www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4 1
April 24, 2012
Expanded Programme on
Immunization, Department of
Immunization, Vaccines and
Biologicals, WHO, Geneva,
Switzerland (E Simons MPH,
A Burton BS, P Strebel MBChB);
Center for Infectious Disease
Dynamics, Pennsylvania State
University, University Park, PA,
USA (M Ferrari PhD, J Fricks PhD);
and Global Immunization
Division, Center for Global
Health, US Centers for Disease
Control and Prevention,
Atlanta, GA, USA
(K Wannemuehler PhD,
A Anand MBBS)
Dr Peter Strebel, Expanded
Programme on Immunization,
Department of Immunization,
Vaccines and Biologicals,
20 Avenue Appia, World Health
Organization, Geneva 1202,
Assessment of the 2010 global measles mortality reduction
goal: results from a model of surveillance data
Emily Simons, Matthew Ferrari, John Fricks, Kathleen Wannemuehler, Abhijeet Anand, Anthony Burton, Peter Strebel
Background In 2008 all WHO member states endorsed a target of 90% reduction in measles mortality by 2010 over
2000 levels. We developed a model to estimate progress made towards this goal.
Methods We constructed a state-space model with population and immunisation coverage estimates and reported
surveillance data to estimate annual national measles cases, distributed across age classes. We estimated deaths by
applying age-specific and country-specific case-fatality ratios to estimated cases in each age-country class.
Findings Estimated global measles mortality decreased 74% from 535 300 deaths (95% CI 347 200–976 400) in 2000 to
139 300 (71 200–447 800) in 2010. Measles mortality was reduced by more than three-quarters in all WHO regions
except the WHO southeast Asia region. India accounted for 47% of estimated measles mortality in 2010, and the
WHO African region accounted for 36%.
Interpretation Despite rapid progress in measles control from 2000 to 2007, delayed implementation of accelerated
disease control in India and continued outbreaks in Africa stalled momentum towards the 2010 global measles
mortality reduction goal. Intensified control measures and renewed political and financial commitment are needed to
achieve mortality reduction targets and lay the foundation for future global eradication of measles.
Funding US Centers for Disease Control and Prevention (PMS 5U66/IP000161).
In 2007, investigators reported that the global goal to
reduce measles deaths by 50% by 2005, compared with
1999, had been achieved.1 Building on this accomplish
ment, in 2008 the World Health Assembly endorsed a
target of 90% reduction in measles mortality by 2010,
compared with 2000. Endemic transmission of measles
virus was interrupted in the Americas in 2002, and four
of the remaining five WHO regions (all except southeast
Asia) have set target dates for measles elimination by
2020 or earlier.2 The establishment of a global measles
eradication goal has been extensively discussed by the
World Health Assembly and advisory committees to
WHO and now hinges on progress towards regional
elimination outside the Americas.3
Monitoring measles mortality is relevant for all part
ners involved in child survival. The fourth Millennium
Development Goal (MDG4) aims to reduce deaths of
children by two thirds by 2015 compared with 1990. The
proportion of children vaccinated against measles was
adopted as an indicator to measure progress towards
MDG4; a rebound in measles deaths would pose a
substantial threat to achieving this goal.4,5
The rapid progress in measles control from 2000 to
2007 was based on implementation of recommen
ded measles mortality reduction strategies, including
increas ing routine immunisation coverage, periodic
supple mental immunisation activities (SIAs—ie, mass
vaccination campaigns aimed at immunising 100% of a
predefined population within several days or weeks),
laboratorysupported surveillance, and appropriate
man agement of measles cases.6 Countries that have
fully implemented and sustained these strategies have
experienced reductions in measles cases of greater than
90%.7 However, not all countries have managed to do so,
and several of the largest recorded outbreaks of the past
decade were during 2009–10.8
Because most measles deaths are in countries where
vital registration systems cannot provide reliable infor
mation on causespecific mortality, WHO has relied
on mathematical models to estimate the global burden
of measles.1,9 Previous models have not objectively
incorporated measles surveillance data and instead relied
on vaccination coverage data as the primary indicator of
local disease burden. Consequently, these models could
neither consistently capture the effects of large outbreaks
on measles mortality where high vaccination coverage
was reported, nor show periods of low mortality between
outbreaks when low vaccination coverage was reported.
