotype Enteritidis resulted from consumption of an
unusual food vehicle (i.e., cake) that had been contaminated from a more typical source (i.e., raw eggs). The bread was stored
at room temperature together with cake on which a mixture made from raw eggs had been poured. The bread was probably
contaminated by contact with the cake.
Conclusion: Food handlers should be instructed that intact fresh eggs can harbor S. enteritidis, foods made from eggs must be
cooked, and prepared food must be stored under refrigeration.

Introduction Methods
On July 13, 2004, a local hospital reported to the Huizhou A case of gastroenteritis was defined as illness occurring in
City Center for Disease Control (CDC) that approximately either a factory worker or in a family member who ate a meal at
70 patients had been admitted that day to the hospital with the factory cafeteria during July 7–16 and who had diarrhea
gastroenteritis characterized by diarrhea, fever, nausea, and (two or more liquid stools per day) or fever >99.5ºF (>37.5ºC)
vomiting. All 70 patients were workers at a local factory that in addition to one of three symptoms: nausea, vomiting, or
manufactured electrical products. Investigators from the abdominal pain. Ill workers were identified from doctors’
Huizhou and Guangdong CDCs visited the hospital and the reports and from an announcement to workers. Stool samples
factory to identify the agent causing this outbreak. Initial were collected from 142 patients (136 workers and six food
interviews and review of medical records indicated that for handlers) and 24 healthy food handlers. Stool samples were
the majority of patients, onset of gastroenteritis occurred dur- inoculated to Shigella-Salmonella agar and common agar cul-
ing a 2-day period (July 12–13). ture of enteric bacterial pathogens, including Shigella, Salmonella,
36 MMWR April 28, 2006

and Escherichia coli. Slide agglutination tests were used to iden- FIGURE. Number* of cases of gastroenteritis among factory
workers, by time† and date of illness onset — Huizhou City,
tify the suspected bacteria and later the serotype of Salmonella. Guangdong Province, China, July 9–14, 2004
As the investigation proceeded, other workers were learned to
have sought treatment at a second local hospital. 14
12
A case-control study was conducted to identify the prob-
10
able vehicle of transmission; 92 ill workers from the two hos-

Number
8
pitals were selected as a convenience sample, and 100 factory
6
workers with no symptoms of gastroenteritis were selected as
4
controls. Controls lived in the same dormitory or were 2
employed in the same workshop as the ill patients. Ill workers 0
and controls shared similar demographic characteristics. Dur- 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1
9 10 11 12 13 14
ing July 13–15, the workers were administered a standard
Jul
questionnaire that collected demographic information and
Time and date
asked to state where and what they had eaten at each meal
during July 11–13. Because patients reportedly ate all their * N = 92.
†1 = 12:00 a.m.–3:59 a.m.; 2 = 4:00 a.m.–7:59 a.m.; 3 = 8:00 a.m.–11:59 a.m.;
recent meals in the factory cafeteria and drank only boiled 4 =12:00 p.m.–3:59 p.m.; 5 = 4:00 p.m.–7:59 p.m.; and 6 = 8:00 p.m.–11:59 p.m.
water, the investigation focused on foods served in the cafete-
ria. Respondents were asked to identify which foods they had days was not associated with illness. Of 92 ill workers, 84
eaten for breakfast, lunch, and dinner and to report symp- (91%) ate breakfast at least once during July 11–13. Time
toms of gastroenteritis. Frequencies of exposure of ill workers intervals between eating breakfast and onset of illness varied
and controls were compared for all meals. For meals associated (range: 12–54 hours).
with illness, exposures of different foods served were compared Eight food items were served each day at breakfast: cake,
among patients and controls. Respondents who did not answer bread, congee (i.e., rice porridge), steam bread, meat-stuffed
a specific question were not included in the analysis for that steam patties, fried rice, fried noodles, and boiled noodles.
exposure. In addition, a field survey was conducted to identify Eating cake, bread, or congee was associated with illness
whether anything had happened in the factory cafeteria that (Table 2). These three foods were not served at either lunch or
was associated with the outbreak, especially in the environment dinner. The same cake was served on all 3 days and had the
of the cafeteria and the food preparation process. highest OR (OR = 19.0; 95% confidence interval [CI] =
7.4–47.0), whereas bread was associated with illness only on
July 12 (OR = 8.9; CI = 2.9–27.0). On all 3 days, the OR for
Results congee was lower (OR = 3.6; CI = 2.0–6.6) than that for cake
Of approximately 2,000 factory workers who worked dur-
ing the 3-day outbreak, 197 (10%) had an illness consistent TABLE 1. Number and percentage of workers exposed who ate
with the case definition. An additional two family members meals at a factory cafeteria, by meal and date — Huizhou,
Guangdong Province, China, July 11–13, 2004
who also ate in the factory cafeteria became ill. Stool speci- No. of respondents % exposed
mens from 44 (31%) of 142 ill workers yielded Salmonella Meal/Date Case Control Case Control OR* (95% CI†)
enterica serotype Enteritidis. Of 92 patients interviewed in Breakfast
the case-control study, the most common clinical symptoms July 11 61 84 72 43 3.5 (1.7–7.0)
were diarrhea (80 [87%]), abdominal pain (72 [78%]), July 12 82 98 90 55 7.5 (3.3–17.0)
July 13 44 83 77 48 3.7 (1.6–8.3)
fever (61 [66%]), nausea (26 [28%]), and vomiting (17 Total 92 100 91 59 7.3 (3.0–18.0)
[18%]). Onset of illness time was clustered from midnight Lunch
July 11 through midnight July 13 (Figure). Patients’ ages July 11 58 82 64 80 0.4 (0.2–0.9)
varied (range: 17–50 years; median: 25 years) and reflected the July 12 79 93 77 77 1.0 (0.2–2.2)
July 13 28 64 61 67 0.8 (0.3–2.1)
age distribution of the factory workforce. Between one and Total 92 100 80 81 1.0 (0.4–2.1)
three cases occurred in each of the factory’s 100 dormitories, Dinner
each of which housed 15–20 workers. July 11 59 76 68 68 1.0 (0.4–2.1)
A comparison analysis indicated that eating breakfast July 12 67 87 73 78 0.8 (0.3–1.7)
July 13 18 56 33 66 0.3 (0.01–0.8)
in the factory cafeteria during July 11–13 was significantly Total 92 100 82 77 1.3 (0.6–2.8)
associated with having gastroenteritis (odds ratio [OR] * Odds ratio.
† Confidence interval.
range: 3.5–7.5) (Table 1). Eating lunch or dinner on these
Vol. 55 / Supplement MMWR 37

