NOT FOR QUOTATION WITHOUT THE PERMISSION OF THE AUTHORS
Targetting Lifesaving:
Demographic ljnkages Between
Population Structure and Life Expectanc
yJames W. k u p e l A n a t o l i I. Y a s h i n
November
1 9 8 5 WP-85-78Working Ftzpers are interim r e p o r t s on work of t h e International Institute f o r Applied Systems Analysis and h a v e r e c e i v e d only limited review. Views o r opinions e x p r e s s e d h e r e i n d o not necessarily r e p r e s e n t t h o s e of t h e Institute o r of i t s National Member Organizations.
INTERNATIONAL INSTITUTE
FOR
APPLIED SYSTEMS ANALYSIS 2361 Laxenburg, AustriaAcknowledgments
W e thank Jan Hoem, Nathan Keyfitz, Thomas Schelling, and Michael Stoto f o r helpful suggestions, and Susanne Stock f o r prompt, meticulous typing.
Life expectancy in a heterogeneous population c a n b e i n c r e a s e d by lowering mortality rates o r by a v e r t i n g d e a t h s a t d i f f e r e n t a g e s , f r o m d i f f e r e n t causes, for d i f f e r e n t groups, as well as by changing t h e p r o p o r t i o n s of individuals in v a r i o u s r i s k groups, p e r h a p s by a l t e r i n g t h e transition rates between groups. Under- standing how s u c h c h a n g e s in population s t r u c t u r e a f f e c t s life e x p e c t a n c y i s useful in evaluating a l t e r n a t i v e lifesaving policies.
Targetting Lifesaving:
Demographic Ihkages Between Population Structure and Life Expectancy
James W. Vaupel a n d A n a t o l i I. Yashin
The individuals comprising t h e typical population of men, mice, or machines f a c e differing mortality chances. This heterogeneity a r i s e s , in p a r t , from indivi- dual c h a r a c t e r i s t i c s t h a t c h a n g e or c a n b e changed, like a g e , b e h a v i o r , occupa- tion, or residence. Alteration of t h e a g e composition, occupational s t r u c t u r e , or o t h e r p a t t e r n of h e t e r o g e n e i t y in a population, p e r h a p s as t h e r e s u l t of some poli- c y intervention, will c h a n g e t h e distribution of mortality c h a n c e s a n d h e n c e change t h e life expectancy of t h e population. In t h i s p a p e r w e develop some formu- l a s f o r analyzing how v a r i o u s kinds of changes in population s t r u c t u r e will a f f e c t l i f e expectancy.
Change in life e x p e c t a n c y i s a measure of t h e number of y e a r s of l i f e saved ( o r l o s t ) by a n a l t e r a t i o n in population s t r u c t u r e and hence i s a useful m e a s u r e for policy analysis. In p a r t i c u l a r , t h i s measure is a p p r o p r i a t e f o r what might b e called t a r g e t analysis. If limited r e s o u r c e s are available f o r lifesaving interven- tions, how should t h e r e s o u r c e s b e t a r g e t e d ? How effective would p r o g r a m s b e t h a t are d i r e c t e d toward d i f f e r e n t a g e groups, diseases, r i s k g r o u p s (like c i g a r e t t e smokers), regions, e t c ? A complete t a r g e t analysis would h a v e
to
include consideration of how difficult i t isto
focus a n intervention on a p a r t i c u l a r g r o u p a n d how r e s i s t a n t t h e g r o u p i sto
change. Nonetheless, understanding t h e benefits of a change, if achieved, in l i f e e x p e c t a n c y gained or life-years saved i s c l e a r l y a key component of any t a r g e t analysis.In addition
to
such policy applications, t h e methods and formulas p r e s e n t e d in t h i s p a p e r are useful in gaining a d e e p e r demographic understanding of howm o r -
tality r a t e s , d e a t h s , r i s k g r o u p s , a n d life e x p e c t a n c y are i n t e r r e l a t e d . How, for instance, d o mortalityrates
c h a n g e if some d e a t h s are a v e r t e d ?Four d i f f e r e n t analytical a p p r o a c h e s are used in t h e p a p e r
to
analyze t h e demographic linkages between population s t r u c t u r e and life expectancy: t h e comparative-statics a p p r o a c h , t h e dynamics a p p r o a c h , computer simulation, and a novel method t h a t w e call t h e "second-chance" a p p r o a c h . The p a p e r provides some discussion and illustration of t h e s t r e n g t h s , weaknesses, and interrelationships among t h e s e a l t e r n a t i v e methods of demographic analysis.LIFE AND DEATH RATES
Consider, f i r s t , a g e
structure
as c h a r a c t e r i z e d by t h e s u r v i v o r s h i p functionwhere p ( z ) r e p r e s e n t s t h e f o r c e of mortality at a g e
I.
(Formula (1) and the r e s u l t s t h a t follow c a n b e i n t e r p r e t e d as pertainingto
e i t h e r p e r i o dor
c o h o r t cal- culations.) A change in p will change this age s t r u c t u r e and h e n c e life expectancyat
b i r t h :where o is a n a g e beyond which no o n e lives.
