2 The financing of hospital care in Europe

Traditionally, most European countries paid for hospital care through payment systems such as global fixed budgets, fixed rates per admission or per diem rates (Cylus and Irwin, Reference Cylus and Irwin2010). Each of these systems leads to different result in the provision of hospital care. Establishing a fix payment per admission may incentivise hospitals to increase the number of admissions, while per diem payments may encourage lengthier hospital stays (Jegers et al., Reference Jegers, Kesteloot, De Graeve and Gilles2002; Smith et al., Reference Smith, Preker, Ligth and Richard2005). In addition, global budgets tend to induce providers to under-provide services because they may aim not to waste their fixed funds by promoting cost control (Kutzin, Reference Kutzin2001). These flat-rate or fixed budget payment systems have contributed to performance issues and declining health outcomes in some countries (Cylus and Irwin, Reference Cylus and Irwin2010). Hospital managers may adopt strategies to ‘cheat’ the reimbursement mechanism in order to achieve more funds, although some tactics are easier than others. For this reason a key element when implementing a reimbursement mechanism is to establish a control system that limits managerial manipulation. Case-based payments try to determine a price for the different hospital services/cases according to their intensity. The most frequently adopted case payment mechanism is the one based on the diagnosis-related group (DRG) classification, introduced in 1983 as the basis for reimbursing providers under the US Medicare system. In general, DRG-type hospital payment systems consist of two fundamental building blocks (Scheller-Kreinsen, Reference Scheller-Kreinsen, Geissler and Busse2009): (1) a patient classification system and (2) a payment rate-setting mechanism that defines cost weights or prices per DRG.

Nowadays, many European countries have incorporated a DRG case-based payment system to reimburse hospitals, although most of them combine it with other mechanisms such as global budgets and payment per procedures (see Table 1). Thus, the reimbursement systems used to finance hospital care in Europe vary across countries. The organisation and structure of the European health systems (Social Security or National Health System) partially explain these differences but cultural and political factors also influence the particularities of the policies adopted by each country. All these factors should be taken into account when comparing the strategies adopted at international level since each policy takes place within a specific context.

Table 1.

Main hospital payment schemes in Europe

DRG = diagnosis-related group.

Source: OECD (2010); Thomson et al. (Reference Thomson, Foubister and Mossialos2009).

Germany is a reference of the countries using the social security model. The hospital sector is made up of publicly owned hospitals and of not-for-profit and private hospitals. The infrastructures of these institutions are funded by state budgets while the operating costs are paid mostly by the insurance contracted by employees. The Statutory Health Insurance Act of 2000 resulted in the gradual introduction of a per-case DRG payment system instead of the previous historically based hospital budgets (using per diem charges as the unit for reimbursement). The main goal of the reform was cost containment, assuming that a more activity-oriented payment system would promote efficiency, quality and transparency in the hospital sector (Wilke et al., Reference Wilke, Höcherl, Scherer and Janke2001). After the implementation of the reform, it was found that the DRG-based payment mechanism was contributing to the accomplishment of these objectives (Quentin et al., Reference Quentin, Geissler, Scheller-Kreinsen and Busse2010).

The French health care system is another example of the social security system although, during the last two decades, it has gradually been supplemented with taxes (Bellanger and Mossé, Reference Bellanger and Mossé2005). The French health care system is a mix of public and private providers. Providers of outpatient care are mainly private and they are financed on a fee-for-service basis. One-third of the hospitals are public and account for two-thirds of the beds. Each public hospital is legally autonomous and manages its own budget. The payment of the operating costs of acute hospitals, that is, excluding mental and tertiary hospitals, is calculated using a DRG-based system. Capital investments are partially funded through specific programmes (Chevreul et al., Reference Chevreul, Durand-Zaleski, Bahrami, Hernández-Quevedo and Mladovsky2010). Three-quarters of the resources come from statutory health insurance funds and the remainder from individuals in the form of direct payments to the hospitals or indirectly via contributions paid for complementary health insurance coverage (Bellanger and Mosséf, Reference Bellanger and Mossé2005). Until 2004, the funding of hospitals depended exclusively on global budgets, which were mainly calculated on a historical basis including some adjustments based on regional case-mix (Bellanger and Tardif, Reference Bellanger and Tardif2006). The percentage of the payment prospectively determined has continually increased from 10% in 2004 to 100% currently (Chevreul et al., Reference Chevreul, Durand-Zaleski, Bahrami, Hernández-Quevedo and Mladovsky2010).