TABLE 2. Number and percentage of workers exposed to salmonellae At breakfast, the cake and bread were placed on the same
who ate breakfast at a factory cafeteria, by type of food eaten and
date — Huizhou, Guangdong Province, China, July 11–13, 2004
serving platter. After breakfast, the cake and bread were
No. of respondents % exposed
placed together in an open container, stored at room tem-
Food/Date Case Control Case Control OR* (95% CI†)
perature for another 24 hours, and served for breakfast on
Cake
July 12. The bread was completely eaten on July 12, but
July 11 27 32 37 6 8.8 (1.7–45.0) leftover cake was served on July 13. Congee was made fresh
July 12 69 53 41 8 8.4 (2.5–30.0) each morning by boiling rice in water for 1 hour and served
July 13 34 39 59 0 ∞ p<0.01
while still hot. No cake or bread remained for culture. All
Total 92 100 54 6 19.0 (7.4–47.0)
Bread eggs used to make the Sela oil were consumed before samples
July 11 27 31 22 10 2.7 (0.5–16.0) could be collected.
July 12 69 53 42 8 8.9 (2.9–27.0)
July 13 34 39 26 10 3.2 (0.8–14.0)
Salmonella enterica serotype Enteritidis was isolated from
Total 92 100 37 10 5.3 (2.4–12.0) the stool specimens of 10 (33%) of 30 food handlers. Of
Congee the 10 food handlers infected with S. enteritidis, six (60%)
July 11 27 78 31 48 3.7 (1.2–12.0)
July 12 69 67 53 43 2.6 (1.2–5.5) had gastroenteritis (attack rate: 20%). Swabs from the hands
Total 92 100 62 31 3.6 (2.0–6.6) of one food handler also yielded S. enteritidis.
Steam bread
July 11–13 92 100 15 10 1.6 (0.78–3.8)
Steam patty Discussion
July 11–13 92 100 12 12 1.0 (0.4–2.4)
Fried rice The investigation indicated that an outbreak of
July 11–13 92 100 20 11 2.0 (0.9–4.4) S. enterica serotype Enteritidis resulted from consump-
Fried noodle tion of an unusual food vehicle (i.e., cake) that had been
July 11–13 92 100 45 50 0.8 (0.5–1.4) contaminated from a more typical source (i.e., raw eggs).
Boiled noodle Sela oil poured on the cake was probably responsible for
July 11–13 92 100 13 15 0.9 (0.4–1.9)
* Odds ratio.
the initial contamination, and the cake was the source of
† Confidence interval. the infection. The high risk for illness from eating both
cake and congee, compared with the absence of risk for
(OR = 19.0; CI = 7.4–47.0). The bread and cake were baked eating congee alone, indicates that the cake was the source of
on the evening of July 10 and first served on July 11; they the infection. The congee might have had a synergistic effect
were then stored together in the same container. The bread on the occurrence of illness. The method of preparing congee
was used up after breakfast on July 12 and replaced with newly (boiling rice in water for 1 hour) would not permit salmo-
baked bread on July 13. Cake that was left over on July 12 nella survival to occur once, let alone in three batches on 3
was served again on July 13. Congee was prepared fresh daily consecutive days. However, congee has a relatively high pH
and was not in contact with the bread or cake. Flies were (6.4) and could neutralize stomach acid, thus permitting sal-
present in the kitchen, and food handlers seldom washed their monellae from cake to reach the small intestine. In addition,
hands while working. congee could speed gastric emptying, leading to more of the
To understand how three different foods could be associ- ingested dose of organisms reaching the small intestine.
ated with a single outbreak, investigators stratified the analy- S. enteritidis was probably transferred from the cake to the
sis of these three foods, comparing them alone and in bread by contact with a common serving platter or utensil, the
combination to a single reference exposure (i.e., workers who hands of a kitchen worker, or flies. Storage for a sufficient time
had not eaten any of the three foods at any breakfast). This at room temperature then facilitated multiplication of
analysis indicated that cake eaten on any of the 3 days and S. enteritidis on the bread. Cross contamination without multi-
bread served on July 12 were associated with illness (Table 3). plication of S. enteritidis probably did not occur because this
Congee alone had no association with illness when eaten with- would have caused infection on any or all of the 3 days that the
out cake but an OR of 53 if eaten with cake. cake was served.
The cake and bread were baked at midnight July 10. Sela oil The key event in the outbreak was the preparation of the
(a mixture prepared from raw eggs, sugar, vinegar, and water) Sela oil. After initial contamination from either raw eggs or
was poured over the baked cake. On July 10 at midnight, the another source, both suitable temperature and nutrition were
Sela oil was heated at approximately 104ºF (40ºC) and stored available for 3 hours for salmonellae to multiply before appli-
at room temperature (77 º F–86 º F [25 º C–30 º C]) for cation to the surface of the cake. Once the oil was on the cake,
3 hours before it was poured on the cake. Then the cake was suitable temperatures for growth of S. enteritidis continued
stored at room temperature for another 4 hours until breakfast. for another 4 hours. By the time the cake was sold at breakfast
38 MMWR April 28, 2006

