A multinational retrospective longitudinal study was performed. Data were collected for all 27 EU countries for the period 2000 to 2019, from 3 databases:

Detailed descriptions of the health expenditure variables can be found in Supplementary Table S1.

The collected YLL, YLD and DALY values were adjusted for a standardised age and for both sexes.

Data treatment was performed using STATA^® (version 14.2) and Microsoft Excel^® 365. First, an exploratory data analysis was carried out which included a weighted average of DALY and YLD, and some graphical representations. For a better analysis of expenditures within the private sector, private expenditure (EPrv) was generated by the difference between PVT-D and the out-of-pocket expenditure (EOOP), and, for uniformity in the reading of the results, a logarithmization of the DALYs referring to each NCD was carried out in STATA^®.

Secondly, a econometric panel data models were used to assess the impact of healthcare expenses on DALYs, through cross-sections (analysis of between countries in a given year) and chronological sequences (analysis of a country over the years). To avoid collinearity issues, the analysis was performed in two steps: the first step consisted in the analysis of fixed effects (FE) and random effects (RE) models for the DALYs of a NCD for country i at time t (DALYit) with the total of health expenditures (ETot) as the single covariate. The FE model can be described by Equation 2,

Where β0 is a constant, β1 is the coefficient of the independent variable, μi are the country-specific effects that are assumed constant over time and verify ∑iμi=0, and εit are the normal error terms. The RE model is given by Equation 3,

Where αi stand for the country-specific effects that are now assumed to be normal random variables with null mean and equal variance, and ϵit are the normal error terms.

The second step was to consider as explanatory variables all possible combinations between EPub, EPrv and EOOP. Both FE and RE models were considered.

The option between the FE and RE models was based on the result of the Hausman test for a significance level of 5%. The BIC (Bayesian information criterion) was also used to obtain a parsimonious selection of independent variables (18–21). Only models with an overall coefficient of determination equal to or higher than 10% were analysed (22).

The evolution of DALYs for NCD, CMND and INJ in the EU from 2000 to 2019 is detailed in Figure 1. The burden of NCDs is significantly higher than the burden of CMND or INJU since the minimum for NCDs (16,800 years per 105 inhabitants in 2019) is more than five times greater than the maximum number of injuries (3350.28 years per 105 inhabitants in 2018) and about ten times greater than the CMND maximum (1719 years per 105 inhabitants in 2009).

Despite the tendency to maintain DALYs, the percentage of these due to YLD has changed. While the CMND and INJU had a decrease in the rate of YLD within the DALYs (50.14 to 38.85% and 49.28 to 37.64%, respectively), the NCDs show an increase in the burden of YLD within DALYs, rising from 41.45 to 54.46% (Figure 2).

When analysing the evolution of DALYs within NCDs (Figure 3), cardiovascular diseases presented the highest DALY values within NCDs. During the period under review, these diseases had a positive evolution with a decrease over time (maximum 5,502 years per 105 inhabitants in 2002, and minimum 2,189 years per 105 inhabitants in 2019).

Neoplasms were the second most impactful NCD, with recorded values exceeding 4,000 years per 105 inhabitants during the initial three years under analysis, and in subsequent years these values consistently remained below that threshold.

Musculoskeletal and mental disorders and other NCDs change their position in terms of rank over time. However, all showed increasing values of DALYs. In 2019, mental disorders were the third NCD, followed by musculoskeletal disorders and other NCDs.

Neurological disorders is the sixth NCD with the most significant effect, with the lowest values in the first three years under analysis. The two highest records are found in the last decade (1,443 years per 105 inhabitants in 2012 and 1,438 years per 105 inhabitants in 2019), which indicates an upward trend.

Conversely, there has been a notable downward trend in the percentage of YLD (Figure 4) in diabetes and kidney diseases (maximum of 65.10% in 2000 and minimum 47.58% in 2019) and substance use disorders (maximum of 79.12% in 2003 and minimum 52.41% in 2010), while chronic respiratory diseases follow an increase in the percentage of YLD (minimum 41.43% in 2000 and maximum 57.22% in 2019). When analysing the variation in the rate of YLD between 2000 and 2019, these same diseases were the only ones with changes exceeding five percentual points.

The total of health expenditures (ETot), public and private, has increased since 2000 (6.90% of GDP), reaching the maximum value in 2009 (8.47% of GDP) and remaining above 8.00% until the end of the study period (cf. Figure 5). The maximum expenditure occurred in 2019 in Germany (11.70% of GDP), while the minimum was in 2000 in Romania (4.21% of GDP).

Public expenditure (EPub) was the ETot component with the most significant impact on health, its evolution over time was similar to the total growth. Thus, EPub attained its minimum in 2000 (5.04% of GDP) and its maximum in 2009 (6.18% of GDP), maintaining approximately 6.00% of expenditure afterwards. When examining the data by country (Figure 6), significant variations in values were observed. Sweden, the country with the highest EPub, presented an expenditure of 9.28% of GDP in 2018 which is three times higher than Cyprus in the same period (2.88% of GDP in 2018).

