Overview of Adult Mortality in the Developing World Project

1
Overview of Adult Mortality in the Developing
World Project

• Kenneth Hill
• July 9th, 2004

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
2
The Global Burden of Disease 2000 Project

• Program Project funded by NIA to provide
  methodological and empirical basis for assessing
  GBD in 2000
• Project PI Christopher Murray components at
  Harvard, Johns Hopkins, University of Queensland,
  WHO, UC-Berkeley and elsewhere
• Projects include
• Communicable disease mortality transitions
• Summary measures of population health
• Risk factor assessments
• Measuring adult mortality in developing world

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
3
The AMDC Project

• Weakest demographic component of 1990 GBD was
  adult mortality in LDCs
• Project focused on
• Assessment and development of methods
• Compilation of appropriate data
• Systematic application of preferred methods
• Estimation for close to 2000 and trends
• Method of choice is evaluation of recorded age
  patterns of mortality

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
4
The AMDC Project (continued)

• Project PI Kenneth Hill, co-PI John Wilmoth
• Advisory group consisting of
• Mari Bhat, Institute for Economic Growth, Delhi
• Hoda Rashad, American University in Cairo
• Ian Timæus, London School of Hygiene and Tropical
  Medicine
• Graduate student RAs who made the whole thing
  possible Yoonjoung Choi, Danzhen You, Bernardo
  Queiroz, Piedad Urdinola
• Spreadsheet development by Pablo Lattes

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
5
Methodology

• BoD required estimates of mortality by age and
  sex for time period close to 2000
• Indirect methods do not provide independent
  information on age patterns of mortality
• Focus has therefore been on death distribution
  methods that evaluate a recorded age pattern of
  (adult) mortality by comparison with population
  age distributions

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
6
Death Distribution Methods

• Necessary connections between age distributions
  and age distributions of death have long been
  recognized
• Brass was first to develop an easily applied
  approach (Growth Balance), though assumed
  stability
• Preston and Coale proposed alternative approach
  assuming stability (Extinct Generations)
• Bennett and Horiuchi generalized EG and Hill
  generalized GB for non-stable, closed populations

AMDC Workshop, The Marconi Center, July 8th
11th, 2004
7
Overview of General Growth Balance and Synthetic
Extinct Generation Methods
AMDC Workshop, The Marconi Center, July 8th
11th, 2004
8
a) Growth Balance Methods
The Growth Balance Identity simply states that
the (rate of) change in population between two
time points is equal to the difference between
the (rates of) entries and the (rates of) exits
during the interval
In a population with no migration, entries e are
births and exits are deaths d. For an open-ended
age group a, entries are ath birthdays and exits
are deaths a and over. Thus in terms of
rates e(a) - r(a) d(a) where e(a)
N(a)/N(a) , r(a) is the growth rate of the
population a, and d(a) D(a)/N(a)
9
Typical data errors include incomplete death
registration and changes in coverage from one
census to another. If deaths are registered with
completeness c (at all ages), D(a)
1/cDo(a) If we assume constant exponential
growth between two censuses, r(a)
1/tlnN2(a)/N1(a) If coverage of the second
census differs from that of the first by a factor
k (also constant at all ages), r(a)
1/tlnN2o(a)/N1o(a)1/k ro(a) -
1/tln(k)
10
Replacing true values with observed values,
(1)
Thus the residual estimate of the death rate in
each open-ended age group should be linearly
related to the observed death rate in the age
group, with slope equal to the completeness of
death recording c and intercept a function of any
change in completeness of population recording
and the length of the interval.
11
b) Synthetic Extinct Generations Method
Vincent (1951) proposed that (with perfect
recording of deaths) the population age a at time
t could be estimated ex post facto by cumulating
the deaths to that cohort after time t until the
cohort was extinct. On a cross-sectional basis,
in a stationary (life table) population, the
survivors to age a are equal to the deaths above
age a
12
In a stable population, the deaths above age a
need to be expanded by the stable growth rate to
reflect change in birth cohort size If death
registration is c complete, constant by age, the
ratio of the estimated population age a based on
the expanded deaths a to the observed population
age a estimates the completeness of death
recording relative to population recording
Where c is the completeness of death recording
(relative to population)
13
Bennett and Horiuchi (19811984) generalized the
method to any closed population by using the
variable growth rates above age a
In words, the population age a can be estimated
from the deaths above that age by applying summed
age-specific growth rates to allow for the
demographic history of the population. The ratio
of the population age a estimated in this way
from the deaths to the observed population age a
estimates the completeness of death recording
relative to population recording.
14

• Results of General Growth Balance (GGB) and
  Synthetic Extinct Generations (SEG) methods
  applied to a non-stable population may be
  affected by
• Omission of deaths
• Age misreporting
• Emigration
• Differential omission of population
• Combinations of these distortions
• Potential effects of these errors need to be
  borne in mind when reviewing applications

15
AMDC Workshop, The Marconi Center, July 8th
11th, 2004
16
AMDC Workshop, The Marconi Center, July 8th
11th, 2004
17
AMDC Workshop, The Marconi Center, July 8th
11th, 2004
18
AMDC Workshop, The Marconi Center, July 8th
11th, 2004