TABLE 3. Stratified analysis comparing cake eaten on any had consumed. Third, because of time constraints, ill patients
of 3 days to bread eaten on July 12 and congee eaten during
July 11–13, 2004 — Huizhou City, Guangdong Province, China
and controls were selected by convenience and might not be
Cake and bread eaten on July 12
representative of all persons. Finally, other possible routes of
Cake Bread Case Control OR* (95% CI†) contamination (e.g., food handlers) could not be excluded.
Yes Yes 9 1 37 (4.4–826.0)
Yes No 41 5 34 (11.0–111.0)
No Yes 20 3 28 (6.9–129.0)
Conclusion
No No 22 91 Reference As a result of this investigation, changes were suggested in the
Cake and congee eaten during July 11–13 methods of preparing Sela oil or any other sauce made from
Cake Congee Case Control OR* (95% CI†)
uncooked eggs. Heating the Sela oil to pasteurization tempera-
Yes Yes 39 2 53 (11.0–343.0)
Yes No 11 4 7.5 (1.9–31.0)
tures (132.8ºF [56ºC]) for 30 minutes during preparation and
No Yes 18 29 1.7 (0.7–3.8) storage of the prepared Sela oil or sauce under refrigeration should
No No 24 65 Reference reduce the risk for salmonellosis. Food handlers should be
* Odds ratio. instructed that intact fresh eggs might harbor S. enteritidis.
† Confidence interval.
Because direct use of raw eggs without cooking might cause sal-
on July 11, it likely was contaminated. Additional incubation monella contamination of food, food handlers should apply these
of the same cake for 24–48 hours is consistent with the con- guidelines to all foods made from eggs and store foods prepared
tinuing association of illness with the cake for 3 days. with eggs under refrigeration. Researchers and health-care work-
The original contamination of the Sela oil with S. enteritidis ers should investigate whether such outbreaks occur more
probably came from the raw eggs. Intact shell eggs are known frequently than has been reported previously.
to become infected with S. enteritidis by vertical transmission References
(1–3). S. enteritidis is the only salmonella in chicken embryo 1. Gast RK, Beard CW. Detection and enumeration of Salmonella enter-
that can be found inside eggs. One infected egg typically con- itidis in fresh and stored eggs laid by experimentally infected hens.
tains 10–20 S. enteritidis bacterial cells, but only a few eggs per J Food Protect 1992;55:152–6.
1,000 or even 10,000 become infected (3). Infected eggs are 2. Humphrey TJ, Baskerville A, Mawer S, Rowe B, Hopper S. Salmonella
enteritidis phage type 4 from the contents of intact eggs: a study
the most likely source of this outbreak. Other less likely pos-
involving naturally infected hens. Epidemiol Infect 1989;403:415–23.
sible sources of infection are a food handler having contami- 3. Henzler DJ, Kradel DC, Sischo WM. Management and environmen-
nated hands, flies, or other ingredients in the Sela oil. tal risk factors for Salmonella enteritidis contamination of eggs. Am J
Transmission of S. enteritidis in association with intact chicken Vet Res 1998;59:824–9.
eggs was first recognized in the United States in 1986 (4,5). 4. Angulo FJ, Swerdlow DL. Salmonella enteritidis infections in the United
Since then, it has been recognized worldwide (6,7). This is the States. J Am Vet Med Assoc 1998;213:1729–31.
5. Angulo FJ, Swerdlow DL. Epidemiology of human Salmonella enterica
first outbreak of S. enteritidis in China that has been attributed
serovar Enteritidis infections in the United States. In: Saeed AM, Gast RK,
to raw eggs. However, salmonella surveillance and other out- Potter ME, Wall PG, eds. Salmonella enterica serovar Enteritidis in
break reports suggest that S. enteritidis outbreaks occur more humans and animals. Ames, IA: Iowa State University Press; 1999:33–41.
commonly than has been recognized previously. At approxi- 6. Molbak K, Neimann J. Risk factor for sporadic infection with Salmo-
mately the same time as this outbreak, gastroenteritis outbreaks nella enteritidis, Denmark, 1997–1999. Am J Epidemiol 2002;156:
attributed to S. enteritidis occurred in two nearby cities (Hong 654–61.
Kong and Guangzhou City, Guangdong Province). Investiga- 7. Food Safety Authority of Ireland. Report on zoonoses in Ireland 2000
& 2001. Dublin, Ireland: Food Safety Authority of Ireland; 2004.
tion of the Hong Kong outbreak implicated a mango pudding Available at http://www. http://www.fsai.ie/publications/reports/zoo
made from raw eggs (8). The Guangzhou outbreak was not noses_report.pdf.
investigated. In Hong Kong, salmonella surveillance has indi- 8. Ka-wing A. A major food poisoning outbreak associated with raw eggs.
cated that S. enteritidis isolations doubled during 1995–2000 In: Proceedings of the Third TEPHINET Global Conference [Abstract].
whereas other serotypes remained stable (9). In 10 of China’s Beijing, China, November 8–12, 2004:177.
31 provinces that have conducted food surveillance, S. enteriti- 9. Yeung ST, Kam KM. Salmonella surveillance in Hong Kong. Public
Health and Epidemiology Bulletin 2001;10:62–7.
dis is among the top seven salmonella serotypes isolated (10). 10. Wang M, Ran L, Wang Z, Li Z. Study on national active monitoring
In 2000, S. enteritidis was a leading cause of foodborne salmo- for food borne pathogens and antimicrobial resistance in China 2001
nellosis in the United States (11,12). [Chinese]. Wei Sheng Yan Jiu 2004;33:49–54.
The findings in this report are subject to at least four limita- 11. Schroeder CM, Naugle AL. Estimate of illnesses from Salmonella
tions. First, no cake or eggs remained available for testing. Sec- enteritidis in eggs, United States, 2000. Emerg Infect Dis 2005;11:113–5.
12. Patrick ME, Adcock PM, Gomez TM, et al. Salmonella enteritidis
ond, certain workers could not recall exactly which foods they
infections, United States 1985–1999. Emerg Infect Dis 2004;10:1–7.
Vol. 55 / Supplement MMWR 39

