population ageing and health expenditure - ministry … ageing and health expenditure: new zealand...

Population Ageing and Health Expenditure:

New Zealand 2002–2051

Public Health Intelligence Occasional Bulletin No 22

Citation: Ministry of Health. 2004. Population Ageing and Health Expenditure: New Zealand 2002–2051. Wellington: Ministry of Health.

Published in October 2004 by the Ministry of Health

PO Box 5013, Wellington, New Zealand

ISBN 0-478-25703-1 (Book) ISBN 0-478-25704-X (Internet)]

HP 3882

This document is available on the Ministry of Health’s website: http://www.moh.govt.nz

Acknowledgements The model was built by John Bryant and Audrey Teasdale (New Zealand Treasury), Martin Tobias and Jit Cheung (Public Health Intelligence, Ministry of Health) and Mhairi McHugh (Funding Policy, Ministry of Health). The report was written by Martin Tobias and peer reviewed both within the Ministry of Health and Treasury and by external experts. The team is grateful for the constructive advice received from the peer reviewers.

Disclaimer Views expressed in this report are solely those of the authors, and do not necessarily reflect the opinions of the peer reviewers or the policy advice of the Ministry of Health or the New Zealand Treasury.

Population Ageing and Health Expenditure: New Zealand 2002–2051 iii

Foreword In approximately a decade from now the population of New Zealand will begin a period of rapid structural ageing as the large ‘baby boom’ cohorts of 1945–75 reach retirement age. This will transform the age distribution of the population, such that by the late 2040s the proportion of the population aged 65 years and over will have doubled – from 12% at present to approximately 24%. At the same time, the proportion of the population aged 85 years and over will increase more than fourfold, from approximately 1.3% to 5.5%. This unprecedented demographic transformation will create a number of challenges, not least for the health sector. Some of these challenges are explored in this joint project of the Ministry of Health and the New Zealand Treasury. The project attempts to answer such questions as: Could improvement in the health status of future cohorts of older people offset ‘ageing pressure’ on health expenditure? How might changes in health relative to life expectancy affect future health care consumption patterns? What will be the impact of population ageing on intergenerational equity of health funding? The key findings are presented from a health perspective in this report, Population Ageing and Health Expenditure: New Zealand 2002–2051. The results are also presented in greater technical detail in a companion Treasury working paper that adopts more of an expenditure perspective. Both reports provide useful insights for policy analysts, health workers and researchers interested in such matters as sustainable funding paths for the health system and the making of investment decisions in health technologies. Comments on this report are welcomed, and should be sent to Public Health Intelligence, Public Health Directorate, Ministry of Health, PO Box 5013, Wellington.

Don Matheson Deputy Director-General Public Health Directorate

iv Population Ageing and Health Expenditure: New Zealand 2002–2051

Contents

Foreword iv

List of Tables vii

List of Figures vii

Executive Summary viii

Introduction 1

Drivers of Health Expenditure 2 Population size 2 Population ageing 2 Health status 3 Interaction between fatal and non-fatal outcomes 6 The complex influence of mortality 7 Coverage and prices 7 Interaction between population ageing, health status, and coverage and prices 8

The Model 12 Framework 12 Projection 13 Sensitivity analysis 14

Operationalisation of the Model 15 Base year 15 Projection period 15 Back-casting period 15 Population size and age structure 15 Mortality 15 Disability 16 Growth in coverage and prices 16 Growth in GDP per worker (labour productivity) 17 Projection scenarios 17 Base costs (health expenditure in the launch year 2001/02) 18

Results 20 Results by scenario 20 Trajectories 20 Expenditure in relation to GDP 20 Distribution of health expenditure across life-cycle stages 20 Drivers of health expenditure 20

Population Ageing and Health Expenditure: New Zealand 2002–2051 v

Results by scenario, for selected years 20 Expenditure in relation to GDP 22 Distribution of expenditure across life-cycle stages 23 Drivers of health expenditure 24 Interaction of health status dimensions: health expectancy and health expenditure 27

Discussion 30 Limitations 30 Key findings 32 Conclusion 34

Appendix 1: International Trends in Disability Prevalence 37 Method 37 Criteria 37 Analysis 38 Results 39

Appendix 2: Model Mathematics 41 Person-years 41 Costs 41 Summation 41

Appendix 3: Key Results by Scenario 43

References 53

vi Population Ageing and Health Expenditure: New Zealand 2002–2051

List of Tables Table 1: Health expenditure, distance from death and disability prevalence, 2001 5 Table 2: Health states included in the model 12 Table 3: Average annual percentage changes, by projection scenario, 2002–2051 (%) 17 Table 4: Summarised results, by scenario, for 2026 20 Table 5: Summarised results, by scenario, for 2051 21 Table 6: Health expenditure as a percentage of GDP, by scenario, 2051 23 Table 7: Share of health expenditure, by life-cycle stage, 2002 and 2051 23 Table 8: Percentage growth in per capita expenditure, by selected age group, scenario E,

2002–2051 24 Table 9: Impact of growth in coverage and prices, 2051 25 Table 10: Impact of change in disability prevalence, 2051 25 Table 11: Impact of reduction in mortality, 2051 26 Table 12: Estimates and projections of life expectancy at birth, 1951–2051, genders pooled 27 Table 13: Selected scenarios, illustrating expansion and compression of morbidity, 2051 28 Table A1.1: Selected studies 38 Table A1.2: Prevalence of disability 39 Table A1.3: Average annual percentage change (exponential) 39 Table A3.1: Scenario A (COV = 1.0, MORT = 1.0, DIS = 0.0) 43 Table A3.2: Scenario B (COV = 1.0, MORT = 1.0, DIS = 0.5) 44 Table A3.3: Scenario C (COV = 1.0, MORT = 1.5, DIS = 0.0) 45 Table A3.4: Scenario D (COV = 1.0, MORT = 1.5, DIS = 0.5) 46 Table A3.5: Scenario E (COV = 1.5, MORT = 1.5, DIS = 0.5) (central scenario / base case) 47 Table A3.6: Scenario F (COV = 1.5, MORT = 1.5, DIS =1.0) 48 Table A3.7: Scenario G (COV = 1.5, MORT = 2.0, DIS = 0.5) 49 Table A3.8: Scenario H (COV = 1.5, MORT = 2.0, DIS = 1.0) 50 Table A3.9: Scenario I (COV = 2.0, MORT = 1.5, DIS = 0.5) 51 Table A3.10: Scenario J (COV = 2.0, MORT = 2.0, DIS = 1.0) 52 List of Figures Figure 1: Older people (65+ and 85+) as a proportion of the population, 2001–2051 2 Figure 2: Annual government per capita health expenditure, by age and service group,

genders pooled, 2001/02 3 Figure 3: Prediction of age specific health expenditure from distance to death and disability,

2001 6 Figure 4: Visualisation of the drivers of per capita health expenditure 9 Figure 5: Diagrammatic representation of the drivers of per capita real health expenditure,

based on the age-cost curve 10 Figure 6: Total health expenditure (constant 2002 dollars) and expenditure as a percent of

GDP 19 Figure 7: Share of health expenditure, by lifecycle stage, selected years 24

Population Ageing and Health Expenditure: New Zealand 2002–2051 vii

Executive Summary From approximately 2010 until 2040 the New Zealand population will undergo rapid structural ageing as the baby boom cohorts reach retirement age. This will inevitably increase pressure on government health expenditure (defined here as Vote Health, which includes disability support services), raise concern about intergenerational equity of such spending, and pose complex ethical issues about investment in disability-reducing versus life-extending interventions. This report, a joint Ministry of Health – Treasury study, updates and extends previous work carried out by both organisations on this issue. The report is based on a macrosimulation model that disaggregates the New Zealand population (from 1951 to 2051) by age, gender and health status. Health status is described along two dimensions – mortality and morbidity (the latter indexed by disability requiring assistance, adjusted for severity). Costs for each age by health state reflect non-demographic drivers of expenditure such as technology (determining the scope of services provided), expectations, government policy, and inflation of input prices (collectively described in the model as ‘coverage and prices’). Initial cost and population values for the launch year (2002) were extracted from the Ministry of Health Expenditure Database, and from Statistics New Zealand 2001 Census and vitals data and the 2001 Household Disability Survey respectively. Plausible ranges for transition rate parameters (growth in coverage and prices; growth in GDP per worker; and trends in fertility, migration, mortality and disability rates) were derived by back-casting from 2002 to 1951, Treasury’s long-term fiscal model, Statistics New Zealand projections, and a systematic review of the international literature on disability trends. The model projects that – under ‘central’ assumptions for the rate parameters – government health expenditure (as defined) as a proportion of GDP may increase by 50%, from 6.2% (2002) to 9.2% of GDP (2051). Growth in coverage and prices (non-demographic drivers) accounts for much of this increase, but ageing pressure does make it more difficult to constrain this growth. Anticipated improvement in the health status of future generations has complex effects, as the net effect of mortality improvement is to increase spending pressure (mortality improvement contributes to population ageing and expands time lived with disability – swamping the ‘distance to death’ effect, which acts to decrease spending). Only if disability prevalence declines rapidly relative to the rate of mortality decline will compression of morbidity be achieved, lifetime health care costs reduced, and ageing pressure on health spending eased. The model indicates that plausible degrees of compression of morbidity could restrict the growth of health spending as a proportion of GDP by up to one-third. The share of total government health expenditure consumed by older people is projected by the model to increase from 40% (2002) to 63% (2051), while the ratio of spending on the average older person to that on the average younger person is projected to decrease.

viii Population Ageing and Health Expenditure: New Zealand 2002–2051

Introduction Over the next 30 to 40 years New Zealand will undergo rapid population ageing, the result largely of the baby boom generation reaching retirement age. Inevitably, health expenditure will tend to rise as a result, at the same time as economic growth will tend to fall (all else being equal). This report develops projections of health expenditure over the next 50 years, to answer (to the extent possible) such questions as: Could improvements in the health status of future cohorts of older people offset the impact of population ageing? How might the relative rates of improvement in disability (non-fatal health outcomes) and mortality (fatal health outcomes) of future birth cohorts affect their health care consumption? How might the distribution of health expenditure between younger and older age groups change? The next section of this paper reviews the links between ageing, health status and coverage and prices – the drivers of per capita health expenditure. This is followed by an outline of the model used to project health expenditure, at the conceptual level (the mathematics are provided in Appendix 2). Then the operationalisation of the model is discussed, including the sources of data and methods used to estimate the model variables. Finally, the key results (model output) are presented followed by a brief discussion of the policy implications. A fuller account is provided in an accompanying Treasury report1 (Bryant et al 2004). The model is essentially an extension of one produced earlier by staff of the Ministry of Health (Johnstone and Teasdale 1999). It differs from previous work in this field internationally mainly by making explicit the complex roles of mortality and health expectancy in determining health expenditure.

Definitions Ageing: An increase in the proportion of the population aged 65+ years.

Health status: The level of health of the population, integrating mortality (length of life) and morbidity (disability or functional capacity). Typically measured in units of health expectancy.

Health expenditure: In this report, expenditure by government on personal health services, mental health services, public health services and disability support services (including residential care). ACC expenditure is not included.

1 The Treasury report (Bryant et al 2004) provides more detail on health expenditure in relation to GDP,

while the Ministry of Health report provides more detail on health expectancy.