The effect of a c h a n g e in p on e o c a n b e analyzed by e i t h e r of two a p p r o a c h e s . In t h e comparative-statics a p p r o a c h , t h e t r a j e c t o r y of p i s assumed
to
changeto
p', where
t h e analyst relates t h e c h a n g e b ( z )
to
t h e change in eo, p e r h a p s a s measured by:In t h e dynamics a p p r o a c h , t h e r e i s some r a t e of change in p ( z , t ) o v e r time t :
t h e analyst relates t h i s
rate
of change p ( z , t )to
t h erate
of c h a n g e in eo(t ):Both a p p r o a c h e s are informative and w e will c o n s i d e r b o t h . F o r notational simpli- c i t y , w e will d r o p t h e a r g u m e n t
t
t h r o u g h o u t a n d w r i t e A x ) r a t h e r t h a n CL(z,t) a n d e o r a t h e r t h a n e o ( t ) .If p ( z ) i s c o n s t a n t o v e r a n i n t e r v a l of time of l e n g t h T, t h e n
Combining t h i s r e s u l t with (3) yields t h e relationship between p a n d 6:
If 6 i s small, t h i s r e d u c e s t o
Hence, r e s u l t s c o n c e r n i n g p ( x ) c a n b e d e r i v e d f r o m r e s u l t s c o n c e r n i n g 6 ( z ) and visa-versa: t h e c o m p a r a t i v e - s t a t i c s a p p r o a c h a n d t h e dynamics a p p r o a c h comple- ment e a c h o t h e r . Note t h a t p ( z ) c a n b e a r b i t r a r i l y l a r g e , as long as T i s small enough.
A c o m p a r a t i v e - s t a t i c s r e l a t i o n s h i p c a n r e a d i l y be d e r i v e d f r o m (1)-(4):
s
In t h e c a s e of a uniform c h a n g e in mortality
at
a l l a g e s , b ( z ) = 6 , a l l z , formula (9) c a n b e r e w r i t t e n as0
F o r small 6 ,
Hence
where
In t h e limit, a s b a p p r o a c h e s z e r o , formula (12) holds exactly. Consequently, i t i s a p p a r e n t t h a t
where p i s t h e uniform
rate
of p r o g r e s s in reducing mortalityrates:
d t
P
= ,
allz .
d z
Thus, for small changes in & t h e comparative-statics a p p r o a c h yields t h e same formulas
as
t h e dynamics a p p r o a c h . Keyfitz (1977) d e r i v e d (14) and noted t h a t H i sa
measure of a g e h e t e r o g e n e i t y ; a s Demetrius (1979) indicated, H c a n b e i n t e r - p r e t e das
t h e e n t r o p y of t h e a g e composition of t h e population.THE SECOND-CHANCE APPROACH
Interventions
to
r e d u c e mortality ( o r equipment f a i l u r e ) work by saving lives, i.e. by a v e r t i n g t h e s c y t h e of d e a t h . Suppose t h a t for some p r o p o r t i o n6
of a c o h o r t ( p e r h a p s a s y n t h e t i c p e r i o d "cohort"), d e a t h is a v e r t e d once. Let 1 ( z ) r e p r e s e n t t h e p r o p o r t i o n of t h e c o h o r tat
a g ez
t h a t is alive and h a s n o t b e e n s a v e d and l e t t + ( z ) r e p r e s e n t t h e p r o p o r t i o n of t h e r e s u s c i t a t e d whoare
a l i v eat
a g ez .
Since t h e p r o p o r t i o n of t h e c o h o r t survivingat
a g ez
i s given byt h e new life e x p e c t a n c y , 8 ;
,
i s given byThe r e l a t i v e change in life e x p e c t a n c y is simply
An e x p r e s s i o n f o r l + ( z ) i s readily developed. Assuming t h a t t h e resuscitated f a c e t h e same f o r c e of mortality
as
t h o s e who have not been saved, t h e probability of survivalto
a g ez
f o r t h o s e whose liveswere
savedat
a g e w i s given bywhere
T
r e p r e s e n t s t h e time of death. Because t h e distribution density of w i s CL(w)l(w),Substituting (20) in (18) yields
Note t h a t t h e H in (21) d e n o t e s t h e same expression as Keyfitz's H in (12) and (14). Hence, (21) p r o v i d e s
a
t h i r d i n t e r p r e t a t i o n of Has a
measure of t h e p r o p o r - tional i n c r e a s e in life e x p e c t a n c y if everyone's lifewere
saved once, o r a l t e r n a - tively,as
t h e p r o p o r t i o n a l i n c r e a s e ina
randomly chosen individual's life span if t h a t individual's life i s saved. For Swedish males in 1982, Hwas
.15 and s owas
72 y e a r s . Consequently,at
1982 p e r i o d mortality r a t e s , a v e r t i n g t h e d e a t h ofa
Swed- ish male would give t h e r e s u s c i t a t e d a b o u t 11 y e a r s of life expectancy.The formula f o r beo/ e o in ( 2 1 ) holds exactly f o r any 6, whereas t h e analogous formula in ( 1 2 ) only holds approximately, f o r s m a l l 6. The r e a s o n c a n b e under- stood by considering some simple diagrams. The model where death i s only a v e r t e d o n c e can b e r e p r e s e n t e d as:
Individuals
are
a l l initially in t h e l e f t box. A proportion 6 of those who would h a v e diedare
saved. but just once: t h e r e s u s c i t a t e d e x p e r i e n c e t h e o r i g i n a l force of mortality A z ) . On t h e o t h e r hand, t h e model where mortalityrates
are d e c r e a s e d by 6 can b e r e p r e s e n t e d as:T H E RESUSCITATED ORIGINAL
COHORT
Because t h e force of mortality in a n y
state
i s (1-
6 ) p ( z ) , t h e o v e r a l l f o r c e of mortality must also b e (1-
6 ) p ( z ). What t h e decomposition into a n infinitestream
ofstates
r e v e a l s i s t h a t a reduction in mortalityrates
may r e s u l t insome
people's'I 'I
6 p ( z )
*
bp(z)
t m m m
-
t v t
(1
-
6)P(r) (1-
~ ) P ( z ) (1-
~ ) P ( Z )SAVED TWICE 8 P ( l
C
-
THOSE SAVED ONCE ORIGINAL
POPULATION
8 ~ ( r rn
lives being saved s e v e r a l times.
