Hospital efficiency analysis in England and Germany. What lessons can be learned from each other?

Utilising Data Envelopment Analysis


Master's Thesis, 2015

85 Pages, Grade: 80


Excerpt


Content

List of Figures and Tables

List of Appendices

1 Introduction

2 Literature Review

3 Research Methodology

4 Data Envelopment Analysis
4.1 Definition and Classification
4.2 Mathematical Fundamentals
4.3 Performance measurement in healthcare

5 Data collection and explanation
5.1 Hospital funding bodies
5.2 Identification of the measurement categories
5.2.1 Input Data
5.2.2 Output Data

6 Data analysis
6.1 Overall Efficiency Distribution
6.1.1 Setup
6.1.2 Findings
6.1.3 Efficiency Plot
6.1.4 Input and Output Relations
6.1.5 Efficiency Patterns
6.2 Efficiency Frontier 'Convex Cone'
6.2.1 Setup
6.2.2 Findings
6.3 Best Performance Hospitals
6.3.1 Reference Set
6.3.2 Potential Improvements

7 Conclusion

8 References

9 Appendix

List of Figures and Tables

Figure 4.1 DEA model classification

Figure 4.2 DEA technologies

Figure 4.3 Components of performance

Figure 4.4 Systems view of a DMU

Figure 5.1 Overview of input and output data

Figure 6.1 Overall hospital efficiency distribution

Figure 6.2 Deaths - Hospital size

Figure 6.3 Deaths - Admissions

Figure 6.4 Deaths - Medical specialists

Figure 6.5 Deaths - Number of wards

Figure 6.6 Deaths - Population density

Figure 6.7 Efficiency frontier: Size

Figure 6.8 Efficiency frontier: Medical specialists

Figure 6.9 Efficiency frontier: Number of wards

Figure 6.10 Reference set frequencies

Figure 6.11 Potential improvements: Admissions

Figure 6.11 Potential improvements: Size

Figure 6.11 Potential improvements: Medical specialists

Table 4.1 Classification of frontier methods

Table 5.1 Average input values

Table 5.2 German mortality rates per hospital

Table 6.1 Classification of input and output factors

Table 6.2 Efficiency plot scores

Table 6.3 Input correlation scores

Table 6.4 Overview: Area and population of England and Germany

Table 6.5 Efficiency patterns of German hospitals

Table 6.6 Efficiency patterns of English hospitals

Table 6.7 Hospital size

Table 6.8 Hospital comparison

Table 6.9 Best performance hospitals

Table 6.10 Characteristics of the efficient hospitals

Table 6.11 Hospitals with the highest number of reference set frequencies

List of Appendices

Appendix 1: Matching hospital wards

Appendix 2: Demographic Environment: Population Density

Appendix 3: Adjusted Mortality rates per sample hospitals

Appendix 4: Overall efficiency distribution scores

Appendix 5: Proposal

Appendix 6: Reflection of the research process

Acknowledgements

First of all I want to thank my supervisor Dr. Eren Demir for his outstanding support and his communicative way of providing guidance during the whole research project. I am very happy that I had the chance to learn from his expertise in the field of healthcare analytics. Moreover, besides his commitment to my research project I want to thank Eren for his advice on my future career development. I highly appreciate it and I hope that I can continue my studies in the field of healthcare analytics in the near future

The second acknowledgement goes to my family, especially my parents who enabled me to study abroad in the first place. I feel privileged for their support and the trust they put into me. Moreover, I want to thank my big brother Jonas for his time and thoughts he spend on checking most of my written coursework and his support and advice he offers me beyond that

Further, I want to thank my fellow student and friend Beway with whom I shared nearly all modules of the MSc. International Business programme last year and who was the first person I met when I arrived on campus. His ambitious way of studying impressed and more importantly infected me

Last but not least I want to thank my fiancée Christine for her encouragement in every situation of my life and her never ending love

Executive Summary

This report analyses the efficiency of hospitals in England and Germany. Data Envelopment Analysis (DEA) is utilised to estimate a best practice frontier and to evaluate the performance characteristics of different hospitals in the two countries. The measurement categories of input factors are represented by the hospital size (amount of beds), number of wards, number of employed medical specialists, sum of total inpatient admissions, demographic environment (population density in population/km²), and one output factor (mortality rates)

