Summary measures of average population health: an introduction

1
Summary measures of (average) population health
an introduction

• John Powles
• MPhil 2004

2
Summary measures of (average) population health
Essence

• Combine information on mortality and non-fatal
  health states to measure levels of health in
  populations
• Intrinsically normative
• In construction and use

3
Uses

• Comparing health levels in populations
• Comparing health levels in 1 population at
  different times
• Measuring inequalities
• Giving appropriate attention to non-fatal
  conditions
• Informing consideration of priorities

4
Unidimensional or multidimensional?

• Money is unidimensional!
• So resource allocation decisions
• even where only implicit
• imply a single (unidimensional) ranking
• (but Sander Greenland argues for multidimensional
  measures in SMPH)

5
Summary measures

• Need to be distinguished from their data inputs
• Instruments to measure health / disability levels
  in the living
• are normative
• Choices and value judgements are used in their
  construction and use

6
Desirable attributes of summary measures of health

• Should be sensitive to all types of health loss
• cf measures using thresholds eg disability-free
  life expectancy
• Should only take account of age and sex
• not eg country of residence
• Should treat like health states as like
• Should use metric of time
• rather than event rates

7

• Why metric of time?

8
Deaths from stroke and RTAs comparisons on 2
measures, East Anglia 19901. Event (crude
death) rates
9
Deaths from stroke and RTAs comparisons on 2
measures, East Anglia 19902. Lost lifetime
Assuming all decedents would otherwise have
survived to 75
10
Deaths from stroke and RTAs comparisons on 2
measures, East Anglia 1990
Assuming all decedents would otherwise have
survived to 75
11
Occurrence measures in public health

• Studying causation
• Metric of incidence typically optimal
• Comparing health levels
• Metric of time typically optimal
• Ie healthy time lived (expectancies) or lost
  (gaps)

12
Life tables summarise current death rates in
terms of average time lived
e0 50
Area under curve represents person-time lived by
those who shared it ie 100000
13
Allowing the clock to run forward
Snapshots in time
14
Summarising time spent at different health levels

• Divides each lifetime into
• A part lived in full health (A)
• A part lived in less than full health (B)

C
B
A
15
2 families of measures

• Health expectancies
• A f (B)
• Where full health 1
• Eg DALE
• Health gaps
• C g (B)
• Where 1 is equivalent to
  death
• Eg DALY

C
B
A
16
Health expectancies

• Active life expectancy
• Disability-free life expectancy
• Disability-adjusted life expectancy (DALE)
• Years of healthy life
• Quality-adjusted life expectancy
• etc

17
Attributes of health expectancies

• Period or cohort
• Calculation method
• Prevalence / double decrement / multi-state
• Definition and measurement of health
• Eg Active le lt- measures of ADL
• Methods used to value health states

18
f (B)
19
Measuring health levels in the living

• Controversial
• Subjective preferences (utilities)
• Consistent with economic authodoxy

20

• But
• Self-reported health can show marked regressive
  bias when used in comparisons between populations

21
Life expectancy in India compared to the US, mid
1990s
Sen, BMJ 2002 324, 861
22
Self-reported morbidity, India (mid-1970s) and US
(mid-1980s)
Sen, BMJ 2002 324, 861
23
Measuring health levels in the living

• Controversial
• Subjective preferences (utilities)
• Consistent with economic authodoxy
• But, especially for group comparisons,
• Self reports subject to serious (usually
  regressive) bias

24
Methods used to value health states

• Whose values
• Individuals in states / general public /
  professionals / household carers
• Type of valuation question
• Standard gamble / time trade-off / person
  trade-off / visual analogue
• How health states are presented
• Range of health states valued
• Valuation process ? deliberative

25
Essence of health gaps (eg DALY)

• Amount of healthy life lost relative to some norm

26
Attributes of health gaps (eg DALY)

• Implied target or norm
• Some vary with mortality level in population
• How health states defined measured
• ? Dimensions
• ? Self-perception vs observation
• Method used to value health states
• Inclusion of other values
• Eg Age-weighting, time-preference, equity weights

27
Age structure dependence

• Health expectancies intrinsically age-independent
• Health gaps
• Age-dependent when expressed in absolute terms
• (DALYs lost in population X in year Y)
• But can be age-standardised

28
Desirable properties of summary measures for
comparative uses

• Criteria for optimality of a measure should not
  be confused with criteria for resource allocation
• Murray et al use Rawlsian veil of ignorance
  approach to specify criteria

29
Criteria for summary measures for comparative uses

• If, for a given cause/health state in any
  given age group, everything else being equal
• Mortality ? ? measure ?
• Measures using gaps relative to current
  population fail
• Prevalence ? ? measure ?
• Incidence-based measures fail
• Incidence ? ? measure ?
• Prevalence-based measures fail
• Remission ? ? measure ?
• Severity (within a given state)? ? measure ?

30
Other desirable properties of health measures

• Comprehensibility and feasibility
• Eg life expectancy does well despite complexity
• Additive decomposition
• Ie Total contribution of (a b c)
• For eg disease groups or risk factors
• All health expectancy measures fail

31

• Losses (gaps) can be attributed to diseases or to
  determinants
• health expectancies can not

C
B
A
32
Calculating the contribution of diseases,
injuries and causes

• Only possible with Gap measures
• Methods
• Categorical
• Eg ICD rules TB with HIV is assigned to HIV
• But ?myocardial infarction in diabetic or liver
  cancer with chronic hep B
• Counterfactual
• Compare current or future with expected under
  specified alternative

33
But health expectancies can be more readily
understood

• and so appeal to journalists and politicians
• DALE used in World Health Report 2000

C
B
A
34
Health expectancies

• Active life expectancy
• Disability-free life expectancy
• Health-adjusted life expectancy (HALE)
• Years of healthy life
• Quality-adjusted life expectancy
• etc

35
Conclusions

• Some measures fail basic tests
• Eg those using internal mortality norms or
  simple dichotomies for less than perfect health
  (eg disability-free life expectancy)
• None simultaneously fulfil prevalence and
  incidence criteria
• Only gap measures (eg DALYs) permit
  decomposition by conditions and causes

36
The marginalist critique (Williams et al)

• Summary measures (totals) not useful for policy
• What we need to know is the next best thing to do
  at the margin
• But
• why not use both totals and marginals?
• Totals also protect against partisan use of
  epidemiology (disease advocacy)

37
Conclusion

• The DALY is in use
• Its derivation and characteristics need to be
  understood by public health professionals