INT J TUBERC LUNG DIS 10(3):259–263
© 2006 The Union
The burden of childhood tuberculosis and the accuracy
of community-based surveillance data
B. J. Marais,*†‡ A. C. Hesseling,*† R. P. Gie,*†‡ H. S. Schaaf,*†‡ N. Beyers*†‡
*Desmond Tutu TB Centre, and †Department of Paediatrics and Child Health, Faculty of Health Sciences,
University of Stellenbosch, Cape Town, ‡Tygerberg Children’s Hospital, Cape Town, South Africa
S U M M A R Y
BACKGROUND: Inadequate surveillance and diagnostic
difficulties compromise the quality of epidemiological
data on childhood tuberculosis (TB).
OBJECTIVE: To document the incidence of childhood
TB and to evaluate the accuracy of community-based
surveillance data in a high-burden setting.
METHODS: This prospective observational study was
conducted from February 2003 to October 2004 at five
primary health care clinics in Cape Town, South Africa.
Comprehensive surveillance was done to ensure that all
children ?13 years of age treated for TB were included.
RESULTS: During the study period, 443 children (?13
years of age) received anti-tuberculosis treatment, of
whom 389 (87.8%) were recorded in the TB treatment
register. The TB incidence calculated from the TB treat-
ment register was 441/100000/year amongst children
and 845/100000/year amongst adults. Fifty-four chil-
dren treated for TB were not recorded in the TB treat-
ment register, including 21/28 (75%) children with se-
DISCUSSION: Children ?13 years of age contributed
13.7% of the total TB burden, but experienced more
than half (52.2%) the TB incidence recorded in adults.
Community-based surveillance data excluded the major-
ity of children with severe disease. The accuracy of sur-
veillance data is an important consideration when de-
scribing the epidemiology of childhood TB or measuring
the success of public health interventions.
KEY WORDS: childhood tuberculosis; burden; surveil-
THE CONTRIBUTION of children to the total tuber-
culosis (TB) caseload is poorly documented, espe-
cially in countries with a high burden of disease.1,2
Official epidemiological data may not reflect the true
burden of childhood TB in these countries, due to the
inadequacy of existing surveillance systems and the
difficulty of diagnosing TB in children. National Tuber-
culosis Programmes (NTPs) focus almost exclusively
on the diagnosis and treatment of sputum smear-
positive disease,1,3 which excludes the vast majority
of children with TB.4–6 For these reasons, the Interna-
tional Union Against Tuberculosis and Lung Disease
(IUATLD) stated in 1991 that reliable information on
the incidence of childhood TB can only be obtained in
Observational studies in children with TB are often
hospital-based, which results in considerable bias, pre-
venting accurate extrapolation to the community
level. The paucity of accurate epidemiological data,1,4
together with the significant TB-related morbidity
and mortality suffered by children in high-burden set-
tings,8 underline the need to obtain a more compre-
hensive picture of childhood TB at the community
level in high-burden settings.
The aim of this study was to document the inci-
dence of childhood TB and to evaluate the accuracy
of community-based surveillance data in a high-burden
This was a prospective observational study conducted
from 1 February 2003 to 31 October 2004 in Cape
Town, Western Cape Province, South Africa.
The study setting included five local clinics in Cape
Town, served by Tygerberg Children’s Hospital as the
referral hospital. Local clinics provide primary health
care services in the community, and coordinate the di-
agnosis and treatment of TB patients. All children
?13 years of age started on anti-tuberculosis treatment
during the 21-month study period were included. Pae-
diatric services were accessible only to children ?13
Correspondence to: Dr Ben J Marais, Department of Paediatrics and Child Health, Desmond Tutu TB Center, Faculty of
Health Sciences, Stellenbosch University, P O Box 19063, Tygerberg, 7505, South Africa. Tel: (?27) 21-938 9155. Fax:
(?27) 21-838 9138. e-mail: email@example.com
Article submitted 22 April 2005. Final version accepted 5 September 2005.
