Determinants of increased health care spending: Colombia in the international context

: Financial sustainability in health refers to the balance over time between income and expenditure, so expenditure is a major fiscal challenge and its determinants require monitoring. The objective of this study is to define and measure the most important determinants that influence increases in health expenditure in 80 countries at different income levels, and to determine how the average level of increase compares with Colombia. A literature review was conducted to define the determinants. Then, a fixed-effects panel regression model was estimated to reveal the contribution of these determinants to the increase in spending. The results indicate that demographic variables and technological innovation have the greatest impact on the increase in health spending (45% and 43%, respectively). In conclusion, the study validates what is reported in the international literature with regard to upper-middle-income countries, where the evidence identifies that demographic and technological factors have the greatest impact on the increase in spending. There is also a need to include institutional and outcome variables to ensure that a fuller array of health spending determinants are considered, quantified, and explored.


Introduction
Financial sustainability in health systems is often the subject of debate in political, economic, and social circles. However, the debate is rarely accompanied by a clear idea of what it means for a health system to be financially viable or how to assess financial sustainability or, indeed, what the economic and policy implications of this problem are (Kutzin, 2008). The probable reason for this is that the meaning of financial sustainability is conceived as self-evident because the presence of an imbalance between the obligations of a health system and its capacity to cover them can shed light on the understanding of its functioning. While this idea is accurate, it is inadequate, as it says nothing about (i) the nature of the problem itself, (ii) the economic, political, and demographic implications of financial sustainability / unsustainability, or (iii) behavior and monitoring of the level at which resources (revenues) and uses (expenditures) should be aligned.
To understand financial sustainability, it is essential to carry out an analysis of the resources that are allocated to health financing and/or to analyze spending behavior (OCDE, 2015).
Although it is not possible to determine how much a country should spend on health (Savedoff, 2007), the pressure is on their governments, a priori and irrespective of other factors, as there is a predominance of public funding in different health systems (WEF, 2012). This is why rising health costs put significant pressure on national budgets, even more so when spending grows faster than economic growth (Global Burden of Disease Healt Financing, 2017).
To assess financial sustainability in health, it is necessary to analyze income and expenditure behavior independently, but most studies concentrate on containing the increase in expenditure (OCDE, 2015); (European Commission, 2014); (Appleby, 2013); (Abellán, 2010) & (WEF, 2012. On the income side, the need for greater priority of government resources and diversification of funding sources is discussed (WHO, 2010). Governments therefore seek to ensure greater prioritization and efficiency in the allocation of resources (OECD, 2015).
On the expenditure side, the impacts of different determinants are considered where, at the international level, technological innovations, the increase in prices and quantities of health services, the level of per capita income, population dynamics (particularly ageing), suboptimal use of resources, awareness of the value of health, lifestyles, population expectations, and insurance coverage are the main determinants contributing to increased expenditure and leading to substantive fiscal pressures on governments 4 (Willemé, P & Dumont, M, 2014); (European Commission, 2014); (Herwartz, H., & Theilen, B, 2014); (Cluter, D, 1995) & (Newhouse, 1992).
From a review of the relevant literature, variables that allow measuring the determinants of the increase in expenditure are selected for this study. These variables include technological innovations in health, demographic changes, price inflation in health, and income growth.
Further, two additional variables are proposed that are rarely measured in the literature, namely an outcome variable (health efficiency) and a policy variable (corruption). These latter two variables are fundamental to assess the increase in spending. Understanding the determinants of increases in health spending is particularly pertinent at the time of writing given the contemporaneous global crisis which has unfolded because of the COVID-19 pandemic which has necessitated increased public funding as well as adjustments and controls in financing policies to respond to the disease (Joe Kutzin, 2020).
Colombia has taken significant steps in the search for universal health coverage, which implies fiscal pressure on the government. Insurance coverage had reached 97% of the population by 2017, there is an equal benefit plan for the entire population, and in addition, out-of-pocket spending is low according to data from the WHO data repository for 2016 (16.4% of total health spending) compared to the average for Latin American countries (31.6%) and the OECD countries (30.7%). However, although significant progress has been made, there is concern about the continuity of the system's achievements and challenges, so it is important to understand the need to contain the increase in health spending pursuant of ensuring the financial sustainability of the system. Despite its importance, a broad and informed debate is not appreciated in Colombia, as there is no consensus on what sustainability really means, how to evaluate it, nor how to follow it up (GES, 2017). Therefore, there is a pressing need to measure the determinants of the increase in health expenditure pursuant of understanding its impact in countries with different income levels. This understanding could then be leveraged to optimize design specifications which can ensure financial sustainability in the near-and long-term. 4 The literature thus far does not address the effects that a pandemic could or would have on planning, organization, and reconfiguration in the For this study, a sample of 80 countries spanning different income levels was chosen, making use of a classification by the World Bank. The selection of countries was determined by the availability of information and the objective herein is to compare the average level of increase in spending across countries with that in Colombia so as to understand and inform the status and possible future trajectories of this Latin American country. The sample period spans 2000-2016, with the terminal year reflecting the lag in the data vis-à-vis the financing of health systems. Therefore, the year 2016 reflects the most recent data on total expenditure on health (% of GDP) at the time of writing. In sum, the results indicate that technological innovation is the main determinant of increased spending, followed by demographic changes, institutional variables, and outcome variables.
Based on this general framework, the following question is posed as a guide to advance this study: How much do different determinants contribute to the increase in health system spending? To answer this question, a panel data econometric methodology with fixed effects and intragroup estimators within dummies is employed.