Varicella Outbreak Among Primary School Students — Beijing, China, 2004

Huilai Ma,1 R. Fontaine2
1Chinese Field Epidemiology Training Program, Chinese Center for Disease Control and Prevention
2Division of International Health, Coordinating Office for Global Health, CDC

Corresponding author: Huilai Ma, Chinese Field Epidemiology Training Program, Chinese Center for Disease Control and Prevention, 27 Nanwei Road, Beijing,
China, 100050. Telephone: 86-10-63022556; Fax: 86-10-83171509; E-mail: huilaima@sohu.com.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other
association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled
use of commercial products or products for investigational use.

Abstract
Introduction: On June 9, 2004, a varicella outbreak was reported in a Beijing primary school affecting approximately 80%
of children in one preschool classroom. An outbreak investigation was initiated to identify factors contributing to the high rate
of transmission and to assess the effectiveness of control measures.
Methods: A varicella case was defined as onset of a generalized, vesicular pruritic rash lasting >4 days in a student at the
school during January 1–June 26, 2004. Parents of all students in the four lowest grades (K–2) were questioned concerning
varicella illness before January 1, 2004. Exposure and vaccination histories of 111 ill students and 120 control students with
no history of varicella were compared.
Results: During January 1–June 26, 2004, of 1,407 students, 138 (9.8%) had varicella; 488 (35%) K–2 students had no
history of varicella before the outbreak. In five classrooms in which attack rates (ARs) were high (>40%), a primary-case
student had remained in school 2 days while ill with a rash. The secondary attack rate (SAR) in these classrooms was 21%,
compared with 1.7% in classrooms in which the first ill student was sent home immediately (risk ratio [RR] = 10; 95%
confidence interval [CI] = 3.7–29.0). A total of 111 (70%) ill students rode the school bus daily, compared with 120 (33%)
control students (odds ratio [OR] =4.9; CI = 2.7–9.0). A total of 73 (33%) ill students had a history of varicella vaccination
before January 1, 2004, compared with 32 (69%) control-students (OR = 0.22; CI = 0.08–0.59).
Conclusion: Students who were not excluded from school on the first day of rash were key contributors to the spread of
varicella in their classrooms. High susceptibility to varicella at school entry indicates that vaccination of susceptible students
might be the only effective measure to control this recurrent problem.

Introduction On June 9, the Chinese Center for Disease Control and

Prevention (China CDC) was notified of a varicella outbreak
Varicella has been a nearly universal disease of childhood.
in a primary school in a suburb of Beijing, affecting approxi-
In temperate climates, nearly 90% of children had varicella
mately 80% of children in certain classes. Since the school
by age 15 years before vaccine became available (1). Before
opened in 1999, varicella outbreaks have disrupted activities
1995, when vaccine was licensed, an estimated 4 million cases
every year, despite a school policy to isolate students with a
of varicella occurred each year in the United States; approxi-
rash at home for 7 days. An outbreak investigation was initi-
mately 100 patients died and 10,000 persons were hospital-
ated to identify factors contributing to the high rate of trans-
ized annually because of varicella and related complications
mission and assess the effectiveness of control measures.
(2). Because varicella is not a notifiable disease in China, sta-
tistics concerning hospitalization and death attributable to
varicella are not available. However, since January 1, 2004, Methods
varicella outbreaks have been reported to the public health
emergency reporting system. During January–June 2004, a A varicella case was defined as a generalized, vesicular pru-
total of 64 varicella outbreaks were reported, accounting for ritic rash in a school student lasting >4 days with onset during
20% of all communicable disease outbreaks reported in China; January 1–June 26, 2004. To identify cases, investigators