The International Journal of Tuberculosis and Lung Disease
years of age, and therefore children were defined as
those ?13 years instead of the customary 15 years of
age. Two of the five local clinics were sites for on-
going epidemiological research, where a concurrent
household contact study (with supervised tracing of
household contacts ?5 years of age) continued dur-
ing the study period.
The study communities rarely utilise private medi-
cal services, and children diagnosed with TB are rou-
tinely referred to the local clinic, where they receive
fully supervised treatment free of charge. The inci-
dence of all TB in Cape Town was 678 per 100000
population,9 while the prevalence of human immuno-
deficiency virus (HIV) infection amongst women at-
tending public antenatal clinics in the Western Cape
Province was 13.1% (95% confidence interval [CI]
8.5–17.7%),10 in 2003.
Data collection and analysis
According to the South African NTP guidelines,11 all
children with suspect symptoms, as well as those ?5
years of age in household contact with a sputum
smear-positive source case, should be routinely evalu-
ated for TB with a tuberculin skin test (TST) and a
chest radiograph (CXR). Once a diagnosis of TB is
made and anti-tuberculosis treatment is commenced,
the child’s name is entered into the TB treatment reg-
ister maintained at the clinic, which provides the only
data source for community-based surveillance data.
For research purposes, a comprehensive prospec-
tive surveillance system was put in place to identify all
children started on anti-tuberculosis treatment during
the study period, irrespective of whether they were re-
corded in the clinic TB treatment register. The inves-
tigator visited each clinic on a weekly basis to evalu-
ate children with suspect symptoms (as defined by the
NTP)11 and those newly diagnosed with TB. In addi-
tion, a study nurse documented all children started on
anti-tuberculosis treatment at the referral hospital.
Two independent experts evaluated the CXRs of all
children treated for TB to assess the accuracy of the
diagnosis. Children were categorised as ‘not TB’ if
they met all three of the following criteria: 1) both ex-
perts agreed that the chest radiograph CXR was not
suggestive of TB; 2) no bacteriological confirmation
was obtained; and 3) no extra-thoracic TB was docu-
mented. All children treated for TB who had radio-
logical signs suggestive of TB, bacteriological confir-
mation or extra-thoracic manifestations of TB were
categorised as ‘probable TB’. Severe disease was de-
fined as disseminated (miliary) disease12 and/or central
nervous system (CNS) involvement.
Unadjusted data from the 2001 national census
were used for the calculation of TB incidence rates.
One clinic was excluded from these calculations, as
the population served by this clinic had increased
considerably since 2001, due to new housing develop-
ments and informal settlements. The populations served
by the other four clinics are stable, as reflected by com-
parable 1996 and 2001 census data and the fact that
these areas are completely built-up, with no recent
housing developments or space for informal settle-
ments. TB incidence was calculated by using the an-
nualised number of entries (excluding those trans-
ferred in) recorded in the clinic TB treatment register
during the 21-month study period as the numerator,
and age-specific population figures for the area as the
denominator. The corrected TB incidence in children
was calculated by subtracting the number of children
categorised as ‘not TB’ from the total number treated
to determine the numerator, and using area- and age-
specific population data as the denominator.
Ethics approval was obtained from the Institutional
Review Board of Stellenbosch University, the City of
Cape Town Health Department and local community
During the study period, a total of 443 children were
treated for TB, of whom 389 (87.8%) were recorded
in the TB treatment register (Figure). The files of four
patients were lost. Table 1 shows the demographics
and clinical characteristics of all 439 children whose
records were evaluated. The sex distribution was
equal; 223 (52.6%) children were ?3 years of age,
and 283 (64.5%) were tested for HIV infection, of
whom 25 (8.8%) were HIV-infected.