Regarding these factors, the main findings reveal that the allocation of inefficient hospitals in England is broadly spread and, therefore, the mean efficiency value of hospitals in England is ≈44% (whereas 100% describes efficient units). In contrast, the mean efficiency value of German hospitals equates to ≈91%. Consequently, the allocation of inefficient German hospitals is less spread and the overall performance of hospital health supply is more efficient in Germany compared to England

Most remarkably, the amount of inpatient admissions is one of the main drivers for efficiency and especially the English hospitals suffer from high numbers of inpatient admissions, which may be reduced due to improvements in the primary healthcare supply provided by general practitioners

1 Introduction

Contemporary healthcare providers in different healthcare markets experience rapid growths in healthcare costs due to increasing complexity and competitiveness (Barnum et al., 2011). Maniadakis et al. (2009) add that especially western countries recorded a substantial trend of increasing healthcare costs during the last four decades and that this trend is expected to continue in the future. Consequently, today's healthcare managers have to face the challenging task of providing high quality care with more and more limited resources (Ozcan, 2009). This is where the efficiency analysis of healthcare providers (e.g. hospitals) comes into play, with the aim to identify best performance hospitals. Moreover, best practices of how these best performance hospitals apply input factors to produce certain measurable output can be derived. Further, the average and low performance hospitals can adapt the identified best practices and learn how to improve their own efficiency of healthcare delivery. Although this research does not directly cover healthcare cost aspects within the efficiency analysis, it can be assumed that there is a correlation between the amount of inputs used and the associated costs. Thus, the more inputs being used to produce a certain output, the more costs can be associated. Barros (2003) emphasises the importance of healthcare efficiency of hospitals as these units represent one important part of the healthcare system. Therefore, improvements in more efficient hospital healthcare delivery may positively affect the overall healthcare system of a nation. Further, Barros (2003) explains that changes in the hospital funding body might increase the efficiency in healthcare delivery, as the motivation of private run hospitals is higher to assess hospital performance in order to gain maximum benefit. Therefore, healthcare managers have to use decision tools (i.e. scientific methods) like Data Envelopment Analysis (DEA) to assess hospital performance (Ozcan, 2009). The benefit of DEA compared to "Other methods is that it identifies the optimal ways of performance rather than the averages" (Ozcan, 2008:16). However, Ozcan (2009) adds that the results of quantitative decision making tools do not alone determine a final decision, as other qualitative factors have to be considered. Consequently, the findings of a hospital efficiency analysis provide a brief overview of the individual performance level, but general recommendations on how to improve the efficiency score have to be reassessed and linked to other qualitative findings.

This report will analyse hospital efficiency in England and Germany and indicate how the particular underlying healthcare system affects the individual healthcare supply. According to the specifics of a healthcare system, it is obvious that the amount and occurrence of stakeholders varies from one system to another and, therefore, the interaction between these players is determined by different regulations. In this context stakeholders can be defined as healthcare suppliers (e.g. practitioners and hospitals), field administration units and the state with its health care policy. One major difference between the healthcare system in England and Germany is healthcare funding. In the United Kingdom (UK) the National Health Service (NHS) provides tax based health supply, which means that every citizen can obtain health supply based on need and not on the ability to pay (NHS England, 2013a). In contrast, the German health supply is mainly based on a membership in the Statutory Health Insurance (SHI), where every member has to pay a monthly fee to get access to health providers (Obermann et al., 2013). Another major difference lies in the spending on healthcare, where Germany spent 34% more on healthcare per capita in 2012 compared to the UK (The World Bank Group, 2014; OECD, 2014b). Moreover, the doctors density per 1,000 population in 2012 in Germany was 4.0 and in England 2.8 (OECD, 2014c). In addition, the NHS Confederation (2014) states that Germany had 8.3 hospital beds per 1,000 people in 2012, compared to 2.8 in the UK. Maybe this is one reason why NHS hospitals fail to meet targets on treatment waiting times (The Guardian, 2014).

Even though the given facts indicate that Germany is likely to provide a higher quality in healthcare "It is unlikely, comparing two health care systems, that one will be superior to another in all components of the average" (McGuire & Bauhoff, 2011:12, in Klusen et al., 2011). According to this, the Commonwealth Fund declared the NHS healthcare system as the most impressive overall, compared with the healthcare systems of ten other countries: Australia, Canada, France, Germany, the Netherlands, New Zealand, Norway, Sweden, Switzerland and the USA (NHS Confederation, 2014). However, when it comes to an international hospital performance ranking in total 11 German hospitals appear before the first English hospital on the ranking list (Hospitals, 2012). Accordingly, the results of the efficiency analysis of this report can be used to either support or contradict this international hospital ranking. In order to set up a comparable benchmarking of hospitals the individual units have to transform the same type of input factors to the same type of outputs (Bogetoft & Otto, 2011).