I. Literature Review
The extant literature in this domain points to a lack of consensus regarding the definition of financial sustainability in health. The Health Economics Group (GES) at the University of Antioquia constructed a definition from a panel of experts for the Colombian case, which considers that financial sustainability concerns the temporal balance between the resources and the needs of the population or the objectives of the health system, in such a way that expected future expenditure corresponds to the available income of the next generations (GES, 2017).
The literature identifies a series of determinants of the increase in health spending, the most important of which are technological changes or innovations, demographic trends, and price and quantity trends in health (Appleby, 2013); (Abellán, 2010); (Newhouse, 1992) & (Cluter, D, 1995). It also identifies a number of determinants that have less of an impact namely income growth, sub-optimal resource utilization, health expectations, healthy habits, and chronic The literature also contains estimates of the contributions of different determinants to health spending increases. Technological innovation is the most important determinant, estimated to represent between 50% and 75% of the increase in expenditure (Newhouse, 1992); (Ginsburg, 2008); (Appleby, 2013). Demographic changes reflecting falling fertility rates and increasing life expectancy increase spending by between 2 to 38 per cent (Thomson, 2009), (WHO, 2006; (Dormont B, 2006); (Seshamani M, 2004); (Seshamani M, 2004), and (Office, 2007). Another factor of great interest is that related to inflationary trends in health, since the increase in prices and quantities can increase spending between 2 and 35%. (Abellán, 2010) & (Cluter, D, 1995).
Although some studies refer to a single determinant to understand and explore health spending In the case of Colombia, some approaches have been developed to measure the determinants of the increase in health expenditure. Gutiérrez (2018) carried out a study for Fedesarrollo and the results indicated that technological change has explained between 35% and 59% of the increase in spending. The study projects that public spending on health will reach 6.2% of GDP in 2020, where 44% of the growth will be explained by technological change, 38% by population growth, 12% by the change in demographic structure (aging), and the remaining 6% by the projected change in the epidemiological profile towards chronic diseases, which implies a fiscal pressure that generates particular challenges in relation to the sustainability of the system.
Bardey & Buitrago (2017) use a fixed-effects panel model to predict health expenditure for OECD countries and Colombia, and also proxy-evaluate the determinants of increased expenditure, identifying that demographic variables are positively related to such increases.
In this context, it is important to emphasize financial sustainability within the framework of the right to health and the regulation of