Eighty-five (19.4%) children were categorised as
‘not TB’. Three had disease caused by non-tuberculous
mycobacteria (NTM): one immune-competent child
had a regional Mycobacterium bovis bacille Calmette-
Guérin (BCG) abcess in the right axilla, and the other
two were HIV-infected with M. bovis BCG and an
unspecified NTM, respectively, cultured from their
gastric aspirates. Fifty-five were asymptomatic house-
hold contacts of an adult index case, while 27 pre-
TB during the 21-month study period. ‘Not TB’ ? both experts
agreed that CXR was not suggestive of TB, no bacteriological
confirmation was obtained and no extra-thoracic TB was docu-
mented. Probable TB ? all children treated for TB who had ra-
diological signs suggestive of TB, bacteriological confirmation
or extra-thoracic manifestations of TB. TB treatment register ?
the clinic TB treatment register, which is the only data source
used to compile community-based surveillance data. TB ?
tuberculosis; CXR ? chest X-ray.
Flow chart of all children ?13 years of age treated for
The burden of childhood tuberculosis
sented with suspect symptoms or possible contact
outside the household. Ten of the 25 (40.0%) HIV-
infected children were categorised as ‘not TB’, com-
pared to 26/253 (10.2%) non-HIV-infected children
(odds ratio [OR] 5.82, 95%CI 2.17–15.56, P ? 0.001).
Table 2 reflects the burden of childhood TB in the
study area. During the 21-month study period 2830
people were treated for TB, of whom 389 (13.7%)
were children ?13 years of age. The proportion of
children ?13 years of age was 18.0% (107/593) in
the two clinics where active contact tracing was well-
supervised, compared to 12.4% (278/2237) in the
clinics where no supervised contact tracing took place
(OR 1.55, 95%CI 1.21–1.99, P ? 0.001). The TB
incidence amongst adults, using routine surveillance
data, was 845/100000/year, compared to 441/100000/
year in children ?13 years of age. The corrected TB
incidence in children was 407/100000/year.
Table 3 reflects the TB manifestations documented
in the 54 children who were treated for TB but not re-
corded in the clinic TB treatment register. In total, 21/
28 (75.0%) children with severe disease were not re-
corded in the clinic TB treatment register, comprising
9/12 (75.0%) with disseminated (miliary) disease, 7/10
(70.0%) with CNS involvement and 5/6 (83.3%) with
both disseminated disease and CNS involvement.
In this high-burden setting, children ?13 years of age
contributed 13.7% of the total TB caseload, which
compares remarkably well with the 15% contribu-
tion estimated for children ?15 years of age in low-
income countries.13 This indicates that although results
from this hyperendemic area may not be generalis-
able, it is reasonable to suspect a similar situation in
other endemic areas. Efforts should be increased to
measure the burden of childhood TB with greater
accuracy in endemic areas.
A striking observation was the high incidence of
childhood TB in the study community. Although chil-
dren ?13 years of age contributed 13.7% of the total
of all children ?13 years of age treated for TB
Demographics and clinical characteristics
Children treated for TB
Total analysed (4 of 443 records lost)
Actively traced contact of an adult index case
Presented with suspect symptoms
Proportion categorised as ‘not TB’
HIV status unknown
’Not TB’ ? both experts agreed that CXR was not suggestive of TB, no bacterio-
logical confirmation was obtained, and no extra-thoracic TB was documented.
TB ? tuberculosis; HIV ? human immunodeficiency virus; CXR ? chest X-ray.
data, and data corrected for both misdiagnosis and absence from the clinic
TB treatment register*
The burden of childhood TB in the study area using community-based surveillance
Community-based surveillance data
Total number recorded in the TB treatment register
Proportion of children ?13 years of age
Proportion of children recorded in the TB treatment register in clinics
with supervised household contact tracing
without supervised household contact tracing
Total population (unadjusted national census data from 2001)†
Proportion of children ?13 years of age†
All-TB incidence‡ in
the total population†
adult population (?13 years of age)†
children (?13 years of age)†
Corrected TB incidence§ in children ?13 years of age†
*TB treatment register ? this refers to the clinic TB treatment register, which is the only data source used to compile
community-based surveillance data.