Therefore, individual input and output factors (e.g. hospital size, amount of employed medical specialists, number of wards, population density, total sum of admissions and the mortality rates) have been defined to conduct this analysis. Most notably the availability of sufficient data sets from English and German hospitals determined the selection of appropriate input and output factors. Bogetoft and Otto (2011) indicate that modern benchmarking analyses use best practice frontiers to highlight best performance units and explain variations of average and low performance units from this frontier. Moreover, the term benchmarking can be defined as "The systematic comparison of the performance of one firm against other firms" (Bogetoft & Otto, 2011:1). Further, Bogetoft and Otto (2011) declare that in the benchmarking literature the use of multiple inputs and outputs is inherent. Therefore, the DEA is very suitable to conduct a modern benchmarking as it allows the selection of multiple input and output factors (Ozcan, 2008). Besides the most obvious purpose why healthcare providers pursue benchmarking activities, by determining how they are doing relative to their competitors, Ozcan (2008) reminds that benchmarking can also be used to detect changes from one period to another and to investigate deviations from plan. This concludes, the need as in today's "Health markets no hospital can afford to be an average performer" (Ozcan, 2008:16).

2 Literature Review

The need for efficient health supply in any healthcare system becomes more and more important. Henke and Schreyögg (2004) emphasise that the main reason for the growth in health expenditures is based on changes in the demographic characteristics, as higher life expectancy and lower birth rates result in an aging population. In addition, Debatin (2011) argues that the healthcare market is very special and its commodities unique. Therefore, measuring efficiency and analysing performance of healthcare providers is complex because of its nature. Consequently, several efficiency analyses of hospital health supply were developed from different views to describe hospital performance patterns. Hussey et al. (2008) reviewed existing healthcare efficiency measures in order to find a common ground about the sufficiency of these methods. Thus, available input factors can be distinguished between physical and financial inputs, whereas the use of physical inputs helps to analyse whether the output could be produced faster with less resources (e.g. working hours, supplies). Further. Hussey et al. (2008) state that common efficiency analyses used hospital discharges, procedures, and physician visits as output factors. Kerr et al. (1999) add that managers and policymakers pursue the improvement of efficiency due to an adjustment in the use of resources and, likewise, assure clinical quality. However, against the frequent assumption that cost cutting will negatively affect the quality outcome, Valdmanis et al. (2008) indicate that both targets can be met by improving the efficiency. The main findings of their research conclude that high quality hospitals show a higher overall efficiency compared to the other hospitals of their study. Moreover, another study by Nayar and Ozcan (2008) confirm that managerial intentions of improving hospital efficiency is not likely to affect quality. In addition, Clement et al. (2008) found out that efficiency and quality go together. Another study by Jha et al. (2007) conclude that an inverse relationship between mortality rates and hospital performance exists. This means, the higher the individual hospital mortality rate the lower the overall hospital performance.

Concerning the methods used to conduct efficiency analyses, Hollingsworth (2008) states that the number of publications that apply Data Envelopment Analysis (DEA) is rapidly growing. Additionally, the most common methods to carry out hospital efficiency analyses are DEA and Stochastic Frontier Analysis (Hussey et al., 2008).

However, while science management scientists are convinced that DEA is a superior method, econometricians argue the opposite (Maniadakis et al., 2009). Sherman (1984) was the first scientist using DEA to analyse overall hospital efficiency (Ozcan, 2008). Since then Liu et al. (2013) noticed an enormous growth of DEA research and Cook et al. (2014) predict that the literature will grow to at least double its current size. However, Maniadakis et al. (2009) emphasise that the choice of the appropriate efficiency concept is important for the quality and explanatory power of an analysis. In contrast, Normand and Shahian (2007) argue that the data quality is more important in this context. Consequently, the quality of any analysis depends on the choice of the appropriate concept and the use of corresponding data.