The British National Health Service (NHS) is the main reference for general taxation funded systems. Specialised care is provided in government-owned hospitals organised as ‘trusts’. The number of public hospitals has fallen dramatically during last decades while, at the same time, the health expenditure has increased (Boyle, Reference Boyle2011; OECD, 2011). Nowadays, the United Kingdom has the lowest rate of bed per inhabitant in Europe (Boyle, Reference Boyle2011). One of the basic features of the NHS is the division between purchasing (commissioning) and care delivery functions. The NHS consists of organisations that deliver services (service providers) and organisations that contract services, mainly Primary Care Trusts (PCTs). PCTs are responsible for contracting hospital services from a range of providers, mainly in the public sector but including private and voluntary providers. In 2003, the Department of Health (DoH) introduced its ‘payment by results’ system as the main way of financing hospitals, supplemented with global budgets. The introduction of this system was motivated by policy objectives to increase the efficiency, activity and quality of care in hospitals (DoH, 2002). Before payment by results, PCTs paid hospitals a fixed amount through locally negotiated block contracts (Epstein and Mason, Reference Epstein and Mason2006). With the establishment of a per-case reimbursement system, hospitals can retain surpluses. Some studies evidence a reduction in the costs of care after the implementation of the payment by results system due to reductions in the length of stays and an increment in the proportion of day cases treated (Farrar et al., Reference Farrar, Yi, Sutton, Clakley, Sussex and Scott2009).

In Finland, health care services are also mainly provided by the public sector and financed through general taxation. The principal particularity of the Finnish system is its high level of decentralisation. Municipalities have the responsibility for the organisation and financing most of the health care services (Vuorenkoski, Reference Vuorenkoski2008). The municipalities are grouped into hospital districts for the provision of specialised care, although each municipality can purchase services from private providers. The main source of funding for health care services is taxation, but hospitals also receive funds from state subsidies and user fees. In addition, there is a secondary public finance scheme, the National Health Insurance (NHI), based on compulsory insurance fees. The NHI partially reimburses public services as well as services that are provided by the private sector (Häkkinen, Reference Häkkinen2010). In the absence of national regulations or guidelines, each hospital district determines the payment method used to reimburse its hospitals. The hospital reimbursement mechanism has been consistently moving away from per diem towards per case methods. Nowadays, 13 of the 21 districts use DRG-based pricing, although the principles and rules applied vary greatly (Häkkinen, Reference Häkkinen2010). The productivity improvements of Finnish hospitals may be more related to the process of decentralisation of competencies to municipalities than to the effect of the introduction of a performance-based reimbursement. Most of the efficiency gains have been achieved by reducing the average length of stays after cutting the bed capacity of most hospitals even when they are still reimbursed on a bed-day basis (Linna, Reference Linna2000).

The Andalusian health care system constitutes a further example of a national health care system because the provision of specialised health care is mainly provided through public hospitals funded by general taxation and the SAS assumes the deficits of public hospitals. The reimbursement system adopted by the SAS, like most of the European systems, excludes from the funding formula the capital investments of public hospitals that are allocated through independent programmes. The SAS considers the complexity of the activity, using a DRG classification, when it calculates the budget assigned to each hospital according to the level of activity carried out. The main particularity of the reimbursement system implemented in Andalusia is that activity is not directly linked to the payment because there is not a tariff per case but per inhabitant. The activity carried out is used to determine the population assigned to each hospital for funding purposes. This means that the same medical act may result in a different payment depending on where the patient comes from. How the system works exactly is explained in Section 3.