Population Ageing and Health Expenditure: New Zealand 2002–2051 1

Drivers of Health Expenditure

Population size At the most fundamental level, health expenditure is simply a function of population size. However, population ageing has an effect on spending that can be examined separately from any simultaneous change in population size. As explained below, the ageing effect results from a shift in the age distribution of the population, such that a higher proportion of the population comes to be in the older – and costlier – age groups.

Population ageing In New Zealand a period of heightened fertility occurred from (approximately) 1945 to 1975 (peaking in the early 1960s). These ‘baby boom’ cohorts produce a ‘wave’ in the population pyramid as they pass through it. From (approximately) 2010 to 2030 the proportion of the population aged 65+ years will therefore increase rapidly (Figure 1). Figure 1: Older people (65+ and 85+) as a proportion of the population, 2001–2051

0%

5%

10%

15%

20%

25%

30%

2002 2007 2012 2017 2022 2027 2032 2037 2042 2047 2052

Percent aged 65+Percent aged 85+

Percent

Population ageing is affected not only by the initial size of the baby boom cohorts (relative to earlier and later cohorts), but also by their rate of extinction. Improvements in life expectancy over the past half century, and projected further declines in mortality over the next half century, will therefore add to spending pressure by exacerbating population ageing. Here we project a range of future mortality trends, each of which influences the size and age structure of the population accordingly.

2 Population Ageing and Health Expenditure: New Zealand 2002–2051

Population ageing affects GDP as well as health expenditure. This is because, with ageing, the proportion of the population that is of working age diminishes, so exerting (downward) pressure on economic output. This is taken into account in the model by projecting GDP on the basis of population size and age structure (influencing projected labour force participation rates) as well as labour productivity (GDP per worker).

Health status Why does population ageing drive health expenditure? Because per capita health expenditure is strongly (exponentially) related to age (Figure 2). Figure 2: Annual government per capita health expenditure, by age and service group,

genders pooled, 2001/02

$0

$2,000

$4,000

$6,000

$8,000

$10,000

$12,000

$14,000

$16,000

0–4 5–9 10–14 15–19 20–24 25–29 30–34 35–39 40-44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85+

Age group (years)

Personal health (including mental health)Disability support servicesPublic health

Yet age itself does not ‘cause’ health expenditure. Rather, age serves as a proxy for health status: both the risk of dying, and the prevalence of chronic disease and disability, increase exponentially with age.


Distance to death Typically, someone who is soon to die (ie, is in his or her last year of life) is sicker, and uses more health care, than someone who is not. Studies overseas using unit record data have shown that one-quarter or more of lifetime health expenditure may be consumed, on average, in the last year of life (McGrail et al 2000, Spillman et al 2000, Yang et al 2003, Miller 2001). So mortality improvement reduces health expenditure by reducing the proportion of people dying each year (ie, by increasing distance to death). In our model, mortality trend projections affect the probability of dying at each age in each calendar year, so altering distance to death and hence the proportion of decedents to survivors in each age–sex cell.

Disability At older ages the main components of health expenditure are disability support services (including long-term and residential care) and personal health services relating to the management of chronic diseases (figure 2). In the absence of comprehensive data on the prevalence of chronic diseases, we use the prevalence of dependent disability (disability needing assistance to carry out everyday tasks), adjusted for severity, to index non-fatal health status. In our model, projections of dependent disability prevalence influence the proportion of chronically ill to healthy survivors (and decedents) in each age–sex cell.

Measuring ‘health’ in the model Health status in our model is conceptualised as a two-dimensional construct. The non-fatal dimension of health (morbidity) is measured using the prevalence of dependent disability (disability needing assistance with everyday routines), adjusted for severity, as a proxy for the prevalence of chronic diseases and consequential functional limitations. The fatal, or survival, dimension of health is measured using distance to death (the probability of surviving at each age – summarised as life expectancy). The model integrates these two dimensions to provide a summary measure of health: independent life expectancy or health expectancy (the average number of years lived free of disability requiring assistance). If the difference between life expectancy and health expectancy decreases, absolute compression of morbidity is occurring (if the difference increases, absolute expansion of morbidity occurs). If the ratio of health to life expectancy increases, relative compression of morbidity is occurring (if this ratio falls, relative expansion of morbidity occurs).


Residual effect of age To what extent does health status explain the effect of age on health expenditure? Using data on distance to death (Statistics New Zealand mortality database) and dependent disability prevalence (Statistics New Zealand postcensal disability survey) for 2001 (Table 1), we fitted a statistical model in which health expenditure in each age group and gender is a function solely of these two variables (Figure 3). The model agrees closely with the observed health expenditures (from the Ministry of Health database), indicating that health status (distance to death together with dependent disability prevalence) does indeed explain most of the age variation in health expenditure.2 Table 1: Health expenditure, distance from death and disability prevalence, 2001

Annual per capita health expenditure

Percent of population in last year of life

Percent of population with disability

Age group

Males Females Males Females Males Females

0–4 1877 1623 0.16% 0.13% 3.7% 2.7% 5–9 723 624 0.02% 0.02% 9.2% 5.8% 10–14 658 585 0.03% 0.02% 9.2% 6.1% 15–19 843 1111 0.09% 0.04% 5.4% 4.2% 20–24 881 1638 0.13% 0.04% 4.6% 4.7% 25–29 930 2022 0.13% 0.05% 5.1% 6.1% 30–34 905 2005 0.14% 0.05% 5.8% 7.3% 35–39 937 1646 0.14% 0.07% 6.5% 8.2% 40–44 986 1262 0.19% 0.12% 8.2% 9.5% 45–49 1218 1403 0.27% 0.20% 10.9% 10.8% 50–54 1442 1551 0.43% 0.32% 13.6% 11.5% 55–59 1772 1773 0.74% 0.51% 16.8% 12.8% 60–64 2349 2199 1.19% 0.82% 20.8% 16.0% 65–69 3519 3123 2.10% 1.27% 25.5% 22.2% 70–74 4903 4219 3.42% 1.95% 32.1% 31.2% 75–79 6840 6303 5.58% 3.29% 42.8% 43.0% 80–84 8976 8985 9.18% 5.98% 59.0% 56.3% 85–89 12,978 13,735 15.24% 10.77% 77.6% 67.0% 90–94 15,573 18,944 25.87% 18.97% 91.2% 78.3% 95+ 18,738 24,738 30.61% 29.47% 97.5% 89.8%

2 The good fit may also reflect correlation with other factors, not included in the model, that actually

drive expenditure.


Figure 3: Prediction of age specific health expenditure from distance to death and disability, 2001

Males Females

$0

$5,000

$10,000

$15,000

$20,000

$25,000

$30,000

0–4

5–9

10–1

4

15–1

9

20–2

4

25–2

9

30–3

4

35–3

9

40-4

4

45–4

9

50–5

4

55–5

9

60–6

4

65–6

9

70–7

4

75–7

9

80–8

4

85–8

9

90–9

4

95+

Age group (years)

ActualFitted

$0

$5,000

$10,000

$15,000

$20,000

$25,000

$30,000

0–4

5–9

10–1

4

15–1

9

20–2

4

25–2

9

30–3

4

35–3

9

40-4

4

45–4

9

50–5

4

55–5

9

60–6

4

65–6

9

70–7

4

75–7

9

80–8

4

85–8

9

90–9

4

95+

Age group (years)

ActualFitted

In summary, as health status improves, so the age–cost curve (Figure 2) flattens and the impact of ageing on health expenditure lessens. Ultimately, if future cohorts of older people remain free of chronic disease and disability until extreme old age when their deaths are preceded by only a short period of illness (which is not intensively treated), the age–cost curve would become almost completely flat (except for a brief spike at the end) and ageing would no longer be a significant driver of health expenditure.

Interaction between fatal and non-fatal outcomes Mortality reduction will tend to reduce spending pressure by increasing distance to death, as explained above. Yet longer survival could also mean more years lived with disability. This will depend on the relative trends in chronic disease (disability) prevalence and mortality incidence (life expectancy); ie, on trends in health expectancy (health expectancy is defined here as independent life expectancy, or expectation of life free from dependent disability). A decline in health expectancy relative to life expectancy (‘expansion of morbidity’) will occur when mortality declines faster than disability, so increasing spending pressure. By contrast, an increase in health expectancy relative to life expectancy (‘compression of morbidity’) will further reduce spending pressure. In this report, we measure compression as a reduction in the percentage of life expected to be lived with disability requiring assistance (adjusted for severity). The effect on spending of a positive trend in health status (ie, later cohorts are healthier than earlier cohorts) will therefore depend on the interaction between trends in fatal and non-fatal health outcomes. In our model, trends in mortality and trends in disability interact to yield nett effects on spending via health expectancy, while mortality also has indirect effects on spending via its contribution to population ageing.


The complex influence of mortality Our model is innovative both in its explicit treatment of health expectancy, and in recognising the complex effects of improvements in mortality on health expenditure. The literature on health expenditure modelling has not previously explicitly recognised that a reduction in mortality risks (ie, improved cohort survival) will have five separate but inter-related effects on health spending pressure, as discussed above and summarised below:

• increase in distance to death, so reducing spending pressure

• increase in population size, so increasing spending pressure

• increase in population ageing, so increasing spending pressure

• increase in time lived with disability (unless offset by faster disability decline), so expanding morbidity and increasing spending pressure

• increase in the prevalence of disability, if the reduction in the risk of dying is relatively greater for those with (severity-adjusted) disability, so increasing spending pressure.

We show, counterintuitively, that the predominant effect of improving mortality will be to increase spending pressure, despite a decrease in the proportion of people in their last year of life, unless the gain in survival is accompanied by relative (as well as absolute) compression of morbidity.

Coverage and prices The final – and arguably most important – driver of health expenditure is the coverage and price3 of health (including disability support) services. At any given time, health expenditure per capita will be determined by the age structure of the population, its (age-dependent) health status, and the coverage and prices of health services. Coverage will be determined by complex interactions between technology (determining the ‘universe of the possible’, or the potential scope of services), consumer expectations (influencing demand), and government policy decisions. Prices reflect inflation of labour and non-labour input costs to the health system in excess of general inflation. Coverage and prices includes: • health care and related technology • organisation of health services • patterns of care (eg, referral patterns, institutional vs community care) • access • demand (eg, consumer and community expectations) • supply (eg, workforce issues, labour productivity, drug prices) • public / private funding mix • government health, social and economic policies.

3 ‘Coverage and prices’, often described in the literature simply as ‘technology’, can also be thought of

as a residual term, including all drivers other than demographic and health status drivers.


In general, advances in health care and related technology will lower spending pressure by improving productivity (eg, new drugs may reduce complications of chronic diseases and hence consumption of expensive surgical interventions; new organisational structures, such as primary health organisations (PHOs), may improve primary health care delivery and so reduce expensive ambulatory sensitive hospitalisations). Yet the predominant effect of technological development may well be to increase spending pressure, by expanding the possibilities for improving health. A good example would be the discovery of insulin, which transformed type 1 diabetes from a rapidly fatal untreatable condition (which involved little health expenditure beyond brief supportive care prior to death) to a chronic condition that involves considerable lifelong health care expenditure. Neonatal intensive care, which has led in some cases to the survival of severely damaged infants requiring lifelong medical and social care, is another example.

