Efficiency Evaluation of Singapore Hospitals

1. Introduction
2. Literature Review
3. Methodology
4. Expected Contribution to Hospital Management

Introduction

The purpose of this research project is to study the operation efficiency in Singapore Hospital industry during the time period from 1989 to 1998 . Both longitudinal and latitudinal analysis of operation efficiency in individual hospital will be employed to identify the changing pattern of Singapore hospital industry as a whole as well as individual. It is also possible to gain some insight into hidden mechanism accounting for the different inefficiency pattern across individual hospital. Sophisticated examination of such result could be able to contribute to our knowledge about Singapore hospital operations and help us to figure out corresponding measures dealing with specific inefficient operations in an effective fashion.

Giant amount of money was kept on being spending on national health care expenditure and was also in increasing trend, amounting to $4 billion dollars in 1997. Health manpower in Singapore constitutes a large workforce. Efficiency in health service delivery should be paid great concern. The possibility of improvement in efficiency of health service delivery is an attractive issue because it could facilitate our more efficient utilization of existing resource and making more output, insight into which problem could help the effort of the Ministry to build a healthy nation in an efficient way.

This research focuses on hospital industry operation due to the fact that hospitals play a vital role in the Singaporean health care system as the key channel of health service delivery. The Service program (Public hospital and institution) consumes the largest part of the Ministry of Health's operating Budget, 76% in 1997. Consequently we have good reason to place our major concerns on the operation efficiency in Singapore Hospital Industry, in an effort to shed light on possibility of improving the health service. Thorough knowledge of hospital operation will certainly add to our knowledge about the whole health care system, strengthen the information basis on which relevant health policy is made. Furthermore, the emphasis of research examination will be laid on the public and restructured hospitals. These two types of hospitals, providing 75% of inpatient hospital services in 1998, are the major components of the hospital sector, the dominance of which justified our perspectives. There are six government and restructured hospitals providing acute general inpatient and specialist outpatient services and 24-hour accident & emergency (A&E) services before 1997: Alexandra Hospital (AH), Changi Hospital (CH), National University Hospital (NUH), Singapore General Hospital (SGH), Tan Tock Seng Hospital (TTSH), Toa Payoh (TPH); after 1997, this set includes: Alexandra Hospital (AH), Changi General Hospital (CGH), National University Hospital (NUH), Singapore General Hospital (SGH), Tan Tock Seng Hospital (TTSH). These hospitals comprise the sample set for this research.

Productive efficiency is composed of two major components: technical efficiency and allocative efficiency. Technical efficiency is concerned about (1) possible increase in output for a given set of inputs or (2) possible decrease in inputs to produce a given set of outputs; while allocative efficiency is concerned about (1) the optimal mix of inputs for observed outputs or (2) the optimal mix of outputs for observed inputs. Price information is necessary for discussion about the allocative efficiency. However, in the context of hospital research, price information is typically distorted by such factors as social concern and government regulation, rather than a direct projection of marginal cost. It is also believable that, hospitals have more profound and diversified objectives than profit-oriented firms do, which is pursuing cost-minimization. Technical efficiency focuses on the process of transmission from inputs to outputs. Physical inputs and outputs information are often available. Thus, this research will concentrate on technical efficiency, which is free from the need of price information.

Since Debreu (1951) and Farrell (1957) defined technical efficiency basically, various approaches have been developed to measure the technical efficiency in production. DEA (Data Envelopment Analysis) is employed in this research to measure the efficiency score of observed hospitals. DEA, as an effective tool in efficiency evaluation, has been applied in nonprofit sector such as government, army and hospitals. With such application, efficiency in nonprofit service involved with multiple output and input can be assessed in a technical term, where profit indicator alone cannot account for the whole story. Significance beyond the cost-revenue relationship can be recognized through analysis. Statistical analysis will be conducted to see through the efficiency score to identify the relationship between efficiency score and other factors that could be associated with operation efficiency.

Literature Review

Measurement and analysis on hospital efficiency, with increasing popularity, has been included in research practice for a long time. The significance of such research has been recognized through various important researches in this area. In literature on health-care management, DEA method has been used extensively and intensively to analyze efficiency in hospital industry.