†Data from one local clinic where the unadjusted national census data from 2001 were considered to be unreliable
were excluded from both the numerator and the denominator.
‡The all-TB incidence was calculated by 1) using the annualised number of entries (excluding transfers in) recorded in
the TB treatment register during the 21-month study period as the numerator; 2) using area- and age-specific popu-
lation figures as the denominator, and 3) expressing this ratio as the number of TB cases/100000/year.
§The corrected TB incidence in children was calculated by subtracting the number of children categorised as ‘not TB’
from the total number treated to determine the numerator and using area- and age-specific population data as the
TB ? tuberculosis.
The International Journal of Tuberculosis and Lung Disease
TB caseload, they experienced a TB incidence that
was 52.2% (441 vs. 845/100000/year) of the adult
(?13 years of age) incidence, using data from the
clinic TB treatment register. The incidence of child-
hood TB remained high (407/100000/year) despite
correcting for children with ‘not TB’, who received
treatment inappropriately. The proportion of paedi-
atric cases was influenced by the diligence with which
active contact tracing was performed, as demon-
strated by the significant difference in the proportion
of child cases recorded in the clinics where active con-
tact tracing was well-supervised, compared to those
without supervised contact tracing.
The fact that significantly more HIV-infected chil-
dren were categorised as ‘not TB’ reflects the tremen-
dous diagnostic difficulties experienced in this group.14
Preventive chemotherapy is advised in HIV-infected
children infected with M. tuberculosis or exposed to
an adult index case, once active TB has been excluded.15
However, because adherence to unsupervised preven-
tive chemotherapy is known to be poor16 and radio-
logical signs are difficult to interpret,14 concerned cli-
nicians often prefer to give fully supervised therapy to
children whom they perceive to be at high risk of de-
veloping TB, such as HIV-infected children exposed
to an adult index case.
The study is limited by the fact that contact tracing
was not well-supervised in all five clinics; however, all
children who presented with suspect symptoms were
screened. In addition, children with severe TB may
have died without presenting to a health care facility,
but access to local clinics and hospitals is good within
the study area, which makes this unlikely. Despite
these limitations, this study provides the best achiev-
able estimate of the true burden of childhood TB in
this high-burden setting.
The most important observation was the frequent
omission of severe cases and the gross under-repre-
sentation (7/28, 25%) of children at the severe end of
the disease spectrum in the clinic TB treatment regis-
ter, although the total number of children recorded
was a slight over-representation of the true TB burden
due to misdiagnosis (389 vs. 358, 108.7%). All 54
children who were not recorded in the clinic TB treat-
ment register were diagnosed at the referral hospital,
from where the majority were transferred to a TB hos-
pital due to the severity of their disease or to social
problems to complete their treatment under hospital
supervision. Those discharged from the referral hos-
pital on anti-tuberculosis treatment were not always
captured in the clinic TB treatment register, although
treatment was continued.
The current policy is that all children discharged
from hospital on anti-tuberculosis treatment should
be recorded in the clinic TB treatment register, but
this policy needs to be re-emphasised to health care
personnel. Children who complete their full duration
of treatment in a TB hospital are currently entered
into a hospital-held register, but these data are not re-
layed back to the local clinic. The most feasible solu-
tion in the study setting is that data from the hospital-
held register should be transferred to the appropriate
clinic register at the time of electronic data entry. The
electronic register should then provide all-inclusive
community-based data. However, current international
TB surveillance categories preclude the identification
of these more serious forms of TB, a limitation that
we think needs to be addressed to determine the true
burden of childhood TB.
In conclusion, children with severe forms of TB are
inaccurately reflected in community-based surveil-
lance data. The accuracy of surveillance data must be
taken into consideration when reporting on the epide-
miology of childhood TB or when determining the
effectiveness of public health interventions.