Interaction between population ageing, health status, and coverage and prices In summary, per capita health expenditure in real terms is driven by the interaction of three factors – population ageing, health status, and coverage and prices. In reality, these factors are not independent of each other: mortality is both a dimension of health status and of population ageing, and investment in service coverage or health care technology impacts favourably on health status (both mortality and morbidity). These interactions are taken into account by our model, directly in the former case and through restrictions on the sensitivity analysis (see later) in the latter. One visualisation of the model is shown in Figure 4. This clearly identifies the drivers, and how they interact to influence health expenditure, but says little about the shape of their relationships to health expenditure. (Note that nett migration is not shown in the diagram although it is included in the model – it has relatively little effect).


Figure 4: Visualisation of the drivers of per capita health expenditure

Birth rates

Age structure Distribution of

population by age and health status

Disability prevalence rates

Death rates

Coverage and prices (technology, etc)

Health expectancy

Per capita health expenditure

An alternative way of visualising these interacting determinants of health expenditure growth is to consider that population ageing determines the weight to be given to each age group,4 while coverage and prices determine the level of the age-specific cost curve (best seen on a log scale), and (non-fatal) health status determines its shape (specifically, its upward slope with age) (Figure 5).

4 This can also be visualised by appropriately varying the width of the bars in a bar chart version of

Figure 5.


Figure 5: Diagrammatic representation of the drivers of per capita real health expenditure, based on the age-cost curve

(a) Population ageing

$0

$5,000

$10,000

$15,000

$20,000

$25,000

$30,000

0–4

5–9

10–1

4

15–1

9

20–2

4

25–2

9

30–3

4

35–3

9

40-4

4

45–4

9

50–5

4

55–5

9

60–6

4

65–6

9

70–7

4

75–7

9

80–8

4

85–8

9

90–9

4

95+

Age group (years)

0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

8.0%

9.0%Cost weights, 2002 (left axis)Age distribution, 2002 (right axis)Age distribution, 2051 (right axis)

Population ageing shifts the age distribution to the right (dotted line vs light solid line), so increasing the proportion of older people: total health expenditure rises provided the age–cost curve (dark solid line) remains unchanged.

(b) Coverage and prices

1

10

100

1,000

10,000

100,000

0–4

5–9

10–1

4

15–1

9

20–2

4

25–2

9

30–3

4

35–3

9

40-4

4

45–4

9

50–5

4

55–5

9

60–6

4

65–6

9

70–7

4

75–7

9

80–8

4

85–8

9

90–9

4

95+

Age group (years)

Cost weights, 2002Cost weights, 2051

The age–cost curve shifts uniformly upwards (on a log scale): total health expenditure increases even though the population age structure remains unchanged.


(c) Health status

$0

$2

$4

$6

$8

$10

$12

$14

$16

0–4

5–9

10–1

4

15–1

9

20–2

4

25–2

9

30–3

4

35–3

9

40-4

4

45–4

9

50–5

4

55–5

9

60–6

4

65–6

9

70–7

4

75–7

9

80–8

4

85–8

9

90–9

4

95+

Age group (years)

Cost weights, 2002 (age 0–4 = 1)Cost weights, 2051 (age 0–4 = 1)

Health status improvement changes the shape of the age–cost curve, flattening the rise associated with older ages: total health expenditure decreases (offsetting to some extent any increase resulting from population ageing or growth in coverage and prices).


The Model

Framework The model simulates the New Zealand population (disaggregated by age and sex) over the next half century (see box on page 14).

Definition of model terms

Health state definitions A person who dies during the year is a decedent (ie, is in the last year of life). Everyone else is a survivor. A person who needs assistance with everyday tasks during the year (six months or more) is disabled (generally the result of chronic illness or late effect of an injury). Everyone else is non-disabled.

Change rate definitions The rate at which the incidence of mortality (risk of dying) falls for each subgroup of the population and each calendar year is designated ‘MORT’.

The rate at which the prevalence of disability requiring assistance, adjusted for severity, falls for each subgroup of the population and each calendar year is designated ‘DIS’.

The rate at which coverage and prices increase for each subgroup of the population and calendar year is designated ‘COV’.

The rate at which GDP per worker (ie, labour productivity) grows for each calendar year is designated ‘PROD’.

Population ageing Population ageing is reflected by changing the relative sizes of the total population in each age–sex cell. This is largely predetermined by past fertility patterns, but is also influenced by the mortality reduction rate projected.

Health status classification The population in each age–sex cell is classified into four health states, determined by distance to death (proportion dying) and disability prevalence (Table 2). Table 2: Health states included in the model

Non-disabled survivor Non-disabled decedent

Disabled survivor Disabled decedent


Changes in health status are reflected by changing the distribution of the population in each age–sex cell over these four health states. This distributional change is in turn achieved by projecting the mortality reduction rate (‘MORT’) and the disability prevalence reduction rate (‘DIS’).

Coverage and prices Coverage and prices are reflected by changing the costs for each health state (within each age–sex cell). The growth rate in coverage and prices (‘COV’) in the simplest version of the model is assumed to change uniformly across all health states at all ages. More complex versions of the model allow COV to change more for older than for younger age groups (for example), so altering the shape as well as the level of the age–cost curve. While there is some empirical evidence for such differential growth rates in the past, there is little reason to conclude that any such trend will continue in the future. Hence we report here only scenarios involving uniform growth rates of COV across all ages and health states.

Projection The momentum component of population ageing attributable to past fertility trends is predetermined, and is available directly from Statistics New Zealand. Fertility projections to 2051 used in this study are those of Statistics New Zealand series 4 throughout. Migration also affects the age structure of the population; in this study, Statistics New Zealand’s series 4 nett migration projection is used throughout. The projection challenge therefore consists of projecting future trends in mortality (MORT), disability (DIS) and growth in coverage and prices (COV). This was done by a combination of back-casting (analysis of historical trends) and international comparative studies.

Analysis of historical trends Relatively robust data on health expenditure, population ageing and mortality are available as far back as 1951. We therefore analysed historical trends over the period 1951–2001. Unfortunately, we have no time series of disability prevalence, so assumptions had to be made (ie, that disability prevalence was stable over the historical period5). It was then possible to estimate the historical trend in the coverage and price growth rate (COV) by back-casting (since the values of all other variables in the model were known for the historical period). Historical trend analysis was also used as the basis for projecting mortality (MORT).

5 We lack empirical evidence as to past trends in disability prevalence, so the most parsimonious model

is to assume stability. However, we have also tested situations in which disability prevalence declined slowly over the past 50 or 25 years, analogous to our future projections: COV is relatively insensitive to such variations.


International comparative studies Extrapolation of past trends in COV and MORT do not provide an adequate basis for projecting future trends in these key parameters. Instead, we also examined changes that have occurred in developed countries over the past several decades as an indicator of what future trends may be in New Zealand. This method was used on its own for projecting future disability prevalence (DIS).

Sensitivity analysis Since neither analysis of past trends nor of trends in other countries provides a fully satisfactory basis for projecting MORT, DIS or COV, these analyses were used instead to estimate the plausible range for these parameters over the projection period. These ranges then provided the basis for sensitivity analysis. Sensitivity analysis was restricted to plausible combinations of the ranges for COV, MORT and DIS to allow for the interaction between coverage and prices and health status (ie, higher investment in health technology / greater coverage of health service delivery would be expected to be associated with greater improvement in health status, and vice versa).

How does the model work? The model has two components: a ‘population’ component and a ‘cost’ component.

Population component The population is divided into 160 cells (two genders by 20 age groups by four health states). The initial numbers in each cell are those estimated for 2002. The numbers are changed for each subsequent year to 2051 by applying to each cell the relevant changes in fertility, migration, mortality and disability rates according to each scenario.

Cost component The initial costs for each of the 160 cells are those estimated for 2002. The costs are changed for each subsequent year to 2051 by applying to each cell the relevant change in the growth rate for coverage and prices. In the simplest form of the model, costs change at a constant rate for each year for each cell.

Expenditure Per capita expenditure by age and sex is then calculated for each year by multiplying the cost for each health state by the number of people in that health state in each age–sex cell (in that year). These totals are then summed, and the grand total divided by the total cell population (for that age–sex cell for that year).


Operationalisation of the Model

Base year The base year for the model was 2002, the most recent year for which health expenditure data were available.

Projection period The projection horizon was 50 years, in order to cover the period of rapid population ageing. Thus the projection period is 2002–2051.

Back-casting period Reliable data on population size and age structure, on mortality (but not disability), and on health expenditure,6 were available back to 1951. The back-casting period was therefore 1951–2001.

Population size and age structure Population size by single year of age and gender for 1951–2001 was available from Statistics New Zealand. For the population projections, the ‘medium’ (series 4) projections of fertility and nett migration were used. The mortality projections used are discussed below – these closely correspond to Statistics New Zealand’s ‘low’, ‘medium’ and ‘high’ mortality projections respectively. The projected population for 2002–2051 in turn falls out of these mortality, fertility and migration projections.

Mortality Historically, mortality7 has fallen relatively steadily over the past half century by about 1.5% per year. This is similar to recent trends in other developed countries. On this basis, a plausible range for future mortality trends was considered to be an average annual percentage reduction over the projection period (across all ages and for both genders) of 1.0%–2.0%. Accordingly, three mortality projections are used for the sensitivity analyses: 1.0%, 1.5% and 2.0% annual reductions, corresponding closely to Statistics New Zealand’s ‘high’, ‘medium’ and ‘low’ mortality projections respectively.

6 The historical expenditure series was first made consistent with the current scope of Vote Health as

described in the box ‘ Reconstructing the historical health expenditure series’ on pages 18–19. 7 Non-injury mortality is used to approximate the mortality risks for people with chronic disease

(‘disabled’), while injury mortality is used to approximate risks for people without chronic disease (‘non-disabled’). The same rates of change are assumed to apply to both causes of mortality.


Disability Only two national surveys of disability have been done in New Zealand – in 1996 and 2001. In the absence of a long time series of historical data, projection of disability prevalence has here been based on a comparative analysis of national disability surveys in other countries (as well as the two New Zealand surveys). Full details of the methods and results are included in Appendix 1. Briefly, seven studies were identified that met the inclusion criteria: the 1996 and 2001 New Zealand Postcensal Disability Surveys; the 1986, 1991 and 2001 Canadian Postcensal Disability Surveys; the 1988, 1993 and 1998 Australian Bureau of the Census Surveys of Disability; the 1995 and 2001 Health Surveys for England; and three US studies – the 1982–1999 National Long Term Care Study, the 1977–1994 Framingham Heart and Offspring Studies, and the 1992–96 Medicare Current Beneficiary Surveys. Comparison of the results of these studies showed a consistent picture of decreasing prevalence of moderate (IADL only) disability over recent decades, but a mixed picture for severe (ADL) disability.8 After weighting both levels of disability (according to prevalence and severity), we concluded that the plausible range for ‘dependent disability’ prevalence as used in our model is from stability to a gradual decline. A further increase in severity-adjusted dependent disability prevalence appears unlikely. Accordingly, our sensitivity analysis ranges from an average annual reduction rate in disability prevalence of DIS = 0.0% (high disability), through DIS = 0.5% (medium disability) to DIS = 1.0% (low disability).