58 (91%) varicella outbreaks occurred in schools. Varicella reviewed records kept by the school doctor and absentee records
vaccine has been available commercially since 2003, but China kept by teachers, and teachers and doctors were questioned
has no programs to support routine varicella vaccination. concerning rash illness among students. A questionnaire was
40 MMWR April 28, 2006

sent to the parents of any student identified as having a rash had a secondary skin infection. Cases occurred in all eight
illness to collect information concerning the student’s symp- grades and in 21 (60%) of 35 classrooms. Higher ARs were
toms and the results of any physician consultation. Case reported among students aged 3–8 years in the four lower
determination was based on the results of all three sources. A grades (K–2) (Table). Although cases began to appear in Janu-
primary case was defined as the first varicella case in a student ary, the outbreak was not evident until mid-April, peaked in
with onset of rash to appear in a classroom. A secondary case May, and ended with the closing of school on June 30. Dis-
was defined as any varicella case in a student with onset of tinct peaks of cases occurred at 15-day intervals during April–
rash 11–19 days after the onset of a primary case in the same June (Figure 1). Over the course of this outbreak, teachers’
classroom. A coprimary case was defined as a varicella case in records indicated that 1,090 days of absenteeism resulted from
a student with onset of rash <10 days after occurrence of a varicella illness, equivalent to 8 days of school missed per ill
primary case in the same classroom. Secondary attack rates child.
(SARs) were calculated by dividing the number of secondary Further analysis was limited to 488 (77%) of 635 K–2 stu-
cases by all students with no history of current or previous dents in 15 classrooms who did not have varicella before Janu-
varicella minus those with primary and coprimary cases. SARs ary 1, 2004. The AR in this group was 25%; analysis of ARs
for classrooms in which the student with the primary case by classroom identified two distinct groups: 10 classrooms
had been isolated immediately were compared with SARs for with ARs <15% and five classrooms with ARs that were sub-
classrooms in which the student with the primary case had stantially higher (40%–80%). In all classrooms with ARs
not been promptly isolated. >40%, one or more ill students had remained in school >2
A case-control study was conducted; participants included days while ill with a rash. These classrooms had new teachers
all 123 students with varicella in grades K–2 and 123 controls who were not familiar with the school’s isolation policy. In
selected randomly from all 365 students in grades K–2 with
no history of current or previous varicella. Self-administered TABLE. Number of varicella cases among primary-school
students, by grade levels — Beijing, China, January 1–June 26, 2004
questionnaires were sent to parents of all 635 students in grades
Age
K–2. Questionnaires requested demographic information, ill- group No. of No. of
ness characteristics, varicella vaccination history (including Grade (yrs) students cases AR* (%)
dates and place of vaccination), previous varicella disease his- K† 3–5 63 7 11.0
P§ 5–6 194 68 35.0
tory, and exposures to varicella. Parents were contacted by tele- 1st 6–7 213 16 7.5
phone to obtain missing information. 2nd 7–8 165 32 19.0
Responses were obtained for 111 (90%) of 123 students 3rd 8–9 210 3 1.4
with varicella in grades K–2. The exposure histories of these 4th 9–10 214 7 3.3
5th 10–11 185 3 1.6
students were compared with those of 120 control students. 6th 11–12 163 2 1.2
The number of controls selected for each classroom was pro- Total 3–12 1,407 138 9.8
portionate to the number of varicella cases that occurred in * Attack rate.
† Kindergarten.
the classroom (i.e., frequency matched). In two classrooms § Prefirst (between K and first grade).
with high attack rates (ARs), an insufficient number of con-
trol students was available, and additional control students
were selected at random from other classes in the same grade. FIGURE 1. Number of varicella cases detected among
children in a primary school, by date of rash onset — Beijing,
China, January 1–June 26, 2004
Results 40