L e t r e p r e s e n t t h e e x p e c t e d life y e a r s lived by a n individual in t h e i ' t h s t a t e , i.e., by a n individual whose life has been saved i and only t times:
where ( z ) d e n o t e s t h e probability t h a t a newborn individual i s alive and in
state
i at a g ez .
Note t h a t T: i s equalto
eo, t h e original life e x p e c t a n c y b e f o r e t h e 11- fesaving intervention. Clearly,When
6
i s small, i t i s unlikely t h a t anyone will gain much life expectancy by being saved more t h a n once, i.e., t h eterms
T:, 703, and so onare
unimportant. (We p r o v e and expand o n t h i s intuitively plausible r e s u l t elsewhere, in Vaupel and Y a s h h (1985).) Hence,In t h e
two-state
model, where d e a t h i s only a v e r t e d once,The similarity between (29) a n d (25) sheds light on why Keyfitz's
H
in (12) i s identi- calto
t h eH
in (21).I t i s sometimes e a s i e r
to
analyze t h etwo-state
model t h a n t h e many-state model. Since t h etwo
models h a v e equivalent implications f o r life expectancy in t h e limit f o r small 6, t h etwo-state
model may provide a convenient line of a t t a c k . W e exploit this, and t h e relationship between6
and p discussed e a r l i e r , in s e v e r a l subsequent d e r i v a t i o n s in t h i s p a p e r . W e call t h e method involving t h etwo-state
model t h e "second-chance" a p p r o a c h , in c o n t r a s t with t h e comparative-statics ap- proach and t h e dynamics a p p r o a c h . Although in t h i s p a p e r t h e second-chance ap- p r o a c h i s only usedto
analyze changes in life expectancy. i t h a s more g e n e r a l ap- plicationsto
any situation, including marriage, divorce, abortion, unemployment, t h e r e p a i r of equipment, e t c . , where changingsome rate
c a n b e considered as equivalentto
givingsome
individualsa
second chance.Suppose, as above, t h a t some proportion
6
of d e a t h sare
a v e r t e d once. How will t h e t r a j e c t o r y of mortality r a t e s ,as
given by p ( z ) , c h a n g e ? In brief, how d o e s saving lives a f f e c t mortality r a t e s ? Substituting (20) in (16), taking log deriva-tives, and t h e n simplifying yields:
A t a g e z e r o , when I ( z ) i s one, t h e formula simplifies
to
A s s u r v i v o r s h i p d e c r e a s e s , however, p ' ( z ) a p p r o a c h e s p ( z ) . Thus, reducing d e a t h s by some proportion b
at
all a g e s r e d u c e s t h e f o r c e of mortality by less t h a nb
at
all a g e s a f t e r b i r t h . The distribution of death times, as given byAz)L
( z ) , c h a n g e sto
so
t h a t a reduction in d e a t h s by b l e a d sto a
new distribution of d e a t h times s h i f t e dto
o l d e r ages. S i n c e d e a t h , a s S h a k e s p e a r e put it. "is c e r t a i nto
all", i t i s c l e a r t h a ta
d e a t h a v e r t e d today i san
additional d e a t htomorrow.
The mathematics of t h i s adjustment i s c a p t u r e d b y (26) a n d (28).IF
THE RESUSCITATED A R E DWPERENTThe formulas a n d calculations a b o v e assume t h a t
a
r e s u s c i t a t e d p e r s o n would face t h esame force
of mortality o v e r t h erest
of h i sor
h e r life as a p e r s o n whose l i f e had not b e e n saved. To g e n e r a l i z e t h e formula, i t i s usefulto
c o n s i d e r t h e fol- lowing variation o n t h e model discussed above:Note t h a t now individuals who
are
saved e x p e r i e n c ea
mortality t r a j e c t o r y given by p + ( z ) , ra t h e r t h a n byd z ) .