Growth in coverage and prices Historical values of growth in coverage and prices (COV) were determined by back-casting; ie, using the model in reverse to estimate the value of COV necessary to reconcile the modelled with the observed trend in health expenditure (adjusted for consistency of composition) over the historical period, given the observed trends in health expenditure and mortality and assumptions of stable disability prevalence. It was found that the value of COV varied significantly from year to year, partly reflecting real effects (eg, nett migration flows) and partly artefactual effects (eg, inclusion or exclusion of various cost components). Smoothing over 10-year intervals, it was found that COV varied from a high of 3.3 % in the 1950s to a low of 1.8 % in the 1990s, although it has again risen to 3–4% per year since the late 1990s. The average value of COV over the whole 1951–2002 period was 2.3%.

8 IADL refers to instrumental activities of daily living, such as shopping, house and garden maintenance

and household budgeting. ADL refers to more basic activities of daily living relating to self-care, such as eating, bathing, toileting, transferring and dressing.


Given the long-term trend9 toward a decreasing value of COV, the plausible range for COV over the next five decades was set at 1.0–2.0%. Accordingly, sensitivity analyses were run using three values for COV: low (average annual percentage increase of 1.0% across all age–sex cells), medium (1.5%) and high (2.0%).

Growth in GDP per worker (labour productivity) The Treasury’s Long Term Fiscal Model projects an average annual growth rate in GDP per worker10 of 1.5% over the next 50 years (Woods 2000). This parameter (designated ‘PROD’) was not varied for the sensitivity analyses reported here, although a model run using a value of 1.0% for PROD is presented. Note that for estimation of health expenditure as a percentage of GDP, it is not the absolute values of COV and PROD that matter, but the difference between them (ie, COV–PROD).11

Projection scenarios Sensitivity analyses were restricted to comply with the expectation that greater growth in coverage and prices would yield greater returns in health status. So the following combinations of variables were run (Table 3). Table 3: Average annual percentage changes, by projection scenario, 2002–2051 (%)

Scenario A B C D E F G H I J

COV 1.0 1.0 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0

MORT 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 1.5 2.0

DIS 0.0 0.5 0.0 0.5 0.5 1.0 0.5 1.0 0.5 1.0

The scenarios range from A (low coverage and price growth, high mortality, high disability) to J (high coverage and price growth, low mortality, low disability). Scenario E is considered the ‘central’ scenario or ‘base case’.

9 The underlying assumption is that the current growth rate of 3–4% per year for COV represents a short-

term trend that is unlikely to be sustainable over the projection period. 10 In the model, this parameter is multiplied by the projected labour force participation rate – itself

influenced by population ageing – to yield projections of economic growth (ie, per capita GDP). 11 The Treasury report provides further detail of this parameter, including more extensive sensitivity

analysis of (COV – PROD).


Base costs (health expenditure in the launch year 2001/02) Data for health expenditure in the base year (2001/02) was available from the Ministry of Health database, disaggregated by age and sex. The challenge was to partition the costs by health state.

Public health Costs were assumed to be the same for all health states.

Disability support services Costs were partitioned equally between disabled survivors and disabled decedents.

Mental health services Costs were partitioned equally between disabled survivors and disabled decedents.

Personal health care services Cost for a decedent12 was fixed at $10,000 (based on overseas and limited New Zealand data [G. Jackson, personal communication December 2003, Auckland regional data] on the cost of dying). Cost for a survivor was partitioned equally between disabled and non-disabled survivors until age 50–54. Thereafter, cost for a non-disabled survivor was assumed to remain stable at the age 50–54 level. Cost for a disabled survivor aged 55+ was then estimated as the residual cost (ie, this was estimated by difference, since both the total cost and the costs for the other three health states were now known).

Reconstructing the historical health expenditure series The government pays for a variety of services that contribute in some way to health, ranging from hospitals, to public health campaigns, to medical schools. Not all such services are paid for under Vote Health, the budget allocated to the Ministry of Health. The allocation of services to Votes changes over time. For instance, between 1992/93 and 2001/02 funding for disability support services was transferred out of Vote Social Welfare into Vote Health. These services accounted for about 8% of Vote Health expenditure in 2001/02. When, in our projections, we refer to ‘government health expenditures’ we mean expenditure by the government on the basket of services that were provided through Vote Health in 2001/02. Ideally, we would like to construct an historical expenditure series based on exactly this definition.

12 Sensitivity analysis was done using $15,000 also, with very little difference to the results.


We have two sets of data with which to construct such a series. The first is annual figures on Vote Health in the period 1950/51 to 2001/02, assembled by the Treasury from official year books. The second is annual estimates for ‘total net transfers’ to Vote Health for the period 1993/94 to 2001/02. These transfers indicate expenditure on services moved from other Votes into Vote Health since 1992/93. To construct our historical expenditure series from these data, we have made two assumptions. The first is that the range of services provided under Vote Health remained the same over the period from 1950/51 to 1992/93. The second is that the ratio between expenditure on services provided through Vote Health in 1993/94 and expenditure on services provided through Vote Health in 2001/02 remained the same throughout the period 1950/51 to 2001/02. A different way of stating these assumptions is that growth rates for the two sets of services are the same throughout the period. Our assumptions are unlikely to be fully met in practice. The reconstructed expenditure series, which includes funding for disability support services throughout the study period while excluding funding for ACC, is nevertheless better suited to our purposes than the unadjusted Vote Health series would be, particularly for the 1990s. The graph below illustrates this ‘reconstructed’ series.

Figure 6: Total health expenditure (constant 2002 dollars) and expenditure as a percent of GDP

0123456789

1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001

$billions

0%

1%

2%

3%

4%

5%

6%

7%

Expenditure (left axis) Expenditure as % of GDP (right axis)


Results

Results by scenario Full results are presented for each of the 10 scenarios in Appendix 3. Key results across all 10 scenarios for selected years (2026 and 2051) are summarised in Tables 4 and 5 respectively for ease of reference. The demography implied by each scenario is summarised first, followed by the health outcomes associated with each, and finally their respective implications for health expenditure.

Trajectories In this report we do not provide detail of the trend trajectories.13 As a general finding, however, relatively little change in health expenditure or other parameters occurs in the first half of the projection period (and none at all until at least 2010). Most change occurs from 2026 (or later) to 2051.

Expenditure in relation to GDP The results relating to the share of GDP spent on health services are summarised first. Limited sensitivity analysis around GDP per worker (PROD) is included.

Distribution of health expenditure across life-cycle stages The results relating to the share of total health expenditure for each life-cycle stage are then summarised. Findings for trends in per capita expenditure by age group are also presented.

Drivers of health expenditure Finally, the results of detailed scenario comparisons are presented. These comparisons are intended to illustrate the independent impact of each driver on health expenditure (to the extent that this is possible, given the strong interdependence between all the drivers). Growth in the share of GDP spent on health services is used as the outcome measure, but the results would be similar if other outcome measures, such as per capita health expenditure growth, were used instead.

Results by scenario, for selected years

Table 4: Summarised results, by scenario, for 2026

Scenario 2002 A B C D E F G H I J

Assumptions COV 1.0 1.0 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 MORT 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 1.5 2.0

13 See the Treasury report (Bryant et al 2004) for detail on the trajectory of change under different

scenarios (assumptions).


DIS 0.0 0.5 0.0 0.5 0.5 1.0 0.5 1.0 0.5 1.0

Demography Total population 3.91 4.47 4.47 4.50 4.50 4.50 4.50 4.54 4.54 4.50 4.54 % 65+ 11.9 17.5 17.5 17.9 17.9 17.9 17.9 18.4 18.4 17.9 18.7 % 85+ 1.3 2.2 2.2 2.4 2.4 2.4 2.4 2.7 2.7 2.4 2.7

Health expectation

LE without disability 66.4 67.7 69.2 68.2 69.8 69.8 71.3 70.4 71.9 69.8 71.9 LE with disability 12.0 13.4 11.9 14.1 12.5 12.5 11.1 13.1 11.6 12.5 11.6 Total life expectancy

78.4 81.1 81.1 82.3 82.3 82.3 82.3 83.5 83.5 82.3 83.5

% life disabled 15.3 16.5 14.7 17.1 15.2 15.2 13.5 15.7 13.9 15.2 13.9

Health expenditure

Total expenditure 7.70 13.16 12.34 13.54 12.68 14.27 13.41 14.66 13.76 16.06 15.48 Per capita expenditure

1969 2945 2761 3006 2815 3169 2978 3233 3034 3566 3414

% GDP 6.2 6.7 6.3 6.9 6.4 7.2 6.8 7.4 7.0 8.1 7.8

Expenditure by age group

25–29 1545 1931 1873 1930 1871 2107 2048 2106 2047 2371 230345–49 1359 1719 1654 1717 1652 1859 1794 1857 1792 2092 201665–69 3434 4338 4035 4333 4030 4538 4234 4533 4229 5106 475885–89 13,975 17,595 15,815 17,559 15,780 17,766 15,981 17,730 15,946 19,989 17,942

Ratio 85–89 : 25–29

9.0 9.1 8.4 9.1 8.4 8.4 7.8 8.4 7.8 8.4 7.8

Expenditure by lifecycle stage (%)

0–14 11.8 8.2 8.6 8.0 8.3 8.3 8.7 8.1 8.5 8.3 8.5 15–64 48.3 41.0 42.1 39.9 41.0 41.0 42.2 40.0 41.1 41.0 41.1 65+ 39.9 50.8 49.3 52.1 50.6 50.6 49.1 51.9 50.4 50.6 50.4

Notes: For projection of GDP per capita, PROD = 1.5 Results are pooled for both genders, and (where applicable) for all age groups LE = life expectancy at birth (years) * average annual percentage change * * constant 2002 dollars (total expenditure in $ billions)

Table 5: Summarised results, by scenario, for 2051

Scenario 2002 A B C D E F G H I J

Assumptions COV 1.0 1.0 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 MORT 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 1.5 2.0 DIS 0.0 0.5 0.0 0.5 0.5 1.0 0.5 1.0 0.5 1.0


Demography Total population 3.91 4.61 4.61 4.74 4.74 4.74 4.74 4.87 4.87 4.74 4.87 % 65+ 11.9 22.4 22.4 24.1 24.1 24.1 24.1 25.9 25.9 24.1 25.9% 85+ 1.3 4.3 4.3 5.5 5.5 5.5 5.5 6.9 6.9 5.5 6.9

Health expectation

LE without disability 66.4 68.8 72.1 69.8 73.4 73.4 76.2 74.6 77.7 73.4 77.7 LE with disability 12.0 15.0 11.7 16.6 13.0 13.0 10.1 14.3 11.2 13.0 11.2 Total life expectancy

78.4 83.8 83.8 86.3 86.3 86.3 86.3 88.9 88.9 86.3 88.9

% life spent disabled

15.3 17.9 14.0 19.2 15.1 15.1 11.7 16.1 12.6 15.1 12.6

Health expenditure

Total expenditure 7.70 21.13 18.27 23.54 20.21 25.74 22.40 28.77 24.85 32.75 31.61 Per capita expenditure

1969 4581 3960 4967 4264 5432 4728 5902 5098 6910 6485

% GDP 6.2 7.6 6.6 8.4 7.2 9.2 8.0 10.2 8.8 11.7 11.2

Expenditure by age group

25–29 1545 2477 2332 2475 2330 2968 2823 2965 2820 3776 358845–49 1359 2196 2035 2191 2031 2587 2426 2582 2421 3291 308065–69 3434 5555 4808 5546 4798 6112 5364 6103 5354 7776 681285–89 13,975 22,477 18,092 22,403 18,021 22,955 18,570 22,883 18,501 29,205 23,538Ratio 85–89:25–29 9.0 9.1 7.8 9.1 7.7 7.7 6.6 7.7 6.6 7.7 6.6

Expenditure by lifecycle stage (%)

0–14 11.8 6.2 7.0 5.6 6.3 6.3 7.1 5.7 6.4 6.3 6.4 15–64 48.3 32.0 34.3 28.8 31.2 31.2 33.6 28.0 30.4 31.2 30.4 65+ 39.9 61.8 58.7 65.6 62.5 62.5 59.3 66.4 63.2 62.5 63.2

Notes: For projection of GDP per capita, PROD = 1.5 Results are pooled for both genders, and (where applicable) for all age groups LE = life expectancy at birth (years) * average annual percentage change * * constant 2002 dollars (total expenditure in $ billions)

Expenditure in relation to GDP Scenario E (characterised by a growth rate in coverage and prices of 1.5%, associated with a mortality reduction rate of 1.5%, and a disability prevalence reduction rate of 0.5%) is considered the ‘central’ scenario. Under this scenario, government health expenditure as a percentage of GDP increases over the projection period from 6.2% (in 2002) to 9.2% (in 2051), an increase of almost exactly 50%. However, the range of projections for this outcome is wide, varying from almost no increase (6.6%) in scenario B to an almost 90% increase (11.7 %) in scenario I.