On the other hand, private sector expenditures (Private expenditure (EPrv) and out-of-pocket expenditure (EOOP)), had a similar evolution over time. On average, both types of expenditure reached their minimum recorded at the beginning of the period under study, although the maximum of EPrv occurred in 2019 (0.60% of GDP) and of EOOP in 2014 (1.78% of GDP). By country, the Netherlands had the highest EPrv in 2017 (2.49% of GDP), while for EOOP, Bulgaria had the highest in 2012 (3.60% of GDP). The countries with the lowest EPrv and EOOP were, respectively, Slovakia in 2004 (0.003% of GDP) and Luxembourg in 2019 (0.52% of GDP).

In short, all health expenditures have an upward trend with stabilisation in the last decade. By analysing the typology of health expenditure (Figure 6), it is possible to observe that Cyprus was the only EU country with an EPub mean lower than 50% of ETot (being 46.27% of these expenditures by EOOP), followed by Bulgaria and Latvia with ana EOOP higher than 40%. Most countries have an EPrv and EOOP lower than 30% (17 countries), thus there is a higher expenditure on health by public agencies. EOOP represent more than 50% of non-public expenditure in most countries, with a median of almost 20% of ETot. France (9.15%), Netherlands (10.16%) and Luxembourg (11.65%) have the lowest percentages of EOOP.

The panel data models were found to model the evolution of the DALYs of each NCD with ETot, as the only explanatory variables are detailed in Supplementary Table S2. Four of the NCDs presented a low overall r² (chronic respiratory disease, substance use disorders, diabetes and kidney disease and other non-communicable disease). Mental disorders despite an overall r² of more than 30%, did not present statistically significant for ETot. As for the other NCDs, all had a significant favourable evolution, except for musculoskeletal disorders, where, according to the fitted model, a 1% increase in ETot increased DALYs by 0.26%. The evolution was especially favourable in cardiovascular and digestive diseases.

To determine the effect of the expenditures EPub, EPrv and EOOP on the evolution of the DALYs of each NCD, panel data models were fitted for every combination of the explanatory variables. The most parsimonious model, according to the BIC criteria, was chosen and its parameters are presented in Table 1 (results for diseases with an overall r² less than 10% were omitted). Supplementary Table S3 shows the complete results.

Public health expenditure has a significant effect on all NCDs except for skin and subcutaneous diseases. However, in musculoskeletal disorders, the increase in public expenditure does not have a positive impact on DALYs (a 1% increase in EPub increases its DALYs by 0.32%).

Upon analysing both the private and public sector expenses, it becomes clear that these expenses have a significant impact on NCDs, with cardiovascular diseases showing the coefficients with the greatest improvements for EPub (10.37%), EPrv (45.32%) and EOOP (18.54%). Comparing the types of expenditure, the impact of the increase in private sector expenditure (EPrv and EOOP), in percentage terms, always showed an expected variation greater than that of EPub, in all models.

In this study, we assessed trends in DALYs and health expenditure across the 27 EU countries over 20 years (2000–2019), and analysed the effect of health expenditure in the DALYs evolution.

According to GBD 2019 Diseases and Injuries Collaborators (2020), with the increase in the sociodemographic index, there is an inversion of the burden from CMND to NCDs, where the contribution of YLD to DALYs becomes greater. This research also observed that NCDs present a more significant burden for health systems compared to CMND and injuries, as well as the trend of increasing disability, measured by YLD, for these diseases.

Despite the higher burden of NCDs, there was a maintenance of DALYs in the EU over time, through a downward trend in diseases such as cardiovascular and neoplasms and an upward trend for musculoskeletal and mental disorders. Daroudi et al. also observed a maintenance of DALYs for NCDs between 2000 and 2016 (worldwide), while Liu et al. observed a downward trend worldwide in DALYs for musculoskeletal disorders between 1990 and 2019 (23).

The drop of more than 1900 years per 105 inhabitants in cardiovascular diseases in 2019 should be carefully analysed since the mean in 2018 is lower than the three lower minimum values found for the same variable over the period analysed. This fall needs further studies and analysis in subsequent years.

Regarding the contribution of the YLD in the evolution of the DALYs, the NCDs have shown different performances. DALYs concerning cardiovascular, digestive, and sensory organ diseases decreased, and the YLD percentage remained similar over time. DALYs due to diabetes and kidney disease fell, but the YLL percentage increased. On the other hand, skin and subcutaneous diseases, chronic respiratory diseases, other NCDs and neurological, musculoskeletal and mental disorders had a trajectory of increased DALYs with YLD percentage showing low variation. However, disorders due to substance use increased DALYs and decreased the YLD percentage. Thus, most NCDs maintain the percentage of DALYs components over time, except for diabetes and kidney disease and disorders due to substance use, where there is an increase in premature deaths, and chronic respiratory diseases, with an increase in disability. Moreover, in the study by GBD 2019 Diseases and Injuries Collaborators, worldwide, an increase in YLL for disorders due to substance use was observed, justified by the inadequate prescription of opiates or fentanyl abuse. On the other hand, Kotwas et al., in a study in central Europe, observed an increase in DALYs for type 2 diabetes mellitus with an increase in YLDs (24).

