Health systems sustainability is one of the major challenges public policymakers in developed countries face. Since the second half of the 20th century, health expenditure has been increasing at a faster pace than national income levels all over the world 1. The worldwide 2008 financial crisis made evident the financial difficulties of health systems, and provided an opportunity to bring healthcare rationing to the political agenda 1. Currently, health care rationing is no longer a discussion issue, but rather how it should be performed in a transparent and explicit manner. At the micro-level, rationing or priority setting (terms that will be used interchangeably) is particularly complex and controversial, for it involves bedside treatment decisions about denying, on the grounds of scarcity, a potentially beneficial treatment to patients. On what basis should these priorities be established? This question remains unanswered among main social actors involved.
In response to this situation, several studies have proposed appropriate and conflicting ethical principles to take bedside healthcare rationing decisions 2-9, which can be divided into three groups: (i) economic (direct consequences of treatments, such as increasing life expectancy or improving the quality of life, or indirect consequences, such as long term care costs, both financial or the burden it represents to other people involved, either physical or emotional); (ii) medical (severity of illness), and (iii) patient based (age, life-styles, economic condition, among others).
Somehow, according to a fourth principle, explicit rationing is unethical, as no one should be forced to decide who is going to live or to die. Based on the 'fair chances' reasoning (including lottery and first come, first served basis) everyone has a fair chance in terms of receiving an appropriate treatment 10.
In addition, an increasing number of empirical studies l have been conducted around the world in order to find out which of these criteria are supported by the general public, on whose behalf health care rationing decisions are being made. Whereas medical based principles seem to be generally considered as valid for priority setting 11-14, there is not enough information on whether economic 13-19 or patient-based criteria 18-23 are acceptable by the general public for this purpose.
However, it is not possible to draw any detailed conclusion from these studies, since their results vary depending on aspects such as the countries that were surveyed, the operationalization that was chosen, and the research design 22. In fact, it seems that healthcare rationing preferences differ not only at individual levels, but also at cultural levels, thus a successful policy in health care priority setting must consider the values and cultural setting of the population for which it is intended.
In this context, the aim of this study is to explore the preferences of Spanish citizens regarding the criteria on which healthcare rationing decisions at the micro-level should be based. This research is particularly relevant for Spain for two reasons. First, the Spanish health system has faced great challenges due to the severe austerity measures that have been imposed for obtaining the bailout from the European Financial Stability Facility [for more detailed information on the Spanish health system reforms see Gallo and Gené-Badia 24 and Cervero-Liceras et al. 25. Second, to the best of our knowledge, so far there are not studies addressing the preferences of the Spanish population regarding the health care priority setting principles that should be considered as the basis for the allocation of scarce healthcare resources at micro-levels.
Therefore, this study it's the first of its kind assessing, in a systematic way, people's beliefs and perceptions about healthcare rationing in the Spanish context.
MATERIALS AND METHODS
Questionnaire design
The questionnaire consisted of three questions. The first two questions addressed a hypothetical rationing exercise designed elsewhere 20. The scenarios used for said hypothetical exercise described four patients, each one distinguished by their personal features and health status (Table 1). Respondents were asked to act as decision makers and indicate: (i) among the four patients who should be treated if there were enough resources to treat them all, and (ii) what should be the order of health care assistance or treatment provision in the case there are not enough resources to provide treatment to all the patients at the same time. On the other hand, the third question was composed of 14 statements, designed elsewhere 13, related to the criteria that should be considered in order to define treatment provision priorities between patients. In this case, respondents were asked to express their level of agreement with each statement through a five-point Likert scale (1="strongly disagree" and 5="strongly agree").
Data collection and participants' description
Data were collected through a self-completion questionnaire made available in several social networks (Face-book, Twitter, LinkedIn, and Google Plus) and via email during four months. Only Spanish residents over 18 years were allowed to complete the survey. The questionnaire was designed to be complete anonymously and was preceded by a sheet in which the scope of the study was explained, and the permission to use the data obtained through it was requested.
A total of 546 respondents completed the questionnaire. A detailed description of the random sample is shown in Table 2. On average, participants were younger and had a higher schooling than the average Spanish population 26.
Data analysis
Quantitative analysis was made through descriptive statistics. Then, the 14 statements were subjected to principal components analysis with varimax rotation to identify the main rationing principles regarding patients' treatment prioritization. Factorability of the factorial analysis was confirmed through several pre-tests, including the determinant of the correlation matrix, Cronbach's alpha coefficient, a Kaiser-Meyer-Olkin (KMO) test of sampling adequacy and a Bartlett's test of sphericity. In addition, multinomial logistic regressions were used to test for differences in the priority order that was assigned to the hypothetical patients, taking into account the main rationing principles. The whole analysis was performed with SPSS (version 24) and STATA (version 14) software.
RESULTS
Selecting patients and ranking treatment priority
The pattern of responses to the first (first column) and second questions (second column) of the questionnaire is presented in Table 3. Only two respondents refused to choose what patient should be treated and to set a treatment priority order, while all of the other participants gave priority to at least one patient. In the first scenario (where only one patient could be treated), the child (Pedro) was the one chosen by most respondents (65.3%), followed by the elderly patient (Rosa) (19.9%), the young adult (Juan) (11%) and María, the former drugs user (3.8%). Likewise, from the mean and median of the priority ranking exercise in the second scenario (where all patients could be treated but not at the same time) it seems that the pattern of responses was consistent between both questions, as in this scenario, in average, Pedro was chosen as the one to receive the top or joint-top priority, followed by Rosa and Juan, and María was the one to be assigned the lowest priority.