To assess progress towards the 2010 global measles
mortality reduction goal, we developed a new model that,
unlike previous models, uses surveillance data objectively
to estimate both incidence and the age distribution of
cases, accounts for herd immunity, and uses robust
statistical methods to estimate uncertainty.
Estimating annual measles incidence
For 65 countries with adequate vital registration data
(≥85% of estimated deaths of children younger than
5 years registered and coded), we used the reported
number of measles deaths. These deaths accounted for
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4
less than 0·01% of global measles mortality, according to
vital registration data and estimated mortality.10
For 128 remaining countries with inadequate vital
registration data, we estimated countryspecific measles
deaths through a threestep process. We estimated annual
measles incidence on the basis of reported measles cases
for each country, then we distributed estimated incidence
across age groups, and finally we calculated the number
of deaths in each age class by applying agespecific and
countryspecific measles casefatality ratios (CFRs).
Measles cases and vaccination coverage are reported
annually to WHO by all member states through the WHO/
UNICEF Joint Reporting Form.11 WHO derived coverage
estimates for the first routine dose of measlescontaining
vaccine (MCV1) from reported coverage data and survey
results by use of computational logic.12 We sent input data
(full list in the appendix) to national immunisation
programme managers to identify updates and corrections.
Measles cases reported through surveillance systems
typically represent a fraction of the true number of cases
because many children do not present for medical
attention and, when medical care is sought, cases can be
misdiagnosed or not reported to central authorities.13
However, national measles surveillance data can show
underlying measles incidence trends over time.
Statespace models provide a probabilistic framework
to predict the unobserved elements of a dynamic process,
such as true measles incidence, when provided with
observed elements of the dynamic process, such as
reported measles cases and vaccination coverage.14–17 Our
statespace model is characterised by two interrelated
sets of equations with unknown parameters: a process
model that represents the evolution of true disease
incidence through time as a function of infection risk
and vaccination coverage and an observation model that
represents the observation, or reporting, of cases through
the national measles surveillance system.
The technical aspects of the statespace model are
described elsewhere.15 In general, the process model
describes the trend in measles cases in a country over
time as a function of births, deaths, measles virus
transmission, and vaccination. This is an annualised,
linear approximation to a dynamic susceptible, infected,
recovered (SIR) model of disease transmission.18 In a
particular year, we added infants that are born and neither
die from other causes of death nor are immun ised
through routine vaccination services to a pool of
susceptibles. The pool of susceptibles was then depleted
by SIAs and infection, which might result in death or
recovery and immunity. We assumed the first vaccination
was 84% effective if given before age 12 months and 93%
effective at or beyond 12 months, the second vaccination
was 99% effective irrespective of age at administration,19
and vaccineinduced immunity was lifelong.20
We assumed the annualised rate of infection increases
from zero in a completely immune population to 100% in
a completely susceptible population, which is consistent
with the findings that greater than 90% of a completely
susceptible population is likely to become infected if ex
posed to measles.18 Although we did not include a nonzero
threshold of population susceptibility below which trans
mission ceases, the reduction of the annualised infection
rate with increasing population immunity is compatible
with herd immunity. In accor dance with the equation
below, population sus cept ibility reduces the annualised
rate of infection:
θ1 is the transmissibility estimated from the surveillance
data separately for each country (estimation method
described below), S is the number of susceptibles, and
N is the total population. The median estimated value for
θ1 across all countries was 3·8 (IQR 1·3–8·0). With a θ1 of
3·8, the percentage of susceptibles that would be infected
at 10%, 50%, and 90% population susceptibility would be
26%, 77%, and 93%, respectively.
The observation model describes the relation between
the expected number of measles cases estimated by the
process model (with the infection rate above) and the
number of reported measles cases. Measles cases are
estimated to be underreported at a baseline rate θ2 that
is independent for each country: reported cases in
year t = θ2 * true incidence in year t.