According to the different hospital funding bodies, Ozcan et al. (1992) found out that state run hospitals show high percentages of the overall efficient hospitals. In the contrary, Valdmanis et al. (2008) conclude that for profit hospitals (i.e. private run hospitals) represent the best performance hospitals. Either the hospital healthcare market experienced a clear shift in the way these two hospital funding bodies operate or the used variables and their measurements used in these two studies differ, as defining appropriate variables is challenging due to the nature of healthcare supply (Ozcan, 2008). Even though the different variations of hospitals (e.g. size and specialty) make it hard to compare and identify individual performance levels (Ozcan, 2008), the process of benchmarking is necessary to identify best performance units and to document best practices (Bogetoft, 2012).

3 Research Methodology

The research methodology of this report is characterised by an inductive research approach, since the nature of the research question can be described as explorative.

Further, the methodological choice of this research was based on a mono quantitative research methodology to conduct a benchmarking of in total 140 NHS run hospitals in England and, comparably, 100 German hospitals. Therefore, five input factors, which are represented by the hospital size (amount of beds), number of wards, number of employed medical specialists, sum of total inpatient admissions, demographic environment (population density in population/km²), and one output factor (mortality rates) were identified in order to set up a benchmarking of best performance hospitals in England and Germany. The specifics of every input and output factor are highlighted in chapter 5.2 Identification of the measurement categories. The data collection process was based on an archival research strategy, whereas the English data sets were provided by the NHS and the Germany data sets by the Federal Statistical Office of the nation. In addition, the time horizon of this research can be described as cross- sectional (Saunders et al., 2012), as the individual data sets reflect a snapshot of the year 2012. Concluding, the data analysis was carried out by utilising the DEA technique, within the Frontier Analyst programme, in order to identify best performance hospitals and to explore efficiency patterns of English and German hospitals. Besides the overall efficiency scale of hospitals in England and Germany, different efficiency plots where used to highlight particular correlations between the input/output factors. Moreover, the DEA technique was utilised to build a convex cone of efficient units. Remarkably, the convex cone required input minimisation as optimisation mode and the selection of the constant returns scaling mode. Further, only one input factor and two output factors had to be selected to pursue this analysis. Therefore, the hospital size was selected as input factor and the mortality rates and the sum of total inpatient admissions were chosen as output factors in order to conduct the calculation of a convex cone of efficient units. Consequently, the individual efficiency patterns of best performance hospitals in England and Germany were recorded and the identification of the characteristics of least efficient hospitals eased the process of formulating recommendations. In the end, the most significant findings were interpreted and linked to other research findings in the field of hospital performance measurement.

4 Data Envelopment Analysis

The theoretical approach of Data Envelopment Analysis (DEA) will be introduced within this chapter. In detail this chapter covers a brief definition and classification of DEA in benchmarking literature. In addition, the mathematical fundamentals of DEA will be discovered as well as the identification of its special characteristics this benchmarking tool holds within a healthcare context.

4.1 Definition and Classification

The DEA represents a performance evaluation method among others like Ratio Analysis (RA), Least-Squares Regression (LSR), Total Factor Productivity (TFP) and Stochastic Frontier Analysis (SFA), which can be used to build comparative performance analysis (Ozcan, 2008). According to Bogetoft and Otto (2011) 'state-of- the-art' benchmarking methods are a combination of two research traditions:

1. Management science, mathematical programming and operations research,
2. Economics- and Econometrics-oriented.

In which the DEA models refer to the first approach (1.) and the SFA models apply to the second approach (2.). However, the integration of these two research traditions is still delimited from a methodological view, but more and more scientists and consultants use both types in practice (Bogetoft & Otto, 2011). Ozcan (2008) adds, that scientists are currently developing stochastic and other variants of DEA. The following definition of DEA gives a solid description of DEA's characteristics.

"DEA provides a mathematical programming method of estimating best practice production frontiers and evaluating the relative efficiency of different entities" (Bogetoft & Otto, 2011:81).

In contrast to the SFA approach, where all firms are supposed to be not efficient, the DEA approach assumes that not all firms are efficient (Ozcan, 2008). Therefore, the best performing units will set a benchmark in terms of an efficiency frontier to enable the evaluation of substandard units.

According to Cooper et al. (2006) the DEA obtained its name from the way it 'envelops' data to discover a 'frontier' which can be used to assess observations representing the performances of all units that belong to the benchmarking analysis.