Rolf Fare et al (1992) employed a Malmquist output-based productivity index to examine the productivity developments in Swedish hospitals. The operation data across a 15-year-long period was put into the analysis. The malmquist index bears a close relationship to the CCR output-oriented DEA model, allowing the authors to identify the changes in efficiency as well as in technology. Both regress and progress were found in the changing patterns. Considerable variation across individual hospitals in their sample was found. The authors gave corresponding measures against the specific efficiency types in various areas such as capital utilization, personnel deployment and so on. Patricia Byrnes et al (1992) gauged the performance of U.S. hospitals. They employed linear programming model to compute overall cost minimizing efficiency and its components, technical efficiency and allocative efficiency. Input-oriented CCR DEA scores were decomposed into different inefficiency sources to show their possible cause in operations. The efficiency measures based on DEA model and Farrel classification were validated by comparing the measurement with more traditional measurement of cost performance. Based on their result, hospital manager could identify their operation benchmark and determine their own competence compared with their competitors. Possible improvement was indicated in the result.

Banker et al (1986) studied 114 North Carolina hospitals using DEA method. Their research demonstrated the possibilities of returns to scale in individual hospitals. Prior regression-based studies of the same populations reached the conclusion that no returns to scale were present. The findings in this perspective can provide convincing argument for or against expansion, merger or split of individual hospital.

Jon A. Chilingerian et al (1990) explored why some physicians hospital practices were more efficient. The author used DEA method inside hospitals to measure the highly complex and divergent activities involved in the provision of clinical services in hospitals. This research, as a pioneering study of physician efficiency inside hospitals, showed that DEA could be adapted to address many problems associated with measuring physician efficiency. It was found that the physician who used resources more efficiently did so with equivalent case-mix complexity and severity, and average patient age. Key factors associated with physician efficiency were demonstrated in the interpretation of DEA measurement. James F. Burgess et al. (1996) compared the technical efficiency achieved under different ownership structures in the US hospital industry. Their research, in an industrial organization perspective, could provide insight on hospital industry restructure to achieve more efficient operations. Paul W. Wilson et al (1998) analyzed the variation across US hospital inefficiency, which could provide guideline for possible improvement. Another interesting research was conducted by Eulalia Dalmau-Matarrodona et al (1998). They related the DEA inefficiency scores of hospitals with health care market structures in a Spanish circumstance. They concluded competition would contribute to better efficiency

His-Hui Chang (1998) employs Data Envelopment Analysis to evaluate the efficiency of central government-owned hospitals in Taiwan. Inputs and outputs in the author's model were selected according to the specific characteristics of hospital operations. Efficiency score is estimated with DEA measurement. Multiple regression statistical model is put forward to discern the possible impact of other factors on hospital operations efficiency. DEA score serves as the dependent variable. Various factors are proposed to explain the score. The independent variables are chosen to capture hospital operation characteristics. The independent variables include: service complexity, occupancy rate, proportion of retired veterans, and anticipatory impact of National Health Insurance (NHI) program/time of study. This paper remind the readers that the score itself is not the whole story. The author makes an effort to explore the determining factors that is hidden behind the operating phenomenon. Moreover, local context characteristics are incorporated successfully into consideration, which makes the research practical local management and foreign researcher and enlightening for researchers who are going to conduct such applied research.

Methodology

Data Envelopment Analysis

Efficiency measurement has long been a tough task for management to undertake. The problem became more intriguing but vexing in those sectors whose concerns go beyond profit maximization such as government, health care, social service and defense. Multiple outputs are desired and expected from the consumption of multiple inputs. Some of the outputs and inputs cannot be measured in monetary dimension. The objectives of the trade cannot be reduced to simple revenue-cost tradeoff. There is, unfortunately, no single overriding objective to take the place of the profit motive, which presides private sector. Some way is in demand to accommodate the multiple measures to give comprehensive assessment of operation efficiency.