Let r ' ( 2 ) b e t h e remaining life e x p e c t a n c yat
a g eTHE RESUSCITATED ORIGINAL
COHORT
.
b p ( z
*
z
of t h e r e s u s c i t a t e d :w h e r e
Because t h e density
at
a g ez
of t h e distribution of ( f i r s t ) d e a t h i s given by I.l(z)l(z), t h e value of beo must b e given byHence
w h e r e
If p + ( z ) e q u a l s A z ) , so t h a t individuals are, in e f f e c t , saved from d e a t h once, t h e n
H +
equals H. If p + ( z ) equals ( 1-
6 ) p ( z ) ,so
t h a t d e a t hrates are
r e d u c e d uniformly for e v e r y o n e , r e g a r d l e s s of whether t h e y h a v e b e e n r e s u s c i t a t e dor
not,H+
will b e close in value to Has
long as b i s small. Consequently,w h e r e
This expression
for
H, which i s equal in valueto
Keyfitzss e x p r e s s i o n f o r H,was
d e r i v e d by Vaupel (1986) d i r e c t l y from Keyfitz's formula. The e x p r e s s i o n c l e a r l yindicates how t h e e f f e c t of saving lives on life expectancy depends on t h e number of d e a t h s
at
various a g e s and on t h e number of additional y e a r s of lifea
resusci- t a t e d p e r s o n might have.SAVING
THE
OLD BEBORgTHE
YOUNGA s Vaupel (1986) discusses
at
length, if d e a t hrates
a r e reduced by some pro- portion6
between a g e sa
and @, then for small6,
where
Correspondingly, if p r o g r e s s is being made
at a rate
p against mortality between a g e sa
and @, t h e nThe values of H a p for various five y e a r a g e categories f o r Swedish males and f e males in 1982 are given in Table 1. Remarkably, i t is f o r males 70 t o 75 and f o r f e males 75
to
80 t h a t H a p is largest. A o n e p e r c e n t reduction in mortality in t h o s e a g e c a t e g o r i e s would i n c r e a s e life expectancyat
b i r t h bymore
t h a n twiceas
muchas a
o n e p e r c e n t reduction in mortality in infancy and e a r l y childhood.AVERTING NEOPLASTIC DEATH IN VENICE
Let p C ( z ) r e p r e s e n t t h e f o r c e of mortality from c a n c e r , o r more generally any specified c a u s e of death. Suppose t h a t f o r some proportion
d
of individuals who would h a v e died from c a n c e r , t h i s ( f i r s t ) d e a t h from c a n c e r is a v e r t e d . Furth-er
suppose t h a t t h e s e r e s u s c i t a t e d individuals t h e n h a v e t h esame
remaining life expectancyas
o r d i n a r y individuals. Using t h e second-chance a p p r o a c h and t h esame
kind of reasoning employedto
d e r i v e formulas (31)-(33), i t i s c l e a r t h a tTable 1. Values of
Ha,,
f o r Swedish males and females in 1982.Age Period Males Females
H
(i.e.,total
f o r .I5270 all a g e s )SOURCE:
Vaupel (1986).If
6
i s small, i t is unlikely t h a t a n individual would b e saved from c a n c e r d e a t h more than once. Hence, (39) holds approximately f o ra
reduction6
in c a n c e r mor- tality r a t e sas
long as6
i s small. I t follows t h a twhere p i s t h e
rate
of p r o g r e s s in reducing c a n c e r mortalityIf c a n c e r is independent of o t h e r causes of death, then i t is possible
to
derive a n alternative expression f o r Hc t h a t is similar t o Keyfitz's formula f o r H in (13).Let l:(z) r e p r e s e n t t h e proportion of people in t h e population who
are
aliveat
a g ez
and who have been saved once from c a n c e r death (at any a g e p r i o rto
2 ) . By analogyto
(20), letting w denote t h e a g eat
which c a n c e r death w a s a v e r t e d , i t fol- lows t h a twhere l c ( z ) can b e i n t e r p r e t e d as t h e survival function when c a n c e r is t h e only cause of death
Hence, by t h e same logic used
to
d e r i v e (21).Keyfitz (1977) d e r i v e s formula (44) using a different a p p r m c h . In addition, h e p r e s e n t s some illustrative examples. For instance, f o r Italian females in 1964, Hc f o r d e a t h s from neoplasms
was
0.0300, compared witha total
H of 0.1631. Thus,a
o n e p e r c e n t reduction in c a n c e r mortality would increase life expectancyat
b i r t h by about t h r e e p e r c e n t of one p e r c e n t , o r by about 8 days given Italian fe- m a l e life expectancy of 72.9 y e a r s in 1964. By way of comparison, Hc f o r d e a t h s from cardiovascular diseasesw a s
0.0564, almost twiceas
highas
t h e Hc f o r d e a t h s from c a n c e r , whereas Hc f o r deaths from influenza, pneumonia and bronchitisw a s
0.0122, o r less than halfas
g r e a tas
t h e Hc f o r d e a t h s from c a n c e r .MALES
GOFLRST
Consider now a population t h a t i s s t r u c t u r e d according
to
r a c e , s e x , socio- economic s t a t u s , region or some o t h e r classification. Adopting t h e line of a t t a c k of t h e second-chance a p p r o a c h , suppose t h a t a proportion d i of t h e f i r s t d e a t h s in g r o u p iare
a v e r t e d . What will t h e e f f e c t b e on t h e life e x p e c t a n c y of t h e e n t i r e population? Letting & (z), Li (z), and ei (z) denote t h e f o r c e of mortality,sur-
vivorship function, and remaining life expectancy a t a g ez
of t h e i -th group, t h e nw h e r e mi (0) i s t h e initial p r o p o r t i o n of t h e population
in
t h e g r o u p i . Hence.a n d
w h e r e
and
The
U.S.