This 'frontier' is described by Ozcan (2008) as 'best-practice frontier' as this border line illustrates best units' practice pattern.

Table 4.1 Classification of frontier methods

illustration not visible in this excerpt

(Bogetoft & Otto, 2011:18)

Table 4.1 shows a classification of frontier methods, divided into deterministic/ stochastic and parametric/non-parametric characteristics. Therefore, the DEA can be identified as a deterministic and non-parametric method, invented in the late seventies. This means for the non-parametric classification, that the model structure is not defined a priori and it may be formed by its particular data sets and for the deterministic classification, that individual observations may be not affected by random noise (Bogetoft & Otto, 2011).

However, Bogetoft and Otto (2011) emphasise, that for each of the given methods, like DEA, exist a large number of types according to various assumptions, e.g. about the production technology. Therefore, the DEA model can be classified into different envelopment models illustrated in Figure 4.1.

Figure 4.1 DEA model classifications

illustration not visible in this excerpt

(Ozcan, 2008:24)

The DEA model can mainly be distinguished based on the input or output orientation. Further, Ozcan (2008) differentiates two types of DEA models1: Constant Return-to- Scale (CRS) and Variable Return-to-Sale (VRS). The CRS type assumes constant changes in the relation of input and output factors, e.g. higher input leads to higher output and vice versa. In contrast the VRS type considers variable changes in the relation of input and output factors, e.g. higher input can lead to lower output and vice versa. Figure 4.2 illustrates these two technologies. The CRS frontier on the left side represents a line through origin and entity A and B can be assumed to be efficient as they lay on this frontier.

Figure 4.2 DEA technologies

illustration not visible in this excerpt

(Bogetoft & Otto, 2011:87)

The VRS frontier reveals entities A, B and C to be efficient as they build the best practice frontier. Comparing these two charts of CRS and VRS shows that the CSR chart suppresses entity C and paints a 'bigger picture' of the efficiency frontier. Cooper et al. (2006) argue, that the CRS technology, although technically correct, is misleading as returns-to-scale can vary. Concluding, Ozcan suggests to use the CRS model "If one can assume that scale of economies do not change as size of the service facility increases" and to use VRS "If one cannot assume that scale of economies do not change as size of the service facility increases" (Ozcan, 2008:23). According to this, the choice of the appropriate DEA model depends on the object of investigation and the specifics of its entities.

These different DEA models enable the use of DEA in a wider set of units, including business companies as well as government and non-profit organisations like schools, hospitals or military units (Cooper et al., 2006). Bogetoft and Otto (2011) state that these different units are typically called Decision-Making Units (DMUs) in the DEA literature.

Cooper et al. (2006) add, that a DMU is regarded as the unit responsible for turning inputs into outputs and whose performances are to be evaluated.

4.2 Mathematical Fundamentals

The mathematical fundamentals of DEA refer to linear programming, which has its foundation in the operations research literature (Ruggiero, 2011). Further, the analysis of performance is based on the production theory, meaning the transformation of input factors to output factors (Bogetoft & Otto, 2011). Moreover, Bogetoft and Otto (2011) add, that each DMU has a common 'underlying technology' (T), which can be described as follows:

illustration not visible in this excerpt

This setting involves K DMUs, which use m inputs to create n outputs. Whereat [illustration not visible in this excerpt] describes the inputs used and [illustration not visible in this excerpt] specify the outputs by DMU (k, k=1,...,K). The given definitions of inputs and outputs lead to the assumption, that the DEA allows multiple inputs and outputs. In connection, the notation [illustration not visible in this excerpt] in (4.1) denote that inputs and outputs have to be non-negative. However, Zhu and Cook (2007) introduce the translation invariance property of DEA models, which 'translates' negative input/output into non-negative data. One way to apply this approach is to add the value of the smallest input/output variable to all input/output variables of the same type. Therefore, "The obtained results are exactly the same as if the original data set were analyzed" (Zhu & Cook, 2007:67).

The DEA for a particular observation refers to the following linear programming formula (4.2), in which the 'underlying technology' (4.1) of every DMU will be processed to determine an efficiency frontier by identifying 'best performance' DMUs (Cooper et al., 2007).

illustration not visible in this excerpt

and accordingly

illustration not visible in this excerpt

The individual input (ݔ௞ ሻ and output ሺݕ௞ ሻ factors can be weighted to specify the analysis (Ozcan, 2008). However, this may be appropriate if the appearance of certain input factors seem similar, e.g. Headcount of nurses and doctors as input factors. In this case one might give doctors a higher weight than nurses to reflect the input of higher quality.