Because measles surveillance sensitivity can increase
when measles outbreaks happen,21 we assumed that years
with large outbreaks had more complete reporting. We
defined years with outbreaks as those with docu mented
outbreaks, annual case reports that had a large effect on a
linear regression of reported measles incidence against
time (dfβ >0·37), or when estimated measles mortality
exceeded 20% of all child mortality with the baseline
reporting rate (θ2). We assumed that the years defined as
outbreaks had an increased reporting rate of θ2+θ3. We
then fitted the model to all countries, with θ2 and θ2+θ3
estimated according to the applicable countryyears of
We used a recursive algorithm called an extended
Kalman filter,22 which has shown valid approximations of
unobserved measles virus transmission dynamics,15
to estimate the parameters θ1–θ5 (θ1–θ3 described above, θ4
and θ5 are variance parameters described in the appendix),
which resulted in a predicted number of cases with the
highest likelihood in view of the observed surveillance
data. The extended Kalman filter provides point estimates
of the unobserved incidence in each year, It, and SEs of
those estimates, SEt.
During validation exercises we noted that model
accuracy was low in two circumstances: consistently high
population immunity to measles and low levels of all
cause child mortality. We consequently did not use the
statespace model to estimate reporting rates for countries
where measles was eliminated, where 95% or greater
coverage with two routine doses of measles vaccine was
See Online for appendix
Percentage of susceptibles infected annually
1 – e
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4 3
reported for 5 years or longer, or where child mortality fell
in the lowest quartile of 2009 national childmortality rates
(27 countries). Instead, we estimated measles incidence by
applying a 20% (with 5–40% upper and lower bounds)
reporting rate to reported cases, based on published
reporting rates in highperforming surveillance systems.13
We extracted data from WHO’s measles casebased
reporting system for age at infection for 172 191 measles
cases reported from 121 countries from 2000 to 2009. Age
was expressed as a categorical variable (<1, 1–4, 5–9,
10–14, and ≥15 years) and predicted as a function of two
explanatory variables with multinomial logistic regres
sion. The explanatory variables were the 5 year moving
average of estimated MCV1 coverage, in categories of less
than 60%, 60–84%, and 85–100%; and geographical
region based on a modification of the classification used
by the Global Burden of Disease project.23 To account for
the clustering of cases within countryyears we used the
Taylor series method of variance estimation.24
We took countryspecific measles CFRs for children
aged 1–4 years from a published review of community
based studies,25 which identified 58 publications providing
102 different measles CFR estimates in 29 countries
during the period 1974–2007. The investigators of the
review noted that models of CFRs, attempted with several
covariates, led to implausible results for many countries,
because of low predictive value of the variables available
for all member states. Consequently, expert opinion was
used to develop a set of countryspecific CFRs for children
aged 1–4 years.25 We revised this set of CFRs to include
new data from India and Nepal (appendix).26,27 Relative to
the CFRs for children aged 1–4 years, we assumed that
CFRs were equal for infants, half for children aged
5–9 years, and zero for children older than 10 years.
With a Monte Carlo algorithm, we derived 1000 values of
the annual cases for each country from a normal distri
bution with mean It and variance SEt. Next, we derived
1000 age distributions from the multivariate normal distri
bution with mean equal to the point estimate and variance
equal to the estimated variancecovariance matrix. The
product of estimated cases, age distribution, and the age
specific CFRs resulted in 1000 values of measles deaths in
each countryyear. We took the 2·5th and 97·5th quantiles
of this distribution as the lower and upper bounds of the
estimated number of measles deaths.
To test the robustness of the results to variable assump
tions, we assessed measles mortality under various
alternative assumptions and compared the results with
basecase estimates (ie, univariate sensitivity analysis).
The variables we tested were vaccine effectiveness,
temporal change in CFRs, threshold for defining outbreaks
in reported case data, and age distribution of cases.
Role of the funding source
The sponsor of the study had no role in study design,
data collection, or data interpretation. Two authors are
employees of the sponsor institution and were involved
in data analysis and reviewing the report, but were not
2000 2010 Mortality
Number of doses
deaths (95% upper
and lower bounds)
(95% upper and
Africa520 10256% 337 000
(216 600–653 000)
(29 400–82 300)
(25 700–84 900)
(70 700–107 800)
(347 200–976 400)
63% 197 900186 67576% 50 000
(53 600–78 800)
(71 200–447 800)
36% 569 300 85%527
27310 072132 500 79%
37 42191% 0% 270030 62595%0 290087%52
39 723 76% 9%59 90020 457 79%8%96 600 78%169
38 83555% 16%224 600 29 808 74% 47%208 800 26%12
Western Pacific177 05285% 2% 48 000 49 460 97%2% 53 30076% 382
Worldwide total853 480 72%100%623 100327 305 85% 100%1 066 900 74%1756
SIAs=supplemental immunisation activities. *Not all SIAs are reported to WHO. †We estimated that measles mortality in the Americas was too low to allow reliable measurement of mortality reduction between
2000 and 2010.