During January 1–June 26, of 1,407 students, 138 (10%)

had a rash illness consistent with the case definition of vari- 30 Outbreak
investigation
cella. Cases were identified in three ways: 1) the school doctor begins
Number

20
examined 77 students and recorded their temperatures; 2) a Summer
vacation
doctor in the community examined another 42 students; and Winter vacation begins
10
3) 19 students were identified from responses to parent ques-
tionnaires. In addition to a characteristic rash, 64 (46%) stu- 0
dents had temperature >99.5ºF (<37.5ºC); of these, 36 (56%) 1 11 21 31 10 20 1 11 21 31 10 20 30 10 20 30 9 19 26
had a temperature >100.4ºF (>38ºC). Four (2.9%) students Jan Feb Mar Apr May Jun
Date
Vol. 55 / Supplement MMWR 41

each of these five classrooms, outbreaks began with a student was 21% (34 of 163) compared with 1.7% (four of 235) in
who was allowed to stay in school for >2 days while having a seven classrooms in which the first student with varicella rash
vesicular rash. Lax isolation continued during the second gen- was isolated immediately (in three classrooms, no cases
eration of cases in these classroom outbreaks, and tertiary cases occurred) (risk ratio [RR] = 10; 95% confidence interval
followed (Figure 2). [CI] = 3.7–29.0). In three classrooms in which a single stu-
In the five classrooms in which the student with the dent with a primary case was not isolated, the SAR was 26%
primary case was isolated only after >2 days of rash, the SAR (29 of 111) (RR = 12; CI = 4.4–34.0) compared with those

FIGURE 2. Number of varicella cases for five primary-school classrooms with attack rate (AR) of >40%, by date of rash onset —
Beijing, China, April 1–June 20, 2004

20 20
AR = 76% AR = 65%
15 15
Number

Number
10 10

5 5

0 0
1 11 21 1 11 21 31 10 20 1 11 21 1 11 21 31 10 20
Apr May Jun Apr May Jun
Date Date

20 20

AR = 56% AR = 46%
15 15
Number

Number

10 10

5 5

0 0
1 11 21 1 11 21 31 10 20 1 11 21 1 11 21 31 10 20
Apr May Jun Apr May Jun
Date Date

