1.1 Background

The gradual ageing of the population in developed countries is a reality that is posing major challenges, unknown to date, related to the sustainability of public finances [1], in areas as varied as pensions [2], savings [3], health spending and long-term care [4], that requires timely decision-making to ensure the viability of public accounts in the future. Within this framework, health economics has analysed for years the effect of population ageing on health systems and their use that is [5-7], where abundant empirical research has been generated. The growing trend over the last few decades in both health expenditure and the number of elderly people has led to search for the relationship between the two.

As said before, there has been a significant increase in public health expenditure in developed countries in recent years. In Spain, measured in relative terms, public health spending grew by 84.7% between 2002 and 2018.* Measured in per capita terms, Spanish public health expenditure in 2018 reached 1523 euros, the highest figure in the last decade, when in 2002 it was 921 euros. In relation to gross domestic product (GDP), public health expenditure in Spain in 2018 was 5.9% of GDP, being 5.1% in 2002.

Developed countries are also undergoing a rapid process of population ageing. The Spanish population over 65 years of age accounted for 19.26% of the total population in 2019†; only 20 years earlier, this proportion was 16.76%. The number of elderly people has risen from 6.7 million to 9 million in that period. The “ageing of ageing” is also a fact: people over 75 years of age accounted for 49.5% of the total elderly population in Spain in 2019 (in 1999 they were 42.2%), with an increase of 1.6 million people in that period. This trend, far from stagnating, looks set to continue to increase in the coming years in developed countries and many developing countries, due to an increase in life expectancy and a decrease in fertility rates, with important macroeconomic implications, including health spending.

These circumstances essentially pose a problem for the sustainability of public health and long-term care systems. In this area of analysis, part of the empirical research has focused its efforts on finding out how the progressive increase in the elderly population affects the increase in healthcare and long-term care expenditure [8-10], since it is to be expected, at least from a theoretical point of view, that an increase in the elderly population, due to gains in longevity and the increase in associated health and dependency problems, such as disability, the presence of chronic diseases and poly-morbidity, will generate significant increases in public health spending, which in turn will cause significant strains on public finances.

In this context, the analysis of the effect of the elderly population on the consumption of health resources and health expenditure constitutes an important area of analysis, given the unstoppable demographic trend towards ageing populations in developed countries.

Spain has a public, universal and free Beveridge-type health system, with funding from public budgets. Administratively, Spain is divided into 17 Autonomous Communities, with a high degree of decentralisation, where each is directly responsible for both the expenditure and the management of the health systems in its region. The Ministry of Health establishes a basic portfolio of common services for all the Autonomous Communities; but each of them has the power to set additional services in its portfolio, as well as to choose the organisational model that they consider most appropriate, always within the limits set under the laws of the Spanish state. 99.1% of the Spanish population has free public health coverage. The Spanish healthcare system does not have any type of copayment for acute or chronic healthcare, if presenting a copayment for nonhospital pharmaceutical consumption, which in the case of the elderly is very limited (in 2017, the maximum total monthly copayment for elderly pensioners ranged from 8.23 euros to 18.52 euros depending on income, except for incomes above 100,000 euros per year, rare in Spain, whose limit was 61.75 euros. In addition, there are total exemptions for certain cases, such as dependency cases or the receipt of noncontributory pensions).

This study aims to analyse the effect that aging may have on the use of acute care services. A second objective is to determine whether aging may have an interaction with medical technology growth, which has often identified as the main driver of increased healthcare spending [11-13].

1.2 Conceptual framework

The analysis of the effect of aged on public health expenditure in this paper is based on the model proposed by Meijer et al. [14] developed from Andersen and Newman [15], which incorporates other determinants of health expenditure, identified as societal determinants, into the original model, focused on the consumption of health resources, so that we move from a model of consumption of health resources to a model of health expenditure.

In this model, the determinants of healthcare services use and health expenditure depend on individual factors and social factors. Individual factors are further divided into three groups: predisposing factors, which are those that reflect the predisposition to use health services, including gender, living and working conditions, health behaviour, coresidence status, socioeconomic status and the main independent variable in this study, age.