Table: Reported measles cases, measles vaccination coverage, estimated measles deaths, and children reached by SIAs by WHO region
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4
involved in the decision to fund this study. The corres
ponding author had full access to all the data in the study
and had final responsibility for the decision to submit
The statespace model and adjusted surveillance data sug
gested that, from 2000 to 2010, annual estimated measles
incidence aggregated across all countries fell 66% from
4·6 to 1·6 cases per 1000 total global popu lation. During
the same period, global MCV1 coverage increased from
72% to 85% and the reported number of measles cases
declined 62% from 853 480 (140 per million total pop u
lation) to 327 305 (48 per million total population; table).
In developing countries in the prevaccine era, roughly
70% of children were infected with measles virus by
age 5 years.28 Results from the multinomial regression
model suggest that children younger than 5 years account
for greater than 60% of cases in most countries with less
than 60% MCV1 coverage (figure 1). However, at 85% or
greater MCV1 coverage, cases were predicted to be
predominantly in children older than 5 years and adults.
The shift in age distribution was most noticeable for
regions with the longest history of high coverage, such as
eastern Europe, central Asia, and Asia Pacific, where the
proportion of cases before age 5 years was predicted to be
38–51% at low MCV1 coverage and 13–20% at high MCV1
coverage (figure 1).
Global measles mortality was estimated to have decreased
74%, from 535 300 deaths (95% CI 347 200–976 400) in
2000 to 139 300 (71 200–447 800) in 2010 (figure 2).
Compared with estimated mortality assuming the complete
absence of measles vaccination, 9·6 million deaths were
averted by measles vaccination during 2000–10.
We estimated that most measles deaths (79%) were in
Africa and India during 2000–10. Measles mortality de
creased by 85% in Africa, from 337 000 to 50 000, during
2000–10 (table). Large scale SIAs began in the south
ern African countries of Botswana, Lesotho, Malawi,
Namibia, South Africa, Swaziland, and Zimbabwe during
1996–99 and in remaining countries of the WHO African
region during 2000–06. Estimated mortality had already
de creased by more than 90% before 2000 in southern
African countries compared with mortality before the im
ple men tation of SIAs and then decreased by more than
90% dur ing 2000–07 in remaining countries of the
African region. The estimated reductions in mortality by
2007 were largely maintained, but mortality did not
decrease further there after. As a whole, the African region
accounted for 36% of global measles mortality in 2010,
down from 63% in 2000. India’s small decline in measles
mortality (26%) led to an in crease in the country’s share
of global measles mortality from 16% in 2000 to 47% in
2010. We estimated that measles mortality decreased by
78% during 2000–10 in the re maining ten countries in
the WHO southeast Asia region.
The WHO eastern Mediterranean and western Pacific
regions accounted for 9–11% of estimated global measles
mortality during 2000–10, and we estimated that measles
mortality fell 79% in the eastern Mediterranean and
76% in the western Pacific region. Although the WHO
European region continues to have large outbreaks of
North Africa and the Middle East
West and central Europe
Asia Pacific and Australia
Eastern Europe and central Asia
Proportion of measles cases
Proportion of measles cases
Proportion of measles cases
1–4 years5–9 years≥10 years
Figure 1: Predicted age distribution of measles cases by MCV1 coverage
and region, 2000–09
MCV1=first routine dose of measles-containing vaccine.
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4 5
measles, because of very low CFRs, the region accounted
for less than 1% of global measles mortality.
Of the alternative variable assumptions we tested by
uni variate sensitivity analysis, low vaccine effectiveness
and low outbreak threshold resulted in higher mortality
esti mates compared with the base case, whereas indexing
CFRs to underfive mortality rates, an alternative
age distribution of cases, high vaccine effectiveness, and
high outbreak threshold resulted in lower mortality
esti mates (figure 3). All the sensitivity analyses led to
annual mortality estimates lying within the uncertainty
bounds of the basecase scenario and none resulted in a
90% or greater reduction in measles mortality over
Our findings suggest that the goal of reducing measles
mortality by 90% from 2000 to 2010 has not yet been met.