male population, for example, might b e classified as white and nonwhite. The value of Hi f o rU.S.
nonwhite males in 1950 w a s a b o u t 0.038. S o reducing nonwhite male mortality byone
p e r c e n t would a d d a b o u t 9 d a y sto
t h e o v e r a l lU.S.
male life e x p e c t a n c y of 65.5 y e a r s . By comparison, t h i s r e d u c t i o n in nonwhitemale
mortality would a d d a b o u t 75 d a y sto
nonwhitem a l e
life e x p e c t a n c y . The d i f f e r e n c e i s l a r g e l y explained b y t h e proportion of nonwhitesat
b i r t h , a b o u t 12.6 p e r c e n t .The U.S. population as a whole can b e divided into male and female groups.
The value of H f o r males
at
1980 mortality r a t e s w a s 0.193, t h e value f o r females w a s 0.155. If t h e two g r o u p s are given equal weight, t h e n H f o r t h e e n t i r e popula- tion is 0.179 and Hi i s 0.096 f o r m a l e s and 0.077 f o r females. Suppose t h e r eare
t h r e e a l t e r n a t i v e interventions. The f i r s t reduces m a l e mortality by 2 p e r c e n t , t h e second r e d u c e s female mortality by 2 p e r c e n t , and t h e t h i r d r e d u c e s t o t a l mor- tality by 1 p e r c e n t . The male s t r a t e g y would save a b o u t 11 p e r c e n t more l i f e y e a r s t h a n t h etotal
s t r a t e g y which, in t u r n , would save a b o u t 15 p e r c e n t more life y e a r s t h a n t h e female s t r a t e g y .Suppose t h a t
a
population consists oft w o
subpopulations with a g e s p e c i f i c mortalityrates
k ( z ) a n d @(z), where @(z)>
k ( z ) and where t h et w o
g r o u p s might b e r e s i d e n t s of u r b a n vs. r u r a l a r e a s , smokers vs. non-smokers, blue-collar workers vs. w h i t e c o l l a r workers. people in t h e south ofa
country vs. people in t h e north, people whoare
overweight vs. people who a r e not, e t c . How will changesin
t h e mix of t h e population between t h e s et w o
g r o u p s a f f e c t life expec- tancy?Consider a n intervention t h a t changes n(z), t h e p r o p o r t i o n of t h e population in t h e high-risk g r o u p , by
some
p r o p o r t i o n6 at
all a g e s a f t e r s o m e initial a g ezo:
I t i s convenient
to
c o n s i d e r a g ez o
t h e a g eat
"birth", s o t h a t e o r e f e r sto
remain- ing life expectancyat
a g ez o
andz
r e f e r sto
y e a r s of a g e s i n c ezo.
The f o r c e of mortality f o r t h e population asa
whole is given byand
Hence.
I t follows t h a t
If d is small,
--
tit e 0- p H , .
where
and
A s a n example of t h e u s e of t h e s e formulas, suppose t h a t t h e population con- s i s t s of non-smokers a n d smokers, a n d t h a t t h e population i s being studied s t a r t i n g
at
a g e 35 (so t h a t e o r e f e r sto
life expectancyat
a g e 35). F u r t h e r suppose t h a t t h e f o r c e of mortality f o r non-smokers i s . 0 0 1 s . ~ , ( t being a g e minus 35). t h a t t h e f o r c e of mortality f o r s m o k e r s i s twiceas
high, and t h a t half t h e population smokesat
a g e 35. Remaining life e x p e c t a n c y f o r non-smokers in t h i s c a s e i s a b o u t 40.8 y e a r s and remaining life e x p e c t a n c y f o r smokers about 34.2 y e a r s . ThenH I
t u r n s o u tto
equal 0.077. If t h e p r o p o r t i o n of t h e population t h a t smokes i s r e d u c e d by 1 p e r c e n t , t h e n life e x p e c t a n c y ( a t a g e 35) will i n c r e a s e by 0.077 p e r c e n t ,or
by a b o u t 11 days, given t h e a v e r a g e remaining life expectancy f o r t h e population asa
whole of 37.5 y e a r s .More generally, i t is i n t e r e s t i n g t o investigate t h e values of HI, and of ex- p e c t e d d a y s of life saved,
at
d i f f e r e n t s t a r t i n g a g e s , i.e.,at
different a g e s of in- tervention. Table 2 p r e s e n t s some sample calculations. Note t h a t H1 i n c r e a s e s with age: a reduction in smoking yields a g r e a t e r proportional i n c r e a s e in life ex- pectancyat
t h e a g e s with t h e highest mortalityrates.