EU adhere to the principle of universal access to healthcare, which is achieved through compulsory funding for the public sector, and there is no country in the EU (and very few worldwide) in which the private sector is the only source of access to health (25). Therefore, in this study, all EU countries financed their health systems through the public and private sectors, where, on average, there was an upward trend in all types of expenditure over time, with EPub being the ones that most contribute to ETot, followed by EOOP and EPrv. Note that Cyprus was the only country with a EPub component below 50%, as in WHO et al., being the value justified by the inability of the Cypriot health system to cover 10% of the population, motivating a significant reform in the health system in 2019 (26).

Thus, public entities were the major funders of health systems in EU countries, reaching over 80% in Croatia, Sweden, Denmark, Czechia and Luxembourg.

Figure 5 shows a sudden increase in health expenditure in 2009. This increase, as described by OECD & European Commission (2020), is due to a contraction in GDP due to the 2008 financial crisis and not to increased funding for health.

Both this study and WHO et al. observed a higher burden with EOOPcompared to EPrv, except in France, Slovenia and the Netherlands. These results may be due to its quick access, the provider’s choice or better facilities provided by the private system. However, it also shows a deficit in the articulation of health subsystems and the private sector, since EOOP are borne directly by users, withdrawing income or savings from households. In the EU, on average, 1/5 of total health expenditure is paid out-of-pocket, mainly for pharmaceutical, dental and other long-term healthcare services (26).

The NCDs received, until 2019, a residual investment, mainly compared to the expenses in diseases such as AIDS, tuberculosis, malaria and neonatal, child and maternal health (27). However, it is estimated that a 1% increase in per capita health expenditure reduces DALYs for all causes by 0.24%, and in countries with a high development index [as in the 27 EU countries (28)] the decrease in DALYs reaches 0.27% (29). This highlights the importance of analysing the impact of the health expenditure (30) in the DALYs for each category of disease as its potential benefit has been previously reported in other studies (31, 32). Our findings suggest that benefits regarding neoplasms, cardiovascular and digestive diseases are significantly higher than the estimated benefit of 0.27%, for all causes, estimated by Daroudi et al.

Most health expenditures are related to public organisations, reflecting fewer changes in DALYs compared to the private sector. All increases in EPrv and/or EOOP decrease the DALYs of NCDs, while the increase in EPub always shows less significant improvements compared to the private sector or even the increase in DALYs (as for musculoskeletal disorders). On the other hand, despite EOOP, in most countries, being responsible for more than 50% of expenditures in the private sector, the health outcomes for increasing these expenditures are always lower than the results with increasing EPrv.

Neurological disorders increased DALYs over time, however, only EPub significantly contributed to the decrease in their impact. Thus, a deeper private sector involvement should be considered in the future. Conversely, skin and subcutaneous diseases presented an increase in their DALYs and its evolution was only influenced by the private sector. A bigger contribution from the public sector would be important to face this increase.

Cardiovascular diseases showed the most significant effect on health systems through DALYs and simultaneously had the most favourable evolution when there was a 1% increase - in any health expenditure type - translating into a decrease in DALYs between 10 and 45%! These results show considerable attention to this pathology, justifying the downward DALY trend over time. Otherwise, musculoskeletal disorders showed the worst increasing trends for the period under study. As for the musculoskeletal, it was the only NCD in which the increase in ETotand/or EPub did not show positive results in the health of the population. Mental diseases showed a poor relation between DALYs and expenditures (overall r2 equal to 1%). This result shows that investment, mainly public, is not responding to the needs of the population since, as advocated by GBD 2019 Diseases and Injuries Collaborators (2020), there is little development of strategies for these diseases, given the low mortality (main focus of health policies at a global level). Singh et al., in southeast Asia, found better results with public expenditure, compared to private expenditure: a 1% increase in public expenditure reduced NCD mortality by 0.6%, while private expenditure increased mortality by 0.15% (33).

Several studies have already addressed the problem of assessing the impact of healthcare expenditure on health outcomes (e.g., mortality rate) (15, 16, 22, 33). However, to our knowledge, this is the first study in which the health outcome of interest are DALYs and it is important to conduct further research in this area.

This study had some limitations. Firstly, the study was designed as a second analysis of GBD data, and its limitations have already been published (such as the availability of primary data and the case definition or measurement method). Secondly, only the main categories of diseases within the NCDs were analysed without considering each disease, which may bias the results. Thirdly, only health expenditures were analysed as contributing to the DALYs, and the literature points to a multifactorial impact [risk factors such as poor diet, obesity and high blood (33, 34); socioeconomic and demographic structure of the populations and health inequalities (12)]. This is particularly clear in the cases where models have a very low overall coefficient of determination r² (e.g., mental disorders). Finally, the study did not consider the typology of health systems within the EU, relying only on definitions of health expenditure.