Our conclusion was sustained under all alternative
scenarios we assessed. Estimated global measles mor tality
declined substantially from 2000 to 2007, associated with
increases in routine MCV1 coverage as well as the delivery
of more than a billion doses of measles vaccine through
SIAs. However, from 2008 to 2010, estimated global
measles mortality did not diminish further and large
outbreaks in southern Africa in 2009 and 2010 resulted in
a small increase in estimated mortality for the WHO
African region. Although SIAs substantially affected
measles mortality, the highly infectious nature of measles
virus requires maintenance of very high levels of population
immunity through high routine coverage and timely
implementation of SIAs to address immunity gaps.
Measles remains widespread in India because of
delayed implementation of SIAs and restricted improve
ment in MCV1 coverage. We expect planned SIAs
targeting 134 million children and the introduction of a
routine second dose in some states of India during
2011–13 to substantially reduce measles mortality by
2015.29 The decrease in estimated mortality (78%) during
2000–10 in the remaining ten countries in the WHO
southeast Asia region is attributable to comprehensive
improvements in immunisation: roughly 169 million
children were vaccinated in SIAs, MCV1 coverage rose by
11 percentage points, and six of the ten countries
introduced routine MCV2.
To our knowledge, our work presents the first use
of a statespace framework to assess the global burden
of a disease (panel). The key advantage of this approach
for estimating measles mortality is the objective and
comprehensive integration of surveillance data in the
estimation process, which allows the results to show the
episodic nature of measles outbreaks and periods of low
virus transmission between outbreaks. A model that
tracks fluctuations in measles incidence is better suited
than basic natural history1,9 or proportionate mortality30
approaches to monitor progress as measles mortality
declines to low levels.
2002 2004 2006 20082010
Estimated measles deaths (1000s)
CFR indexed to U5MR in 2000
Low vaccine effectiveness
Low outbreak threshold
Base-case point estimate
Alternate age distribution
High vaccine effectiveness
High outbreak threshold
Figure 3: Global measles mortality re-estimated with alternative variable assumptions
The appendix provides details on assumptions. CFR=case-fatality ratio. U5MR=under-five mortality rate.
20002002 20042006 2008
2001 2003 2005
Estimated deaths (1000s)
Estimated measles deaths with vaccination
95% CI of estimated measles deaths
Estimated measles deaths in absence of vaccination
Deaths averted by measles vaccination
Figure 2: Global estimated measles mortality and measles deaths averted
Panel: Research in context
We searched for country-level information on annual measles incidence through annual
surveillance reports, monthly case-based measles reporting, and a process of country
consultation with WHO member states. To develop age distributions, we analysed the
largest available global dataset of line-listed reported measles cases.11 We systematically
searched for the new information on case-fatality rates published after the latest systematic
review of case-fatality ratios25–27 and used the most comprehensive global set of vital
registration data available,10 which we used in place of modelled estimates for countries that
registered more than 85% of estimated deaths. We used sampling methods to combine
variables and estimate uncertainty.
Compared with past efforts to model measles burden, we estimated fewer measles deaths,
but similar rates of mortality reduction, since 2000 because of advancements in methods
and updated input data. Unlike past efforts to model measles burden, our study accounts for
herd immunity, uses robust statistical methods to estimate uncertainty, and uses case-based
surveillance data to estimate age distribution of cases, and aggregate surveillance data to
estimate incidence. As the first attempt to objectively incorporate surveillance data in
modelling the global burden of measles, our study provides a basis for more in-depth
research to understand the complexities and importance of disease surveillance data.
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60522-4
4 UN. Road map towards the implementation of the United Nations
Millennium Declaration: report of the SecretaryGeneral. New York,
NY: United Nations, 2001.
van den Ent MM, Brown DW, Hoekstra EJ, Christie A, Cochi SL.
Measles mortality reduction contributes substantially to reduction
of all cause mortality among children less than five years of age,
1990–2008. J Infect Dis 2011; 204 (suppl 1): S18–23.
WHO. Progress in global measles control and mortality reduction,
2000–2007. Wkly Epidemiol Rec 2008; 83: 441–48.
Otten M, Kezaala R, Fall A, et al. Publichealth impact of accelerated
measles control in the WHO African Region 2000–03. Lancet 2005;
WHO. Measles outbreaks and progress towards meeting measles
preelimination goals: WHO African Region, 2009–2010.