Considerable research has been expended in the development of more sophisticated tools of evaluating the efficiency of a unit in relation to the other units in its grouping. The breakthrough came in the research work by Charnes, Cooper and Rhodes(CCR). The CCR research reported in 1978 is the basis for all subsequent developments in the non-parametric approach to evaluating technical efficiency. In a later paper, Charnes and Cooper gave their formal definition of efficiency as follows,

"100% efficiency is attained for a unit only when;

(a) None of its outputs can be increased without either (i) increasing one or more of its inputs, or (ii) decreasing some of its other outputs; (b) None of its inputs can be decreased without either (i) decreasing some of its outputs, or (ii) increasing some of its other inputs."

CCR used the generic term "Decision Making Unit (DMU)" to designate the collection of firms, departments, divisions or administrative units which have common inputs and outputs and which are being assessed for efficiency. The term encompasses organizational units in both the public and private sectors. The term Data Envelopment Analysis has been used since the CCR paper describe their approach to efficiency evaluation. The emphasis was put on decision making by nonprofit entities by CCR. This meant they could concentrate on multifactorial need of problems and could circumvent the necessity of converting various factors into common monetary basis. It can incorporate multiple inputs and multiple outputs into both the numerator and the denominator of the efficiency ratio. Thus the DEA measure of efficiency explicitly accounts for the mix of inputs and outputs and consequently is more comprehensive and reliable than a set of operating ratios or profit measures. The DEA method makes possible and feasible measuring the efficiency of nonprofit pursuit.

DEA is a linear programming model that attempts to maximize a service unit's efficiency, expressed as a ratio outputs to inputs, by comparing a particular unit's efficiency with the performance of a group of similar service units delivering the same service. In the process, some units achieve 100 percent efficiency and are referred to as the "relatively efficient units", while other units with efficiency ratings of less than 100 percent are considered inefficient units. A frontier of efficiency can be discerned, which can constitute the basis for benchmarking for operation. For each inefficient DMU, those that lie below the frontier, DEA identifies the sources and level of inefficiency for each of the inputs and outputs. The efficient DMUs constitute the basis of benchmarking. The level of ineffciency is determined by comparison to a single referent DMU or a convex combination of other referent DMUs located on the efficient frontier that utilize the same level of inputs and produce higher level of outputs, or utilize lower level of inputs and produce the same level of outputs. Thus each inefficient DMU is compared with its potential. Consequently the decision maker is informed of the possible scope of improvement in operation.

The DEA method does not require explicit specification of the form of production function that relates inputs to outputs. Multiple inputs and outputs are accommodated in DEA without artificially assigning weights to various outputs and inputs. Conversely, the algorithm give internal evaluation of outputs and inputs for each DMUs. For a particular DMUs, based on the data, those outputs or inputs that are utilized or produced more productively are given high weights and others are given low weights. Consequently, the algorithm takes into account each DMU' peculiar endowment and capability.

DEA can serve as a new way of analyzing and organizing data. Both managerial and theoretical insight can be gained through the application. It should be noted that DEA calculations

1. focus on individual observations in contrast to population averages. 2. Produce a single aggregate measure for each DMU in terms of its utilization of input factors (independent variable) to produce desired outputs (dependent variables); 3. Can simultaneously utilize multiple outputs and multiple inputs with each being stated in different units of measurement; 4. Can adjust for exogenous variables; 5. Can incorporate categorical (dummy) variable 6. are value free and do not require specification or knowledge of a priori weights or price for the inputs and outputs. 7. Place no restriction on the functional form of the production relationship; 8. Can accommodate judgement when desired; 9. Produce specific estimates for desired changes in inputs and/or outputs for projecting DMUs below the efficient frontier onto the efficient frontier; 10. Are Pareto optimal; 11. Focus on revealed best-practice frontiers rather than on central-tendency propeties of frontiers; and 12. Satisfy strict equity criteria in the relative evaluation of each DMU

Efficiency Measurement

The measurement of hospital operation efficiency is supposed to be conducted in a Data Envelopment Analysis framework. Input-based or Output-based CCR DEA model can be applied in this case. The former model evaluates minimal input usage with constant outputs, whereas the latter evaluates maximal outputs production with constant inputs. However, the demand for medical service is not highly elastic. Thus the hospital has more control on inputs than outputs. Inputs-based model should have more weights than output-based model. But both models will be employed to ensure the robustness of this research.