II. Estimating the determinants of increased health care spending
The methodological design is based on the construction of a panel data model in Stata 14, which includes a sample of 80 countries by income level for the period 2000-2016. The main objective of this method is to capture unobservable heterogeneity among countries. On the basis of a Hausman test, a fixed effects model is estimated, rather than its random effects counterpart (Baltagi, B., 2008); (Cameron, A. C. and Trivedi, P. K, 2005).
The World Bank's classification by country group and loans is used to define the income level of each country. Herein, 30 high-income, 21 upper-middle-income, 23 lower-middle-income, and 6 low-income countries are taken. This selection was informed by data availability, selecting those countries for which complete information is available for the requisite variables. Each of the 80 countries is observed for 17 years, thus n = 1360.
Given the criterion that sample countries must not be associated with missing information, in this sense, the panel data method is consistent in measuring the determinants of increased health expenditure for the set of countries ( It is complex to find specific data for each of the countries with respect to the determinants.  For the demographic variables, the Population Growth Rate (PGR) and an aging index (Envj) are constructed. The PCR is a state indicator that shows the situation of each country in terms of population evolution. It allows measuring the growth or decrease of a population in a given period based on the changes that a society experiences due to three demographic phenomena: birth rate, mortality, and migration (Marble, M, 2008). The birth, death, and migration rates are b, d, and m, which are a function of the total number of individuals in a country for a given period. Where N(t) represents the number of individuals alive at time t, analogously for N(t+Δt) and using the obtained rates: (1) Taking limits when Δt tends to zero: (2) The following mathematical model is obtained: Equation 4 represents the number of individuals in a time period, which is the solution of a differential equation. This equation is complemented by an initial condition N(0)=N0 that corresponds to the number of individuals in an initial instant: Equation 5 is a homogeneous linear differential equation of order one and when complemented with the initial condition, a single solution is obtained: Equation 6 is defined as the exponential growth rate of a population. Where r is the intrinsic rate of population growth In Equation 7, r reflects the population growth rate that measures the increase in a country's health expenditure given demographic change. Therefore, if r is greater than zero, the population grows. If r is less than zero, the population decreases, and if it is equal to zero, the population remains constant over time, and therefore health spending is expected to grow, decrease, or remain constant as the population growth rate varies.
To measure the aging index (Envj), the ratio of people 60 years and older to people under 15 years of age is calculated as (ECLAC, 2012). This makes it possible to appreciate the intergenerational changes resulting from the ageing process, as it highlights the changes in social demands generated by the elderly (ECLAC, 2012).
A health outcome variable is also constructed (Vrsalud), which makes it possible to establish efficiency ratios for the health system, since it is expected that the better the health outcomes of countries, the lower the expenditure (OECD, 2015). This variable is defined by taking the under-five mortality rate per 1,000 (U5MR), maternal mortality per 100,000 live births (MMR), and life expectancy at birth for each country (LER).
Then, by taking rates of variation, the health outcome variable can be decomposed as the product of nTMI, MMR, and VTE rates over a period of time: Obtaining equation 8 in periods t and t-1, and dividing it by Vrsalud (9) Since the rate of life expectancy at birth runs counter to the rates of infant and maternal mortality, the reverse applies to the latter. C , 2013).

The institutional variable (CORRUP) is taken from the corruption perception index developed
by Transparency International, which evaluates three characteristics in the administration to control corruption risks: visibility, institutionality, control and sanction. With this information, the institutional variable can be related to the increase in health spending, and countries can be classified according to their level of risk to corruption (Table 2). Therefore the estimated model is as follows: Where αi is the individual effects and µit is the error term. When using equation (12), it is assumed that within individual units (countries), the explanatory variables (determinants) can be affected or biased, so fixed effects allow the impact of these time-invariant characteristics to be eliminated in order to assess the net effect of the determinants on the increase in health expenditure for the set of countries.
To determine the average level of health expenditure in each country and compare it with Colombia, a panel data model with fixed effects and intragroup estimators within dummies is used; these binary variables measure the variation between the units of analysis (countries) assuming that the slopes remain constant (Stock, J & Watson, M, 2003). This means that this model provides a mechanism for understanding the pure effects of the determinants on the increase in spending as a whole, controlling for unobserved heterogeneity (Cameron, A. C. and Trivedi, P. K, 2005). It also allows for a comparison of the average level of increase in Colombia's spending with respect to other countries. In other words, the model would take the following form: Where αi is the effect of the first individual (in this case Colombia) which is used as the base country, dj are the binary variables which take the value 1 if the data corresponds to individual j (another country different from the base) and zero in another case (base country), αj are the coefficients of the dummy variables and represent the degree to which the values of the country intercepts differ from the base country, X'i represent the explanatory or determining variables of the increase in expenditure.
In this way, the determinants of increased spending can be estimated for 80 countries by income level. This will make it possible to answer the question concerning the contribution of the determinants for all countries. In addition, the average level of increase can be compared with respect to Colombia.