Wkly Epidemiol Rec 2011; 86: 129–36.
Stein CE, Birmingham M, Kurian M, Duclos P, Strebel P.
The global burden of measles in the year 2000—a model that uses
countryspecific indicators. J Infect Dis 2003; 187 (suppl 1): S8–14.
10 Department of Health Statistics and Informatics. WHO mortality
data base. Geneva: World Health Organization, 2011.
11 WHO. WHO vaccinepreventable diseases: monitoring system.
Geneva: World Health Organization, 2010.
12 Burton A, GacicDobo M, Karimov R, Kowalski R. A computational
logicbased representation of the WHO and UNICEF estimates of
national immunization coverage. http://www.doc.ic.ac.uk/~rak/
papers/wuenic.pdf (accessed April 14, 2012).
13 Harpaz R. Completeness of measles case reporting: review of estimates
for the United States. J Infect Dis 2004; 189 (suppl 1): S185–90.
14 Breto C, He DH, Ionides EL, King AA. Time series analysis via
mechanistic models. Ann Appl Stat 2009; 3: 319–48.
15 Chen S, Fricks J, Ferrari MJ. Tracking measles infection through
nonlinear state space models. J R Stat Soc Ser C Appl Stat 2011;
16 Ionides EL, Bretó C, King AA. Inference for nonlinear dynamical
systems. Proc Natl Acad Sci USA 2006; 103: 18438–43.
17 Shumway RH, Stoffer DS. Time series analysis and its applications.
New York, NY: Springer, 2000.
18 Anderson RM, May RM. Infectious diseases of humans: dynamics
and control. Oxford: Oxford University Press, 1991.
19 Uzicanin A, Zimmerman L. Field effectiveness of live attenuated
measlescontaining vaccines: a review of published literature.
J Infect Dis 2011; 204 (suppl 1): S133–48.
20 Moss WJ, Scott S. The immunological basis for immunization
series—module 7: measles update 2009. Geneva: World Health
21 Mette A, Reuss AM, Feig M, et al. Underreporting of measles an
evaluation based on data from North RhineWestphalia.
Dtsch Arztebl Int 2011; 108: 191–96.
22 Elliott R, Aggoun L, Moore J. Hidden Markov models: estimation
and control. Berlin: Springer, 1995.
23 Harvard University. Global burden of disease study operations
Manual_Jan_20_2009.pdf (accessed Sept 1, 2010).
24 Binder DA. On the variances of asymptotically normal estimators
from complex surveys. Int Stat Rev 1983; 51: 279–92.
25 Wolfson LJ, Grais RF, Luquero FJ, Birmingham ME, Strebel PM.
Estimates of measles case fatality ratios: a comprehensive review of
communitybased studies. Int J Epidemiol 2009; 38: 192–205.
26 Joshi AB, Luman ET, Nandy R, Subedi BK, Liyanage JBL,
Wierzba TF. Measles deaths in Nepal: estimating the national
casefatality ratio. Bull World Health Organ 2009; 87: 456–65.
27 Sudfeld CR, Halsey NA. Measles case fatality ratio in India: a review
of community based studies. Indian Pediatr 2009; 46: 983–89.
28 Black FL. Measles antibody prevalence in diverse populations.
Am J Dis Child 1962; 103: 242–49.
29 WHO. Progress in implementing measles mortality reduction
strategies, India 2010–2011. Wkly Epidemiol Rec 2011; 86: 439–44.
30 Morris SS, Black RE, Tomaskovic L. Predicting the distribution of
underfive deaths by cause in countries without adequate vital
registration systems. Int J Epidemiol 2003; 32: 1041–51.
The estimates we present are lower than previous
estimates of measles mortality,1 although the previous
estimates lie within the 95% CI of the new estimates
(appendix). The decrease in mortality is largely due to
downward revision of population and childdeath esti
mates, reduction of CFRs for infants, and constraining
measles mortality to less than 20% of total child mortality.
We did not incorporate several aspects of measles virus
transmission dynamics in our statespace framework
because of restricted data availability, restricted effects
of these refinements on broad patterns in annual
measles incidence, or both. These transmission dynamics
include the effect of population agestructure on age
specific incidence, generation of infections in one time
step from cases in the previous time step, likelihood of
importations or interruption of transmission, and contact
rates between subpopulations with differing levels of
immunity. Within very populous countries, such as China
and India, measles burden could be more accurately
estimated with subnational data showing local, within
country variation in measles incidence.