Input and output variable

This research defines four separate proxies for hospital outputs: outpatient specialist clinic attendance, A&E attendance, day surgery and short-term inpatient care. Service complexity is an important concern. However, because the sampled hospitals have the same scope of service, it is reasonable to assume the compositions of the service they provided are roughly homogeneous. In past literature, weighted inpatient days are extensively used to account for the different characteristics of different short-term inpatient care specialty. This method will be considered. However, the specification of weight associated with each types of care is determined subjectively, which might induce doubt on its scientific rigor. The separation between A&E attendance and specialist attendance is reasonable because A&E service is on a general basis unlike specialist service.

Input is defined by five measures, the full time equivalent employment of doctors (including dental officers): the full time equivalent employment of nurses, the full time equivalent employment of paramedical/pharmacist staff, full time equivalent employment of administrative/clerical staff, and the number of staffed bed. The operating capital is not included because it could be distorted by various factors such as accounting practice, inflation and so on. The number of staffed bed can partly reflect the size and scale of a hospital.

Window Analysis

DEA Window analysis will be employed in this research as well as normal DEA measurement. In window analysis, the efficiency is measured on a rolling basis. The operation data is grouped in a three-year period. Efficient frontier is constructed for each window, which can be presented graphically to render intuitive recognition of progress in operation efficiency in the whole industry over time. The movement of frontier indicates either progress or retrogression. A trend of efficiency change can be discerned.

DEA result analysis

Efficiency measurement itself is far away from the whole picture we want to visualize through the research. Both qualitative reasoning and quantitative analysis will be employed to help understanding the scores we extract from DEA calculation. Significant change in efficiency of both industry and individual hospital will be given attention to find out the reason. The effect of health policy, hospital restructure and important events such as Asian Financial Crisis can be evaluated against DEA result.

Statistical analyses conducted in this research.

Technology progress has been playing increasingly important role in modern society. It is beneficial to recognize the relationship between IT investment and operation efficiency. It is also hypothesized that high-technology medical equipment, high-tech non-medical supporting equipment can lead to higher efficiency. Staff training is also important for good medical service. Multiple regression models will be constructed between efficiency score and IT investment, high-technology equipment investment and staff training expenditure. To accommodate the difference in scale across hospitals and inflation, the annual expenditures in IT, high-tech equipment facility and training are expressed in the form of the proportion of the two items to the whole annual operating expenditure of each hospital. Because both technology investment and staff training can affect hospital operation in long run, this research is going to test the existence of relationship between the efficiency score and the IT and training expenditures of the preceding year. It should be noted that, it is unreasonable and lopsided to evaluate the benefit of such effort for operating unit in on purely financial basis, the knowledge on the relation between operation efficiency and such effort might be a good attempt. Quality control is a popular and important issue in hospital management. Like staff training effort, QC effort cannot be evaluated based on only financial return. It is also valuable to identify the possible contribution QC effort could make to operation efficiency. QC participation rate in hospital (the ratio of employees involved in QC to the total employment) and employee average QC training hours can both be used to measure QC effort.

It will be beneficial for the hospital industry to examine the role played by staff morale in operation. Morale is a complex concept reflecting the quality of the organization culture, working atmosphere and leadership. It is hypothesized that the condition of morale is positively related to efficiency score. However, it is difficult to operationalize this construct. The employee turnover rate is employed as a proxy measure. This research is also trying to examine the possible relationship between efficiency and employee stability. The ratio of newly recruited employees to the total employment is used as a proxy variable for employee stability. This research is the first case in DEA application to incorporate concern in human resource aspects.

Expected Contribution to Hospital Management

This research is expected to help hospital management in the following aspects.

Knowledge of the cause in hospital operation efficiency can be obtained through the analysis. Recommendation can be made based on such information. For example, underutilization of specific resources can be identified from the results. Possible improvement in operation can be indicated. Hospital benchmarking information can be available from such research result to show the competence of specific hospitals.

Changing pattern of hospital inefficiency over time can be found to obtain knowledge of the history of hospital industry in Singapore. Such information will be helpful to assess the effectiveness of both national health policy and individual hospital management decision in a long-term period, which can add to experience on health care management for future decision making.