Downloaded from jech.bmj.

com on 28 October 2009

Infant mortality is not an adequate summary

measure of population health
C D Mathers, J A Salomon and C J L Murray

J Epidemiol Community Health 2003;57;319

doi:10.1136/jech.57.5.319

Updated information and services can be found at:

http://jech.bmj.com/cgi/content/full/57/5/319

These include:
References This article cites 3 articles, 1 of which can be accessed free at:
http://jech.bmj.com/cgi/content/full/57/5/319#BIBL

Rapid responses You can respond to this article at:

http://jech.bmj.com/cgi/eletter-submit/57/5/319

Email alerting Receive free email alerts when new articles cite this article - sign up in the box at the
service top right corner of the article

Notes

To order reprints of this article go to:

http://journals.bmj.com/cgi/reprintform

To subscribe to Journal of Epidemiology and Community Health go to:

http://journals.bmj.com/subscriptions/
 Downloaded from jech.bmj.com on 28 October 2009

COMMENTARY 319

Infant mortality purpose, given the considerable policy

................................................................................... interest for many populations as to
whether—and to what extent—gains in

Infant mortality is not an adequate life expectancy have been accompanied

by improvements in health status.12 The

summary measure of population World Health Reports use healthy life

expectancy as a summary measure of the
level of population health in member
health countries in order to provide a compara-
tive assessment of average levels of
C D Mathers, J A Salomon, C J L Murray health, supported by more detailed esti-
mates of mortality and burden of disease
...................................................................................
by age, sex, and cause. Over time, succes-
Measuring the quantity of interest sive reporting on healthy life expectancy
will provide evidence of progress towards
achieving global goals for improving

T
he World Health Organisation that matter, female IMR correlates very health, fatal and non-fatal, for people of
(WHO) annually reports infant mor- highly with male IMR across the 191 all ages.
tality rates (IMR), child mortality WHO member states in 2000 (r=0.996); J Epidemiol Community Health 2003;57:319
rates, adult mortality rates, average life by the argument of Reidpath and Allotey,
expectancies, and healthy life expectan- we could use male IMR as a good indica-
.....................
cies for all 191 member states.1 2 IMR tor for women’s health.
Although there may a high correlation Authors’ affiliations
correlates highly with HALE across these C D Mathers, J A Salomon, C J L Murray,
191 member states in 2000 (r=0.93). between infant mortality and other Evidence and Information for Policy, World
Reidpath and Allotey argue that IMR is components of health cross sectionally, Health Organisation, 1211 Geneva 27,
an acceptable proxy measure of popula- time trends may differ substantially, and Switzerland
tion health because of this high this is of great importance for the evalu-
Correspondence to: Dr C D Mathers;
correlation.3 ation of health systems and policy. In [email protected]
Inspection of the country specific esti- Australia, the Aboriginal IMR has de-
mates of HALE and IMR reveals that, clined 50% faster than non-Aboriginal
REFERENCES
despite the high correlation, there are IMR over the past 20 years, at the same
1 World Health Organiation. The World
substantial variations both in adult mor- time that the gap in total life expectancy Health Report 2001. Mental health: new
tality and in the average loss of full at birth has widened from 16 years to 20 understanding, new hope. Geneva: WHO,
years.6 7 In sub-Saharan Africa, infant 2001.
health at any given level of IMR. In the 2 Lopez AD, Ahmad OB, Guillot M, et al.
nine countries with an IMR in the range mortality rates have declined by 17% World mortality in 2000: life tables for 191
of 6–7 per 1000, for example, there is a over the past decade, whereas adult mor- countries. Geneva: WHO, 2002.
tality (the probability of dying between 3 Reidpath DD, Allotey P. Infant mortality rate
range of 71 to 80 in life expectancy at as an indicator of population health. J
ages 15 and 60) has increased by 37%
birth and a range of 10% to 15% in the Epidemiol Community Health
reflecting the devastating impact of the 2003;57:344–6.
proportion of total life expectancy lost
AIDS epidemic.1 8 Clearly IMR is not an 4 Mathers CD, Stein C, Tomijima N, et al.
through living in years of less than full Global burden of disease 2000: version 2
adequate summary measure for moni-
health. As a more specific example, methods and results. Geneva: WHO (GPE
toring trends and differentials in popula- Discussion Paper no 50), 2002.
females in India and Zimbabwe both had
tion health. 5 World Bank. World development indicators
IMRs close to 63 per 1000 in the year [CD-ROM]. Washington, DC: IBRD, World
Good science entails measuring the
2000, but the life expectancies at birth quantity of interest rather than using a Bank, 1999.
for these two populations were 46.0, and 6 Mathers CD, Douglas RM. Measuring
proxy with an additional error term that progress in population health and wellbeing.
62.7 years, respectively. is unknown (until we measure the In: Eckersley R, ed. Measuring progress: Is life
Correlation of variables in itself is no quantity of interest). Without measuring getting better? Collingwood, Victoria, CSIRO
argument to use one as a proxy for the Publishing, 1998:125–155.
average levels of population health, the 7 Australian Institute of Health and
other in scientific analyses. For example, most important outcome of the health Welfare. Australia’s health 2002. Canberra:
breast cancer mortality rates in women system, it is impossible to assess whether AIHW, 2002.
correlate highly with age (r=0.95 for health policies are working—whether 8 Murray CJL, Lopez AD, eds. The global
Australia, New Zealand, and Japan using burden of disease: a comprehensive
levels of health are improving and assessment of mortality and disability from
five year age groups).4 Should we stop inequalities are being reduced. Health diseases, injuries and risk factors in 1990 and
monitoring breast cancer mortality and policy is not aimed only at reducing projected to 2020. 1st edn. Cambridge:
monitor female age instead to determine Harvard University Press, 1996.
mortality. Substantial resources are de- 9 Murray CJL, Frenk J. World Health Report
if prevention programmes are effective? voted to reducing the incidence and 2000: a step towards evidence-based health
Healthy life expectancy has a correlation prevalence of conditions that cause ill policy. Lancet 2001;357:1698–700.
of 0.76 with average telephone lines per health but not death, and to reducing 10 Field MJ, Gold GM, eds. Summarizing
population health: directions for the
capita across 164 countries.5 At what their impact on people’s lives. So it is development and application of population
level of correlation should we cease to important to capture both fatal and non- metrics. Washington, DC: National Academy
measure healthy life expectancy and fatal health outcomes in any summary Press, 1998.
11 Murray CJL, Salomon JA, Mathers CD, et al,
focus instead on telephone lines? For measure of population health (SMPH). eds. Summary measures of population health:
Broad interest and use of SMPH in the concepts, ethics, measurement and
policy arena9–11 demonstrates the recog- applications. Geneva: WHO, 2002.
12 van de Water HP, Perenboom RJ, Boshuizen
The acronym for healthy life expectancy was nition of their value at the practical level HC. Policy relevance of the health expectancy
changed from DALE to HALE in the World for monitoring population health. Mor- indicator: an inventory of European Union
Health Report 2001.1 tality indicators are not adequate for this countries. Health Policy 1996;36:117–29.

www.jech.com