Moreover, DEA researchers suggest a 'rule of thumb' concerning the number of DMUs

(K) and the amount of inputs (ݔ௠ሻ and outputs (ݕ௡ሻ (Bogetoft & Otto, 2011).

illustration not visible in this excerpt

Therefore, the number of DMUs should exceed three times the number of inputs and outputs as well as the product of the number of inputs and outputs. Bogetoft and Otto (2011) continue by saying that Investigations, which include too many input and output factors compared to the number of DMUs will tend to make many firms efficient. Consequently, the analysis will lose its capability to differentiate best performers from the rest.

4.3 Performance measurement in healthcare

The performance measurement of an individual DMU should reveal strengths and weaknesses of its transformation process, in which inputs are used to create certain outputs. Ozcan (2008) defines performance as an interaction between effectiveness and efficiency (see Figure 4.3). Further, Ozcan (2009) emphasises that performance measurement in healthcare has to be distinguished from other manufacturing industries.

Figure 4.3 Components of performance

illustration not visible in this excerpt

Adapted from Ozcan (2008:4)

Moreover, "The relationship between efficiency and quality of care had mixed results in prior studies" (Ozcan, 2008:5). In this context quality of care relates to the effectiveness of medical treatments. While the view of efficiency concentrates on the input factors, with the aim to minimize inputs for a certain given number of outputs, effectiveness refers to the evaluation of output factors and whether the inputs are being used to produce the best possible outputs (Ozcan, 2008). Accordingly, "A hospital can be efficient, but not effective; it can also be effective, but not efficient. The aim is to be both” (Ozcan, 2008:4). In this report the expression of best performance hospitals refer to the most efficient operating hospitals, as indications about the effectiveness of hospital health supply are not covered within this research approach.

In a healthcare context the characteristics of each patient with individual disease pattern, different response to medical treatments, and the ability for regeneration is vitally important as it directly affects the medical outcome.

According to this, the performance measurement e.g. based on 'successful' treatments with a positive outcome is relative as it is highly dependent on the patient's condition at the moment of the medical treatment and its characteristics in general. Barros (2003) points out that hospital treatments appear to be ineffective if patients' characteristics show low chances of survival. However, considering "Hospital output as a change in the distribution of survival probabilities" (Barros, 2003:226) enables the separation of hospital performance from patients' characteristics.

Figure 4.4 illustrates the systems view of an DMU, or in other words the 'underlying technology' introduced in (4.1).

Figure 4.4 Systems view of a DMU

illustration not visible in this excerpt

Adapted from Bogetoft and Otto (2011:14)

Bogetoft and Otto (2011) accentuate that the transformation of inputs to outputs is affected by the management and other exogenous factors within every particular DMU. Therefore, in addition to the inputs controllable variables (e.g. employment of labour) and uncontrollable variables (e.g. population density) determine the outputs. It is to assume, that the individual patients' characteristics, will take an impact on the output quality. In this context, these patients' characteristics may be described as exogenous factors, which are not controllable by the management. Moreover, Ozcan (2009) emphasises the intangible nature of healthcare outputs by stating that healthcare treatments cannot be tested before and the evaluation of certain treatments is highly dependent on patients' conditions (e.g. healing processes). Therefore, the difficulties in measurement and evaluation of healthcare outputs stems from its challenging nature.

4 Data Envelopment Analysis Page | 20

Maniadakis et al. (2009) add that it is impossible to calculate the possible maximum output of a hospital nor the necessary minimum amount of inputs to achieve certain outputs. This, once more, underlines the reasonableness of an DEA approach within a healthcare context, as a frontier of 'best-practice' hospitals will identify efficiency in a particular investigation. Further, the identification of the sources of inefficiency of DMUs can be provided as the benchmarking reveals the best configuration of input to output factors (Cooper et al., 2007). As shown above, the measurement and evaluation of effectiveness is challenging within a healthcare context. In comparison, the measurement of efficiency seems to be less challenging.