We thank the primary health care clinics involved for their kind
assistance, and Dr I Toms and Dr V Azevedo (City of Cape Town
Health Department) for their constructive input. The manuscript is
in partial fulfilment of a PhD thesis.
1 Walls T, Shingadia D. Global epidemiology of paediatric tuber-
culosis. J Infect 2004; 48: 13–22.
2 Nelson L J, Wells C D. Global epidemiology of childhood
tuberculosis. Int J Tuberc Lung Dis 2004; 8: 636–647.
with probable TB who were absent from the clinic TB
The TB manifestations recorded in children
Total with probable TB (439–85 ‘not TB’*)
Proportion absent from the TB treatment register†
Intra-thoracic TB only
Ghon focus or primary complex
Lymph node disease
Disseminated (miliary) disease
Extra-thoracic TB only
Peripheral lymph node involvement
Both intra- and extra-thoracic TB
TBM and miliary TB
TBM and lymph node disease on CXR
Total with severe disease‡ absent from
the TB treatment register
*’Not TB’ ? both experts agreed that CXR was not suggestive of TB, no
bacteriological confirmation was obtained, and no extra-thoracic TB was
†TB treatment register ? this refers to the clinic TB treatment register, which
is the only data source used to compile community-based surveillance data.
‡Severe disease ? CNS and/or disseminated (miliary) TB.
TB ? tuberculosis. CNS ? central nervous system; TBM ? tuberculous men-
ingitis; CXR ? chest X-ray.
The burden of childhood tuberculosis
3 Donald P R. Childhood tuberculosis: out of control? Curr
Opin Pulm Med 2002; 8: 178–182.
4 Starke J R. Pediatric tuberculosis: time for a new approach.
Tuberculosis 2003; 83: 208–212.
5 Eamranond P, Jaramillo E. Tuberculosis in children: reassess-
ing the need for improved diagnosis in global control strate-
gies. Int J Tuberc Lung Dis 2001; 5: 594–603.
6 Zar H J, Hanslo D, Apolles P, Swingler G, Hussey G. Induced
sputum versus gastric lavage for microbiological confirmation
confirmation of pulmonary tuberculosis in infants and young
children: a prospective study. Lancet 2005; 365: 130–134.
7 Hershfield E. Tuberculosis in children: guidelines for diagnosis,
prevention and management (a statement of the scientific com-
mittees of the IUATLD). Bull Int Union Tuberc Lung Dis 1991;
8 Chintu C, Mudenda V, Lucas S, et al. Lung diseases at necropsy
in African children dying from respiratory illnesses: a descrip-
tive necropsy study. Lancet 2002; 360: 985–990.
9 Caldwell J, Azevedo V. Cape Town TB Control. Progress Report
1997–2003. Durban, South Africa: Health Systems Trust, 2004.
www.hst.org.za/publications/618 Accessed November 2005.
10 South African Department of Health. Report. National HIV
and syphilis antenatal sero-prevalence survey in South Africa
2003. Pretoria, South Africa: DOH, 2003. www.doh.gov.za/
docs/reports/2003/hiv/p1-23.pdf Accessed November 2005.
11 South African Department of Health. The South African Tuber-
culosis Control Programme—Practical Guidelines 2000. Preto-
ria, South Africa: DOH, 2000: pp 32–37. www.capegateway.
gov.za/text/2003/tb_guidelines2000.pdf Accessed November
12 Marais B J, Gie R P, Schaaf H S, et al. A proposed radiologic
classification of childhood intra-thoracic tuberculosis. Pediatr
Radiol 2004; 33: 886–894.
13 Nelson L J, Wells C D. Tuberculosis in children: considerations
for children from developing countries. Semin Pediatr Infect
Dis 2004; 15: 150–154.
14 Graham S M, Coulter J B S, Gilks C F. Pulmonary disease in
HIV-infected children. Int J Tuberc Lung Dis 2001; 5: 12–23.