The presence of age as a predisposing factor is essential in the analysis, because it indicates that age by itself is not necessarily a determinant in the consumption of health resources and may be relevant to the extent that it is related to the determinants of the second subgroup, the need factors, which indicate the reasons why an individual, given all other factors, makes use of health services. The main ones are health status, disability and mortality. Indeed, there is abundant empirical evidence that the effect of age on healthcare expenditure underlies the presence of chronicity or dependency conditions, as well as poor health status [7, 16-19].

The third subgroup, called enabling factors, is the group of resources available to meet the need for health resources consumption. These include the availability of informal care, which is mainly influential in the consumption of long-term services, personal or family income, and the presence of health insurance.

The social factors identified as determinants of increased health spending are national income, technology, health policy and healthcare system, prices and wages. Although national income is shown as a factor strongly related to the increase in health spending, when income is analysed at the individual level, the impact is much more limited [20, 21], an issue that can be explained by the presence of collective health insurance for Bismark-type health systems or, as in Spain, the presence of a Beverdige-type public health system with universal coverage.

As we previously pointed out, medical technology progress is often mentioned as the most important driver of increased healthcare expenditure, mainly in acute care [22, 23], essentially related to a significant increase in use [24]. The evidence further shows that medical technology interacts strongly with age and health, that is, population aging reinforces the influence of medical technology on health expenditure growth and vice versa. The present study delves into the relationship between ageing and health technology use to determine whether there is an interaction between both.

Prices and wages are the other two drivers identified by the model. As healthcare systems are labour-intensive, healthcare labour costs may be influenced by the so-called Baumol's cost disease [25], and there is a large literature on this effect in the healthcare sector [26, 27].

2.1 Sample and variables

The data used in this study have been taken from the adult module (over 15 years) of the 2017 Spanish National Health Survey (ENS 2017). The ENS is a national cross-sectional study. The data used in this study correspond to the survey conducted throughout 2017 in a total of 23,860 households and collects health information regarding the Spanish resident population, with a total of 23,089 individuals.

For the present study, and given that the aim is to analyse the effect of aged on public health expenditure, those individuals who did not have public health insurance at the time of the survey in any of the existing modalities in Spain were excluded, with a total of 204 cases being eliminated (0.88% of the sample) as well as those who, having public health coverage, also had private health insurance (3040 individuals, 1.45% of the sample), so that all the sample units analysed only had public health coverage at the time of the survey, totalling 19,845 individuals. In this way, the presence of the so-called enabling factors that directly affect the aims of the present study is controlled.

2.2 Statistical analysis

First, a univariate logistic regression was performed for each of the 11 dependent variables on the main independent variable age, to evaluate the isolated effect of ageing in each of the dependent variables used. Subsequently, multivariate logistic regressions were performed, with the statistical objective of modelling the probability of the occurrence of a given dichotomous event (the dependent variables) in the presence of a series of factors, described in the previous section.

We also run logistic regressions for each of the 11 dependent variables on the main independent variable age, controlled by pooled factors (need, enabling and predisposing). The purpose of this additional evaluation was to check the robustness of the previous results, analysing the effect of the main independent variable controlled only by the (pooled) variables of each of the three-factor groups described in the theoretical framework.

2.3 Collinearity

Given the possibility of collinearity, especially between the variables that make up the need factors, we performed three checks:

First, the correlation matrix of the estimation of the parameters for the model terms, for each of the different models carried out, was analysed, without finding evidence of high correlations between the need factor variables, except relatively high values for the pair Disability-SRHS.

Additionally, a study of collinearity of the variables in all the multivariate models was carried out through the variance inflation factor (VIF), from which it was concluded that there was no collinearity among the selected variables.

Lastly, the models were run consecutively removing the variables (need factors) on which there was suspicion of collinearity (always keeping the SRHS variable in the models), without appreciating significant changes in the regression coefficients. Therefore, the models used were validated.

The results of each regression are presented as odds ratio (OR), with the aim of facilitating their interpretation.

Three levels of significance were taken: <0.1, <0.05 and <0.01.