Our work has highlighted a crucial lack of information
on the measurement and interpretation of changes in
surveillance system performance over time. We were
unable to develop a simple and yet reliable quantitative
indicator of how changes in surveillance practices (eg,
introduction of casebased reporting or modification of
case definitions), funding, and staffing affect surveillance
sensitivity; we instead settled on the algorithm described
earlier. The development of clear assessment criteria for
surveillance sensitivity would aid routine interpretation of
surveillance data for monitoring immunisation programme
performance, which will only become more important as
measles control efforts move towards global eradication.
ES drafted the report, analysed and interpreted data, and contributed to
study design. MF was involved in the model development, data analysis,
and review or the report. JF was involved in the model development and
programming. KW was involved in data analysis and review of the
report. AA was involved in data analysis and review of the report. AB
contributed to the model design. PS coordinated the project, analysed
and interpreted results, and reviewed the report.
Conflicts of interest
We declare that we have no conflict of interest.
We would like to thank the many immunisation programme and data
managers who reviewed the input data and draft results, members of
QUIVER for their constructive feedback, and WHO/IVB staff, including
Robert Perry and Marta GacicDobo. ES, AB, and PS are employees of
the World Health Organization. The authors alone are responsible for
the views expressed and they do not necessarily represent the decisions,
policy, or views of the World Health Organization.
1 Wolfson LJ, Strebel PM, GacicDobo M, Hoekstra EJ, McFarland JW,
Hersh BS. Has the 2005 measles mortality reduction goal been
achieved? A natural history modelling study. Lancet 2007;
2 Strebel PM, Cochi SL, Hoekstra E, et al. A World Without Measles.
J Infect Dis 2011; 204 (suppl 1): S1–3.
3 WHO. Meeting of the Strategic Advisory Group of Experts on
Immunization, November 2010—summary, conclusions and
recommendations. Wkly Epidemiol Rec 2011; 86: 1–16.
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60638-2 1
Measles: the burden of preventable deaths
Measles has been, and remains, a major killer of
children around the world. Despite the introduction
of the measles vaccine in 1963, measles caused
an estimated 2·6 million deaths in a single year as
recently as 1980.1 In The Lancet, Emily Simons and
colleagues2 estimate that, after more than 45 years of
measles vaccine availability, the disease caused nearly
140 000 deaths in 2010.
Even in industrialised countries, complications,
includ ing pneumonia, diarrhoea, encephalitis, and sub
acute sclerosing panencephalitis, lead to substantial
morbidity and mortality.3,4 However, it is in developing
countries where measles exacts its greatest health
burden. A review of communitybased measles studies5
showed a median casefatality ratio of 3·91% (mean
7·40%, range 0–40·15%).
Through global measles prevention efforts, great
progress has been made in measles control. Elimination
of indigenous transmission of disease has been
achieved in the WHO Americas region.1 Five of the six
WHO regions have set goals to eliminate measles by
2020. At present, there is a worldwide goal of a 95%
reduction in measles mortality by 2015 compared with
2000 estimates. Measles eradication is biologically
feasible and, although no formal eradication goal has
yet been set, progress toward the mortality reduction
goal will lead to consideration of an eradication goal.1,6
Measles is one of the most contagious vaccine
preventable diseases,7 and is one of the best indicators for
problems in vaccination programmes because of its high
communicability and recognisable rash. Outbreaks of
measles with complications and deaths can be a greater
motivating force for change than immunisation coverage
data gaps and the theoretical potential for outbreaks.8 This
was the case in the USA, where a resurgence of measles in
1989–91 led to major investments in, and strengthening
of, the overall National Immunization Program.
If immunisation programmes fail to immunise new
susceptibles added to the population daily through
births and migration, enough susceptibles will accumu
late to fuel another measles outbreak. For example, since
2008, after substantial reductions in measles mortality,
measles has resurged in Africa.9 It is crucial to maintain
high immunity levels and immunise all children at
How can we best monitor the progress of global
immunisation programmes to guide corrective actions
if needed? Measuring measles vaccine coverage provides
some information but does not directly translate into
effects on health burden. Global disease surveillance
systems are at present unable to capture measles case
numbers accurately enough to monitor deaths directly.
