Ageing and future health care demand

Introduction

The phrase ‘demography is destiny’ refers to how population size and composition impacts future prosperity, and is as true today as it was in the 19th century. In today’s society, it provides a useful lens to explore how population ageing may impact health service needs.

Ageing of the population

The Australian Bureau of Statistics (ABS) refers to generation cohorts using the following generational monikers:

• Interwar, born 1945 or earlier
• The Baby Boomer Generation, born 1946–1965
• Generation X, born 1966–1980
• Millennials (Generation Y), born 1981–1995
• Generation Z, born 1996–2010
• Generation Alpha, born 2011–2025.¹

In 2021, Baby Boomers make up 21.5% of the Australian population, marginally greater in absolute terms than Millennials.²

Figure 1: Population size by birth cohort, Australia, 2021

Australia’s population has changed markedly in the past 30 years. In 1991, the median age was 32.4 years increasing to 38.4 years in 2021.³ Corresponding figures for Queensland are 31.8 years (1991) and 38.4 years (2021). Median age also varies geographically, ranging from 52.1 years in Hinchinbrook local government area to 24.5 years in Northern Peninsula Area in Queensland.⁴

The visualisation below shows how Queensland’s population has changed over the past approximately 50 years.

Figure 2: Changes in age distribution in Queensland population

Health care expenditure impacts due to population ageing

While much of the increased costs associated with an ageing population will be borne by the aged care sector, healthcare will also be impacted. Nationally, health, aged care and the National Disability Insurance Scheme are the top three fastest growing national government payments. National health spending is projected to increase from 2022–23 to 2062–63 from:

• 4.2% to 6.2% for GDP
• $4,000 to $8,677 for real spending per person.⁵

The ageing of the population is a main driver—the number of people 65 years and over is expected to more than double, and more than triple for those 85 years and over from 2022–23 to 2062–63. Currently, people in the 65 and over age group account for about 40% of health spending⁵ and demand for health services will increase as this cohort ages.

But these increases are not solely due to the ageing population—both age-specific and age-standardised rates have increased. From 1993–94 to 2019–20, national hospitalisation rates increased:⁶

• 39% for those under 65 years
• 72% for those 65 to 74 years
• 94% for those 75 years or older.

Figure 3 below shows how rates have changed in Queensland since 2002–03.

Figure 3: Changes in age-specific hospitalisation rates in Queensland

Exploring population ageing and healthcare impacts

Queensland hospitals

To explore how the ageing of the Queensland population could impact health care services, trend and projection analyses were conducted using the Queensland Hospital Admitted Patient Data Collection (QHAPDC). These projections are based on a set of assumptions and are not intended as exact forecasts but will give an indication of what might be expected if those assumptions were to apply over the projected time frame. Projections for the following conditions have been analysed:

• All cause hospitalisations
• All injuries
• Chronic obstructive pulmonary disease (COPD)
• Coronary heart disease (CHD)
• Dementia
• Diabetes (type 1 and 2)
• Mental and behavioural disorders
• Stroke
• 7-lifestyle-related chronic conditions—a combined group of 7 preventable chronic conditions: CHD, stroke, lung cancer, colorectal cancer, breast cancer, COPD and diabetes.
• Person awaiting admission to residential aged care service
• Chapter Z of The International Statistical Classification of Diseases and Health Related Problems, 10th Revision, Australian Modification (ICD-10-AM): Factors influencing health status and contact with the health system.

Inpatients waiting on placement to a residential aged care facility have been widely acknowledged to cause significant bottlenecks in the hospital system due to commonly longer lengths of stay.⁷ The ICD-10-AM code of Z75.11 Person awaiting admission to residential aged care service attempts to quantify these episodes.

The biggest contributor to all cause hospitalisations is ICD-10-AM chapter Z, Factors influencing health status and contact with health services. Chapter Z identifies why the patient is receiving services when a disease or disorder is not the reason the services were needed.

Modifiable risk factors

Modifiable risk factors are associated with several of these conditions and trend and projection analyses are used to forecast how these would change due to upcoming demographic shifts. The Queensland preventive health survey (QPHS) risk factors analysed were:

• daily smoking
• overweight and obesity
• obesity
• insufficient physical activity.

Methods summary

When trends in hospitalisation counts and prevalence of selected risk factors meet certain criteria, they can be projected—providing a reasonable estimate of the prevalence of these conditions into the future. Queensland Health uses the Acute Care Estimates (ACE) model for official health service projections.⁸ The ACE approach was broadly used for this exploration of ageing effects on hospitalisations, with analysis limited to episodes for Queensland residents only as is standard across this report. The period of historic data used to project trends was also aligned with the ACE method. Broadly, the approach was:

1. Generate counts by age group for each condition and apply the appropriate estimated residential population (ERP).
2. Conduct trend analysis to determine whether it’s appropriate to extrapolate into the future and, if so, apply population projections.
3. Calculate a projected age-specific rate based on the historic trend using linear or Poisson regression, or 3-year weighted means depending on the conditions characteristics, or use Poisson for risk factors.
4. Generate population level projected counts by applying age-specific rates to the population projections and then summing across age groups.
5. Finally, calculate crude rates using the above results.