The absolute i n c r e a s e in life expectancy, however, as measured by days added, falls off with age. Because i t falls off slowly,at
l e a s t b e f o r e a g e 55or
65, it may b e optimalto
t a r g e t anti- smoking interventions toward o l d e r people-if i t i s relatively e a s i e rto
induce old- e r smokersto
quit. The calculations in Table 2are
merely illustrative, b u t some empirical analysis of t h i ssort
could s h e d light on t h e effectiveness of t a r g e t i n g v a r i o u s kinds of h e a l t h p r o g r a m s toward individuals in different a g e classes.Table 2. Values of HI, life expectancy. and days added
to
life expectancy if t h e p r o p o r t i o n of a population t h a t smokes i s reduced by o n e p e r c e n t ,at
various ages.Days added
to total
life e x p e c t a n c y Remaining life e x p e c t a n c y (in y e a r s ) f o r : if p r o p o r t i o n t h a tsmoke i s r e d u c e d Age H1 Non-smokers Smokers Total population by o n e p e r c e n t
35 .077 40.8 34.2 37.5 10.5
45 .095 31.4 25.2 28.3 9.9
55 .I20 22.6 17.1 19.9 8.7
65 .I50 14.9 10.5 12.7 7.0
75 . l a 4 8.8 5.6 7.2 4.8
85 .215 4.5 2.7 3.6 2.8
INHIBITING IM.BIBITIONS
The r e s u l t s in t h e p r e v i o u s s e c t i o n c a n b e g e n e d i z e d
to
t h e case where t h e population consists ofN
subpopulations with age-specific mortalityrates
~4 ( 2 ) a n d in p r o p o r t i o n s ni ( z ). wherea n d
A s before, l e t t h e f i r s t subpopulation b e t h e healthiest, % ( z )
<
& ( z ) for a l lz
and f o r allt >
1, and l e t6(
denote t h e change in proportionsat
s t a r t i n g a g e 0:Clearly,
For simplicity, assume
6 , = d , all i > l
.
Then i t i s not difficult ta show
-
'~ ( 2 )
-22) =
d(*(z)- 2 z ) )
This formula i s identical ta (53). Consequently,h e 0
-
w d H l,
f o r s m a l l d ,=
0and
where HI i s defined as b e f o r e by (57) and
P
=
8t , a l l i > l.
~ ( 2
That HI i s t h e same
as
b e f o r e may,at
f i r s t glance, seem puzzling but, on c l o s e r thought, i t i s r e a s o n a b l e because t h e assumptions group t h e sub-populations intot w o
p a r t s . Other formulas can b e readily derived f o r o t h e r special cases.A s
an
illustration of t h e use of (66), considera
population of males with a high prevalence of alcoholism.In
p a r t i c u l a r , assume t h a t 5 0 p e r c e n t of t h e population d r i n k moderatelyor
notat
all, t h a t 30 p e r c e n t drink heavily, and t h a t t h e remain- ing 20 p e r c e n t d r i n k v e r y heavily. F u r t h e r , assume t h a t t h e heavy d r i n k e r s h a v e twice t h e mortality and t h e v e r y heavy d r i n k e r s have f o u r times t h e mortality of t h e f i r s t group. Finally,as
in t h e previous example, suppose t h a t t h e population i sbeing considered s t a r t i n g
at
a g e 35 and t h a t t h e f o r c e of mortality follows a Gom- p e r t z c u r v e with a=
0.001 and b=
0.1 for t h e h e a l t h y subpopulation.Remaining l i f e e x p e c t a n c y
for
t h e t h r e e g r o u p s t u r n s outto
b e 40.8, 34.2, and 27.9 y e a r s and, f o r t h e populationas
a whole, 36.2 y e a r s . The value ofHi
i s 0.108;a o n e p e r c e n t r e d u c t i o n in t h e p r o p o r t i o n of heavy and of v e r y heavy d r i n k e r s , would add
t w o
weeksto
t h e population's life e x p e c t a n c y .STARTING STOPPING
Now c o n s i d e r a population t h a t consists of v a r i o u s subpopulations, with indlvi- duals making t r a n s i t i o n s f r o m o n e subpopulation
to
a n o t h e r , such t h a t t h e transi- tionrates are
changingor can
b e changed. F o r instance, t h e population may con- s i s t of s m o k e r s and non-smokers, withsome
s m o k e r s who s t o p and some non- smokers who start. If e i t h e r of t h e s e t r a n s i t i o n s could b e influenced, what would t h e e f f e c t b e o n life e x p e c t a n c y ? This question i s similarto
t h e question con- s i d e r e d in t h e p r e v i o u st w o
sections, e x c e p t now t h e policy l e v e ror
c o n t r o l p a r a m e t e r i s n o t t h e p r o p o r t i o n of t h e population who smoke, but t h e t r a n s i t i o nrates
between t h e non-smoking and smokingstates.