This does not allow for variation in the average rate of productivity growth (PROD), which has been held constant at 1.5%. If we assume that GDP per worker grows more slowly (PROD = 1.0% instead of 1.5%), then government health expenditure as a proportion of GDP rises to 11.7% by 2051 under scenario E (rather than to 9.2%), with a range from 8.4 to 14.9% under the different scenarios (Table 6). Table 6: Health expenditure as a percentage of GDP, by scenario, 2051

PROD A B C D E F G H I J

1.5% 7.6 6.6 8.4 7.2 9.2 8.0 10.2 8.8 11.7 11.2 1.0% 9.7 8.4 10.7 9.2 11.7 10.2 13.0 11.2 14.9 14.3

PROD = growth rate in GDP per worker. Although we do not report detailed results on trajectories of change here (because of space limitations), it is important to note that little change in the percentage of GDP spent on health occurs for the first decade (ie, until approximately 2012). The rate of increase remains relatively slow until approximately 2026, after which it accelerates. For example, under central assumptions a one percentage point increase in the health share of GDP occurs from 2002 to 2026, compared to a two percentage point increase from 2026 to 2051.

Distribution of expenditure across life-cycle stages Under all scenarios, a major shift in the share of resources going to each life-cycle stage appears inevitable. Indeed, the key finding is how little these proportions are affected by varying the projected rates of mortality or disability improvement (Table 7). Table 7: Share of health expenditure, by life-cycle stage, 2002 and 2051

Share (%) 2002 2051 central scenario (E)

2051 range across all scenarios

0–14 11.8 6.3 5.6–7.1 15–64 48.3 31.2 28.0–34.3 65+ 39.9 62.5 58.7–66.4

Not surprisingly, variation in the life-cycle shares of total health expenditure is unaffected by changes in COV (since we have assumed a constant rate of growth in coverage and prices across all age groups for all years), but instead reflects the relative rates of decline in mortality and disability (ie, health expectancy). For example, an increase in the mortality reduction rate of 0.5% raises the expenditure share of older people by four percentage points if disability remains stable. By contrast, an increase in the disability reduction rate of 0.5% lowers the expenditure share of older people by about three percentage points if mortality remains stable. Shares consumed by different life-cycle stages are summarised graphically below (Figure 7). The figure also includes results for 1951 (obtained by back-casting), to illustrate the extent of historical change.


Figure 7: Share of health expenditure, by lifecycle stage, selected years

1951 2002 2051 (E)

19.5%28.7%

51.8%

11.8%39.9%

48.3%

31.2%

62.5%

6.3%

Ages 0–14 Ages 15–64 Ages 65+

It is important to note that the increasing share of health care resources consumed by older people results from population ageing, not from steepening of the age–cost curve. In fact, under the central scenario, the percentage growth in per capita expenditure over the projection period is lower for older than for younger people (reflecting the flattening of the age–cost curve as the disability prevalence rate declines) (Table 8). Table 8: Percentage growth in per capita expenditure, by selected age group, scenario E,

2002–2051

Age group Growth (%)

25–29 92 45–49 90 65–69 78 85–89 64

That is, while intergenerational equity worsens in absolute terms (in that the share of total health expenditure consumed by older people as a group increases), it improves in relative terms (the ratio of per capita expenditure on the average older compared to the average younger person declines – by approximately 25% under the central scenario).

Drivers of health expenditure One approach to estimating the effect of each driver is to set all other drivers to zero and vary the driver of interest. We have not followed this approach because it allows combinations of drivers that we do not believe are realistic. Instead, we examine selected scenarios within our set of plausible scenarios (labelled A through J in Figure 6), chosen to illustrate plausible variation of the driver of interest against a background of realistic values for the other drivers. This allows us to estimate the increase in health share of GDP by 2051 (or other outcome variable) per unit increase or decrease in the value of the driver of interest within a realistic context.


Impact of growth in coverage and prices (COV) To quantify the potential impact of growth in coverage and prices (COV) over the next 50 years we compare scenario E (the central scenario) with scenarios D and I, scenarios that are realistic yet differ from the central scenario only in their value for COV (Table 9). Table 9: Impact of growth in coverage and prices, 2051

Scenario D E I

Assumptions: COV 1.0 1.5 2 MORT 1.5 1.5 1.5 DIS 0.5 0.5 0.5

Expenditure % GDP 2051 7.2 9.2 11.7

In the lower half of its range, a 0.5 percentage point increase (or decrease) in COV changes health expenditure as a percentage of GDP by 2.0 percentage points. In the upper half of its range, a similar change in COV alters the share of GDP by 2.5 percentage points. Within the plausible ranges of all drivers, therefore, COV has an elasticity of approximately 4.5 (averaging across the whole plausible range14). That is, on average, a one percentage point change in COV leads to a 4.5 percentage point change in health expenditure as a percentage of GDP in 2051, all else being equal.

Impact of (non-fatal) health status improvement (DIS) It is more difficult to assess the impact of health status improvement. We do not allow scenarios in which the other drivers remain unchanged while disability prevalence (DIS) changes across its full range (ie, from 0.0% to 1.0%). Instead, we assume that the upper half of this range (ie, values of DIS from 0.5% to 1.0%) can only be achieved in scenarios in which the rate of growth in coverage and prices is 1.5% or greater and the rate of mortality improvement is 1.0% or greater. To quantify the potential impact of DIS over the next 50 years, therefore, we compare scenario E (central scenario) with scenario F, and scenario A with scenario B. This gives us the full plausible range for DIS, but set in contexts in which the associated values of the other drivers are likewise deemed plausible (Table 10). Table 10: Impact of change in disability prevalence, 2051

Scenario E F A B

Assumptions: COV 1.5 1.5 1.0 1.0 MORT 1.5 1.5 1.0 1.0 DIS 0.5 1.0 0.0 0.5

Expenditure % GDP 2051 9.2% 8.0% 7.6% 6.6%

14 Average of 2.0 / 0.5 and 2.5 / 0.5.


Increasing DIS by 0.5 percentage points decreases the share of GDP in 2051 by 1.2 percentage points or 1.0 percentage points in the upper and lower halves of the plausible range respectively. Within the plausible ranges of all drivers, therefore, DIS has an elasticity of approximately 2.2 (averaging across the full plausible range). Although comparison of elasticities is imprecise because of interaction effects, we can nevertheless conclude that – at plausible values for all drivers – growth in coverage and prices is approximately twice (4.5 / 2.2) as strong a driver of health expenditure as is non-fatal health status improvement.

Impact of mortality reduction (MORT) To quantify the potential impact of mortality improvement (MORT) over the next 50 years we compare scenario E (central scenario) with scenario G, and scenario B with scenario D. These scenarios were again selected to provide the full range of variation in MORT (from 1.0% to 2.0%) while ensuring that the context for such variation remained plausible (Table 11). Table 11: Impact of reduction in mortality, 2051

Scenario G E D B

Assumptions: COV 1.5 1.5 1.0 1.0 MORT 2.0 1.5 1.5 1.0 DIS 0.5 0.5 0.5 0.5

Expenditure 2051 10.2% 9.2% 7.2% 6.6%

Decreasing MORT by 0.5 percentage points in the upper half of its plausible range decreases the share of GDP by 1.0 percentage points, and the same change in the lower half decreases the share by 0.6 percentage points (note that the direction of effect is the opposite to that of disability reduction). Within the plausible ranges of all drivers, therefore, MORT has an elasticity of approximately 0.8 (averaging across the full plausible range), with opposite sign to that of DIS. We conclude that mortality reduction is a weak driver of health expenditure relative to disability prevalence reduction and (especially) growth in coverage and prices (average elasticities of 0.8, 2.2 and 4.5 respectively). We also conclude that, since across all plausible scenarios the nett effect of a reduction in mortality is to increase health expenditure, the distance-to-death effect is small relative to the other effects of mortality reduction, and is readily trumped by the contribution of mortality reduction to population ageing and morbidity expansion.

Conclusions on the impact of drivers (acting in isolation) In summary, across plausible ranges for all drivers:

• a 0.5 percentage point change in coverage and prices changes health expenditure in 2051 by 2% to 2.5% of GDP

• a 0.5 percentage point change in disability prevalence (non-fatal health status) changes health expenditure in 2051 by 1% to 1.2% of GDP


• a 0.5 percentage point change in mortality changes health expenditure in 2051 by 0.6% to 1.0% of GDP (note that this change is in the opposite direction to that of disability).

The key conclusions to be drawn are:

• health expenditure is most sensitive to growth in coverage and prices

• plausible improvement in non-fatal health status could potentially offset a proportion of the increase in health expenditure that would otherwise occur (as a result of ‘ageing pressure’)

• mortality improvement will increase health expenditure as its effects on ageing and morbidity expansion overwhelm its distance-to-death effect.

Interaction of health status dimensions: health expectancy and health expenditure

Life expectancy Under the central scenario (annual rates of mortality decline, at all ages and for both genders, of 1.5%), life expectancy at birth increases to 86.3 years by 2051. This turns out to exactly equal the actual extension of life expectancy between 1951 and 2002, at 8.6 years. Statistics New Zealand, like most national statistical offices, assumes that gains in life expectancy will slow in future, and projects (series 4) a value of only 84.5 years in 2051. This more closely resembles our scenarios in which mortality declines at only 1.0% per year (giving a life expectancy at birth in 2051 of 83.8 years). While the issue is controversial, the weight of demographic opinion appears to be turning against the assumption of progressive slowing in the rate of mortality improvement at older ages (Oeppen 2002). On the other hand, our remaining scenarios, in which mortality declines at 2% per year, giving a life expectancy at birth in 2051 of 88.9 years, may well be overly optimistic (Table 12). Table 12: Estimates and projections of life expectancy at birth, 1951–2051, genders pooled

MORT 1951 2002 2026 2051

1.0% 69.1 77.7 81.1 83.8 1.5% 69.1 77.7 82.3 86.3 2.0% 69.1 77.7 83.5 88.9

Note: MORT is the annual rate of mortality decline (all ages, both genders) from 2002 onwards.