More detailed methods are available under Methods and technical notes.

Projected rates and counts over time

Hospitalisations

Figure 4 presents:

• actual numbers of hospitalisation for 2016–17 and 2022–23 and projected hospitalisations for 2026–27 and 2032–33 by condition
• percentage increases from 2022–23 for 2026–27 and 2032–33.

Overall, hospitalisations are estimated to increase by 34.0% from 2022–23 to 2032–33. A large component of this is due to increasing population and the increasing age of the population. Results for specific conditions show that from 2022–23 to 2032–33:

• the largest increases (81.5%) will be for dementia followed by Waiting for aged care placement (74.4%)
• conditions estimated to decrease over time are CHD (20.1% decline) and COPD (40.2% decline) and the group of 7-lifestyle conditions (9.9% decline).

Figure 4: Projected hospitalisations by year and cause and percent change to 2032–33

Changes in proportion of total hospitalisations

The proportion of total hospitalisations explores the future makeup of hospitalisations by condition. For example, some conditions may make up proportionally more of total hospitalisations over time where other conditions may decline. Figure 5 shows changes from 2016–17 to 2032–33 for:

• the number of hospitalisations and the proportion of all cause hospitalisation
• the crude hospitalisation rate.

Using the filter to select specific hospitalisation causes, for example, estimated changes from 2022–23 to 2032–33 are:

• Factors influencing health status increases from 24.9% to 25.9% of all cause hospitalisations, which increases the number of hospitalisations from approximately 721,000 to 1,004,000 and the rate from 13,542 to 16,296 per 100,000
• Waiting for placement to aged care is estimated to increase from 0.3% to 0.4% of all cause hospitalisations with episodes increasing from approximately 9,000 to 16,000
• CHD is estimated to decrease from 1.2% to 0.7% (approximately 33,000 to 27,000 episodes) while COPD is estimated to decrease from 0.6% to 0.3% (approximately 17,000 to 10,000 episodes).

Figure 5: Past and projected hospitalisation patterns by condition

Figure 9 shows similar information for modifiable risk factors as for Figure 5 in the hospitalisation section above.

Figure 9: Current and projected prevalence of modifiable risk factors

The later years of life are typically characterised by high health care costs that increase as death is approached. A risk of extending years of life lived without addressing years lived in poor health is that the period of higher health care costs due to ill health is increased. Extending age only is a scenario that can result in the highest health care costs, as demonstrated by the estimated increases of $7,439 in 2015 to $9,594 per person by 2035.¹⁰

The effects of addressing ill health on health care costs in later years can be estimated using different approaches as shown below.¹⁰

• One approach is to delay the onset of ill health as life expectancy increases. This keeps the total period of ill health constant by delaying morbidity onset by the same amount of time as the increase in life expectancy. This approach decreases costs per adult over age 65 from $6,951 to $5,994.
• A criticism of the above scenario is the assumption that all gains in life expectancy are for years in optimum health. It may be more reasonable to assume that a proportion of those additional years is also spent in poor health. Under that assumption, health care costs per adult over 65 year increased from $7,220 to $7,719.

While it is beyond the scope of this report to model how changes in risk factor prevalence may influence hospitalisation rates, numbers and costs, the potential contribution of addressing modifiable risk factors by extending years lived in full health is included in the Health-related quality of life section. Analyses showed that years lived in full health could be extended by 2.5 years (males) and 1.8 years (females) at a population-level if smoking was entirely absent from the population. Similar gains were observed for obesity with more modest, but nevertheless important, gains for fruit and vegetable consumption. Assuming that the compressed period lived in poor health translated to commensurate reductions in hospitalisations illustrates how addressing modifiable risk factors can contribute to wider health care sustainability.

Methods and technical notes

Projection process

The process to calculate projections uses the following steps:

• Obtain counts by age group for each selected condition.
• Apply the appropriate denominator to each condition from the relevant year ERP.
• Analyse the trend for each condition and age group to determine if it is appropriate to extrapolate this into the future.
• Obtain future population counts by age groups from the population projections file available from the Queensland Government Statistician’s Office (QGSO).
• Use Linear/Poisson regression or 3-year weighted mean depending on the condition’s characteristics to calculate a projected age-specific rate from the historic trend. Additional methodological information for Poisson regression for risk factors is available in the Trends in preventive health risk factors report.
• Calculate the future count of the condition by age group by applying the age-specific rate to the population projection.
• Sum across age groups to obtain a population level projected count of the condition.
• Calculate a crude rate based on the population projection.