Support of rationing principles
Table 4 presents the respondents' level of agreement with the statements regarding the criteria to be used to prioritize patients. These results revealed that Spanish respondents seem to support all the criteria that have been discussed in the available literature in this regard. By order of importance respondents seemed to support medical based (priority assignment to patients with an increased life-threatening risk -ES5), fair-chances based (all patients deserve equal health care provision priority-ES1), person-based (priority is given to children for they have not lived enough and they are more vulnerable than other population groups-ES6 and ES7) and economic based (priority should be given to more effective treatments-ES14) criteria when healthcare rationing is necessary.
Notes: SDIS = strongly disagree; DIS = disagree; NAND = neither agree nor disagree; AGR = agree; SAGR = strongly agree
To identify the main rationing criteria, all statements were analyzed altogether by using the factor analysis method. Factor analysis was performed through a principal component analysis with varimax rotation. Correlations between variables were suitable to be performed as indicated by the Kaiser-Mayer-Olkin (KMO = 0,884) test and the statistical value of the Bartlett's test (c2 = 260,315; p = 0,000). Table 5 summarizes the main results of the analysis. Three distinct factors emerged, accounting for 67.251% of the total variance of the data. All items loaded highly onto the factors and no item loaded onto more than one factor, supporting the independence of the dimension. The most important factors were patients' age (representing 32% of the variance), followed by economic criteria (18.3%) and fair chances (16.9%).
Relationship between the healthcare priority setting criteria and the choice of the patient to treat
Multinomial logistic regressions were applied to test for differences in the top priority that was assigned to hypothetical patients (first question: choosing the only patient to be treated - first column of Table 3) in relation to the stated relevance of each of the three main rationing criteria (Table 5). Regressions were run for all patients; coefficients and marginal effects are shown in Table 6. Results reveal that the age of the patient (personal-based criterion) was statistically significant in choosing Pedro and Rosa to be attended first. Age factor increased the probability of selecting Pedro to be treated first by 11 percentage points (p.p) and Rosa by 9.2 p.p. In contrast, the economic criterion was only statistically significant when the young adult (Juan) was chosen. The economic criterion increased the probability of providing Juan with treatment first by 3.9 p.p. Finally, the fair chances criterion was not statistically significant in choosing any of the patients, something to be expected in accordance with its definition.
*Extraction method: Principal Component Analysis. Rotation Method: Varimax with Kaiser normalization
DISCUSSION
An increasing number of studies have suggested that, when asked, citizens are able to make decisions involving patients' prioritization 18,19. However, given the plurality of views on rationing principles, no simple criteria can be defined to support patients' priority setting. Therefore, the purpose of this study was to explore Spanish respondents' attitudes concerning the principles on which health care microallocation decisions should be based. The results obtained here show that, despite the fact all respondents considered that all patient deserved equal healthcare provision, they showed their support for medical, economic and person-based criteria when prioritization of patients is necessary. However, regarding person-based criteria, only the age of the patient was considered as relevant.
When all the criteria used to prioritize patients were analyzed, three emerged as the most important: (i) patients' age; (ii) economic based criteria and (iii) fair chances. The age-related preferences of the respondents seem to be caused by their aversion to lifetime health inequalities alongside with their protective instinct towards the most vulnerable individuals. This would explain why Pedro and Rosa were given priority by most of the respondents. The same pattern of preferences was reported in Brazilian and Portuguese respondents 27,13. One possible explanation of this concern phenomenon in relation with the elderly in Latin cultures may be the strong concept of family that still prevails in these societies.
The support Spanish respondents gave to the economic criterion was evidenced by their concern regarding the treatments outcomes. This way, choosing Juan to be treated first may be justified by this reasoning. Besides, respondents seem to value the quality of life more than life expectancy, which is in agreement with findings reported around the world 18,19.
Overall, the findings reported here suggest that if patients cannot be treated equally, their age and economic related aspects (such as treatment outcomes) should work as a guide for healthcare priority setting. These results also allow to state that Spanish respondents are concerned about the distributive principles to be applied in healthcare rationing, which is interesting, considering that in recent years the Spanish health system has experienced sizeable cutbacks that have had a direct impact on the users enrolled in it, the benefits that are covered, and what share of the costs of the service provision is paid by the general population.
In this context, policy makers in Spain have the opportunity to start a public and open debate on this matter, in order to define systematic criteria to choose who should receive healthcare. Likewise, this public discussion is of great importance more than ever as decisions involving patients' prioritization may become unavoidable due to the increasing scarcity of resources. However, any health system reform involving ethical values should be discussed with the general population and health professionals in order to achieve their support, otherwise, a serious reform is unlikely to be developed or agreed upon.
The results of this study should be interpreted with proper caution, given the nature of the sample, as it is not representative of the Spanish population and therefore these findings cannot be generalized to the rest of the country's population. Furthermore, data on preferences were collected through an online-based questionnaire; somehow, despite the wide limitations of using online questionnaires, there has been an increasing interest in collecting preferences data online in recent years 27,28 and the few studies that assessed the impact of how social preferences are elicited have described broadly similar responses across the different modes of questionnaires administration 29. We hope these drawbacks are overcome by the contributions made by the present study.
In future studies, it would be useful to replicate this work to a representative sample of Spanish citizens. Also, it would be interesting to explore and compare the opinions of other parties involved, like politicians and health professionals. This is a matter of procedural justice, in facilitating transparency and helping to ensure that open and fair decision making processes are followed 30. There is some evidence suggesting that the opinions of medical doctors and the general public on the appropriate ethical principles for healthcare priority setting may differ 31