3 The SAS and its hospital payment system

Andalusia, a region in the south of Spain has a population of more than 8 million, higher than 11 of the 27 countries in the European Union (EU). The Andalusian Health Service, Servicio Andaluz de Salud (SAS), is the public entity in charge of managing the Spanish National Health System in this region and, in terms of population attended, is one of the biggest public health services in Europe. The SAS has 29 acute hospitals. It has similar functions and competences to other European Health Services, such as the National Health Services of the United Kingdom, Italy and the Scandinavian countries, in which the provision of specialised health care is mainly provided through public hospitals funded by general taxation. Andalusia implemented a hospital choice initiative in 1997. Table 2 provides descriptive figures of the SAS hospitals for 2003 and 2006, and the variation during the three-year period. The Annex presents the information distinguishing between urban and rural hospitals.

Table 2.

Descriptive figures of SAS hospitals

SAS = Servicio Andaluz de Salud; FTE = full-time equivalent; MOO = major outpatient operation.

(1) Planned + emergency operations.

*Adjusted by case-mix.

**Population for reimbursement purposes.

Figures correspond to the 27 of the 29 SAS hospitals included in the analyses.

VAR 06 − 03 is the result of: (Mean2006 − Mean2003)/Mean2003.

In 2003, the SAS concluded the implementation of a new payment mechanism, which was described as a way to promote hospital efficiency, to ensure an optimum provision of services and to avoid wasting resources (SAS, 2007). The SAS asserts that the mechanism adopted is based on the capitation strategy (SAS, 2007) because it allocates the overall budget to hospitals on the basis of the population. However, as will be described below, the mechanism has some characteristics of an activity-based funding model since, for funding purposes, the population is assigned to each hospital according to the activity carried out in the previous year. The years 2000 to 2002 were a transition period in which the capitation strategy was gradually implemented. Prior to the beginning of the financing reform, the SAS hospitals were funded through budgets based on historic costs plus some specific funds according to the particularities of each hospital. The system was changed because of its failure to link activity and funding and because of the lack of an efficiency benchmark. In order to ensure the viability of the public system of provision of health care service, the new reimbursement system was expected to promote efficiency improvements by SAS (2006):

• • Introducing incentives for internal competition, understood as the best way in which each hospital solves the health problems of citizens.

• • Ensuring financial sufficiency for the functioning of hospitals and the efficient production of their services.

• • Controlling costs, to avoid rising beyond the financial capacity of the SAS.

Internal competition is promoted by the way in which the global budget of the SAS for hospital care is allocated to hospitals while efficiency and cost control are promoted by the way in which the global budget of the SAS for hospital care is set. In year n, during the elaboration of the budget for year n + 1, the SAS knows the operative costs of hospitals for year n − 1 and an estimation of the evolution of costs for year n. The SAS uses this information to elaborate the total budget for SAS hospital operative costs for year n + 1. In the elaboration of the budget of year n + 1, the SAS considers both the adjustment of non-authorised costs and avoidable costs for year n and the trend and evolution of costs for year n + 1. In this way, the SAS tries to elaborate a ‘plausible’ budget because it takes future trends into account, but also an ‘efficient’ budget because it eliminates costs that are considered unnecessary. It is worth highlighting that the objective of the SAS is not to reduce expenditure, but to improve the efficiency and the quality of the services. The gain in efficiency must be directed to provide services for those needs not yet covered (SAS, 2007). Data about hospital and global SAS budget for the period analysed are shown in Table 3.

Table 3.

Public expenditure in the SAS (2002–2006)

SAS = Servicio Andaluz de Salud.

Source: SAS annual reports for 2002 to 2006.