Changing t h e transitionrates
will change t h e p r o p o r t i o n s a n d h e n c e life e x p e c t a n c y .F o r a c o h o r t , t h e c h a n g e in t h e p r o p o r t i o n of individuals in
state
( o r group) jat
a g er
i s given by t h e equation:where
Xu
( r )are
t h e t r a n s i t i o nrates
fromstate
ito state
j a t a g e r , with t h e ini- t i a l p r o p o r t i o n s rrj (0) given.In t h e simplest
case
of atwo-state
population with mortalityrates
h ( r ) and& ( r ) a n d t r a n s i t i o n
rate
X ( r ) fromstate
1to state
2, U l e p r o p o r t i o n n ( r ) of indi- viduals instate
2 i s t h e solution of t h e following equationwith rr(0) given. Let t h e
rate
of p r o g r e s s in r e d u c i n g X(r) b e given by p ( r ) :Straight-forward calculations show
w h e r e q ( z ) is t h e solution of t h e d i f f e r e n t i a l equation
with q ( 0 )
=
0. Note t h a t t h i s equation h a sto
b e solved t o g e t h e r with equation ( 7 0 ) f o r n ( z ) . If t h erate
of p r o g r e s s in d e c r e a s i n g X(z) d o e s not depend o n a g e , t h e n ( 7 2 ) r e d u c e sto
w h e r e
SIMULATION AS A SLEDGEHAMMER
Solving ( 7 5 ) for
H A
i s n o t e a s y , s i n c e q ( p ) i s t h e solution of a d i f f e r e n t i a l equation (73) t h a t d e p e n d s o n a n o t h e r differential equation ( 7 0 ) . When mathemati- c a l solutions g e t as complicatedas
this, t h e y may not only l o s e elegance but a l s o usefulness for e i t h e r insightor
computation. I t may t h e n b e f r u i t f u lto
t a k e a dif- f e r e n t t a c k and r e l y o n numertcal, computer simulation.C o n s i d e r , f o r i n s t a n c e , t h e following i l l u s t r a t i v e model:
N O N S M O K E R S .
k(i:
*(i:1
*(1:1
( d e a t h ) X12(z
*
The population i s divided i n t o t h r e e groups-non-smokers. s m o k e r s , a n d q u i t t e r s . T h e s t a r t i n g p o i n t of t h e a n a l y s i s i s a g e 10:
z
r e p r e s e n t s a g e minus 10. F o r non- s m o k e r s , t h e f o r c e of mortality i s given byf o r s m o k e r s i t i s
S M O K E R S
&
a n d f o r q u i t t e r s ,
& ( z )
=
1 . 5 h ( z ).
To b e g i n with a l l individuals
are
non-smokers:nl(0)
=
1 ,n2(0)
=
n3(0)=
0.
The t r a n s i t i o n i n t e n s i t i e s
are
xl,(z)=
.o6r-Sh ,=.028.O* ,
X32(2 )
=
.5 8 -aoa ,A Z 3 ( ~ )
C
*
X32(zQUITTERS
These transition intensities imply:
-
about 6 p e r c e n t of non-smokersstart
smoking a t a g e 10. a b o u t 2 p e r c e n tat
a g e 20, and l e s s t h a n 1 p e r c e n tat
a g e 30;-
t h e p r o p o r t i o n of s m o k e r s who quit smoking r i s e s f r o m about 2 p e r c e n t p e r y e a rat
a g e 1 0 to 1 0 p e r c e n t p e r y e a rat
a g e 50 and 22 p e r c e n t p e r y e a rat
a g e 70;-
t h e recidivismrate
of q u i t t e r s resuming smoking falls from 5 0 p e r c e n t p e r y e a r at a g e 1 0to
33 p e r c e n tat
a g e 30 and 15 p e r c e n tat
a g e 70;-
1 0 p e r c e n t of q u i t t e r s become non-smokers e a c h year. implying t h a t i t takes t e n y e a r s , on a v e r a g e , for a f o r m e r smokerto
r e t u r nto
t h e health s t a t u s of a non-smoker.The following formulas a n d approximations c a n b e used
to
analyze t h i s model:where
where
where t h e pi
Cf),
t h e p r o p o r t i o n s of t h e original c o h o r t t h a t are instate
iat
time1 ,
are given bywhere
and
With t h e p a r a m e t e r values given above, remaining life expectancy a t a g e 10 t u r n s o u t t o b e 61.5 y e a r s . The p r o p o r t i o n of t h e surviving population t h a t smokes r i s e s t o 33 p e r c e n t a t a g e 30 and t h e n f a l l s off
to
23 p e r c e n tat
a g e 50 a n d 6 p e r - c e n t a t a g e 70.The model c a n b e used
to
e x p l o r e various kinds of interventions. If no o n e e v e r smoked, o r if t h e health h a z a r d s of smokingwere
eliminated, life e x p e c t a n c y would i n c r e a s e b y 1.4 y e a r s . If t h erate at
which people beganto
smokewere
c u t in half, life e x p e c t a n c y would i n c r e a s e b y 0.6 y e a r s . If t h erate at