20

AR = 41%
15 No. of days students with case
isolated after rash onset
Number

<1 day after rash onset

10
1–2 days after rash onset
>2 days after rash onset
5

0
1 11 21 1 11 21 31 10 20
Apr May Jun
Date
42 MMWR April 28, 2006

classes for which cases were isolated immediately. In the two acquire varicella at school or in the community and subse-
classrooms with several coprimary cases, the SAR was 9.6% quently infect these more sequestered toddlers and preschoolers
(five of 52) compared with the classrooms with only isolated at home. Moreover, child care centers and other concentra-
cases (RR = 5.2; CI = 1.5–19). Finally, the five classes in which tions of susceptible infants and toddlers are rare in China.
a single student with a primary case was not isolated did not Under Chinese law, mothers receive 6 months of paid
differ from other classrooms regarding crowding, availability postmaternity leave to care for their infants. Thereafter, work-
of handwashing, activities involving close personal contact, ing parents customarily entrust their children to grandpar-
or the sharing of items that might act as fomites (e.g., towels, ents or other older relatives until age 3 years, when the children
eating utensils, and cups). are old enough to attend kindergarten. Consequently, the
Complete responses were obtained from 111 (90%) stu- possibility of exposure of preschool children to natural vari-
dents with varicella and from 120 (98%) control students. cella in the home or community is substantially reduced,
The case-control study in grades K–2 demonstrated that 78 resulting in high prevalence of susceptibility at school entry.
(70%) ill students used the school bus every day, compared Similar high rates of susceptibility to varicella might be wide-
with 39 (33%) control students (odds ratio [OR] = 4.9; CI = spread among students in lower grades in China. Vaccinating
2.7–9.0); 24 (33%) ill students had a history of varicella vac- susceptible students at school entry might be the most effec-
cination before January 1, 2004, compared with 22 (69%) tive control method. To be effective, vaccination coverage will
controls (OR = 0.2; CI = 0.1–0.6). This excludes responses need to be >95% because transmission might be sustained in
for 38 ill students and 88 controls whose parents could not schools with vaccination coverage of 67%–95%.
provide the date and place of varicella vaccination or were Other recommended measures include home isolation of
otherwise uncertain. Stratification of bus riding and vaccina- students with a rash (5,8). In this outbreak, prompt isolation
tion by classrooms or grade resulted in adjusted ORs that were was associated with lower ARs. However, a high SAR has been
not appreciably different from the crude OR and that indi- reported in a school despite universal isolation from class at
cated no difference in effect between high- and low-incidence rash onset; in that investigation, because ill students in all
classrooms (classroom-adjusted OR = 4.3; CI = 2.3–8.5 for classrooms were isolated promptly, the effect of isolation could
bus riding, and 0.24; CI = 0.1–0.7 for vaccination). Parent not be compared with the absence of isolation (9). Isolation is
questionnaires indicated that only two ill students were known only a temporary measure that simply delays but does not
to have had contact with an ill neighbor before onset of vari- prevent varicella illness (1).
cella rash. To stop this outbreak, investigators recommended adding
vaccination to isolation. However, because the investigation
was conducted near the end of the term, the school could not
Discussion implement this recommendation before the school year ended.
High ARs (40%–80%) in classes from which students with Because varicella outbreaks in schools are a common occur-
rash were not sent home immediately are characteristic of trans- rence in China, a more effective plan might be to vaccinate
mission among siblings in the home setting and in primary susceptible students when they begin school rather than to
schools before availability of vaccination (1). In this outbreak, isolate them after an outbreak has begun. Beginning in 2005,
two factors contributed to high ARs. First, 77% of K–2 stu- students at this school will be vaccinated at the beginning of
dents had no evidence of protection by natural immunity. the school year.
Second, ill students were not isolated promptly, leading to In 2003, the Chinese Food and Drug Administration
continuing exposure to varicella for several days and ARs licensed the Oka-strain varicella vaccine for children aged
>40%. The estimated 77% susceptibility for children aged >12 months (11). However, varicella vaccine is not included
3–8 years in this school is consistent with the 30% prevalence in national or provincial vaccination programs. A major
of varicella infection in children in Shanghai aged 3–6 years obstacle to widespread varicella vaccination in China is the
in 1998 (3). high cost of this vaccine, which is 10 times more expensive
A factor that could contribute to high susceptibility for than rubella vaccine and 75 times more expensive than measles
varicella illness is the number of families with only one child vaccine. As of 2002, only 131,730 (23%) children in Shang-
as a result of China’s one-child-per-family policy, which has hai aged <6 years had received varicella vaccine (10). Another
been in effect since 1979. By 2001, of approximately 291 possible contributing factor is that parents might not be aware
million families in China, approximately 130 million (44%) of the availability of vaccine or the need to isolate ill children
had only one child, and 102 million (35%) had two children at home.
(4). Preschool children often lack older siblings who could
Vol. 55 / Supplement MMWR 43