Three factors are included in the formula applied to determine the allocation of payment to hospitals: population, tariff and case-mix:

$$Hospital\,funding\,{\rm{ = }}\,population\,\times \,tariff\,\times \,case - mix$$

The first factor, population, is assigned to each hospital on the basis of the number of patients attended to in the previous year. Each hospital receives an amount of money per person that is estimated that would go for care to that hospital.

In a capitation system, the population is the key factor to determine the final amount of money received. The larger the population in the ‘list’, the greater the reimbursement. Hospitals that are able to increase their population assigned for funding purposes will increase the percentage of the SAS budget they receive to the detriment of those losing population. This way, some kind of competition is generated. The SAS reimbursement mechanism assigns the population to each hospital for year n according to the patients attended in each of the five functional lines during year n − 1. This mechanism takes into account where the patients reside in order to allocate the population of each of the eight Andalusian provinces to hospitals. Therefore, the same medical act may represent a different allocation of population, that is, different payment, depending on the health status of the population in the province of residence for that functional line. The total number of cases treated from each province, for the five functional lines in which the SAS classifies cases, is divided by the population of that province giving a ‘weight’ to each case. Then, the number of cases from that province in each hospital is multiplied by that weight. This is calculated for each hospital, considering the eight provinces and the five functional lines. The five groups (functional lines) are: (1) Hospitalisation, (2) Outpatient surgery, (3) Outpatient visits, (4) Emergencies and (5) Other cases. Therefore, the total population of Andalusia is allocated between hospitals for each of the five functional lines, which are weighted by the percentage of costs that each functional line represents in analytical accounting. The weightings for each functional line are:

The second factor of the formula, tariff, is calculated on the basis of the operative costs of all the hospitals. Some specific costs, such as training and education and capital investments are not included in the calculus of the tariff and are additionally reimbursed. Once the budget for SAS hospitals is set, it is divided by the population of Andalusia that is attended by the SAS, thus obtaining the tariff per inhabitant.

The third factor of the formula, case-mix, is an adjustment for including in the reimbursement mechanism the differences in the services provided by hospitals because they do not all provide the same range of services. Some complex cases and thus more costly cases can only be treated in big hospitals. The complexity of the cases is measured using the DRG classification (SAS, 2007). The SAS uses the All Patient DRG classification. Each patient discharged is classified into a DRG with its corresponding weight and this is converted into DRG points. The case-mix index of a hospital is the result of dividing the total number of DRG points by the number of patients discharged. The more complex the cases treated, the higher the case-mix index. From the classification of patients into functional lines and DRG points, the SAS calculates a case-mix index for each hospital. Because some hospitals have a case-mix index higher than ‘1’ and others lower than ‘1’, the reimbursement formula takes into account the complexity of the cases treated. The average case-mix for all SAS hospitals is ‘1’, so the overall hospital budget previously calculated does not change for the inclusion of this element in the formula.

One main flaw of the previous reimbursement system was the lack of a clear link between the reimbursement system and the measurement of hospital performance, that is, efficiency. In the new system, the income of each hospital, the initial budget, is considered a benchmark for cost efficiency. Hospitals that stay within the budget are considered efficient while those that exceed the budget are initially considered inefficient. This inefficiency is subsequently analysed to identify the causes of the deviations between the real and expected budget since they can arise from an excess in the consumption of resources for the activity predicted or from a higher level of activity than predicted. Rather than rewarding or punishing hospitals, the idea is to have a clearer view of hospital performance. The SAS, as owner of all the public hospitals of the system, assumes any deficit and inefficiency incurred by the hospitals. The final financial responsibility of the SAS reduces the incentives of the managers to stay within the budget, but negative deviation affects their reputation and part of their remuneration. Moreover, since managers are selected by the SAS, they can be removed from their posts if goals are not achieved.