Instead, progress has been assessed through changes in
estimated annual measlesattributed deaths. As noted
by Simons and colleagues,2 65 countries have adequate
vital registration data, which allow the measurement of
actual deaths. However, for the remaining 128 countries
where most deaths from measles occur, vital registration
data are inadequate and necessitate the estimation of
The accuracy of estimates depends on the assump
tions and data used in modelling exercises. Traditionally,
it was assumed that all susceptible people acquired
measles, so the number of cases depended on vaccine
coverage and effectiveness. Once cases were estimated,
age distributions were inferred on the basis of coverage,
and agespecific casefatality ratios for a particular
region were estimated and applied to the number of
cases to estimate the number of deaths.10 Although
this approach has been useful for monitoring the
progress of measles mortality reduction efforts, there
is a potential bias toward overestimating deaths
since it does not account for herd immunity, which
April 24, 2012
Science Photo Library
www.thelancet.com Published online April 24, 2012 DOI:10.1016/S0140-6736(12)60638-2
is likely to decrease incidence of measles and deaths
indirectly. Simons and colleagues2 attempt to take this
into account by incorporating a decrease in the rate of
infection among susceptibles as population immunity
rises, and by using actual surveillance data to modify
the estimates of cases and mortality (along with other
adjustments).2 In so doing, they estimated that that
there were 535 300 deaths from measles in 2000,
27% lower than the previous estimate of 733 000.11
Although substantially lower, this estimate still high
lights that far too many children are dying from this
readily preventable disease. And, in 2010, they estimate
139 300 deaths (382 deaths per day) despite substantial
improvements in immunisation coverage.
Most importantly, perhaps, Simons and colleagues’
report highlights crucial gaps in available data to guide
prevention programmes—surveillance and vital record
registrations are inadequate in much of the world.
What is most needed is not more advanced ways to
estimate mortality, but the direct measurement of
mortality. As measles is considered for eradication,
it will be crucial to improve surveillance to the point
that deaths and cases will actually be measured, not
*Walter A Orenstein, Alan R Hinman
School of Medicine and Emory Vaccine Center, Emory University,
Atlanta, GA 30322, USA (WAO); and Center for Vaccine Equity,
Task Force for Global Health, Decatur, GA, USA (ARH)
We declare that we have no conflicts of interest.
1 Strebel PM, Cochi SL, Hoekstra E, et al. A world without measles. J Infect Dis
2011; 204 (suppl 1): S1–3.
Simons E, Ferrari M, Fricks J, et al. Assessment of the 2010 global measles
mortality reduction goal: results from a model of surveillance data. Lancet
2012; published online April 24. DOI:10.1016/S01406736(12)605224.
Strebel PM, Papania MJ, Dayan GH, Halsey NA. Measles vaccine. In:
Plotkin SA, Orenstein WA, Offit PA, eds. Vaccines, 5th edn. Philadelphia:
Saunders Elsevier, 2008: 353–98.
Orenstein WA, Papania MJ, Wharton ME. Measles elimination in the United
States. J Infect Dis 2004; 189 (suppl 1): S1–3.
Wolfson LJ, Strebel PM, GacicDobo M, Hoekstra EJ, McFarland JW,
Hersh BS. Has the 2005 measles mortality reduction goal been achieved?
A natural history modelling study. Lancet 2007; 369: 191–200.
Bellini WJ, Rota PA. Biological feasibility of measles eradication. Virus Res
2011; 162: 72–79.
Fine PEM, Mulholland K. Community Immunity. In: Plotkin SA,
Orenstein WA, Offit PA, eds. Vaccines, 5th edn. Philadelphia: Saunders
Elsevier, 2008: 1573–92.
Orenstein WA. The role of measles elimination in development of a
national immunization program. Pediatr Infect Dis J 2006; 25: 1093–101.
Moss WJ, Griffin DE. Measles. Lancet 2011; 379: 153–64.
10 Wolfson LJ, Grais RF, Luquero FJ, Birmingham ME, Strebel PM. Estimates of
measles case fatality ratios: a comprehensive review of communitybased
studies. Int J Epidemiol 2009; 38: 192–205.
11 Centers for Disease Control and Prevention. Global measles mortality,
2000–2008. MMWR Morb Mortal Wkly Rep 2009; 58: 1321–26.