which people gave up smoking doubled, t h e gain would b e 0.4 y e a r s . If t h erate
of recidivism could b e c u t in half, 0.3 y e a r s would b e gained; if recidivism could b e eliminated, t h e i n c r e a s e in l i f e e x p e c t a n c y would b e 0.7 y e a r s . If t h e duration of t h e lingering e x c e s s r i s k s f a c e d b y f o r m e r smokers could b e c u t from a n a v e r a g e of 10 y e a r sto
a n a v e r a g e of 5 y e a r s , 0.3 y e a r s would b e addedto
life expectancy. Finally, if t h e e x c e s s r i s k of smokingwere
c u t In half, so t h a t ~ 1 2 equaled 1.5~. r a t h e r t h a n 2 p , about halfa
y e a r would b e gained.This example p r o v i d e s a simple illustration of how micro-simulation c a n s h e d light o n models t h a t
are
difficultto
analyze formally. More e l a b o r a t e , more r e a l i s - t i c models f o r t a r g e t analysis c a n b e handled in t h esame
g e n e r a l way.CONCLUSION
The life e x p e c t a n c y of individuals ( o r units) in a heterogeneous population c a n b e i n c r e a s e d b y numerous s t r a t e g i e s , including
-
lowering o v e r a l l mortality ( o r f a i l u r e ) rates.-
reducing mortalityrates
in s p e c i f i c a g e c a t e g o r i e s ,-
a v e r t i n g d e a t h s ,-
lessening mortalityrates
f r o m some c a u s e ,-
diminishing mortalityrates
in some regionor
f o rsome
population g r o u p ,-
decreasing t h e p r o p o r t i o n of individuals in high-risk g r o u p s , and-
changing t r a n s i t i o n rates between r i s k groups.A s t h e various formulas d e r i v e d ln t h i s p a p e r i l l u s t r a t e , t h e s e changes a f f e c t life expectancy in d i f f e r e n t ways.
The formulas, a n d v a r i o u s extensions or adaptations of them, may b e useful
to
policymakers in t a r g e t analyses of t h e benefits of a l t e r n a t i v e interventions intend- e dto
s a v e lives. In addition, t h e formulas d e s c r i b e t h e linkages t h a t e x i s t between population s t r u c t u r e and life expectancy. Individuals d i f f e r o n numerous dimen- sions t h a t are r e l a t e dto
mortality chances. including a g e ,sex. race.
s o c i e economic s t a t u s . occupation. p l a c e of residence. and personal behavior. A c h a n g e in population s t r u c t u r e along any of t h e s e dimensions will change life expectancy.Four d i f f e r e n t a p p r o a c h e s
were used to
analyze t h e impact ofa
change in po- pulation s t r v c t u r e o n life expectancy: t h e comparative-statics a p p r o a c h . t h e dynamics a p p r o a c h . t h e method w e called t h e "second-chance" a p p r o a c h . and com- p u t e r simulation. The f i r s t t h r e e a p p r o a c h e s yield analytical solutions t h a tare
g e n e r a l a n d t h a t may f a c i l i t a t e insight. In t h e limit. when d i s small. t h e t h r e e ap- p r o a c h e s p r o d u c e equivalent formulas,so
which a p p r o a c hto
a d o p t i sto some
ex- t e n ta matter
oftaste
a n d convenience. The t h r e e a p p r o a c h e s . however. may not b e equivalent when d i s not small. a n d e a c h a p p r o a c h may yielda
d i f f e r e n t insight a n d providea
d i f f e r e n t p e r s p e c t i v e . Computer simulation is useful in a t t a c k i n g complex models t h a t d o not yieldto
t h e o t h e r t h r e e a p p r o a c h e s . The answers p r o - duced by simulation p e r t a i nto
p a r t i c u l a r realization ofa
model in which t h e coef- f i c i e n t sare
specified: t h e answersare
t h u s not g e n e r a l o r elegant. b u t t h e yare
answers.REFERENCES
Demetrius, Lloyd (1979) Relations Between Demographic P a t t e r n s . Demography 2(16):329-338.
Keyfitz. Nathan (1977) Applied Mathematical Demography. N e w York: John Wiley a n d Sons.
Yaupel, James W. (1986) How Change in A g e S p e c i f i c Mortality Affects Life Expec- tancy. P o p u l a t i o n S t u d i e s , March.
Vaupel, James W. a n d Anatoli I. Yashin (1985) How Many Times Has Your Life Been Saved? Forthcoming Working P a p e r . Laxenburg. Austria: International Insti- t u t e for Applied Systems Analysis.