Quality improvements are closely linked to the reimbursement mechanism. Because Andalusian patients have the possibility of choosing the public hospital in which to be treated, SAS hospitals must provide a better service in order to maintain their reputation and not to lose patients. The opinion of the patients about the service received is a critical issue for hospital managers as well as for the SAS itself. Annual surveys are carried out in Andalusia to find out patients’ opinions about hospital care. The results of hospitals in patient satisfaction are compared against the mean of all SAS hospitals as part of their performance evaluation. Hospitals have to elaborate plans in the aspects where satisfaction scores do not reach a minimum level. The results of the annual patient survey are disclosed through the SAS website with an easy format in the hope that will help patients in their decision-making. The website also discloses the number of urgent readmissions within 31 days and the number of admissions after a major outpatient operation (MOO) as indicators of the clinical quality of hospitals. It is expected that patients will most likely choose best performing hospitals, resulting in more funding for those hospitals in future periods. In addition, the SAS does not compute readmissionsFootnote ¹ as new cases and hence they do not result in more funds to hospitals. Initially, General Practitioners assist patients in choosing the hospital in which they want to be treated. Additionally, patients can request more information and change hospitals by using a phone number that the SAS has made available for that purpose. The SAS assumes all the necessary bureaucratic procedures to formalise the change. During the period 2003–2006, the variation of the population assigned to each hospital for reimbursement purposes changed, on average, by more than 10%, which reflects a relatively low flow of patient between hospitals (see Table 2).

4 Methodology

4.1 Analysis of efficiency improvements: the Data Envelopment Analysis (DEA) methodology and the Malmquist index

In order to evaluate the evolution of the efficiency for the period considered, we apply the DEA methodology to measure the efficiency of the hospitals for the years 2003 and 2006, separately, and then we calculate the Malmquist index. The DEA (Charnes et al., Reference Charnes, Cooper and Rhodes1978; Cooper et al., Reference Cooper, Seiford and Tone2000) is an extended methodology applied in prior studies on hospital performance that provides a comparative efficiency indicator of the units to evaluate. Hollingsworth et al. (Reference Hollingsworth, Dawson and Maniadakis1999), Hollingsworth (Reference Hollingsworth2003), Hollingsworth (Reference Hollingsworth2008) and O'Neill et al. (Reference O'Neill, Rauner, Heidenberger and Kraus2008) provide recent revisions of the literature using DEA in hospitals. The method is especially appropriate for evaluating the efficiency of non-profit entities like public hospitals, which operate outside the market since, for them, measures of efficiency such as income and profitability do not work satisfactorily.

DEA provides a comparative efficiency indicator of the units to evaluate (in this case hospitals), which are called decision-making units (DMUs). The relative efficiency of a DMU is defined as the ratio of the total weighted output to the total weighted input. DEA does not require any external production function to be taken as a reference and the actual inputs and outputs observed are used to estimate a benchmark production frontier. The solution to the DEA problem calculates the weightings for the inputs and outputs of the hospital being evaluated that maximise its productivity in comparison to the other hospitals of the sample. If a hospital achieves a score of 100%, it means that it is impossible to find other DMUs or a composite DMU outperforming that hospital. If the score is less than 100%, the hospital will not form part of the production frontier because more efficient hospitals have been found in the sample. From a management perspective, DEA is especially useful to know which hospitals are operating significantly below the production frontier (score <100%), because this signals that improvements are possible. The units with a 100% score are considered efficient, but they are not necessarily the optima, because the results cannot be extrapolated outside the sample. From the different specifications of the model, we apply the input-oriented model and the constant returns to scale (CRS) approach, which is consistent with the DEA literature on health care (see O'Neill et al., Reference O'Neill, Rauner, Heidenberger and Kraus2008). Moreover, CRS is especially appropriate for working with reduced-sized samples (Prior, Reference Prior2006; Vitikainen et al., Reference Vitikainen, Street and Linna2009).