15 Centers for Disease Control and Prevention, MMWR Recomm
Rep. 1998; 47: 1–58.
16 Starke J R, Jacobs R F, Jereb J. Resurgence of tuberculosis in
children. J Ped 1992; 120: 839–855.
R É S U M É
CONTEXTE : La qualité des données épidémiologiques
concernant la tuberculose (TB) infantile est compromise
par une surveillance inadéquate et par les difficultés de
OBJECTIF : Documenter l’incidence de la TB de l’enfant
et évaluer la précision des données de surveillance basées
sur la collectivité dans un contexte à haute prévalence.
MÉTHODES : Cette étude prospective d’observation a été
menée entre février 2003 et octobre 2004 dans cinq
polycliniques de soins de santé primaires à Cape Town,
Afrique du Sud. Une surveillance complète a été menée
pour s’assurer que tous les enfants âgés de moins de 13
ans et traités pour TB avaient bien été inclus.
RÉSULTATS : Au cours de la période d’étude, 443 en-
fants âgés de ?13 ans ont reçu un traitement antituber-
culeux. Parmi ceux-ci, 389 (87,8%) avaient été inscrits
dans le registre de traitement de la TB. L’incidence de la
TB, calculée à partir du registre de traitement de la TB,
a été de 441/100.000 par an parmi les enfants et de 845/
100.000 par an parmi les adultes. Cinquante-quatre en-
fants traités pour TB n’étaient pas inscrits dans le regis-
tre de traitement de la TB, et parmi ceux-ci se trouvaient
21 des 28 (75%) enfants atteints de TB grave.
DISCUSSION : Les enfants âgés de ?13 ans ont représenté
13,7% du fardeau total de TB. L’incidence chez les en-
fants est supérieure à la moitié (52%) de celle enregistrée
chez les adultes. Les données de surveillance basées sur
la collectivité n’ont pas inclus la majorité des enfants at-
teints d’une maladie grave. La précision des données de
surveillance est une considération importante lorsqu’il
s’agit de décrire l’épidémiologie de la TB de l’enfant ou de
mesurer les succès des interventions de santé publique.
R E S U M E N
MARCO DE REFERENCIA : La calidad de los datos epide-
miológicos sobre la tuberculosis (TB) infantil se dete-
riora debido a la vigilancia inadecuada y a las dificulta-
OBJETIVO : Verificar la incidencia de TB en la infancia y
evaluar la precisión de los datos de vigilancia epidemi-
ológica de base comunitaria, en un entorno con alta
carga de morbilidad por TB.
MÉTODOS : Fue este un estudio observacional prospec-
tivo realizado entre febrero de 2003 y octubre de 2004
en cinco consultorios de atención primaria de salud en
Ciudad del Cabo, Sudáfrica. Se llevó a cabo una vigilan-
cia exhaustiva para verificar la inclusión de todos los
niños ?13 años tratados por TB.
RESULTADOS : Durante el período del estudio 443 niños
?13 años recibieron tratamiento antituberculoso, de los
cuales 389 (87,8%) se declararon al registro de trata-
miento de la TB. La incidencia de TB calculada con base
en el registro fue de 441/100000 por año en los niños
y de 845/100000 por año en los adultos. Cincuenta y
cuatro niños tratados por TB no fueron declarados al
registro de tratamiento de la TB ; entre ellos 21 de los 28
niños con una enfermedad grave (75%).
DISCUSIÓN : La tuberculosis en niños ?13 años repre-
sentó el 13,7% de la carga total de morbilidad por TB.
Sin embargo, los niños presentaron una incidencia de
TB correspondiente al 52,2% de la incidencia de los
adultos. Los datos de vigilancia de base comunitaria ex-
cluyeron la mayoría de niños con enfermedad grave. La
precisión de los datos de vigilancia constituye una con-
sideración importante cuando se describe la epidemi-
ología de la TB infantil y cuando se evalúa el éxito de las
intervenciones de salud pública.