Health expectancy An important finding of this study is that health expectancy (expectation of life without severity-adjusted disability) increases over the projection period (2002–2051) under all plausible scenarios (A–J). However, the increase varies widely, from a low of 2.5 years under scenario B to a high of 11.3 years under scenarios H and J; the increase is 7.0 years under scenario E, the ‘central’ scenario. The gain is highest under those scenarios in which both mortality and disability decrease the most, and least under scenarios in which mortality reduces only slowly and disability prevalence remains stable at 2002 levels.


Life expectancy with disability At the same time, under most (but not all) scenarios, expectation of life with severity-adjusted disability also increases (ie, absolute expansion of morbidity occurs). The range is from –1.9 years (absolute compression) under scenario F to +4.6 years (absolute expansion) under scenario C. An increase of 1.0 years occurs under scenario E. So the absolute number of years expected to be lived in a ‘disabled’ state (as defined) increases under scenarios where mortality reduces rapidly but disability remains stable or declines only slowly. By contrast, absolute compression of morbidity is seen only where mortality reduces slowly and disability reduces rapidly.

Proportion of life expected to be lived with disability Of more relevance to health expenditure is relative rather than absolute compression or expansion of morbidity. This can be measured in several ways (eg, the ratio of health to life expectancy), but perhaps the most intuitive measure is the percentage of life expected to be lived in a disabled state. This key parameter varies across scenarios from a 3.6 percentage point decrease (relative compression) under scenario F to a 3.9 percentage point increase (relative expansion) under scenario C. Scenario E shows virtually no change (15.3% in 2002, 15.1% in 2051). Again, whether relative compression or expansion of morbidity occurs is entirely determined by the relative rates of decline in mortality incidence and disability prevalence. If disability declines rapidly relative to the rate of mortality reduction, the percentage of life years lived with disability drops from 15% towards 12%. If disability declines slowly or remains stable while mortality declines rapidly, the percentage of life years lived with disability rises from 15% towards 19%.

Relationship of relative compression / expansion of morbidity to health expenditure The relationship between relative compression or expansion of morbidity and health expenditure can be most clearly seen by comparing scenarios G, E and F – illustrating varying rates of disability decline against varying rates of mortality reduction within the plausible range for both drivers (and with COV and PROD held constant at their most likely values, such that COV – PROD = 0) (Table 13). Table 13: Selected scenarios, illustrating expansion and compression of morbidity, 2051

Scenario DIS MORT COV – PROD % life lived disabled

Growth in per capita

expenditure

Expenditure as % GDP

G 0.5 2.0 0 16.1 200% 10.2 E 0.5 1.5 0 15.1 176% 9.2 F 1.0 1.5 0 11.7 140% 8.0

Note: All outcomes refer to the full projection period (2002–2051).


Relative expansion of morbidity (exemplified by scenario G and characterised by rapid mortality improvement yet slow disability improvement) results in an increase in health expenditure (per capita or as share of GDP) relative to that expected under the central scenario (scenario E). By contrast, relative compression of morbidity (exemplified by scenario F and characterised by modest mortality improvement accompanied by rapid disability improvement) results in a decrease in health expenditure relative to that anticipated under the central scenario. Note that under the central scenario, neither relative compression nor expansion of morbidity occurs over the 50-year projection period: the percentage of life expected to be lived with severity-adjusted disability is 15.3% in 2002 and 15.1% in 2051 (scenario E), a state described as ‘dynamic equilibrium’, in which mortality and disability move in step (Manton 1982). The selected scenarios indicate that plausible degrees of morbidity compression could be associated with reductions in growth of the health share of GDP of approximately 1 percentage point. This corresponds to roughly a 10–15% relative reduction in the health share of GDP ([9.2 – 8.0] /9.2, or [10.2 – 9.2] / 10.2). Put another way, up to approximately one-third15 of the anticipated increase in total spending pressure could potentially be offset through plausible degrees of morbidity compression (whether achieved by the special case of disability reduction alone in the context of stable mortality, as discussed previously, or by the more general case of relatively faster reduction in disability than mortality, as illustrated in Table 13, scenario F).

15 For example, growth in the health share from 6.2% to 8.0% (the outcome given relative compression

under scenario F), rather than to 9.2% (the outcome given neither relative compression or expansion under scenario E).


Discussion

Limitations Models of this type are all limited by the quality of the data available – both historical information and information about possible futures. Our model is also limited by the health state classification used, though to have a health state classification at all is an improvement on much of the earlier literature.

Baseline costs The costs by age, gender and service category are robust. However, the split between different health states (disabled/non-disabled and decedent/survivor) are ‘best estimates’ only. The partitioning of costs across health states has been based on two main assumptions: • that disability support service and mental health service costs do not vary by survival status • that the cost of dying does not vary by disability status. Neither of these assumptions are based on solid New Zealand evidence (though there is some domestic and international evidence for the cost of dying assumption).

Projection of coverage and prices (COV) Projecting the rate of growth in coverage and prices is difficult, because this depends on government policy, international developments in health care and related technologies, and other factors. We chose to estimate COV by back-casting, which suggested values in the range of 1.0–2.0%, depending on the historical period selected. In fact, historically this parameter has not tended toward any particular value, making projection highly uncertain. We therefore chose a range of 1.0–2.0%, which we consider has a good chance of covering the plausible range over long time periods.

Projections of mortality (MORT) Projecting mortality is relatively straightforward, as mortality has been declining reasonably steadily for over a century now, and there is no reason to believe this will not continue. However, as mortality falls – especially in younger age groups, where it is already very low – the rate of decline may slow. The progression of the obesity and associated diabetes epidemics will also have some effect, especially at middle and older ages, so we anticipate that the future rate of mortality decline may be as low as 1% per year, but could be as high as 2%. These rates imply period life expectancies at birth (genders pooled) in 2051 of 84 years and 89 years respectively. We are confident that this captures the plausible range of future mortality.


Projections of disability prevalence (DIS) By contrast, projecting future disability prevalence is again difficult. The only New Zealand data available come from two national surveys only five years apart in the mid-1990s to early 2000s. Comparison of these two surveys shows little change in disability prevalence, especially once weighted for severity. Furthermore, the international data are conflicting, with some studies (mainly from the US) suggesting rapid rates of decline in disability prevalence in recent years (up to 2% per year or even more), and other studies (mainly from Australia and the UK) suggesting little improvement at all (and possibly even a deterioration in severe disability) (see Appendix 1). Our overall assessment is that severity-adjusted disability prevalence is unlikely to increase, but could remain stable or decrease slowly, hence our plausible range for annual reduction in severity-adjusted dependent disability prevalence of 0.0 to 1.0%. It is worth noting that our model is simplistic in considering only disability prevalence, rather than full disability dynamics. A more sophisticated model would focus on disability incidence and remission, and estimate disability prevalence as an outcome of these processes interacting with mortality. However, we have few data on disability incidence or remission rates, or excess risks of mortality.

Health state classification Given the importance of distance to death as a driver of health expenditure, the classification of people as ‘decedents’ and survivors’ is appropriate. However, the classification of non-fatal health outcomes is less satisfactory. We have chosen to focus on chronic disease, since this category accounts for the majority of health expenditure associated with non-fatal outcomes. Yet a significant share of expenditure, especially in younger age groups, is unrelated to chronic disease, but instead reflects acute illness and injury events. And of course maternity costs are driven by completely different drivers. Even among chronic diseases (physical and mental), age patterns and other risk factors are not uniform. Furthermore, we do not measure the prevalence of chronic disease directly, but index it through disability (adjusted for severity). This is useful in terms of the correlation with expenditure on disability support services, but further weakens the link between the classification of non-fatal health states and personal health care expenditure. Nevertheless, the use of severity-adjusted disability as the criterion for classification of non-fatal health states is common in the literature (Cutler and Sheiner 1998, Freund and Smeeling 2002) and seems to work well enough.


Key findings

Health expenditure and GDP Government health expenditure is a year-by-year decision set by government based on priorities within and outside health, as well as overall affordability within the government’s budget constraint. The projections reported here do not attempt to model that process and therefore are not forecasts of future spending. They do, however, reflect the possible range of future pressures on expenditures in the health sector. They can also be thought of as ‘what if’ scenarios; ie, given a particular set of assumptions about the drivers of health expenditure and GDP growth, what percentage of GDP might be spent on health and disability support services in 2026 or 2051? Or alternatively, given a ‘target’ percentage of GDP that government is willing to spend on health, what rate of growth in coverage and prices can be accepted over time? In the central scenario, health expenditure (as defined) is projected to increase from 6.2% of GDP at present to 9.2% in 2051, with an output range across the plausible set of scenarios of 6.6–11.7%. Thus the central scenario projects a 50% relative increase in the share of GDP going to health care over the next half century. This is a substantial increase in the share of national resources devoted to government health spending. The projection that government health expenditure will reach 9.2% of GDP by 2051 is based on the central values of COV – PROD, MORT and DIS. However, if GDP grows more slowly, coverage and prices grow more rapidly, mortality reduction accelerates, or disability reduction stalls, then the health share could potentially be as much as 15%. On the other hand, under the most favourable projections – in particular sustained growth in labour productivity and rapid improvement in non-fatal health status – the health share of GDP could stabilise below 7%. Our conclusion is that alarmist concern about a looming ‘crisis’ in health expenditure resulting from retirement of the baby boom generation is misplaced. We anticipate an increase (from 2002 to 2051) in government health expenditure as a proportion of GDP of about 50% in relative terms. Although this is a large increase, the model clearly shows that health expenditure growth is far more sensitive to assumptions about future growth rates of coverage and prices than to assumptions about trends in population ageing and health status. That is, it is the scope, volume and cost of health and disability support services – much of which is independent of ageing pressure – that will largely determine overall government health spending in the future. Undoubtedly, nett ageing effects will make it harder to limit future increases in spending, yet growth in coverage and prices (ie, spending increases that are not explained by demographic or health status changes) will be the key driver. Our results show that this was the case historically (ie, from 1951 to 2002), and – despite greater ageing effects and anticipated steeper trends in health status – this will continue to be the case for the next half century.


Age distribution of health spending We find a dramatic shift over the study period in the share of resources consumed by different age groups. Despite likely flattening of the age–cost curve as a result of (non-fatal) health status improvements, the dramatic shift in the age structure of the population overwhelms the shape change. In the central scenario, the share of expenditure going to older people (65+ years) is projected to reach 63% by 2051 (versus 40% at present). If we could double the rate of disability decline, this share parameter would still only reduce to 59%, showing that this parameter is relatively insensitive to trends in health status. Yet, at the same time, the ratio of per capita health expenditure on an older person to that on a younger person will steadily decline, according to our model. From this perspective, therefore, we can look forward to an improvement in intergenerational equity of expenditure on health care – again, the consequence of a flattening age–cost curve. This result is similar to recent historical trends in the United Kingdom, but not Australia, Canada or, indeed, in New Zealand (Seshamani and Gray 2003).