The selection of the variables to be included in the model is critical in the development of DEA. In general, the sum of inputs and outputs considered should not exceed one-third of the number of DMUs. The inclusion of a large number of variables reduces the discriminatory power of the model because it increases the number of efficient DMUs (Cooper et al., Reference Cooper, Seiford and Tone2000). This is because the number of dimensions in which a particular DMU can be unique (and, thus, with no partners to be compared with) increases. Johnes and Johnes (Reference Johnes and Johnes1993) argue that a DMU can sometimes achieve 100% efficiency simply because no one else is competing in that niche of activity. We work with variables that have been widely used in previous studies, according to the revision of O'Neill et al. (Reference O'Neill, Rauner, Heidenberger and Kraus2008), as representative of the principal input and output indicator categories in hospitals. The inputs included are the number of beds as a proxy of the size and capacity of the hospitals and full-time equivalent (FTE) physicians and FTE nursing staff as a proxy of the human resources. O'Neill et al. (Reference O'Neill, Rauner, Heidenberger and Kraus2008) reported that the majority of studies use the number of beds as a proxy for the infrastructure of hospitals. Many studies have introduced some indicator of the number of clinical staff as a proxy for labour costs (e.g. Pina and Torres, Reference Pina and Torres1996; Prior, Reference Prior2006; Nayar and Ozcan, Reference Nayar and Ozcan2008; García-Lacalle and Martín, Reference García-Lacalle and Martin2010).

As outputs, we have used indicators that reflect both the inpatient activity – operations (planned + emergency operations) and diagnoses – and the outpatient activity carried out in hospitals – outpatient visits, emergencies and MOOs. Operations and diagnoses are adjusted by case-mix in order to reflect the complexity of these activities. We do not consider any variable representative of inpatient-stays as, with the implementation of PPS, hospitals have incentives to reduce their duration. Indeed, the objective of any health care system should be the reduction of the length of stay and of the services provided to hospitalised patients. Furthermore, bed reductions and the promotion of outpatient activity have the objective of reducing the number of stays. By reducing stays, hospital costs are also reduced. This represents a significant shift from the ‘inpatient day’ to the ‘case’ as the primary means of hospital reimbursement (O'Neill et al., Reference O'Neill, Rauner, Heidenberger and Kraus2008). Anyway, we acknowledge that ‘inpatient days’ may be reckoned as an important hospital output. However, when using the DEA methodology, this indicator may be biased by inefficiency. Hospitals might have patients hospitalised longer than needed, thus having more days of stay, resulting in better DEA results. O'Neill et al. (Reference O'Neill, Rauner, Heidenberger and Kraus2008)'s review of DEA analysis in health care reports that the use of ‘inpatient days’ as an output category has been steadily decreasing since 1985. The main descriptive figures of the inputs and outputs included in our analyses and the two above mentioned quality indicators are included in Table 2.

The Malmquist index allows us to carry out a dynamic analysis of efficiency growth during a period, using the solutions of the DEAs. As we have worked with an input-oriented model, indexes lower than ‘1’ indicate an improvement during the period; greater than ‘1’ indicate a worsening, and equal to ‘1’ indicate no efficiency variations during the period. The precise formulation of the Malmquist index is as follows:

$${{M}_I}({{x}^t}, {{y}^t}, {{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )\, = \,{{\left[ {\frac{{D_{I}^{t} ({{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )}}{{D_{I}^{t} ({{x}^t}, {{y}^t} )}}\frac{{D_{I}^{{t\, + \,{\rm{1}}}} ({{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )}}{{D_{I}^{{t\, + \,{\rm{1}}}} ({{x}^t}, {{y}^t} )}}} \right]}^{{\rm{1/2}}}} $$

where x and y stand for input and output vectors of the sample unit, respectively, whose efficiency we attempt to measure and t stands for the time periods under analysis. From an input-oriented perspective, an efficiency growth is achieved if M < 1 and an efficiency regression if M > 1.