The findings in this report are subject to at least three limi- disruption of school activities would be valuable in assisting
tations. First, information on vaccination and previous his- provinces, educational authorities, and individual schools in
tory of varicella disease is subject to recall bias and other developing a vaccination policy.
reporting errors on the part of the parents. Vaccination his-
tory in particular suffered from a high nonresponse rate or Acknowledgments
responses of “can’t remember” or “unknown.” Accordingly, Xiaodong Sun, MBD, assisted with the initial interviews and data
only the protective effect is discussed and not vaccine effec- collection. Zhijie An, MBD, assisted with data analysis. James
tiveness or effects on classroom-specific ARs. Second, because Mendlein, PhD, assisted with revising the text. Jin Zhong Yang
94% of controls and all ill students participated in the study, facilitated the investigation in the school.
selection bias was likely minimal. Finally, why two classrooms References
in different grades had multiple coprimary cases is unknown. 1. Weller TH. Varicella–herpes zoster virus. In: Evans AS, Kaslow RA,
Exposure in these classrooms might have led to an earlier, mild eds. Viral infections of humans; epidemiology and control. New Haven,
primary case that was missed by the teachers and thus was not CT: Plenum Publishing; 1997:865–86.
2. CDC. Evaluation of varicella reporting to the national notifiable dis-
identified. Mild varicella with sparse rashes can occur in vac-
ease surveillance system—United States, 1972–1997. MMWR
cinated children; these mild infections can be transmitted to 1999;48:55–8.
others and have been implicated in transmission in other school 3. Sun HY, Yuan J, Wang S. Prevalence of varicella–zoster virus infection
outbreaks (5,7). When cases are missed by school staff and and immunogenicity of varicella virus vaccine in healthy people. Chin
parents, isolation measures are not taken. J Vaccine Immunization 2000;6:28–30.
4. State Family Planning Council. The third statistical bulletin of family
planning in 2002 [Chinese]. Beijing, China: State Family Planning
Council; 2002.
Conclusion 5. CDC. Outbreak of varicella among vaccinated children—Michigan,
Although no deaths or severe complications occurred in this 2003. MMWR 2004;53:389–92.
outbreak, varicella is a highly contagious disease that can be 6. Buchholz U, Moolenaar R, Peterson C, Mascola L. Varicella outbreaks
after vaccine licensure: should they make you chicken? Pediatrics
fatal (11). In this outbreak, the low level of naturally acquired
1999;104:561–3.
immunity was the principal factor underlying the high ARs 7. Galil K, Lee B, Strine T, et al. Outbreak of varicella at a day-care center
recorded. Failure to isolate ill students also contributed to high despite vaccination. N Engl J Med 2002;347:1909–15.
ARs in five classrooms. Vaccination was available and protec- 8. Lavanchy D. Chickenpox/herpes zoster. In: Heymann DL, ed. Con-
tive, but the decision to vaccinate was left to parents. trol of communicable diseases manual. 18th ed. Washington, DC:
Universal vaccination at school entry might be the only American Public Health Association; 2004:102–7.
9. Moore DA, Hopkins RS. Assessment of a school exclusion policy dur-
reliable method of controlling varicella. In addition to pre- ing a chickenpox outbreak. Am J Epidemiol 1991;133:1161–7.
venting deaths, vaccination might also reduce hospitalizations 10. Lina T, Jiguang Z. The epidemiologic analysis of varicella outbreaks in
from other varicella-related complications (12). More parents Shanghai since 2001–2002 [Chinese]. Sh J Prev Med 2004;16:13–4.
would be likely to have their children vaccinated if they 11. State Food and Drug Administration. The licensure of freeze-dried
understood the effectiveness of vaccine and the consequences live attenuated varicella vaccine [Chinese]. Beijing, China: State Food
and Drug Administration; 2003.
of varicella illness. A stronger policy to encourage varicella
12. CDC. Prevention of varicella: recommendations of the Advisory
vaccination in schools in China should be developed. Careful Committee on Immunization Practices (ACIP). MMWR 1996;45
assessment of the cost of vaccine compared with the cost of (No. RR-11).
 MMWR

The Morbidity and Mortality Weekly Report (MMWR) Series is prepared by the Centers for Disease Control and Prevention (CDC) and is available free of charge
in electronic format and on a paid subscription basis for paper copy. To receive an electronic copy each week, send an e-mail message to listserv@listserv.cdc.gov. The
body content should read SUBscribe mmwr-toc. Electronic copy also is available from CDC’s World-Wide Web server at http://www.cdc.gov/mmwr or from CDC’s
file transfer protocol server at ftp://ftp.cdc.gov/pub/publications/mmwr. To subscribe for paper copy, contact Superintendent of Documents, U.S. Government
Printing Office, Washington, DC 20402; telephone 202-512-1800.
Data in the weekly MMWR are provisional, based on weekly reports to CDC by state health departments. The reporting week concludes at close of business on
Friday; compiled data on a national basis are officially released to the public on the following Friday. Address inquiries about the MMWR Series, including material
to be considered for publication, to Editor, MMWR Series, Mailstop K-95, CDC, 1600 Clifton Rd., N.E., Atlanta, GA 30333; telephone 888-232-3228.
All material in the MMWR Series is in the public domain and may be used and reprinted without permission; citation as to source, however, is appreciated.
All MMWR references are available on the Internet at http://www.cdc.gov/mmwr. Use the search function to find specific articles.
Use of trade names and commercial sources is for identification only and does not imply endorsement by the U.S. Department of Health and Human Services.
References to non-CDC sites on the Internet are provided as a service to MMWR readers and do not constitute or imply endorsement of these organizations or
their programs by CDC or the U.S. Department of Health and Human Services. CDC is not responsible for the content of these sites. URL addresses listed in
MMWR were current as of the date of publication.

✩U.S. Government Printing Office: 2006-523-056/40033 Region IV ISSN: 1546-0738