Health status and population ageing A commonly asked question is whether anticipated improvements in the ‘health’ of the next generation of older people will offset (at least in part) the impacts of population ageing on health expenditure. To answer this question, let us consider the impacts of improvement in mortality, disability and health expectancy in turn. It is important to note that mortality improvement has a two-way effect. Improvement in mortality rates increases health expenditure (living longer means an expansion of morbidity, all else being equal, and an older population). On the other hand, improvement in mortality also decreases the proportion of people dying in each year, so acting to reduce health expenditure (the distance-to-death effect). Therefore, the nett impact of mortality improvement is an empirical question. Our model suggests that, under plausible ranges for all drivers of expenditure, the distance-to-death effect is relatively small and is swamped by the other effects of mortality, with the result that mortality reduction typically increases spending pressure (albeit less strongly than changes in disability prevalence and – especially – coverage and prices). This finding of a relatively weak distance-to-death effect is in contrast to some studies reported in the international literature (Lubitz and Riley 1993, Miller 2001, McGrail et al 2000). The explanation for this discrepancy appears to be that New Zealand includes long-term care spending within the health care sector, whereas most countries only count (acute) medical costs, which are much more sensitive to distance-to-death effects. Our approach is more comprehensive, including as it does the full range of government expenditure incurred by people becoming sick, needing support or dying, and so may provide a more balanced perspective on distance to death.


Given this complex two-way effect of mortality, and the fact that mortality improvement is itself a driver of ageing, the question asked earlier is better framed in terms of the effect of improvement in non-fatal health status on total spending pressure. We find that plausible reductions (eg, 0.5% per year under the central scenario) in disability prevalence could offset a substantial proportion of the anticipated increase in total spending pressure, the exact proportion depending both on the growth in coverage and prices and – more importantly – on the simultaneous rate of mortality improvement. So the nett effect of ‘health status improvement’ on spending pressure depends on the interaction between mortality and disability trends – that is, on health expectancy. Our research shows that health expectancy will improve in absolute terms under all plausible scenarios. Yet the growth in health expenditure will only be restricted when health expectancy increases relative to life expectancy – and such relative compression of morbidity is anticipated only under some scenarios (ie, those in which disability falls rapidly relative to mortality). Under plausible scenarios in which such relative compression occurs (eg, scenario F), the growth in the share of GDP going to health may be up to one third less than would otherwise be the case under the central scenario (scenario E, in which neither relative compression nor expansion occurs).

Conclusion The complex political economy effects of population ageing are not examined here. However, a substantial increase in the proportion of health care resources consumed by the 65+ age group, relative to younger age groups, appears inevitable. Nevertheless, we show that, when expressed as the percentage increase in per capita expenditure, improvements in health status will lead to intergenerational health spending becoming more – not less – equitable as the age–cost curve flattens. More generally, we conclude that population ageing – despite associated improvements in (non-fatal) health status – will make a substantial contribution to future health spending pressure. However, this source of spending pressure will not be nearly as great as that arising from the anticipated growth in coverage and prices. The ageing of the population and trends in (fatal and non-fatal) health status may be important modifiers, but the key driver of health expenditure will continue to be the scope, volume and cost of health and disability support services that this (older) population demands. This confirms the earlier New Zealand study (Johnstone and Teasdale 1999), and is in keeping with most international literature (Jacobzone et al 2000, National Research Council 2000). Nevertheless, achieving a plausible degree of compression of morbidity could partially mitigate ageing pressure and so restrict the total increase in health expenditure as a percentage of GDP by up to one-third – in terms of the central scenario, reducing it from a three to a two percentage point increase over the next half century. On the other hand, expansion of morbidity would have the opposite effect, plausibly raising the anticipated increase in the health share of GDP over the half century from three to four percentage points.


Summary of key findings over the period 2002–2051, under ‘central’ assumptions • % GDP spent on health will increase from 6% to 9%. • Older people’s share of health expenditure will increase from 40% to 63%, yet the ratio

of spending on the average older versus younger person will decrease (by approximately 25%).

• Growth in coverage and prices, not population ageing, will continue to be the key driver of health expenditure.

• However, ageing will increase upward pressure on spending (especially from about 2026), so making it more difficult to constrain spending growth.16

• Yet relative compression of morbidity (if it can be achieved) will reduce lifetime health care costs and so ease ageing pressure on health spending, constraining total health expenditure growth (by up to 30% of what it would otherwise have been).

Key messages Policies aiming to maximise the benefits to society of the ageing population should focus on interventions to reduce disability rates over time. Reduction in the prevalence of disability (adjusted for severity) will flatten the age–cost curve and so improve intergenerational equity of health expenditure in relative terms. Such interventions are compatible with the Government’s policy of ‘Positive Ageing’. Improvements in health status will have complex effects on the ageing pressure exerted on health spending, depending in particular on trends in health expectancy relative to life expectancy. Cost-effective interventions that compress morbidity relative to mortality have the potential to decrease life-time health care costs, and so ease ageing pressure on health expenditure. Investing in such interventions does not necessarily entail a trade-off against life extension – both disability reduction and life extension are desirable and can be achieved, yet it may prove possible to (cost-effectively) accelerate progress in the former relative to the latter. Policies aiming to reduce the cost of the public health system should focus on managing the growth in scope, volume and cost of health and disability support services over time. Coverage and prices are by far the major drivers of health expenditure.

16 Ageing will increase pressure on Vote Health not only in a direct sense, as discussed in this report,

but also in an indirect sense, since increased superannuation spending – even if partially offset by decreased education spending – will put pressure on all other votes, including Vote Health. Such political economy effects of ageing are beyond the scope of this report.


Managing coverage and price growth in future will need to take account of many factors (some of which have been briefly mentioned in this report), and will be made more difficult by the simultaneous increase in ageing pressure. One major factor will be the increasing globalisation of the health workforce, which will put increasing upward pressure on wages and salaries. Greater vigilance will also be necessary in monitoring service coverage and the diffusion of new technologies. While New Zealand has done well in comparison with other similar countries in controlling capital expenditure and (to a lesser extent) expenditure on pharmaceuticals, we have been less successful in assessing and managing the introduction of other technologies, such as diagnostic tests and surgical procedures. More effort in ‘horizon scanning’, technology trialling and promotion of evidence-based guidelines for clinical practice may be worthwhile. Such efforts should aim, wherever possible, at re-directing investment toward cost-effective interventions that will reduce the incidence and improve the management of chronic, disabling diseases and conditions. While population ageing will inevitably make it more difficult to control health expenditure in the future than has been the case in the past, such investment has the potential to constrain spending growth while simultaneously achieving additional population health gains.


Appendix 1: International Trends in Disability Prevalence Only two national surveys of disability have been done in New Zealand – in 1996 and 2001 (Statistics New Zealand 1998, 2003). In the absence of a long time series of historical data, projection of disability prevalence was based on a comparative analysis of national disability surveys in other developed countries (as well as the New Zealand surveys). The method and results are briefly summarised below.

Method A systematic review of the literature on trends in disability prevalence in developed countries was undertaken. The search methodology included a MEDLINE search of the literature (from which key review articles were identified) and a search of selected government health and statistical websites (from which key surveys and their results were identified). Note that different definitions of disability used in different studies mean that levels cannot be compared across countries, but trends can.

Criteria Criteria for selection of studies were: • cohort study with aged-in cohorts, or serial cross-sectional prevalence survey • national coverage • institutionalised population included • long duration, including at least five years in the 1990s • outcome measures analysable as ADL (severe) and IADL (moderate) limitations • minimal drift in instruments or field methods • high participation rate / low rate of loss to follow-up (cohort studies only) • minimal use of proxy responses • English-speaking country. Not all studies included met all the criteria. Specifically, the Framingham Heart Study was included despite being non-national and excluding institutionalised people, because of its long follow-up period and use of observational rather than self-reported measures of disability. On the other hand, the US National Health Interview Survey (including its Supplements on Aging) was excluded as it proved difficult to express its endpoints in terms of severe (ADL) and moderate (IADL only) disability. In all, seven studies were included, comprising five serial cross-sectional prevalence surveys and two longitudinal (cohort) studies. The latter, both of which are US studies (the National Long Term Care Study and the Framingham Heart Study) are considered more reliable and valid than the former, both because of their design and because of the extremely high standard of follow-up, measurement and analysis achieved in these two cohort studies.


Table A1.1: Selected studies

Country Study Waves References

New Zealand Postcensal Disability Survey

1996, 2001 SNZ 1998, 2003

Canada Postcensal Disability Survey

1986 HALS, 1991 HALS, 2001 PALS*

Statistics Canada 2002a, 2002b

Australia ABS Survey of Disability

1988, 1993, 1998** Australian Bureau of Statistics 1990, 1996; Davis et al 2001

England Health Survey for England

1995, 2001 HSE95, HSE00-01

US I National Long Term Care Study

1982–1999 Manton and Gu 2001, Freedman et al 2002 (review)

US II Framingham Heart Study

1977–1994 Framingham

Allaire et al 1999, Freedman et al 2002 (review)

US III Medicare Current Beneficiary Survey

1992–1996 Medicare CBS

Waidmann et al 2001, Fredman et al 2002 (review)

* Major method drift from earlier waves. ** Method drift from earlier waves rectified post hoc (Davis et al 2001).

Analysis A standard template was developed and used to extract key data from the published results of the selected studies. Initially study results were converted into estimates of prevalence of moderate (IADL only) and severe (ADL) disability, if required. This was done for each 10-year age group included in the study (there was variable age restriction across studies) and both genders. Then a summary table was prepared, showing age-standardised prevalence of moderate and severe disability for the total adult population of each country (standardised to the WHO world population), with genders pooled. We pool genders because the underlying dynamic in disability prevalence should not be gender specific, at least in terms of direction of change. The only exception to this is New Zealand, as indicated. Finally we converted the change between extreme periods for each study to an average annual percentage change, assuming exponential behaviour.


Results

Table A1.2: Prevalence of disability

1996 2001

Moderate 8.0 8.4 New Zealand Severe 2.4 (males)

3.4 (females) 2.6 (males) 2.8 females)

Gender difference

1986 1991 2001

Moderate 6.6 6.7 7.5 2001 not comparable Canada Severe 2.0 2.1 2.0

1988 1993 1998

Moderate 8.7 8.0 8.8 1998 comparability ‘rectified’

Australia

Severe 4.4 4.4 5.5

1995 2001

Moderate 13.5 12.0 UK Severe 4.0 5.0

1982 1989 1999

Moderate 13.6 13.4 12.4 Only first, middle and last waves shown

United States I (NLTCS)

Severe 5.7 4.8 3.2

1977 1994

Moderate 22.6 13.0 Non-inst pop of one town only

United States II (Framingham)

Severe 1.1 0.8

1992 1996

Moderate 13.7 12.0 Very short interval United States III (MCBS) Severe 4.3 5.0

Table A1.3: Average annual percentage change (exponential)

New Zealand

Canada* Australia United Kingdom

United States I

United States II

United States III

Mod +0.9 0.0 0.0 -2.4 -0.6 -3.5 -3.3 Sev -1.1 0.0 +2.3 +2.5 -4.2 -5.5 +3.8

* HALS 1986–1991 only; HALS not comparable with PALS.