The variation in productivity over the time may be a consequence of a different relative position of the hospital with respect to the production frontier because (i) the hospital is operating closer (further) to the maximum level that it is expected or (ii) the capacity for producing has been expanded (contracted) for all the systems. Färe et al. (Reference Färe, Grosskopf, Norris and Zhang1994) factorised the original expression of the Malmquist index (1) identifying the two sources of efficiency change: technical change and technological change

$${{M}^{}} _{I} ({{x}^t}, {{y}^t}, {{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )\, = \,\frac{{{{D}^{}} _{I}^{{t\, + \,{\rm{1}}}} ({{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )}}{{D_{I}^{t} ({{x}^t}, {{y}^t} )}}{{\left[ {\frac{{{{D}^{}} _{I}^{t} ({{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )}}{{{{D}^{}} _{I}^{{t\, + \,{\rm{1}}}} ({{x}^{t\, + \,{\rm{1}}}}, {{y}^{t\, + \,{\rm{1}}}} )}}\frac{{{{D}^{}} _{I}^{t} ({{x}^t}, {{y}^t} )}}{{{{D}^{}} _{I}^{{t\, + \,{\rm{1}}}} ({{x}^t}, {{y}^t} )}}} \right]}^{{\rm{1/2}}}} $$

The first factor (outside the square brackets) represents the technical change, that is, the evolution of the distance of a hospital from the production frontier. The factor that determines the technical change of a hospital is the evolution of its performance in relation to the other hospitals, that is, it depends on the ability of each hospital to reduce inputs and/or increase outputs. The second factor (between the square brackets) measures changes in technology over two time periods, it is to say whether the frontier is shifting out over time. The technological change measures the movement in the production frontier defined by the best practices of the hospitals of the sample in each period, and is determined by the performance of all hospitals of the sample. The principal factors that explain the displacements of the production frontier are technological progress or organisational changes affecting all the hospitals in the system. Figure 1 illustrates a simple case in which a single input ‘x’ produces a single output ‘y’. One hospital is observed at time ‘t’, showing the combination of inputs and outputs (x ^t, y ^t) and, again, at time ‘t + 1’, showing the combination (x ^{t + 1}, y ^{t + 1}). S ^t and S ^{t + 1} represent the production frontiers for each year. This hospital has improved its productivity in the period observed since, in ‘t + 1’, it produces more outputs using less inputs. This improvement has been due to the expansion of the frontier in the period ‘t + 1’, representing the technological change. Nevertheless, the technical change has been negative, since in the year ‘t’ the hospital was on the frontier, whereas in ‘t + 1’ it is operating below its capacity.

Figure 1

The input-based productivity change.

Reforms in hospital reimbursement methods may lead to both technological and technical changes. A more effective allocation of resources would contribute to extending the technological frontier of the system by enabling hospitals to achieve higher levels of production. The method may also encourage hospitals to reduce cost inefficiencies to operate closer to the frontier, resulting in an improvement of technical efficiency (Mckay and Deily, Reference Mckay and Deily2008).

5 Analysis of the results

The results of the DEA (Table 4) show high scores for most of the SAS hospitals, which indicate that they are operating close to the efficiency frontier in both years. Each of the two DEA, one for 2003 and one for 2006, includes all the hospitals in our sample, that is, rural and urban hospitals, which allows to compare their performance. However, as stated in Section 4.1, a score of 100% does not mean that a hospital is operating efficiently, but that no other hospital in the sample shows a better performance considering the combination of inputs and outputs. The lack of an objective to be compared with only allows us to assert that the system is performing homogenously, but not that their performance is optimum. This is probably because the SAS is still, in part, a centralised system, which influences the allocation of resources to hospitals. Nevertheless, the results are consistent with prior studies carried out using DEA methodology. The literature reviews carried out by O'Neill et al. (Reference O'Neill, Rauner, Heidenberger and Kraus2008) and Hollingsworth (Reference Hollingsworth2003) show an average efficiency of 90% for European hospitals and 86% for the United States. In 2003, rural hospitals perform, on average, more efficiently and more homogeneously than urban hospitals, but, in 2006, the gap between rural and urban hospitals disappears and the latter even show greater efficiency. In addition, in 2006, urban hospitals perform more homogeneously than their rural counterparts. The reduction of the standard deviation of DEA scores from 2003 to 2006 points to a homogenisation in the performance of hospitals.