Age groupings

Age groupings for hospitalisations include 0 to 4, 5 to 14, 15 to 44, 45 to 69, 70 to 84 and 85 and over. This aligns with age groupings used elsewhere in Queensland Health for hospitalisation projections.⁸ Stratifying older age groups was deemed important due to the inter-generational report highlighting the impact of the population 85 years and older into the future. This is also reflected in the growth of this age group in the QGSO population projections.⁵

Due to their low number, the following adjustments were made to age groups:

• dementia: those under 45 years excluded
• waiting for admission to aged care: those under 15 years excluded
• coronary heart disease: those under 15 years excluded.

Age groups for risk factors included 18 to 29, 30 to 44, 45 to 64 and 65 years and over. As the risk factor prevalence come from a stratified sampling frame, estimates for the over 85 age group may be based on a sample size that is too small for the purposes of projection, therefore estimates for the older population is limited to those 65 years and older.

Why don’t we use age-standardised rates?

Age-standardised rates are preferred when comparing disease outcomes between two populations with different age distributions. When these outcomes are compared using crude rates, the differences between the populations will likely be due to differing age structures, as age is by far the biggest predictor for poorer health outcomes. The age-standardisation process will convert observed disease outcomes to what would have been expected if an area had a “standard population” (currently the Australian population as at 30 June 2001 is used).

The primary aim of the projection modelling, however, is to determine the number of cases that we are likely to be observing in future rather than from a hypothetical standard population. Calculating actual numbers and crude rates will better enable us to prepare for actual future population ageing and the impacts across the health and aged care sectors.

Denominators

The denominator used for historical hospitalisations data comes from latest ERP for the relevant year as provided by Statistical Services Branch (SSB), Queensland Health. Population projections used for anticipated denominators in the future comes from QGSO. An important consideration on the choice of population is the Australian Statistical Geography Standard (ASGS) version. Because the QGSO populations projections are based on ASGS2021 historical ERPs, analysis was limited to this population version. To align with the ACE methodology, trend analysis on historical hospitalisation data is limited to seven years from 2016–17 through to 2022–23.

Historical denominators used for survey estimates from the QPHS are derived from the population weighted survey estimate and its standard error. This will provide the appropriate level of historical precision of the estimate, which needs to be considered when projecting into the future. Population projections from QGSO are used as an anticipated future denominator in the same way as hospitalisations.

Validation of historical age group trends

Trends for each indicator by age group are fitted to see if counts are of a reasonable size (at least 10 per year) and to see if there is a consistent trend over the historical period. When counts are low, that age group is removed (as in dementia for the under 45-year-olds and waiting for admission to aged care for the under 15-year-olds), this will have a minimal impact on the overall projection. Inconsistent trend is assessed using a quadratic term. When it is inconsistent, a three-year weighted average of the age-specific rate in the most recent years is used to inform the projection.

Validation of regression methods

For hospitalisations, linear and Poisson regression is used to analyse the age-specific rates. Rates have the characteristic of being constrained to zero (meaning that a negative rate is not possible), Poisson regression is ideal in this situation as it will not allow a rate prediction to go below zero. Linear regression can be applied when hospitalisation rates are increasing as it is unbounded because an individual can be hospitalised multiple times.

Observed and expected values from the regression models are graphed and visually checked to ascertain the goodness of fit of the models. A common rule of thumb is to use equal-length time periods for both the historical trend and the projection. This modelling technique used in this feature article broadly follows the ACE methodology and closely aligns with results for all cause hospitalisations.

Methods for analysing trends for risk factors from the QPHS can be found in the Trends in preventive health risk factors Queensland 2002 to 2013 report.

Definition of hospitalisations

Hospitalisations from public psychiatric hospitals and those flagged as unqualified newborns, organ donors or boarders are excluded and aligns with reporting by Queensland Health and the Australian Institute of Health and Welfare. Hospitalisations for interstate and overseas residents are also excluded as the scope of all reporting throughout the Chief Health Officer report is for the health of Queenslanders.

A hospitalisation is defined as when someone is separated (discharged) from their admission to a hospital. A person can be separated multiple times for their treatment in hospital for example when treatment of their acute stage of their hospital stay ceases and transitions to rehabilitation the patient will be separated and therefore count as multiple hospitalisations even though the patient may not have left the facility. Also to note is that hospitalisations are not a count of individual people as a person can be admitted to hospital on more than one occasion throughout the year. More information regarding hospitalisation definitions can be found at Queensland Hospital Admitted Patient Data Collection (QHAPDC).

Definitions of risk factors

Metadata for risk factors used in these analyses is available from supporting documentation included with the Queensland survey analytic system which is the dissemination arm for the QPHS. QPHS results were used to model overweight and obesity, because the self-report data are more granular and more suited to projection modelling even though it is likely an underestimate.