Conclusions Results consistently show a decrease in prevalence of moderate (IADL only) disability over recent decades (mainly the 1990s), although this is less evident in the Australasian surveys. With the exception of the two major US studies and the New Zealand study, results are also consistent in showing an increase in severe (ADL) disability. Given the greater reliability of these US studies, and assuming that they may foreshadow what will happen later elsewhere, it seems reasonable to conclude that severe disability prevalence is unlikely to increase in New Zealand but will probably stabilise or even decrease. It is interesting that no study shows the anticipated decline in severe disability compensated by an increase in moderate disability (predicted under the ‘dynamic equilibrium’ theory of Manton (1982)), except for the New Zealand study.

Plausible range for modelling Because we are interested in severity-weighted disability prevalence over the 50 years from 2002 to 2051, we need to weight the estimated trends for severe vs moderate disability. The former involves at least twice the level of consumption of health care and disability support resources as the latter. On the other hand, the prevalence of moderate disability is at least twice that of severe disability (eg, approximately 8% vs 3% overall in New Zealand). Hence equal weighting of the two series seems an appropriate choice. Given this weighting, the possible rise in severe disability is more than compensated for by the very consistent fall in moderate disability, with the result that the plausible range may be concluded to vary from stability to a gradual decline. A future increase in severity-adjusted disability prevalence appears unlikely based on this analysis. The range in ‘DIS’ (average annual relative change in disability prevalence) selected for this study was therefore 0.0 to –1.0% per year. The sensitivity analysis has been done using average annual reduction rates of 0.0% (high disability prevalence), 0.5% (medium disability prevalence) and 1.0% (low disability prevalence) respectively, for all age-sex cells.


Appendix 2: Model Mathematics A simplified version of the model mathematics is provided here for interested readers. A full version is provided in the Treasury report (Bryant et al 2004).

Person-years Let pij(t) be the total population of gender i in age group j in year t pd

ij(t) be the corresponding population of decedents pdu

ij(t) be the corresponding population of disabled decedents pdw

ij(t) be the corresponding population of non-disabled decedents psu

ij(t) be the corresponding population of disabled survivors psw

ij(t) be the corresponding population of non-disabled survivors and mij(t) be the mortality rate of gender i and age group j in year t mu

ij(t) be the corresponding (non-injury) mortality rate of the disabled mwij(t) be the corresponding (injury) mortality rate of the non-disabled zij be the proportion of deaths attributed to injury in the 2001 mortality database uij(to) be the corresponding severity-adjusted disability prevalence rate in the launch

year (2002) h be the annual rate of change in severity-adjusted disability prevalence then mw

ij(t) = zij mij(t) for age < 65 years mw

ij(t) = 0.5zij mij(t) for age > 65 years17 uij(t) = uij(to)[1 – h]t – to pdu

ij(t) = 0.5muij(t)pij(t) + 0.5mu

ij(t+1)pij(t+1) pdw

ij(t) = 0.5mwij(t)pij(t) + 0.5mw

ij(t+1)pij(t+1) psu

ij(t) = uij(t) pij(t) – pdij(t)

pswij(t) = pij(t) – pd

ij(t) – psuij(t)

Costs Let cdu

ij(to) be the cost weight for disabled decedents in the launch year (2002) cdw

ij(to) be the corresponding cost weight for non-disabled decedents csu

ij(t) be the corresponding cost weight for disabled survivors csw

ij(t) be the corresponding cost weight for non-disabled survivors g be the annual rate of change in coverage and prices (cost weights) then cdu

ij(t) = cduij(to)[1 + g]t–to

cdwij(t) = cdw

ij(to)[1 + g]t–to csu

ij(t) = csuij(to)[1 + g]t–to

cswij(t) = csw

ij(to)[1 + g]t–to

Summation Health expenditure in year t for each five-year age group by gender by health state is given by the product of the relevant person-year and cost weight terms as defined above.

17 Reduced by half to reflect the high prevalence of chronic disease and disability in older age groups.


Summing over both genders, all age groups, and all health states, total health expenditure in year t is given by the expression:

C(t) = ΣiΣj[cduij(t)pdu

ij(t) + cdwij(t)pdw

ij(t) + csuij(t)psu

ij(t) + cswij(t)psw

ij(t)] where the cost weight and person-year terms are as defined above. Notes Injury and non-injury mortality rates are used to approximate mortality risks for non-disabled and disabled individuals respectively. Only the simplest form of the model, in which rates of change are constant across all age by health state cells, is shown above. The model is a deterministic macrosimulation model. Uncertainty is quantified through sensitivity

analysis.


Appendix 3: Key Results by Scenario Key results are presented for each of the 10 scenarios, for selected years; ie, the start (2002), midpoint (2026) and end (2051) of the projection period. Selected demographic results are presented first, followed by health status results, then health expenditure results. Table A3.1: Scenario A (COV = 1.0, MORT = 1.0, DIS = 0.0)

Variable 2002 2026 2051

Demography Total population (million) 3.91 4.47 4.61 % older people 65+ 11.9 17.5 22.4 85+ 1.3 2.2 4.3

Health expectancy Life expectancy without disability 66.4 67.7 68.8 Life expectancy with disability 12.0 13.4 15.0 Total life expectancy 78.4 81.1 83.8 % life spent disabled 15.3 16.5 17.9

Health expenditure Total expenditure ($ billion) 7.70 13.16 21.13 Per capita expenditure ($ ages pooled) 1969 2945 4581 Expenditure as % GDP 6.2 6.7 7.6

Distribution of expenditure by life-cycle stage (%) 0–14 11.8 8.2 6.2 15–64 48.3 41.0 32.0 65+ 39.9 50.8 61.8

Per capita expenditure by selected age group 2–29 1545 1931 (25%) 2477 (60%)45–49 1359 1719 (27%) 2196 (62%)65–69 3434 4338 (26%) 5555 (62%)85–89 13,975 17,595 (26%) 22,477 (61%)Ratio 85–89 : 25–29 9.0 9.1 9.1

Notes: For projection of GDP per capita, PROD = 1.5.Results are pooled for both genders, and (where applicable) for all age groups.


Table A3.2: Scenario B (COV = 1.0, MORT = 1.0, DIS = 0.5)

Variable 2002 2026 2051






Notes: For projection of GDP per capita, PROD = 1.5. Results are pooled for both genders, and (where applicable) for all age groups.


Table A3.3: Scenario C (COV = 1.0, MORT = 1.5, DIS = 0.0)

Variable 2002 2026 2051








Table A3.4: Scenario D (COV = 1.0, MORT = 1.5, DIS = 0.5)

Variable 2002 2026 2051








Table A3.5: Scenario E (COV = 1.5, MORT = 1.5, DIS = 0.5) (central scenario / base case)

Variable 2002 2026 2051






Notes: For projection of GDP per capita, PROD = 1.5. Results are pooled for both genders, and (where applicable) for all age groups,


Table A3.6: Scenario F (COV = 1.5, MORT = 1.5, DIS =1.0)

Variable 2002 2026 2051








Table A3.7: Scenario G (COV = 1.5, MORT = 2.0, DIS = 0.5)

Variable 2002 2026 2051








Table A3.8: Scenario H (COV = 1.5, MORT = 2.0, DIS = 1.0)

Variable 2002 2026 2051








Table A3.9: Scenario I (COV = 2.0, MORT = 1.5, DIS = 0.5)

Variable 2002 2026 2051








Table A3.10: Scenario J (COV = 2.0, MORT = 2.0, DIS = 1.0)

Variable 2002 2026 2051








References Allaire SH, LaValley MP, Evans SR, et al. 1999. Evidence for decline in disability and improved health among persons aged 55 to 70 years: the Framingham Heart Study. American Journal of Public Health 89: 1678–83.

Australian Bureau of Statistics. 1990. Disability and Handicap, Australia, 1988. Canberra: Australian Bureau of Statistics.

Australian Bureau of Statistics. 1996. Disability, Ageing and Carers, Australia, 1993. Canberra: Australian Bureau of Statistics.

Bryant J, Teasdale A, Tobias M, et al. 2004. Population Ageing and Government Health Expenditures in New Zealand 1951–2051. Wellington: New Zealand Treasury (working paper).

Cutler D, Sheiner L. 1998. Demographics and Medical Care Spending: Standard and non-standard effects. Working Paper 6866. Cambridge, MA: National Bureau of Economic Research.

Davis E, Beer J, Gligora C, et al. 2001. Accounting for Change in Disability and Severe Restriction, 1981–1998. Working Papers in Social and Labour Statistics. Canberra: Australian Bureau of Statistics.

Freedman VA, Martin LG, Schoeni RF. 2002. Recent trends in disability and functioning among older adults in the United States: a systematic review. Journal of the American Medical Association 288: 3137–46.

Freund D, Smeeding TM. 2002. The future costs of health care in aging societies: is the glass half full or half empty. Ageing Societies: Responding to the policy challenges. Sydney: University of New South Wales.

Health Survey for England 1995 and 2000-01. URL: http://www.official-documents.co.uk/cgi-bin

Jacobzone S, Cambois E, Robine JM. 2000. Is the Health of Older Persons in OECD Countries Improving Fast Enough to Compensate for Population Ageing? Paris: OECD.

Johnstone G, Teasdale A. 1999. Population ageing and Health Spending: 50 year projections. Occasional Paper No 2. Wellington: Ministry of Health.

Lubitz JD, Riley GF. 1993. Trends in Medicare payments in the last year of life. New England Journal of Medicine 328: 1092–6.

Manton K. 1982. Changing concepts of morbidity and mortality in the elderly population. Milbank Quarterly 60: 183–244.

Manton KG, Gu X. 2001. Changes in the prevalence of chronic disability in the United States black and non-black population above 65 years of age from 1982 to 1999. Proceedings of the National Academy of Sciences of the United States of America 98: 6354–59.

McGrail K, Green B, Barer M, et al. 2000. Age, costs of acute and long-term care and proximity to death. Age and Ageing 28(3): 249–53.

Miller T. 2001. Increasing longevity and medicare expenditures. Demography 38(2): 215–26.

National Research Council. 2000. The health of aging populations. Preparing for an Aging World: The case for cross-national research. Washington DC: National Academy Press.

Oeppen J, Vaupel JW. 2002. Broken limits to life expectancy. Science 296(10 May): 1029–30.



Seshamini M, Gray A. 2003. Health care expenditures and ageing: an international comparison. Applied Health Economics and Health Policy 1: 9–16.

Spillman BC, Lubitz J. 2000. The effect of longevity on spending for acute and long term care. New England Journal of Medicine 342(19): 1409–15.

Statistics Canada. 2002a. A New Approach to Disability Data: Changes between the 1991 Health and Activity Limitation Survey (HALS) and the 2001 Participation and Activity Limitation Survey (PALS). Ottawa: Statistics Canada.

Statistics Canada. 2002b. A Profile of Disability in Canada, 2001. Ottawa: Statistics Canada.

Statistics New Zealand. 1998. Disability Counts. Wellington: Statistics New Zealand.

Statistics New Zealand. 2003. The 2001 Household Disability Survey. Wellington: Statistics New Zealand. URL: http://www.statistics.govt.nz

Waidmann T, Liu K. 2001. Disability trends among the elderly and implications for the future. Journal of Gerontology 55: S298–S307.

Woods J. 2000. Manual for the Long Term Fiscal Model. Treasury Working Paper 002. Wellington: The Treasury.

Yang Z, Norton EC, Stearns SC. 2003. Longevity and health care expenditures: the real reasons older people spend more. Journal of Gerontology 58B(1): S2–S10.

population ageing and health expenditure - ministry … ageing and health expenditure: new zealand...

Documents