Ozcan (2008) introduces four different efficiency concepts in DEA: Technical, Scale, Price, and Allocative Efficiency. According to this, the technical concept defines the best achievable efficiency of a DMU and allows measurements against this best achievable efficiency in different reporting periods (Ozcan, 2008). In contrast, scale efficiency describes a DMU that operates at the optimal size and therefore reached the point of long-term equilibrium (Maniadakis et al., 2009). Moreover, price efficiency processes price or cost information within the efficiency evaluation (Ozcan, 2008). Closing, allocative efficiency identifies input cost minimisation in order to achieve a given output (Maniadakis et al., 2009). Ozcan (2008) adds that health managers use the allocative efficiency to calculate the appropriate mix of inputs. Consequently, technical efficiency is not feasible within a healthcare context, as the total amount of possible outputs cannot be defined (Maniadakis et al., 2009).

Now that the theoretical background of the DEA approach is set and the use of this technique within a healthcare context introduced, the next chapter explains the selection of appropriate input and output factors in order to pursue the efficiency analysis of English and German hospitals.

5 Data collection and explanation

This chapter gives an overview of the collection of basic data for executing the DEA within this report. In detail, the selection and characteristics of the different DMUs will be introduced in 5.1 Hospital funding bodies. Moreover, chapter 5.2 provides an identification of the measurement categories and the main input and output factors for this investigation will be defined in the following chapter 5.2.1 Input data and 5.2.2 Output data.

5.1 Hospital funding bodies

The different types of hospital funding bodies describe different 'levels of state- involvement', in which the degree of state-involvement decreases from public to private run hospitals (Rothgang et al., 2010). Consequently, the type of hospital funding body relates to the exogenous factors in Figure 4.4, as state-involvement represents a non- discretionary factor. In total 140 NHS run entities in England and 100 hospitals from Germany describe the individual DMUs within this report. The NHS provides in total 230 acute trust hospitals in England (NHS England, 2014). However the majority of these hospitals already gained a Foundation Trust status, which sets them free from central government control. Further, NHS Foundation Trust hospitals represent non- profit, public benefit corporations and own greater freedom to decide on their own strategy (Department of Health, 2014). In addition to this, the Health and Social Care Act 2012 was introduced to give every NHS Trust hospital the Foundation Trust status (Department of Health, 2012). Barros (2003) adds that the change from public run hospitals to private run hospitals can be described as a 'natural reform measure'. This report covers 60 NHS Trust and 80 NHS Foundation Trust hospitals. The German hospital funding bodies can be differentiated into public-law, independent non-profit, and private organisational forms (Federal Statistical Office, 2014a). In 2012 the total distribution of 1,930 registered German hospitals into these categories equates to 30% (public-law hospitals), 36% (independent non-profit hospitals), and 35% (private hospitals), however, the number of private hospital beds remained growing constantly (Federal Statistical Office, 2013). The German data set within this report consists of 46 public-law, 31 independent non-profit, and 23 private hospitals.

Although the English data set does not cover private run hospitals, the 23 private run hospitals from the German data set will be included in this investigation with the aim to discover deviances in efficiency measurement of these DMUs, which may be explained by the effect of exogenous factors.

5.2 Identification of the measurement categories

It is reasonable that the selection of appropriate basic data can be seen as the backbone for every comprehensive investigation most notably in the context of benchmarking studies, with the aim to set up a reliable best practice frontier. Hence, special attention was drawn to select comparable data of English and German hospitals. In this context, Ozcan (2008) emphasises that the choice of appropriate data is difficult within a healthcare environment, due to the nature of the services provided. Furthermore, in this connection Nayar et al. (2013) add that quality is an 'intangible output'. According to this, the measurement of quality within a healthcare context can be seen as very challenging.

The first research results have shown that the year 2012 would provide the best data quality. Mainly due to the fact that the majority of data from German hospitals were provided in the latest issue of hospital quality reports, which cover the year 2012 (Federal Statistical Office, 2014a). In total six factors were identified as basic data for every DMU to carry out this analysis (see Figure 5.1).

Figure 5.1 Overview of input and output data.

illustration not visible in this excerpt

Recent research articles in which DEA was applied show the use of size, staff, and demographic environment as measured variables (Nayar et al., 2013).

Further, Cook et al. (2014) covered wards and admissions as indicators and Clement et al. (2008) examined the labour input of nurses.

Further back, Ozcan and Luke (1993) already analysed the efficiency of hospitals in urban markets. Although the DEA approach allows the use of multiple output factors, Hartz and Kuhn (1994) emphasise that hospital comparisons are limited to one outcome as the performance of several outcomes is not consistent. Thus, different measured outcomes refer to different characteristics in health supply (Hartz & Kuhn, 1994).