Table 4.

DEA scores (%) for 2003 and 2006

DEA = data envelopment analysis.

Malmquist results (Table 5) show a positive evolution in the efficiency of the hospitals during the period 2003–2006. In particular, improvements tend to be better in the hospitals that, in the 2003, achieved the lowest DEA scores. Although the improvements are generalised, they are greater in the case of urban hospitals. The decomposition of the Malmquist index into technical and technological changes shows that SAS hospitals have improved in both aspects, particularly in the latter. The technical factor is related to the individual behaviour of the hospitals and the technology factor determines movements of the production frontier. Both rural and urban hospitals have improved in terms of technological change, but only urban hospitals have improved in terms of technical change, that is, improvements due to internal factors. Therefore, the overall improvement of urban hospitals is better than the evolution of rural hospitals, which are harmed by the negative evolution of their technical change.

Table 5.

Malmquist indexes and variations in quality

¹The Δ of the quality indicators is calculated as the score in 2006 (%) minus the score in 2003 (%).

²Significance (bilateral) in brackets; **Significant at 0.01; *Significant at 0.05.

The U-test shows significant differences between rural and urban hospitals in the Malmquist (overall) index and in the technical change. Urban hospitals have had significantly better overall efficiency improvements than rural hospitals during the period. The variations in technology have affected both groups of hospitals similarly, but urban hospitals have improved their efficiency more significantly than rural hospitals in technical aspects. Therefore, the better overall improvement of the urban hospitals is due to internal factors of the hospitals rather than external factors derived from the different impact of technological progress.

As for quality, Table 2 shows that there has been a worsening in the clinical quality indicators of SAS hospitals from 2003 to 2006, although the variation is very small. Urban hospitals have had a negative evolution of their clinical quality in terms of the inpatient activity, as all urban hospitals but one present a decrease of the non-readmission rate (Table 5). On average, urban hospitals improve the clinical quality of outpatient services, although only half of them improve in this aspect. Rural hospitals have had a negative evolution for the two clinical quality indicators, although their behaviour is less homogeneous than urban hospitals. In particular, half of the hospitals improve their admission after MOO rate and the other half have a negative evolution of this ratio, as it happens with urban hospitals, but the mean of the group is especially influenced by the Ru9 hospital. The U-test does not show significant differences between the two groups of hospitals in clinical quality. Both groups of hospitals improve their perceived quality, although the improvement is, on average, greater in urban hospitals. The U-test does not show significant differences between the two groups.

The correlation analyses (Table 6) show no significant relationships between the evolution of efficiency and clinical quality. As for the value of the coefficients, all of them are lower than 0.2 with a significance far from the 0.1, which is the value that usually marks the limit for the statistical significance of the correlation. Technical change and the evolution of non-readmissions are the variables with the closest relationship because they have a correlation coefficient greater than 0.2. Its positive sign would suggest that improvements in hospital efficiency are negatively related to inpatient clinical quality. The perceived quality of SAS hospitals remains steady during the period (Table 2) and the correlation analyses show no significant relationship between the evolution of efficiency and perceived quality.

Table 6.

Spearman correlation results between the evolution of efficiency and quality (clinical and perceived)

Significance in brackets.

Because of the interpretation of the Malmquist input-oriented index, a positive correlation represents a negative relationship between productivity and quality evolution and vice versa.