INTRODUCTION TO LIFE TABLE ANALYSIS

1
INTRODUCTION TO LIFE TABLE ANALYSIS

• with applications to summary measures of
  population health

2
QUESTIONS

• How long will we live, given a series of
  age-specific mortality risks? How many years
  would we gain if we could eliminate a particular
  cause of death?
• How long will we live in good health, given a
  series of age-specific mortality and morbidity
  risks? How many years in good health do we
  currently loose as a result of a particular
  disease?
• How can life table analysis be used to summarize
  the health of a population? How useful are these
  summary measures of population health?

3
CONTENTS

• Simple life table calculation of life
  expectancy at birth
• Cause-elimination life-table calculation of
  life expectancy at birth after elimination of a
  specific cause of death
• Health expectancy calculation of life
  expectancy at birth, taking into account years
  spent in bad health
• Disability-adjusted life years calculation of
  years in good health lost

4
SIMPLE LIFE TABLE (1)7 COLUMNS
5
SIMPLE LIFE TABLE (2) FIRST LINE
6
SIMPLE LIFE TABLE (3) HOW TO GET FROM COLUMN
2 TO COLUMN 7?
7
ABRIDGED LIFE TABLE, MEN, THENETHERLANDS, 1980 -
1984 (1)
8
ABRIDGED LIFE TABLE, MEN, THENETHERLANDS, 1980 -
1984 (2)
9
SIMPLE LIFE TABLE (4)SURVIVAL CURVE
10
SIMPLE LIFE TABLE (4) PERIOD VERSUS
COHORT LIFE TABLE

• Most life tables are period life tables,
  calculated from age-specific mortality risks
  observed in a population during one period of
  time
• Can only be interpreted as expectation of life,
  if one assumes that these cross-sectional
  mortality risks will continue to apply during the
  life-time of the cohort of 100000
• Alternative cohort life-table, calculated from
  longitudinally observed mortality risks

11
COHORT LIFE TABLES Dutch men and
women 1891 - 1965
12
LIFE EXPECTANCY AT BIRTH (IN YEARS), MALES, 2003
13
CAUSE-ELIMINATION LIFE TABLE (1)

• What would be life expectancy at birth if a
  specific cause of death would be eliminated?
• Used for- quantification of potential effect of
  interventions
• - quantification of (relative) importance of a
  cause of death

14
CAUSE-ELIMINATION LIFE-TABLE (2)

• Modify qx
• e.g. qx-i qx - qxi
• in which
• qx-i mortality risk in interval starting at x
  after elimination of cause i
• qxi mortality risk in interval starting at x
  due to cause i only

15
CAUSE ELIMINATION LIFE TABLE (3)
16
CAUSE-ELIMINATION LIFE TABLE (4)

• What difference would it make if cancer were
  suddenly eradicated as a cause of death?
• Ca. 25 of all deaths but only 2.5 years
  (3.5) gain in life-expectancy Tauber paradox

17
CAUSE-ELIMINATION LIFE TABLE (5)

• Explanation average age at death from cancer is
  high - death rates from other causes among those
  saved from cancer are high
• In year after elimination of cancer, crude
  mortality rate may be up to 25 lower but in
  following years crude mortality rate will
  gradually increase again when survivors start to
  die from other causes
• Actually, effects on life expectancy of
  eliminating cancer will be smaller still, due to
  common risk factors between cancer and other
  causes of death

18
SUMMARY MEASURES OF POPULATION HEALTH (1)

• Health expectancy measures- Healthy life
  expectancy (HLE)- Disability-free life
  expectancy (DFLE)- Health-adjusted life
  expectancy (HALE)- Disability-adjusted life
  expectancy (DALE)
• Health gap measures- Healthy life years (lost)
  (HeaLY) - Disability-adjusted life-years (lost)
  (DALY)

19
SUMMARY MEASURES OF POPULATION HEALTH (2)
20
HEALTH EXPECTANCY MEASURES

• Usually calculated with Sullivan method,
  combining life table with prevalence of disease
  or disability- Take simple life table-
  Modify column 5 (years lived during interval)-
  For years spent in good health, multiply by (1
  prevalence) of bad health - For years
  adjusted for bad health, multiply by (1
  prevalence) of states of bad health , weigh
  years with severity weight between 0 and 1, and
  add
• Alternative multistate life tables in which
  transitions from good health to bad health to
  death are modelled directly, i.e. combining life
  table with incidence of disease or disability

21
DISABILITY-ADJUSTED LIFE EXPECTANCY,
2000DISABILITY WEIGHTS USED IN CALCULATION
Source Murray Lopez 1997
22
DISABILITY-ADJUSTED LIFE EXPECTANCY, 2000MAIN
RESULTS FOR 8 WORLD REGIONS
Source Murray Lopez 1997. EMEEstablished
Market Economies, FSEFormer Soviet Economies,
CHNCHINA, LACLatin America and Caribbean,
OAIOther Asia and Islands, MEC Middle Eastern
Crescent, INDIndia, SSASub Saharan Africa
23
HEALTH GAP MEASURES

• Inspired by measures of years of life lost due to
  premature mortality (e.g. Potential Years of Life
  Lost)
• Measure difference between current situation and
  a health target (or norm), e.g. perfect health
  until the age of 80 years
• Can (more easily than health expectancy measures)
  be used for attribution to risk factors

24
DISABILITY-ADJUSTED LIFE-YEARS (DALYs)CALCULATIO
N

• DALY YLL YLDYLL Years of Life LostYLD
  Years Lived with Disability
• YLL N x L N Number of deathsL Life
  expectancy at age of death in years
• YLD I x DW x LI Incident casesDW
  Disability weightL Average duration of case
  until remission or death
• (Additional) social preferences can be built in
  explicitly, e.g. by giving less weight to years
  lived and lost at older ages

25
DISABILITY-ADJUSTED LIFE-YEARS (DALYs)DIFFERENCE
S BETWEEN WORLD REGIONS
Source Lopez et al. 2006
26
DISABILITY-ADJUSTED LIFE-YEARS (DALYs) PROJECTED
CHANGES IN WORLD-WIDE RANK ORDER FOR 15 LEADING
CAUSES
Source Lopez Murray 1998
27
CONCLUSIONS

• Life table analysis and its extensions
  (cause-elimination, health expectancy,
  health gaps) provide extremely powerful
  techniques for analysing population health
• Important requirements for data collection (e.g.
  vital statistics, epidemiology of diseases and
  their consequences, severity weights), and
  necessary assumptions (e.g. age weights) should
  however not be overlooked