Moreover, Rosenthal (1997) adds that mortality rates are extensively used to assess hospital output. Consequently, this proves that the chosen basic data (input and output factors) for this investigation are appropriate. Moreover, the 'rule of thumb' introduced in (4.4) holds, with KENGLAND=140; KGERMANY=100;Input factors xENGLAND/GERMANY=5; Output factors yENGLAND/GERMANY=1:

KENGLAND>3(xENGLAND+yENGLAND)→140>18 KENGLAND> xENGLAND·yENGLAND)→140>5 KGERMANY>3(xGERMANY+yGERMANY)→100>18

KGERMANY> xGERMANY·yGERMANY)→100>5

5.2.1 Input Data

The size of a hospital is specified by the number of beds. Therefore, an increase of hospital beds can be associated with higher input. According to NHS England (2014), the average rate of beds per hospital amounts to 738 (103,357beds/140hospitals≈738). In contrast, this average rate in Germany equates to 593 (59,272beds/100 hospitals≈593) and demonstrates a significant difference between the English and German hospital structure (Federal Statistical Office, 2014a).

Number of Wards

The number of wards expresses the hospital diversity and reflects its medical supply and treatment options. It can be assumed that in connection to a higher number of existing wards the hospital infrastructure and its environment gains complexity.

Moreover, the increasing amount of processes and scheduled activities e.g. allocation of operating rooms, may lead to a higher degree of difficulty in hospital management.

The Office for National Statistics (2012a) reports that in English hospitals up to 75 different wards exist. In contrast, the German hospital diversification in terms of its different wards equates to only 41 (Federal Statistical Office, 2014a). This major difference may exist due to a deeper breakdown of wards in England. Therefore, it is necessary to build a matching table in which a particular number of wards combines the different expressions of English and German wards to find a common ground (see Appendix 1). The sample data set reveals that the average rate of wards per hospital in England amounts to 17 (2,433wards/140hospitals≈17) and in Germany to 11 (1,116wards/100hospitals≈11) based on the matching table.

Medical specialists

The input of medical specialists refers to the expression of labour input. Moreover, medical specialists can be distinguished from other hospital staff e.g. nurses, as medical specialists completed a specified training and show high expertise in particular wards. It can be assumed that the employment of medical specialists improves the outcome of medical treatments. Employment data about medical specialists in England are provided by the Health & Social Care Information Centre (2013a). In contrast, comparable German data had to be requested at the Federal Joint Committee (2014). Moreover, the data has to be filtered out of the individual hospital reports, which made the research process even more complex. The sample data set shows an average rate of 35 (4,910med.sp./140hospitals≈35) medical specialists per hospital in England and a comparable average rate of 116 (11,564med.sp./100hospitals≈116) in Germany.

The amount of admissions per hospital represents the major input factor of this investigation as the individual patient treatment of every hospital expresses basic data for further analysis and consequently enables performance measurement. The type of admissions can be differentiated into inpatient and outpatient admissions (Federal Joint Committee, 2014). Within this investigation the amount of inpatient admissions is covered. Moreover, this input factor is directly linked to the mortality rates (output factor) per hospital, as the appropriate mix of treatments per admission is used to avoid death.

[...]


1 Bogetoft and Otto (2011) introduce a further range of DEA models. However, for this investigation the CRS and VRS model is sufficient. For further reading see Bogetoft & Otto (2011).

Excerpt out of 85 pages

Details

Title
Hospital efficiency analysis in England and Germany. What lessons can be learned from each other?
Subtitle
Utilising Data Envelopment Analysis
College
University of Hertfordshire
Grade
80
Author
Year
2015
Pages
85
Catalog Number
V317821
ISBN (eBook)
9783668169012
ISBN (Book)
9783668169029
File size
1571 KB
Language
English
Keywords
hospital efficiency, Data Envelopment Analysis, healthcare systems, NHS
Quote paper
Lucas Fandrey (Author), 2015, Hospital efficiency analysis in England and Germany. What lessons can be learned from each other?, Munich, GRIN Verlag, https://www.grin.com/document/317821

Comments

  • No comments yet.
Look inside the ebook
Title: Hospital efficiency analysis in England and Germany. What lessons can be learned from each other?



Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free