Analysis and results
A statistical analysis of the variables suggests that there is greater variation within the sample as a whole than between or within countries. Given that the variables exhibit a positive relationship with the increase in health expenditure, which implies a direct relationship between this and the level of countries' income (Figure 1), the problem of growth in expenditure above the rate of economic growth is evident (Global Burden of Disease Healt Financing, 2017). By estimating the model with an intragroup estimator within dummies, a 94% correlation coefficient and a statistically significant F test are obtained (p<0.05); therefore the hypothesis that all estimators are simultaneously equal to zero is rejected. Therefore, the model absorbs the pure effects of the determinants on the increase in spending and the differential values between individual units (countries) taking Colombia as a reference.
This can be seen from the model developed in equation 13, where each country represents a unit of analysis. The first country is the basis for comparison of the average level of increase in health expenditure. Thus, Colombia is the reference value or base for the study; all the other country coefficients are the differentials that provide information with respect to the base. In addition, since this is an average growth value for the study period, the increase in annual health expenditure for Colombia and the other countries can be obtained.  With regard to the contributions of the various determinants of the increase in spending, technological innovation in health (TIIH) is statistically significant. Specifically, for each unit increase in health spending across the countries as a whole, this variable contributes 43% of the increase. When the variables that represent demographic change are analyzed, it is discerned that the Envj variable is statistically significant and its contribution to the increase in spending is 13% for each unit of spending increase. Next, TCP increases health spending by 34%.
Analysis of the outcome and institutional variables shows 32% of each one unit increase in health spending can be attributed to the CORRUP variable, given the low level of transparency and high risk of corruption in many countries. While the Vrsalud variable generates a reduction or containment of the increase in spending by 21% due to the good results achieved by some countries, especially high-income countries (OECD), both variables are statistically significant.
The inflation of prices and quantities in health (CPI) was found to be an insignificant determinant of health spending dynamics in this particular empirical context. The determinant of income growth (GDPpc) is statistically significant, but it does not generate effects on the increase of expenditure. This estimate identifies the income elasticity of spending as being between 0 and 1, which shows that for countries as a whole, health is a basic or necessary good (Baltagi, B & Moscone, F, 2010).
When the medium level of expenditure increase is arranged with reference to Colombia, the problem of the fiscal pressure faced by countries, especially those with high and medium-high incomes, is validated (WEF, 2012). In turn, this fiscal pressure can be understood in terms of the determinants on the increase of health expenditure. Most high-income countries have higher growth rates than Colombia, excepting Greece, Hungary, Israel, Poland, and Austria. Moving on, only three upper middle-income countries have higher growth rates, namely Bulgaria, Ecuador, and Romania. Colombia exceeds all other upper-middle income, lower-middle income, and low-income countries except Malawi, ranking 30th among the 80 countries in the sample.
This validates the hypothesis that countries with different income levels face a trend of increasing health expenditure, since when looking at Table 1 all countries tend to increase the GTS (% of GDP) and this trend is also evident for Colombia.
Finally, the effectiveness of the model is evaluated. Specifically, statistical inference tests are performed on cross-dependence or contemporary correlation, which allows us to infer whether the residuals are correlated with the units of analysis 6 . This test indicates that the null hypothesis that the residuals are not correlated with the units of analysis cannot be rejected (P = 0.399>0.05). In this sense, there is no contemporary autocorrelation (sectional dependence).
Moreover, since we have several units of analysis that are confronting each other, it is quite possible that we are faced with a problem of heteroscedasticity, that is, we are faced with very different variances between countries. Then, to correct this, a robust standard error model can be made to 7 validate the consistency of the values obtained in the results.

III. Discussion and conclusions
The study demonstrates what is confirmed by the literature regarding the problem of increasing health spending in countries according to their income levels (Global Burden of Disease Healt Financing, 2017). This common fact across countries makes it necessary to study the financial sustainability of health systems, because as demand and supply interact, a series of determinants are established that contribute to the increase in expenditure (WEF, 2012). In this regard, countries are making efforts to diversify resources and prioritize funding for health systems (WHO, 2010). But since resources are scarce, priorities must be defined so that a balance is struck between meeting the needs of the current population on the one hand and ensuring fiscal responsibility so that the needs of future generations can also be met on the other hand. Hence why studying the determinants of increased health spending is so important.
In more detail, by identifying, quantifying, and exploring the determinants, health policymakers and other salient decision-makers can monitor and evaluate the extent to which increases in health expenditure are being contained. This is of critical importance because public funding of health services generates a tax burden for the countries (WEF, 2012). Further, it is also necessary to include a series of variables that the literature does not take into consideration to broaden the analysis, such as outcome variables that can provide a reading of the efficiency of the health system, as well as institutional variables that reflect the priority of governments over the health of the population.
These variables are fundamental, particularly nowadays, given the scale and scope of the contemporaneous COVID-19 crisis. This pandemic has necessitated a channeling of resources into health systems, along with adjustment and control interventions. This increased spending can be demarcated to be a function of technological innovation as well as an increase in prices due to the increased quantities of different goods and services which have been required to deal with the surge in demand for healthcare services (Joe Kutzin, 2020 (Table 3).
In this sense, Colombia shows a similar trend to high and upper-middle income countries with respect to increased spending. Therefore, the determinants of the increase in health expenditure have important implications for countries, and it is essential to develop tools that make it possible to contain this increase, especially in relation to technological innovation in health, demographic trends, and institutional aspects. Although progress has been made in drug price control policy and relationships with the pharmaceutical industry in this respect, there is still a lack of clarity as to how far market decisions can influence the health of the population and it is also important to make progress in terms of transparency and reducing corruption.
For Colombia to leverage, and advance on, the achievements already made and thus successfully navigate the challenges of the continued operation of the health system, it is essential to increase negotiation strategies in the assignment of health providers based on the country's various needs, assets, and constraints. Similarly, it is necessary to discourage negotiation strategies based on extreme payment mechanisms and liabilities such as reimbursement. It is also important for the design of benefit plans and prioritization mechanisms to make it widely and transparently clear to the population what their rights and duties are when accessing the health system.